CURSO DE ACTUALIZACIÓN EN LITIGACIÓN ORAL CIVIL Y REFORMA PROCESAL

by identifying it, identifying it, the prompts, identifying the data that I carried to the artificial intelligence, the artificial intelligence tells me that that test is wrong, that the speed is not what the expert says and other details of the accident.

Could it be that this is the other way around, that artificial intelligence serves to question a human pronunciation, specifically technical? That's my question. Yes, the answer is affirmative and right here it will be

the work of the judges when they get these challenges, how they are going to interpret it and I don't want to leave it aside and I repeat again and sorry for being reiterative, but it is the reality that is going to haunt us today and that is going to come upon us from August 1st, is human control. You can do it, you can incorporate this impugnation with those elements, but still you will have to incorporate a expert who will corroborate and

and verify this incorporation generated with the guide.

Yes, perfect, just like that. 208 is the one that gives us the guidelines, that says that for itself it has no probatory value of a test. It serves as an aid, it serves to give clarity to things, but the technical operations, because it is still a technical operation of a specific scientific knowledge, it cannot be, let's say,

completely alien to the realization of a human person. So, that kind of impugnation corroborated by a expert explaining the technical operations, which is surely artificial intelligence,

I always like to clarify how it was done before the IA. It's like we always forget. You say, hey, I have to push.

And we became calligraphic experts, we became experts, and then it is also detected that this pattern has not been followed, such a thing. And in reality, we are worth what an external specialist said or what was ever interpreted in a trial by jurisprudence or doctrine. But we already, I remember having a judge at that time who took you out, passed you all the time saying you would have hired a part expert. You had a expertise that was...

four rows where it was without any kind of scientific foundation, what the doctor says, it was 20 kilometers per hour. And of course, the lawyers armed 20 sheets saying, look, it has not been this scientific term that has been said, you are a specialist expert in this or lawyer. I contacted the judge and took him out to walk saying, I would have put a expert. Here I think that that is solved. Here you are putting with specialized systems, I think that there is back to understand that. They will find themselves at some point, hopefully soon,

As well as we calculate the interests in the bank's calculator, nation's calculator, the lawyer's school's calculator and we trust that calculator that the numbers give the exact interest or some of you verify it manually. Well, here the same thing will happen with specialized calculators for each of the professional schools. At some point, and that's where a little of what has been said, the doctor says, tell me that that generation is a specialist for what you are saying.

Why? Because it's not about uploading this to GPT and saying if it's real or not. It's 20 kilometers or 70. There is a scientific-technical process that, just like the other leg, usually has to be valid by a human. And put it to pass in those trials where all the lawyers have been disobeyed the process, they have changed 200 times of judge. Have you seen those disobeyed processes where there is almost no evidence? Well, the only thing I have is a very specialized algorithm that has, if you ask me, and with all humility,

more or less loaded more than 5,000 human brains at the processing level, has understood this about this situation and should have some initial weight, but does not comply with everything else. I say because imagine all the possible processes that are going to happen. That is, people who are only going to incorporate things with artificial intelligence and I say no, I call a expert, I don't spend. People who are going to start creating their own algorithms. I was going to say, hey, I did, for example, I made one, I'm going to show you later, of voice tests.

Entonces, yo me mando un audio Juan ahora, por ejemplo, y después tengo algo más de Juan, y eso, condiciéndose con todos los estándares científicos internacionales, utiliza un montón de programas, porque realmente esto ya existía por parte de los peritos, y cruza todas las voces, me da un informe especializado. Yo no soy perito, ni perito informático para llamarme eso, ni fonaudiólogo.

But I already enter the process thanks to an algorithm as I have trained it, saying hey, in case Juan comes and denies me that he is his voice, I have already processed this information by specialized algorithms locally. And what is a judge going to say with that? Hey, well, let's check with a professional, right? But look at how you narrow the margin using tools that increase your capabilities. Yes, sorry, doctor, I'm going to repeat the question. Oh, well, there it goes, there it goes. It's that the most it runs, Luciano.

Yes, good morning. In the case that the doctor raised here, specifically with the subject of the examinations, once the transfer of the examinations was carried out, in this case we would be using the tool for the objection, that is, for the examinations, which ultimately

When we founded that impugnation, supported by an IAEA conclusion,

does not have what we are supporting, it does not have a probatory nature, that is, it simply serves us as a tool to object to the expert and tell him, look, for this reason your expert is defective in this or that, and there comes the procedural part in which the expert has to ask for the explanations and that is already

In short, once the impugnation has been answered, it is up to the judge. The judge does not have to go in to evaluate what the IA says or what the IA does not say. He will simply have to evaluate whether the judge will have to evaluate the answer to the expert's impugnation. The judge will draw the conclusion in terms of whether the trial is definitively clear or not, whether it helps or not.

to reveal the facts that are in the case, right?

I think that's where it ends, in this particular case that my colleague was planning, I think that's where the use of the IA ends. It's not that we're not going to object to the foundation, that is, what was founded in the IA, that we are going to have to attribute to what I founded in my impugnation with IA. No, but imagine now that the judge is also going to use that, you know?

Now, we are talking about someone who is going to come, who is going to push something with IA, that was done with IA and that the judge cannot use the same thing to generate an explanation that the parties have not explained. Well, but if the expert also did his expertise with IA, then...

We are like that. You have no doubt and we are advancing that precisely. Well, in fact, if you ask me at some point, if I already gave you a very good technical report, you think that the technical points are key and how many times they are offered. Let's be honest, I mean, tell me if this WhatsApp message is true and well, the expert does his magic.

Now, behind the instruction that you are really giving, is the person who sends it or not, is this number or not, is this metadata, there is much more behind. So the expert has to do a lot of work to interpret legally what that person wanted in that process. Now this is going to give you something much more organized because you should have the opportunity to speak the language of the expert and that he

like a valid checklist. Yes, you're right. Look, no, they're not like that here, brother, because in the context that you're talking about, this one is not understanding what you've put me here. I don't think it's like that. Why? Because this is the human context opportunity. For that, the right and for that we interpret things that are not like that. However, I part the same as you. Yes, they're all going to do it with her, but we don't advance that situation. Why? Because if an attorney does it, it won't be the judge. If the judge does it, it won't be the expert.

The witness comes prepared with an idea, I mean, yes, yes, not only that, but the code validates us in all this, such as even the judges, everyone. So it's 100% allowed. We're going to have more cases anyway. That's how we debate it. Let's not forget that the key will always be in human control. That is fundamental. It is the guide that they have to have.

Now we go to part 210, in terms of the assistance of the editorial office. There is something that catches my attention and I am going to take advantage of the fact that Marcos is here to ask him what he meant. In part 210, he tells us: "lawyers, etc., can use generative data to write, keeping in indelegable responsibility for the fidelity of data.

the legal correction of the arguments and the exact quote of the sources. My doubt arises from the fidelity of data. What I understand and the interpretation that I make is that it is from the data provided to the IAEA.

That's the interpretation I make. I want to know if this is the real spirit of this or not. The input and the output. Perfect. That is to say, if we are here saying, you have to discount and talk about the insumption that this, we told the lawyers, the program that uses what data you have put, I don't know. And part of that too, because of course, here it is very crazy because there is a yellowish thing of it is the fault of the IA. And no, here it says, the report concluded by IA. No, you concluded, brother.

You sign the sentence, you sign the written one. No, for now, this does not have legal capacity. Now, what if it's going to happen to an audience and this is listening? I already have it trained for this specifically, this case and it will say wait, doctor.

And he's going to talk, he's going to talk, he's going to talk, right? That's how we interpret it. It's not that this is pushing, it's not that this is going to happen in the audience summaries. This is going to happen in many things. Tools are being incorporated. Now you continue to be the person who has to corroborate that that jurisprudence, brother, if you wanted to use it for that and well, brother, it's wrong because

First, it's not a library of jurisprudence for what you're using. Probably, if this fails, it was your responsibility to control those data. If this intersects that there were two Juan Rodríguez and one had to give a 30% and another a 25% and then the DNI confused you because you trusted that. And this is how it existed before. You just have to understand that now we are processing data with something that is your responsibility. The data that you enter on that and the data that comes out on that. Good, perfect.

Why did I have this doubt? Because here there is a very fine line between what the client gives you and what you are going to enter into the IA. So what is the responsibility? Well, it was clear that the contributions then, but not on the question of those data that completely escapes the lawyer.

Well, there goes a little intervention from the law school. I saw Rosario at the time says inform your colleague and sign a jury statement that goes to your client, sorry, that you are going to use and to generate to process the data, your file. And I say and if it's the other way around, it comes to me and I fall with all the receipts from the bank that he comes paying. Everything is edited with her in two minutes. I mean, he fights with orders already that the food came ugly, if I edited it with her, the food came perfect. Now the client falls with that and you

arrow, say I have it won, great judgment, how are you going to keep charging and you go with this situation? I think there is a little indifference towards the lawyer in this case, because you are going to have to interpret if the lawyer is wanting to do that or he did it or the client was the one who provided you with all this tool. This audio, which is not from this person, nothing synthetically, it says the same voice as the doctor and I give it to you. So it's going to be very important as a lawyer. We know the obtaining. That's exactly why there's a specific article of the electronic documents.

That part is going to be very important because we are going to live in a synthetic world that does not know that it is real and it is the truth. But hey, it's the lawyer's fault. It's the judge's fault that after the sentence he realized it by another appeal and in court and it came that in reality the first judge had not seen that. It's going to happen a lot. The industry of trial always exists. Imagine that now it's synthetic. We have to deal with all possible tools, but the necessary responsibility on both sides of the counter.

So that the process goes as it should. Now, if they have signed a fake check, it's worth millions of times. I'll pay too, right? Well, to what extent does justice intervene in that? Here it happens the same. You have to see in that aspect how we adapt. Back without Mariano, this process will be carried through this path. If they are processed, it goes down this path as well. Very clear. So, released, which is what I was interested in.

Let's continue. 211, the prohibition of automatic decision and this article goes directly to judges, which opens an interesting door, because it specifies that the jurisdictional resolutions cannot be based exclusively on the use of artificial intelligence that is not specialized or certified. What I understand is that it pretends

open the umbrella to a possible future eventuality that a specialized and certified IA is created that can allow this to the judges. I understand that that is the idea that it has in its final end, that although today it is not possible, it does not mean that tomorrow it begins to be. And how we have automated the written tray, before there was a person who received the written one, he was granted such, right?

We say, no, no one else does it, that a system receives it, I have written it at any time. There are a lot of things that practically, although they are processable issues within the law, tax, whatever, that they pay, we can automate that. They don't stop being a resolution.

I explain now, there are different degrees of resolutions within the process, simple proof, the interlocutor, the one with definitive force, caducity. Let's think about how much human interpretation is in each of the questions and the means to be controlled and to speak in many cases in the executive process or whatever. In short, it will surely be moving forward to have different specialized systems of their own for this type of thing to be done.

Exactly. And article 212 again mentions this issue of progressive adoption of artificial intelligence designed for specific and specialized purposes, which I clearly imagine will be regulated by the court in the future.

and they will be made known at the appropriate time and in addition to indicating, I imagine there will be courses or specific questions that explain how they will be used. And then finally, Article 213, which talks to us about the transcription, summary and preservation of the mediation.

This is key, it will also be key for audiences and being able to have these transcriptions will often facilitate judges later, for example, the preliminary audience for the second one and being able to have that contact beyond having it later in audiovisual support, perhaps a transcription that has to be

control and corroborated by human personnel. Having done this tour already for the interrelated articles of the code, I hope you have taken note. I'm going to invite you to see the sentence. I don't know if they've already been sent by the groups. Well, and no, we have created a parallel group. If they are not there, there is.

-

No se envió al grupo. Me dice que ya se enviaron. ¿En cuál? Comisión B no tiene nada en el grupo, dice. ¿Confirma Comisión B o no? No está Comisión B. Sí llegó Comisión B. Comisión D sí. Acaba de llegar. Ahí está el doctor Falcone. No dice la verdad.

Comisión Ceci. Well, with those two examples, you will have to determine which one is which. At your discretion, at your own discretion, determine which one was made with ia and which one was made with, well, with the hands. And you have 15 minutes, no more than that. Novala would use ia for that. Novala would use it. What? The first one was made? Ah, I've already read it.

foreign

foreign

um

I need the best. I need the best.

-

-

um

um

-

Someone?

Hello, well, I read them very quickly. I think the first one is made with IA, because first of all there is a different language used, it is more sophisticated and makes fewer mistakes than the one with IA. The other one is like, I don't know, maybe in a coma it leaves a space of more or a space of less, there are suspenseful points.

It seems like a matter of human errors. The IA passed it through the IA and then it improved the paragraphs. After the first sentence, it seems to me that it talks about the labor issue that the labor licenses were not requested. I don't know, I didn't see it requested, I feel like the IA added it, the labor license of the progenitors. Someone with the opposite stance?

Which one do you like more?

-

but also the second one is very noticeable in the subjective human character, on the other hand, it is not much more objective and goes much more to legislation than to the subjective criterion. Well, I would like to hear someone from the virtual ones, if anyone wants to enable the microphone.

- - - - - -

Raise your hand if you think that the A is human editing. -Of course. -Placebo works like this. It says A and B. I answered A and B because the commission sent me for A and B. -She wants to talk in the... -Good morning. -Talk, doctor, we're listening. -Yes, I answered A and B because that's how they sent us to the commission.

I don't know if they will be in the same order in the personal exhibition. Silence please. Sorry, doctora López-Zapatina, we hear you now, yes? Yes, I say that the commission passed as A and B, I don't know if it will be ordered in the same way there in the face-to-face. The B, well. They are all nomenclature, A and B. No matter the order, they say A and B.

You doctor say that the B is made by a person, the B file. No, but we are talking to the doctor. To the robot doctor. For me the case B is made by the IA because it is clear how it determines paragraph by paragraph, how it treats, how it uses the uppercase letters, which the IA usually uses and how it differentiates and makes the resolutive shorter.

Doctora Migot, you have your hand up if you want to activate the microphone. Yes, good morning. Well, I actually agreed with the response there in the face-to-face auditorium as to what it is, and I already said it in the chat, it is the file indicated as B, the one that would be carried out with IA, that is, as is the same analysis that was carried out there in the auditorium, the structure itself of the file is already marking you, right? When one already more or less has the eye a little bit

Related to when it comes to the use of the AIA and yes, it does quite the interlining, precisely the ability to summarize and others. Well, I agree with everything you have put out. That was it. There are some who have gone for the formal structure and others for what is in the content, right? If I show you the sentence of someone who is just starting to make sentences, the sentence of a judge, an experienced judge, you will realize why.

but it can be much better written in structure, more beautiful, everything. The first one, the first one, maybe it's perfect, less grammatical errors, even a judge who has put the keyboard, maybe slower with other errors. What is it that we can really identify? Apart from that, I would show you, let's say, before I go to justice and you are not going to detect a sentence that my judge has written without artificial intelligence and that I, as a reporter, have written for him with artificial intelligence and you are not going to detect the difference. It's exactly the same.

Why? Because it is already specialized for the ascendancy with the voice that my judge makes, with the way he writes, the third person does not have interest in such a thing. A thousand questions. So, the doctor says, I imagine that he has given you interpretation guidelines at this time so that you can identify that. But what does this exercise give you? What is the difference, let's say, between the things you have? Distinguish. And what do we prefer? Human factor. Because it seems like bad words to use artificial intelligence.

It seems that we take away the price of using intelligence. That's done with IA. All day my life. You who use IA. The fact of using it, why? Because to a normal person, if you have a 5 or 6 body experience,

It's hard for a normal person to read it. So if you upload all that data or tell the IA that it's a summary, I invite you to a report that helps you with the structure, that doesn't take away the judge's decision. For me, the judge then controls it, adds things to it, takes things out of it. It's a tool.

and as a tool is valuable. The same happens when you were looking for jurisprudence through platforms, they told you that it is not reliable.

That is already extremely reliable. So, trust also that not only that use is a use as a tool, but also in the training that people have to use that tool. That is fundamental and the code marks it. I don't know, Doca, if you go, I don't know where you're going. Very good, very good job they did. Those who considered it to be the other

It is also perfect because this speaks of the training that will be needed and putting these challenges is going to be key for you to start incorporating everything that is guide and detection of it as in your habit.

I also think it is important to highlight that the immediacy that is intended with this code is key and will also be key at the time of the sentences, because beyond the fact that at some point specialized systems for the dictatorship of the people arrive, it is understood that the human eye and the criterion, and although they must fail objectively,

it is understood that the interpretation that judges make of the facts and the evaluation of the evidence and others, can never be left aside, at my discretion. And despite the fact that I love working with IA and that I use it, it seems to me that beyond the incorporation of all these tools, there are things that cannot be left aside.

Now I'm going to show you options that they have, maybe many of you already use them, maybe others don't. Here, very important, the contributions that our colleagues are making in the chat. Reading, controlling, editing, they tell us. They also tell me that the other one that the EA did was the most beautiful and maybe they do detect it more beautiful by the eye or what it says, but the reality is that critical evaluation is key.

-

Maybe many of you already use it. What does this guide allow you? It allows you to avoid hallucinations, it allows you to avoid certain biases. Why do I say certain biases? Because everything we've done is already biased in one way or another.

and crossed by subjectivity when it comes to people who have not written, exhibitions and others. What does this allow you? Create notebooks. Here you realize and what allows you to only use sources that you incorporate. Although it has the possibility of doing web search, what makes it so easy for you is

Just work with what you incorporate, that is, you load two or three files and the work you do will be about that. To look elsewhere unless you ask for it, because there is that possibility. So what gives you this more security? I'm not going to do more introduction on this because I would need hours and hours to learn. I'm also going to support this and I'm going to show what she says.

Apart from making a small glossary of all this, you don't worry about data, algorithm, to see a small one. I'm going to move on to the majority of all. I asked you how you work. Silence, please. You can see it. You can also ask him how it works and he will teach you and he will go on to tell you what to do, how to do, what to do, what it's good for, what it's not good for. You see? Ah, because I put it on the tab.

This is very good, what the doctor shows you, because she shows you or she will show you a suite of tools.

I think it's not the idea, let's say it's like there are tools that incorporate artificial intelligence. There are others that process more volumes of data, they have more context and what the doctor is giving you as a plus to what she gave today of the code process that with these tools can comply and work in a certain way to incorporate your day to day artificial intelligence. Now I'm with Clau, who is from Antropi. It seems to me

Much better for the editorial staff. Yes, it writes well, Claudia. Yes, it writes beautifully. It writes very well. Claudio. Claudio. Claudio. These last few months there has been a sharp drop in all that is the users of OpenAI and ChatGPT and they have shifted a little bit to Anthropic through Cloud. I use it a lot to program, for example. But well, there are different tools. With the doctor we will go a little deeper into that later.

This one, for what we do, is very good, to be honest, to write. It gives you the possibility to convert files, also larger files, even those that the GPT receives, at least in the... All this is in free versions, right?

Obviously, the paid versions of all these tools have advantages. It will depend on the use you give it, if it is worth it or not to pay for the versions. And apart now, now that we have coordinated, we are going to talk about that too, because you sign an electronic contract prior to this, that is, from the moment you used Lex in your study, you sign an electronic contract of the data of the causes that are incorporated there. Here the same thing happens. So it's very good to know in that yes, I accept that we put when we download 200 applications,

Acaba ser muy importante si lo van a utilizar para procesar información y dejemos de compartir la cuenta con nuestro hijito que está haciendo lengua de primer año. No, veamos para qué y cómo son espacios de trabajo y un poco le dice la doctora, ella te muestro la versión gratuita, hay más.

Exactly. As a disadvantage, you could find that at least in this free version, it does not interrelate chats. KChat is new. This is done by ChatGPT with OpenAI. They do develop it this way. So, there could be a disadvantage, but if you use it to consult on various issues, it's an advantage. So, it depends on what service you're going to give it, if you can take it as an advantage or disadvantage.

I use all of them. I use all of them, just like that. Then, well, GPT chat, which is also a free version, offers us two answers to what it does. In general, it is the most used one, so maybe many of you have had contact with it. It depends a lot on how one can configure it. Here it gives you the possibility of creating GPTs and they are personalized chats to which you can upload certain information with which you need to work later.

tends to hallucinate quite a bit for my taste what chatgpt is, so I prefer to be able to avoid that the others. Although you can tell him not to do it, chatgpt tends to want to satisfy the user.

We must also understand that behind all this there is an industry, there is a trade and that obviously they are looking to sell and what I find logical because we are using and we are buying and this version is free, although they are free, it is "free" because the reality is that one pays with data, pays with all the incorporation and training that one is giving in the use. All that information is then used to improve the

-

To answer the hallucinations, when the idea invents hallucinate is human. Now I'm going to explain it. There is a question. How long is the break? Half an hour. 15 minutes. Decide. Yes, thank you.

I wanted to say that I wanted to stay with two concepts that seem key to me in everything you put. One is what you said, doctor, as far as we do not take away the price from the IAEA. I think it is very important. And the other is something transversal that the doctor said repeatedly in terms of the human factor.

that the tool is very wide. I calculate that it will depend a lot on what we can learn each one in terms of the use of the tool, but it will always be crossed by the human factor. So, well, those two concepts seem very important to me in everything that is possible. We are doing very well then. Thank you. We have fulfilled the commitment then. There is still much more to be done because there is much more. The mess is set up afterwards.

And then we have Perplexity, which is quite similar to ChatGPT. Perplexity are all language processors that I'm exposing. And this one, I would tell you, is one of the best search engines with artificial intelligence. I mean, just like you were googling in Perplexity, you can get very good traceable information, it works very well as a search engine. Exactly. And it's the truth that one of the only ones because.

There is a very common mistake that is made, which is to use the AI as a search engine. No, these artificial intelligence systems are not search engines to search, search in Google or in this case, yes, in Perplexity you can do it because it is true that it is traceable and works quite well. In the rest, no, the rest are not search engines and it is a very common mistake to use them that way. So if it works for you, use Google.

use Google, download the sources and use those sources to train these systems. Also, it's very cheap what we're using with IA for what is not being collected, it's very cheap. Likewise, I bet a lot along with Dr. Franco Orellana, who later came to assist me, on the methodology.

The tool from 2023 to today, but radically changed. The one who has been using it understands that this is another world and we give it three more years, but it will continue to be methodology and methodology implies this that the doctor has been explaining with respect to the human factor that the doctor was accepting, etc. But it implies a systematic process. If we work with algorithms, the algorithm is not that once it does one thing and then another systematically extracts information, converts it into zero and one, processes it with its own mind and returns you a result.

Now, that systematization, what we are going to look for within the law is precision. So for me, a very important methodology, because today you drive a FIA 1 with a very hard box. Well, we already have an automatic and the Tesla is coming. Now the driving rules are the same. We continue to try not to run over people. We continue to try to move places. The tool is going to generate a shield. Now the principles are the same.

In that aspect, for me, of methodology, even though now we are going to see the agent part, when you no longer intervene autonomously, something handles your computer and it uploads a script, it does everything for you. We are going to see that now too. And what do we do there?

Where is human self-determination? Can we use that? That's what comes in 2026 and to Argentina. So for me the key is what the doctor says. There are tools for this. Now the methodology. So I'm going to use an element, not to crash people, transfer them safely. I wanted to sleep in a Tesla with wifi. Well, I went to one, it came to you. I recommend the Tesla. And as a last Grock talking about Tesla. Exactly. Well, I think it's part of the Elon Musk ecosystem.

and it also works as a language processor. It has, from what I understand here, its premium version is much superior to the free one. Each one, you try them at home, write them down, ask them the same questions. Now I'm going to try it. And investigate, get involved, use it, because that is the only way to be able to enter this world.

to trial and error, using, seeing how it works, what works best for you. Even introduce the same prompts in all of them. Try to get them back. They introduce the same thing in all of them and you will be able to evaluate your criteria. What is the best result for you? Because let's not forget that here we are all different and that what serves us one may not serve others. So this is the same. That has been my presentation. I hope it has served you.

foreign

Hello, hello, we are entering, please. This way we eat well at noon, we don't go late.

-

I'm going to ask for permission. If you want to ask for permission, they listen to me on the digital, turn off your audio a little bit. I don't know because I took it out so that the PDF is more light and so I don't have anything else. Are we giving you the time, the background so that you can get comfortable? Well,

We start from time to time, if you are grateful, a mate, notice for doubts, from time to time you are grateful. Well, well, first of all, highlight the enormous participation of something so new from Dr. Silvestre Gáster-Galván. I loved it. I loved that there are beginning to exist pro-activist interpretations of the use of the genitive day from many aspects.

uh

That makes it possible for great exchanges to exist. We can have the system we want, the code we want, but this is what generates impact. So, from my point of view, I'm always very grateful for the space you give me. I thank you very much and I also congratulate you for playing it. Because what you are playing is what all of Latin America is going to see.

Here it is not only a substantial and integral change of your file, of the way you transfer the information, of the mediation, of the role of the judge. It is not just that. You are adapting to something that the world is costing you to adapt to, which is that these things are happening and we are having to

Interpretarlas sin herramientas. Y eso es lo que está pasando. Entonces, algunos te mandan a que vaya al Tribunal de Ética del Poder Judicial, otros largan una comunicación interna del Colegio de Abogados. Todos son normativas soft, aisladas, que por ahí cercenan muchas cosas que nosotros buscamos como abogados dentro y fuera de tribunales. Y por otro lado, también, creo yo que no viven la realidad de lo que está pasando.

So here I see that there is a right of reality. For me that is the right of conduct. So first of all, I congratulate you. I congratulate you on a Saturday that you have connected so many people. And well, let's go to what we have come to, which is to be my "quilombo". I have only one requirement, let's say here, that they upload photos. Of course, I come making an effort. My wife has a belly like this, she has a risk pregnancy. The only trip I had to go to Australia now, the 16th, which are process rights days, a thousand more things. But this is really important to me.

This that you are living now is going to be very big, you don't see it today, it's going to be very big, well, in many respects. So, nothing, with that preamble and saying that they follow me, that's why they have scanned, there is always a little QR code to scan, you see, then for the two I put the network. No, it doesn't give you anything, it's not a raffle, no, no. What do I want to start with this? Advance that we don't have to forget, ah, I go back, look, scan, scan, scan, calm. The zoom too, no, the mid, if not, I don't answer questions. Of course, we upload each photo of food, everything, here there is one that...

They come to put their breasts, they post a photo. Thank you. No, they can't. No. Well, no, we'll pass it on to the group. No, no, I'm not going to bother him like that. He doesn't invite me anymore. Second, and before starting, Dr. Franco Orellana is with me. We've already taken it. Let's go back. Dr. Franco Orellana is with me, who is a judicial officer, a lawyer. He has written a book on how to write sentences with artificial intelligence when the world was prohibiting that. We have written a book called

Very good. Algorithmic justice, dictated sentences with generative idea. Dictated, not written, not dictated. And you are about to put to that doctrine, name and surname, and right of reality. So, for me, it's a dream. When you believe in ideas and then they happen, it's crazy. Well, the doctor is coming to accompany me today also for what they want to say in the chat. Reventenlo, as if it were GPT.

Don't forget this. Let's say we go to a process, a process like a dialectic debate where the parties are protagonists, where we have a director of the process and where there are definitely implicit and explicit rules. Because we already know that. Don't forget that. Please, for a chapter of five articles, don't forget that. It's still prevailing. Oh, impressive what we're doing. Free Fire. You're playing Free Fire, it was mine.

The mediation. Many times we think that technology is keeping us away. Here technology brings people closer. That's why we're talking about access to justice. We can't live in a disconnected reality where the client already consulted in chat GPT before coming to us, where the one who just entered the court says these sentences are contradictory, we have to appeal to them.

And where this is happening, this is amplifying the capabilities. Well, if you were good with this, it's going to be great, don't you have a doubt? And if you don't want to use it, don't do it. Follow the same principles as before. Now let's see in the end, in the theory of your case, in the support, in the support of proof, what do you want more in the data or in the story? This brings us much closer to that. So here is that for me it is a distinction. I may not go article by article, because I think the doctor has done a very good job.

a specific interpretation, but you will feel that I am talking about those articles. Because from the moment I am saying, hey, here we do not value artificial intelligence as a phenomenon, but here it is clearly as a tool of the process, there is a very important semantic distinction.

It's going to come and in this case, for example, I represent the intendant of Tucumán in the case of Deep Fake, where there is an influencer who is dancing everywhere, he does it. Well, I raise chevo because I can use my personal biometric data by processing it in an algorithm, ridiculing me in this aspect. That is, the concept that I am doing more than the proof is artificial intelligence because it is the videos, images, etc. I have violence against women, right to the image, I have a thousand more things that are being seen, but it is substantial right.

So let's not forget that part. Let's see, generative, let's see the intellectual property part. Everything that is generative, that has substantial right, is apart. Here it is a 100% tool of the process. So here we do not ask ourselves in the strict sense if he did it or not. If I have, if I make a sentence with generative law and they support me in writing, do I have to put that they support me in writing? No, because we do an indemnization calculation, we say I did it with a calculator. It doesn't work like that.

This is not the protagonist of the action. This is a tool where there are cases where, in order not to violate everything we have been doing in law, I need you to tell me how you have done it, what instruments you have used. So let's clarify this as a tool for the dialectical debate of the parties.

Where are they going to have equality and the judge and the judge tools for what? To say this leg brought me a pendrive. This is going to happen to him. I tell him for an electronic test that will happen. One brings a pendrive, another copies a drive link, another one drive, another dropbox and another one link that disappears.

Well, the test is supported when you go there to solve to see that the link is no longer useful. It will open the pendrive when it has to be solved and it will see that it is burned. You will find 200 questions where you are dealing with it and you will have to interpret a solution. Well, imagine with generative IA. But a thousand more interpretations will have and a thousand solutions to give under the same principles that have already been working. We would not get along with that part. I'm using the microphone, the mate, you see. Does anyone know what the difference is between a board and a computer? It seems like a joke, Gallego, but no.

Thank you very much. Thank you. Doctor, do you know the difference? I'll do it unnoticed. It's thought in the sense that the plan is thought as a single use. The plan serves for that. I like it. The computer serves for a lot of other things. And the two basically work the same way. The two work on an electricity basis. Electric impulses that do or do give a solution to something. What does the plan do? It gives a heat solution. For what? So that Marco doesn't come with the shirt. He'll come with the shirt anyway.

And the computer, what has it started to do? This, electric impulses, but that opens possibilities. So let's think the same. Artificial intelligence, and that's where I go back to the first thing, will continue to exist as the primary, intelligence, intelligence of our species.

So if we start to understand what it is to be really intelligent, we see where artificial can be used. These computerized systems that allow processing large volumes of data to give an answer, whatever a calculator, like our human brain.

without losing sight of what? Once again, the human in the center. All the time this clarifies that situation. So why? Because justice is for people. Let's see the same situation here. We evaluate the income of judges, prosecutors, defenders of many issues, not only in the writing of a sentence. Pass through many filters to that situation. When you are a lawyer, you are much more than a written claim. There is much more behind this and for us to understand that, today a machine does not understand it.

Today, a machine does not understand the context of this change in Catamarca, it does not understand it, it understands parameters that you are going to give and it is going to calculate something probable. If you start to understand the first thing that starts with this, with a glossary of terms, gives conceptual definitions. What is artificial intelligence generated in the judicial process? Not for cancer detection. No, no, no. In the judicial process. That.

makes us have to necessarily have a look at the judicial process of the tool. This, no doubt, will be useful for you for every aspect of your life. You're already using it without knowing it in every aspect of your life. But here in the term, I make a small advance and to generative. If we talk about this, you already start

to understand that in the advance that this was created, it was even processed and the terms of this were debated in a code and with specialists, etc. There is nothing univocal about what intelligence is. Imagine artificial intelligence. What is definitely univocal is that our brain processes data and returns an answer through that stimulus. There are no two without. Well, there's a number in Tokyo. Why? Because they have received a stimulus and their brain has a probable answer.

Look at this situation. I go, yes, I think it's going to open, Franco. Look, there are no two without, ah, you can't see it, you can't see it, yes, yes, yes, I have 200, there are no two without, using the most intelligent model that GPT has, three, right? But there are no two without. You say, why do you keep asking me this? Three. I didn't know this in 2023.

There was no training for our lunfart, that there are no two without three. Where do you think it comes from, there are no two without three? Have you ever questioned those lunfarts that we repeat? We use it for... And well, I already put two bills in the third. I'm going two children, well, in the third. Of course, we want there to be something in the third, let's say. Well, now this understands it. If we, in this exercise...

we would have done it when this probability calculator had not been trained with a context window to understand. I would like that if someone has a chat that fits in their hand, yes, in the chat too, put it like this, there are no two without and try that fast. I love that I use the voice format, Marcia. I've put three of them all. I don't think so. And you know why? Everyone has, raise your hand, the one who didn't answer, there are no two without three.

Look, look, and in the chat, this deterministic element works like our brain. Many of those who perhaps have a more rigorous use and have spoken to him with the Argentine lunfardo, this already knows that there are no two without three, because this begins to have a persistent memory of context that begins to understand a little more your environment. Now, it was not sure that, look, if we talk about a judicial cause where this would have taken a determination, if there are no two without three, perfect, there are three that we have nailed.

Three, I'll put my hand here and I don't know how many in the chat, that I haven't said that. There are not two without, and I'm going to start talking about something else. I don't know if you can tell me someone in the background what I talked about. There it is. There are not two equals. Complete the... Well, if I put... Well, look, here they said, well, that they answer, there are not two without one, there are not two...

Note that in short, the same tool we have does not react the same to the stimulus and we have not written anything about a social science that we know that in a court they will say one thing and on the other side another, that a lawyer will give you a legal recommendation and the other lawyer another. Why a social science? And it's not that one is bad or the other, not necessarily. So why do I make this little feedback of commercial tools like we are using? Why is there a risk of using brains that have no context?

that are calculators that one will interpret that less by less is more than another 20 and so they are calculators, probable words. So what is the concept of processing that we give without being computer scientists? They are systems that process large volumes of data and use algorithms. Let's see what algorithms are to understand what that is and it will give me a good result. Let's see what aspect it starts with. To understand the responsible use. Responsible use. The first thing is to know what you are using, brother. Why does it turn on the oven?

If you don't know how to cook, then the first thing is this. What is it for? To process large volumes of data with algorithms that are housed in a perfect table. I understand that this processes data volumes. I'm doing very well there. Now I start to ask myself, thanks to the articulated one, how, because it is a code, I can delegate something to artificial intelligence. In this case, I tell you, you put the alarm clock and you have to get up. They're going to give a little seat now from three to five. They're going to go play polo.

From 3 to 5. Are you really so nervous that if you have batteries in your cell phone that you are going to wake up at 5, the alarm won't ring? Really so nervous that it won't ring? Have you cognitively delegated an activity that before

produced in your brain space, a concern, in your brain, to use things to do that. Get up in case I fall asleep, call me in case you don't. Do you remember that or not? Well, that doesn't exist anymore. Or well, in some cases, yes.

So here it is good to say, hey, what can I cognitively delegate to these computerized systems that process data? And there we say, hey, if you are going to do it to do a test in this way, in this way, if you are going to add it as a document and so you cannot use it.

So that's good to understand and that's because of how the deterministic tool works. This doesn't understand if you have to use the Mendez formula for the damage calculation or if it's a consumption process or if they don't understand if you understand that and based on that you provide certain parameters. Now you can read the law faster than you do. It's worth more. It can be interrelated to articles as well.

But the assumption of facts to standards to say what happened and what is going to be a very complex process today for this type of calculator. Keep in mind that the calculator evolved from the abaco calculator to the atomic scientific calculator. Well, here the same thing is going to happen. Initially this was like the abaco, it is already being like a scientific calculator. We already have many things that we will be able to delegate here. Let's see which ones we want to.

To understand that, it is also this, there is no Ferrari without oil. You don't go up to anything that will not take you anywhere. And what is the oil of this Ferrari? GPT, Cloud, Perplexity, call it whatever you want, it's Fiat, Toyota, BMW. But its input in any of those are data with what has been trained, with what you can incorporate and with what you instruct. There is no more.

So if all those data, data plus data, information, you understand that they have to be structured and interrelated. You have wondered why the structure of a sentence starts with the date. Well, let's see the criteria in one day, right? A date, right? Who is the judge intervenes?

He turns out to be able to condemn all legally relevant procedural acts so that for the honorary to avoid nullities, to know that the process was followed. Considering that how it starts from the general to the particular. This has been said in parts, so this is controversial, this is not. I'm going to take into account according to such rules of critical science this.

Let's go to the most particular. Ah, about this situation. Are you in agreement? The parts. The accident occurred and the accident occurred and this one came from the right. A van came and this one came in a motorcycle. We agree on that. There is nothing to say. We discuss the speed. About the speed of what test? This, this, this. And how do I determine this? In this way, in this way. And you are narrowing the process, right? That process that you are doing is an algorithm.

Your algorithm to solve a conflict tells you that the hypothetical reasoning of the particular generals that you do is the most reasonable and clear way to say, notice that there are many who complain saying, I advance my criteria. They complain that they do not follow a structure saying, ah, he is already telling me that he is going to nail me and he has not processed me at all.

We don't say I'm so empowered by the gentleman that I already have to pay him 500 bucks. No, but I anticipate him. The object of my demand is such because according to the facts I have done such a thing, because look, I have said such a thing, as I approve, such and such a thing happens. The regulation says this, pay him. But look, it follows a logical order of interpretation. We have naturalized that, but it is an algorithm. They are logical sequences to reach an end. Why do you think that the support has a restriction on the test?

There is an algorithmic reason. We are talking about urgent processes that the test would not have to admit many kilometers. Why? Because it will denaturalize the urgent process, right? So the documentation is very important. Well, we have all agreed socially that today you have a little paper that says that you are the owner and you are the owner.

Well, today things have changed. However, we give a lot of importance. Notice that the first ones regulate the code. Documental test. It comes that it has an order because we give weight. It's an algorithm. Our brain works like this. Tell Luigi. What does he say? Mr. Luis, I don't know if you're going to hear. Where? Here? Luis. When you want, Luis, I'm going to interrupt. Well, that said, if you're going to start understanding.

that you unbind large actions in microprocesses, depending on what you have to do, you already start to understand how you are going to get involved in those microprocesses. So here I think the code plays it by saying general rules, but also particular uses. It was all for throwing away the principles of UNESCO, the indelegability, the human reasoning. That was always, that was always, let's say now.

Give me a little more tools and here it is played. So I want to tell you why. If we understand how I have systems here, I have a brain, a digital brain, a brain that processes things. That brain inside has twerks, we call it algorithm, right? To that brain that has twerks, it is entering information and through those twerks, first it goes over here and the PAM is made.

Right? That is, those are logical sequences to do a judicial process. The same has logical sequences. First, the claim, consultation, opening, test, test, allegation, logical sequence to reach an end. You understand that the judicial process is an algorithm and that this works logically. You will understand that we make an allegation as this is not a logical sequence for our brain. We make an allegation. Music that well, as this is done.

Yes, well, the doctor flies with the gato, he's going to do it to me. But if you understand that this works by small logical sequences, start to understand how to incorporate information, how you are going to ask, why not seek jurisprudence, why not ask for information extraction in certain places where it is not prepared for that.

This learns deeply from which place, it has many learning techniques, but the most gross for me now begins to be what the user can touch, because you are going to have a judicial power that is going to develop tools, a law school too, but today I say, well, what do we do to that? We are going to use tools that will probably be protocolized, they are going to go saying, hey, for example, with the municipality where I work as Director of Artificial Intelligence, we have taken

It is the first resolution, I think, also of that style, where we do not say, well, that's how the generative idea is going to be used, no.

It's a clear guide where it speaks in clear language and says, hey, look, you want to extract information from MP3 and MP4, use this Google tool. You want to use it to extract, use Notebook Lemp, it will serve you for this. It's in public work, use it like this for the accounting part, watch out with this. And start giving an example. For example, ask him to do such a thing. So the guide tells you so that you can practice how to use generative in your place. So here a little bit of this too, I think.

It's going to be very important, the specific training, very important, because they are deterministic ideas. I left the judicial power and I had to stop sharing with Dr. Llanas many generative ideas. Why? Because the vision that I already had inside and outside the court is another one. And now I want to show you why it's going to be very important to count on tools where you agree with how those data are treated. You agree with the traceability that that will have if you need it, but also

I look for the most accessible tool. If I have something that is very complicated and if it is complicated, the Excel formulas, do you remember any? And they know that there are formulas and they ask someone to do an Excel, right? The management systems are complicated. This is not complicated. The tool that you are going to have to start using has to serve you really. If not, you don't use it. Why? Because if you do it in the microwave and you like how it is in the microwave, do it there. If not, well, turn on the oven. If not, turn on the flip. If not, well, there are different ways of cooking information. Here the same thing will happen. But when you do it, you have responsibility.

Understand this last concept that I throw from the glossary, the word, you may have seen, I ran out of tokens, which is like the minimum unit, the token is a token that uses fuel, ultimately artificial intelligence, which is not the same as a word, a word is not a token. What do you use in the end? You use certain ways in which

Has aprendido el abecedario, las palabras y el contexto humano que puede ingresar en el embudo, que va a ser este embedding que yo te digo, de información para que la máquina me entienda. Vos cuando estudias programación, tenés que entender cómo la máquina piensa para vos. Entonces tenés que entender que si vos pones esto, esto, esto, pasa tal cosa. Ahora es al revés. Vos has dado la oportunidad de que la máquina te entienda a vos y no que vos tengas que entender a la máquina para comunicarte.

That is produced thanks to what I am telling you, to the tokenization of words in a specific funnel. What does a specific funnel do? It generates multi-vector spaces of many, many words. For example, if I show it here, this is a tokenization form. If we would do a tool, notice the national constitution is divided into different types of embedding. They say national contribution.

Those are three different tokens. It's a word and maybe, for example, the Constitution has two national one. Why? Because you're going to integrate it into a map of probable words so that when I tell you

If I already have a map of probable words, royalty, and here in royalty is king, queen, prince, princess, castle, right? And I tell you, well, this map of probable words. And here I have humans. And I say, man, with what, if I have to say man of royalty, with what map of words of funnel would it be here that had king, queen, prince, princess?

Con rey o con príncipe, ¿verdad? Bueno, esto se acerca probablemente acá para dar una solución. Entonces, lo que busca esto es cómo podemos asociar esos grupos de palabras para dar algo probable. Funciona como una calculadora. Esto que estoy diciendo es como piensa esto acá atrás. Por eso yo le digo, échale a la ley, échale a la ley. Han asociado unas palabras. O sea, ley, trampa. Voy a decir un alemán esto y te voy a decir que está loco. Échale, échale a la norma.

They have associated a group, their neurons, that's why they are neural networks, their neurons have data, data. I say orange. What have you thought? Depending precisely on their context, depending on how they have structured in their neuron, depending on the stimulus. Understanding this allows you to know a little about how your brain works, to say how these people have achieved that this emulates your brain.

In your neuron is the orange, in your neuron is the fruit embedding, there is the language of the fruit, fruit here and here is color. But hey, I just happened to have seen an orange at breakfast. Our probability is various, they are getting smaller depending on the context. Because if I read the orange fruit and well, we call it the mandarin orange, the thinkers begin to think, others say the orange. Our brain is structured in the way it is. Now, if you understand that, you understand that you are going to say make an allegation.

What is allegation? In law, it's only allegation? Make me write, think, write, describe. You start to understand that our brain also works while you are generating that the neuronal concepts in our brain are shortened. It has a lot of information. But well, we don't stay on that.

We have talked about this, which is also important, because even the concept is the first code or process that puts a word in this way, prompt. And why was prompt decided and not instruction? That is, we use the words in English, because it is the precise technical tool of instruction for the window that makes you enter artificial intelligence, which is a prompt, the stimulus, the instruction that you give. It can be a question, an affirmation, an order, there are no two, yes, what is it?

It's a prompt, but if I have to describe what it is, I say there are no two without a sentence, a sentence, an affirmation that I'm giving a question. Well, this is all that. So here it is not easy that the word prompt has been used, that is, in English. Why? Because here it is talked about how the technique is to communicate with artificial intelligence. So when I see you, I start to say that prompt has been used, it is not only in the character of a jury statement. It doesn't work like this. There are more.

The doctor had spoken of hallucinations that I have just told us. Hallucinating is human, not technically computational. Why? Because you say your leg is crazy. How can I go out with 400 campers? Look at the sun, what does it do? It cannot be that this person is cold given the heat stimulus. He is reacting improbably. He's hallucinating, right? Now this happens the same. I give you some stimulus that I consider because many times the concept of hallucinating

It's a matter of validation. The models by default hallucinate. You're a capo if you believe in what you want. This by default has the possibility of hallucinating. Like in our brain. But of course, they grab you just up and tell you, you're a capo. You're a capo. You're a capo. You're a capo. You're a capo. You're a capo. You're a capo. You're a capo. You're a capo. You're a capo. You're a capo.

Maybe it will not happen very unnoticed in law. Why do I tell you? Because if we use this for medicine, we all bleed the same. If you cut your ligaments, you see the cut of the ligament the same everywhere. But already in Catamarca, in Tucumán, it is the same way of applying the law. We can have the same code if you want, whatever, the same provincial substantive law, but it is interpreted the same. Why? Because in some way this is given to the context. So let's see what we consider to be hallucinating.

Now, if they ask me for a law that doesn't exist, I'm going to tell you the mistake is that you are asking for a law, something that is more likely to require a Taiwan law than the one in Catamarca, than the one in Tucumán, than the one in Buenos Aires, properly said. So we're going to see how to avoid that using a doctorate tool that is good. We're going to see it. Tricks. For me, what worries me the least. What worries me the least is because, mainly, if you put the human in the center, what is the risk of the trick? If you are saying that you make the decision,

What problem do I have with the bias of the IA? If you are telling me that I voted for this and I say no, if I am from such a party, if you are telling me this now, where is the situation? The problem is not going to be in you, the problem is the generations to come. There is the real problem of bias, not the one that has been prepared for the intergenerational change in the use of tools that you already reason with. You have already lived the situation, no, it is not like that. Now, for many people who are getting involved in science, what an IA says is a big word.

So, even if they consider that they make the decision, that decision is made much earlier. Here it is good, I think it does not emphasize that so much because all the time the human decision is centered, not in the bias saying, hey, they are a system where it will tell you if it has to be a preventive embargo to such or such bank. And well, the system decided that this bank, because it considers that in this bank there is more solvency than in the market, pays. Well, there is a bias.

Why? Because the person here moves all their money to Mercopago, which is a virtual wallet that is still under the central bank rules. Well, who's going to know better? The lawyer who is asking for the embargo and the judge who is in a system may have the bias that I am mentioning. Mercopago is 10 years old, the nation is 150, it is a public entity. Well, that's where the bias is, the concern that the system decides for you. And we go specifically to the chapter, let's say, for me, which is generative.

The first thing is for me the object and the scope of the application. It bothers me a lot when processes start to discuss things that you don't have to discuss. It happens all the time, right? Well, I ask you a question. Yes. Could, for example, a bank pay the IAEA so that it does not recommend that embargo is imposed on its accounts or so that yes? Well, look, it's very good. Something that you propose that I propose a lot in terms today of right to oblivion.

You used to have two channels, practically unmistakable, of information bridges. If I tell you to look for information, where did you look for it? Do we have another information highway? Before it was Yahoo, displaced. Google, right? So what's going on? Hey, if everyone is looking for information in your library and you are indexing information, why are you indexing it? And by engine positioning search engine, by organic placement, by payment. Now there is a new positioning that I like a little bit what the doctor asks.

What is the generic positioning, the CEG, you have CEG or CEM, generative, sorry, which is why this person appears here, why this recommends such a thing, how it works so that this brings something. But what happens if we fall into what you're saying, doctor, you're letting something make the decision. So, coimear and it's very difficult for them to get to OpenAI and say, but it can happen to you when, for example, we are making tools. This is recorded, right? Well.

We are manufacturing the tools of the municipality. There is a tourism chatbot. Maybe, eventually, the tourism chatbot of Tucumán can offer you a warehouse that is very good, because I have inserted, at the old school, that you go to a certain warehouse. Now, GPT will not make that situation unmoved. So, I would worry more about that auditability of why I make a decision with these own tools.

Why this recommends such a thing and not such another? Why this restaurant and not another? Why? Because the use is designed for the recommendation. Here technically not. Here technically it is designed for you to process your recommendation of things, your opinion. But hey, every use is in that. However, I would not look for the auditability of systems as large as Cloud, OpenAI, ChaiJPT, Gemini, etc. Given this distinction,

We also have two or three issues that happen in the whole process: text, image, audio. Because we are a computerized system that processes data that can be text, image, audio, right? And reproductions. And we put reproductions in general because they just mentioned, I think, if they could not, I think one of the doctors had mentioned this possibility of reproducing a scenario. But it's worth it, it's worth it, it's a croquis, as Dr. Zabaleta said, to reproduce it, to the technical probability that that happens.

That already happened before. And even when you try to test this code, it also gives you the opportunity of digital representations. I mean, you want to put on some glasses and wear a virtual reality and take me to that space to show me and show me the fear that today gives you to get on an Ubermoto.

Because it has fallen once and I show you from the glasses how you fall and you get hit by being a judge. And well, it's worth it. If I can show you things in a more, perhaps, credible way in that aspect, I believe in the data and not in the story. Now, in the cases of the use of artificial intelligence, yes, use text, image, video, whatever you want. But what is the main point? The real thing.

The main issue is real. I made this croquis that is worth the emergency plans to leave this place, the witnesses who said such a thing. Well, I have made a croquis of what the witnesses have said, who supposedly have passed and the impossibility that these two people have been communicated. Why is it not valid? But I tell you that I made that croquis based on this information here, using these information techniques. They see how

This is perhaps secondary, but main at the same time, because the tool that is going to be created is what is going to be created as a croquis or representation. It is worth data that is there. It's not that I make a croquis that I think something happened. Well, do it, but how is a judge going to value it? I don't know how it was created, I don't know how the calculator has come to the result. Let's say that 5 by 5 is 25, but I need to see it. Why is it 25? And this is perhaps the "what's it for?"

It starts from the trial law. And I think the first code that really introduces the trial aspect, this is super sharp. I mean, I hope, I hope you can see how fast it is to say, hey, come and use artificial intelligence. Hey, in my process, it must be seen in such a way that it is going to be conjugated with very good principles that are, I mean, a code of many principles.

But when we talk about testing, let's see this way. First, let's look at general admissibility issues and then we'll go into the specifics that are in other subsequent chapters. But here he talks all the time about the protection of constitutional guarantees. We don't use generatives because we don't use them for anything, not in any way. I think that especially in oral processes, they will start to exist this use of things. No doctor, look, I have this specialized system that has been created by

The New York Bar that does statistical calculations of environmental damage. Well, I want to show you how we do as a game. We understand that this is being preserved. I can see, you can enter, doctor, the link that we are seeing. Who created it? There is information based on the credibility of this. Now there has to be a bridge where so many probabilities pass. I often see, I say the problem starts before. Do you know how many types of image formats exist? JPG sounds right to you, right? PNG doesn't sound like that. And audio?

mp3, ogg, m4a. If you natively download WhatsApp, look how it downloads. And that's what I want to tell you. We thought an audio was mp3, right? It sounded like that because mp3 is where we listen to music.

The WAP format, exactly. There are different audio formats. And why is this important? Because at the time of conserving the test, that is already important. So I tell you, your system allows you to upload, for example, SAE at the time only allowed MP3. And it happened to me in a particular case over there, that there was a judge who said, you have to shrink the size to MP3, it was M4A or OGG, it was WhatsApp. Shrink the size because it exceeds the 20 megabytes that SAE allowed you to upload.

and pass it as you can to MP3. And you do the same thing I'm telling you, hey, I had come in this glass, right? And you're telling me to put it in this glass, give it to the other one, saying hey, what a process you have, you asked me, if you have done it like that on paper, you have broken it whole, you have reassembled it with tape, it has passed. You, thanks to your process and technology, you are telling me to remove all authenticity techniques from a test. What are we looking for here? That doesn't happen.

That here we can really audit what is going to be processed within certain data. Now I tell you, those problems are going to have it. Hey, I have an audio that I tell you, I see a video of a security camera. Well, the most harmful in terms of cybersecurity is that you take a Prime Drive to court and it connects there, given. Now, are they practical? Yes. Why is it dangerous? We go from a thousand ways, but the main thing is the fact of being able to inject.

to be able to get information, download programs, give me access, burn it to the pendrive, a pendrive is never burned. What if the test was there? You have kept everything there. But hey, the court when it receives it is empty. There will be many things like that. Now, let's solve all these possibilities together under these criteria: protection of constitutional guarantees, right of defense, contradiction, mediation. We are going to see above all what mediation is also in this.

When we talk about the admissibility part, we talk about the same thing again, but more verified. All this, what does Acebo want? That this can be confronted with something, that this can be verified and, in its case, ratified. Let's see why the issue of ratification. I say, Vox, on the one hand, today I think there is a new procedural distinction that I encourage you to...

Put it on, to say now documents, you will have on the one hand electronic documents, right? Those electronic documents will be divided into many representations and things MP3, MP4, written electronic documents in general and subdivide them. If they are, notice, we already have the law of signature and the commercial civil code signed, not signed. We have like a distinction of electronic documents, but how is an electronic document with IA?

Generated by IA. What kind of character is electronic document? Why? Because you have a general principle here that says that any digital representation that you have, you can incorporate into the process. That is, hey, upload the digital representations that you have. The general principle is the electronic document. It doesn't matter what character, what resource you have. You have a video, you have to upload it. You have to be able to upload it. An image too. An audio too. Well, with the generative, how is it?

This you have as an electronic document, then you are going to prepare a process system that allows you to generate authenticity, legitimacy, authorship and all this. You are going to upload it as a screenshot to WhatsApp, but then you will need a clear office to see if this number is a expert to see the metadata. You have understood that this electronic document only serves us, right? Why don't you understand it with documents that are not generative?

You are saying that this alone costs you to generate credibility. Why? First of all, it is very easy to generate things that do not exist, edit images, edit photos. But before I was already saying that an electronic document is volatile. You signed a contract before, your signature is there and there was only one process that we can do. Is your signature or not your signature? Now you can say that video is made by me, but that date is not. And the key was that you are on that date there.

And now I'm going to say that video is not me, it's already saying, look at the amount of ways to push an electronic document that exists. Well, the same thing happens with the guide, because you say, where did you get that from? It happens the same here. If you understand this, you start to understand that it is not very crazy to separate and understand that there has to be a process of evaluation of an electronic document created with artificial intelligence, because there has already been a process of evaluation of electronic documents without artificial intelligence, understanding that today

Practically 90% of the cases you are talking about have electronic documents. They start in the digital world. I send you a WhatsApp, this has happened, I have contracted such service, I have asked for free market. A thousand things like that happen and the proof is in cyberspace. We have given the code the opportunity to protect it. Now,

When it's with artificial intelligence, what are the three requirements for me or characteristics that you have to take into account when it's a document generated with artificial intelligence? First, that this is worked as an intermediate product of something. What does it mean? You have processed facts, regulations, another test and you have generated this. For example, I am a judge and I want to generate conviction in the jury. What is the jury? The probability that something happens. So the parties have come and left

.

of 200 people who are employed more and somewhere there is that is Marco Rossi who has marked at such time, such day, right? But none of the parties has made the statement of this, none of the parties has actually contradicted whether this proof is good or not. Now, as a judge, I cannot process that data, right? That they are still dates, people and numbers in a

endless sheets and tables to say that from June to December this person did not participate a percentage of times that you have said in the demand that you can be there or that there is that processing of information to tell me that it has not complied with the regulation that it had to remain in the establishment. And so I can tell you a thousand more uses where you are going to use the real raw material

the information table with this, the receipts of AVERE, the video, as mentioned in the case of the doctor. I have a case to show you how that would be, where it is a video where six minutes ago a person drowned. The person has a maturation delay, he was in a swimming pool with two teachers, a private teacher and the lifeguard from there. And you have

20 cameras where they show you that they passed over him and you hear in the defense of this situation, 1.90 meters, he measured and had a green, a green green shirt. The water gave him. That's not, it's not data, it's not reasonability. Now, knowing that I have and I do an inspection of this three levels of walking, which are three meters that the water reaches you there, I don't need to do a test for that, right? Now I have cameras that show the time that this person arrived.

And I want to control all that, for example, because I only have the frames, everything is erased, but I want to control it in a video. I can't, but how? I'm going to use, you know, as an intermediate product, the facts. I'm going to use what I already have in the process and you're going to say, look, I've processed this and generated it like this. Why? Because I want to generate a conviction with you that given this rule, no one can stay underwater for more than three minutes.

Look, this person is 6, this is here, this is here. I show you how this person looks according to their physical condition on a 3-meter walk. I show you a general photo. It's strong, it's worth more than strong. But I'm claiming moral damage. What am I claiming? Let's say material damage, that this is no longer happening. Punitive damage. Brother, this is no longer happening in Tucumán. What are we talking about? A person who has not done an electro, nobody knew what. And stop. And he's dead and people are still swimming.

Now, how do I, I mean, yes, the facts sound great, right? In an expedient, how do I tell it? Do I explain where it goes? And it's better with the tools that I have at my disposal. I'm going to show you a little bit of that now, but for you to understand the term of intermediate product, which we specifically use in legislation.

without self-esteem, which the doctor has spoken about, that is, only this has no weight. I, in this case, generate an image of a person, 1.90 meters tall, on a 3-meter walkway, is submerged and that bubbles come out because it is understood that he is drowning and that people are not looking around and I will generate an image. Does that have value? It shouldn't. It's a photocopy, it's the capture of the WhatsApp screen.

Now generate it like storytelling, if you want to do a series. You tell me I want to generate impact as in the case so much and well, do a series with her, there is no problem. But hey, what value does it really have in what we are looking for, which is the right to prove in the possibility of sustaining facts according to the applicable regulations? So this is understood.

It doesn't have an autonomous value in itself, but the fact that you admit this is because this is going to amplify the probatory value of other things or of what you are really saying there. This seems very gross to me because I think no judge is going to say he's right because GPT has said it. It doesn't sound so stupid. Now.

To say it outside, there is no drama. How many things do they have to look for? It happens to me, the car lights turn on, I use the computer vision, I say, what is this little light, brother? And well, I go by the fact that I had to put water later in such a place and that it was the car. But what happens if we delegate certain issues in law? No.

especially in the trial law. For us, it is very rigorous to limit the scope of the test in the use of artificial intelligence so much in general for the judges who interpret the test, as I am mentioning. There is a case that I was just telling the doctor that I liked a lot from 2024, where a North American judge

receives a lawsuit, also a civil lawsuit of contractual breach. They had made some trampolines in the house, a person, some trampolines for the jump, some chanquitos, and tuki tuki. The two chanquitos of him get into an accident. So, the leg goes to the insurance company directly, he sues him and says, look, sorry, he sues the contractor, the contractor says, but I'm insured. The insurance company comes, the contractor comes, there's the part, and the specific clause to know if it covered or not this, was that

He had coverage in all, all application of landscape. And I ask him, do you all know what landscape? Well, none of the parties had explained it. And the judge has to make a decision. Now, is it introduced in the cause? Yes, in the clause that says this. Are the facts of what was done specifically in the patio introduced? Is it introduced? What damage in real terms? Because.

Because before I googled them if you want. How bad can I be? Interrelate this clause with what the parties have said, which is this with the images that I have that have happened, how the trampoline has been broken or whatever, so that you can interpret and have an opinion and an option of what landscape is in the terms of this contract and of these facts that the judge does saying hey, I used it for this. And even though the machine determines that this could be landscape, he nails it, the insurer nails it, the plaintiff says to the plaintiff, he says no.

I consider that this is not a landscape, but he shows you a very dynamic way in which the judge can use tools to process information from there and make his decision. Now, what does that really hurt the due process? If the data are in the file, I'm using something that will assist me, just like it assisted you for this, so that I can make a decision, perhaps of higher quality. So, in relation to what the teacher is saying, introducing the data,

and what is in the test material, can a version generated by IA be obtained that tries to relive the moment in a way and that the other part, introducing the same data, can generate a different one? It's going to happen. In fact, what naturally happens. I mean, today you have something that answers the demand and what is it going to do? The affirmative facts went to the negative, first person to third person instead of actor, and has been sued and so on. Let's think that the manifest rituals are going there now.

Thinking about it that way also implies, as the doctor was saying, being able to contrast facts, as you were saying.

It would be like an allegation. It would be more or less natural. It's an allegation. I grab all the proof that I have in the process with artificial intelligence and I make you this image of what I'm interpreting of the test. And yes, one part of it I will interpret in a way, the other part I will interpret. All this I'm already starting to get into your brain and there is no code, let's say, that regulates artificial intelligence. So if I put the expert in contradiction, look, bro, I have processed this information that according to the Harvard benchmarking, whatever, postdoc compared to you and with all due respect, he knows a lot about this.

And look, it's a little bit incongruent, I think you're saying it's like a little bit stronger impact. Give him that out there. Why do I tell you? Processes will exist with this. There will be a thousand more. But what do I think? We play the right to convince. And if I convince you in oral process, let's not even talk about when we give an oral. How many times do we give an oral saying it was complicated, but Cari, you put on my best suit.

um

The doctor asks why she can't record the audience. Of course, if I had my mechanism, I could record the audience and also that would serve me as a tool for, for example, if I have to make a legato.

Why? Because I will have a fidelity of everything that was said. But if I tell you that I already recorded it in court, why did you record it? But to me that would no longer serve me because it's post.

Why? Because that will be raised later. I need to resolve it in my study before. This is going to happen, lawyer and lawyer, judge and judge. This is going to happen. I'm going to tell the front, hey, my client is there. Where is this going? And what is the judge going to do? Is it going to be the interpretation of the personal protection law? What is he going to interpret? This is the same. It's coming. Yes, doctor.

Even in the article where it is spoken of the audiences, it is said that they are public, so there would be no violation for depriving them of recording them or recording them in audiovisual or audio only, there would be no violation. Well, but recording them with a language model is another thing, because you are doing that data is extracted that is being said there, it is processed in the algorithm and a summary is returned, it is contrasted with the theory of the case. Now, why not? Now I have to warn you.

Do you see the amount of unknowns that are going to arise? Because if you ask me and it's hard to argue with a little guy who has his setup very well prepared, his things to solve with this, he has a tutorial card, the case that he has said here, this test, what the witness has said and everything that has the impugnation, everything that has, he has it now. Now, if he warns him, he does not warn him and a little bit we are going to talk about it with the sentence and with the audience, but we are already moving forward there.

I don't know if you want to say something about the part of ratifying on the front, which is like ratifying on the front, which we also say there, it goes very hand in hand with what comes here, which has no autonomous value, which is an intermediate product and that can be or has to be ratified on the front and it is a bit the conjunction of what we are mentioning here, it goes in harmony with that situation that I showed you, generates an image with artificial intelligence, contact with security cameras, it is a probability, just like what this does, yes, now,

Many of these things, let's say, a forensic accident, what is not done, a model of what happened, that does not exist. The car, you see it, let's say, it has a damage here to the right, then models that came here and show the impact. Is that real? And it does not add probatory value before there is artificial intelligence?

A expert comes and shows you in the trial, says, hey, it was going this way here, this one was going this way, according to the impact that you have here, to what the engine box has marked, whatever, there it goes, therefore, and it shows you, right? And it shows how the dolls do it. Wasn't that done in person, physically, too? Well, if we can't simulate it.

There I go. It's the same, but now you have the opportunity to incorporate quality into what you do, to incorporate this access to much more specialized information. Now, it's going to be ugly if you're really generic, if you really think that because you've uploaded something with artificial intelligence it has value, because the front one can contact everything.

The defense says, "Show me how you got that result." Believe it or not, many times the elections, for example, now in 2027, will be 100%, 100% attacked. Why? Because the guys, who are they going to ask where to vote? Who to vote?

And one will answer him: "Peronism, libertarian." And why do they think he has that? Because it depends a lot on how they say it, how they have worked it, the question, how they have... because he is a determinist. So here the same thing happens. Here in one case you see if the expertise is perfect, and in another, very bad. Take a picture of you dressed like a clown and tell him that you are going to a wedding.

And I'm going to say, hey, how nice, great, you're doing very well, it's very nice to go to the wedding and you're saying I have a clown. This does not contradict your posture and you're going to put it to contradict. I don't think so to use it this way. So it's good to think about it that way. That's why the ratification by the front. And I say we accept 100% synthetic information. What do you think? Well, doctor, I'm a fascist. I don't understand the question. If 100% information created with artificial intelligence. Why?

Because it has nothing to do with the cause. That's it. What we're saying is this. Relate it to me. Why are you using this model? That it's not an ornament. Don't come after me. It happened to me once in a session. It was a very complicated trial. It was my family. I work in courts, but by my own right I could...

And I remember that I took on a new judge and we reached an agreement. Finally, it was a mess. So there was a trial that had been going on for 17 years. I was already on fire. And there was an incompetent, two minors, one of the heirs had died. It was a scandal. In short, when we reached an agreement, I presented a written one.

You remember what legal design is? I present a script with a picture like this saying, look, at this date my grandmother died, at this date my grandfather died, this patina took me here, there is an incapaz, draw it. This is going to go to such a family group, this is going to go to such a group. And you know what response I received from that writer? It happens for life that I answer.

Now, I'm asking you to explain it to you because you're going to have to get 17 new bodies. I already have a kilo of hair, luckily, that I'm fixing and they tell me that my beautiful box is not worth it, I'm going to have to sign digitally. Here, I don't know if it was just the box, I said, hey, look at the cause, all this happened, I'll explain it to you better like that, I redesigned that, legal design. Well, it's always going to start with this. And how is it going to start in a responsible use, for example, in the police station?

Let's go back to the axis of all this. A criminal trial is already alien to law. If I don't have to explain it with a science different from mine, we have put it on the list of experts. The experts are going to be worth it, there is no doubt they should, they should. I often show you, for example, I show you the small system that I made here, I made audios. This system

I'll show you. You're going to see a face, a folder created with things that seem to make no sense. Why? Because I didn't do this like what is called the front end so that you can see a page that says upload this file here and cut it. If not, I did it internally. But what did I do here? I had certain audios. Look, I don't know if it's already there. I have these two audios, these two audios here to compare. Thank you very much. Well, what happened to me? I grabbed and said, oh, well, look.

I show it to you if you want to see how it works. I haven't opened it in a long time. Here, for example, you saw that I put Codex there. You want to see this, I can't zoom in. Here it is. What I'm saying here is I'm going to talk to you as if it were from ChatGPT to my folder that is there. So I put it. What audios does the folder have? This is where it is programmed. I opened my terminal. Why do I want to show you this? This, despite the fact that in reality I would have to use it to program, I'm asking you questions like that GPT of what it has there.

I'm going to list the current files in the folder and it's going to go looking at it. Hey, it has this. If you want, I can. Well, what is it for? I want to show you what it is that ends up solving me and why. This is called by coding. It's making code with natural language. Speaking. I'm making code here. If not, I would have to have put it like that. The same says there and start working with the list, as it says here.

I'm going to list this in such a way and put explore, enter such a part, tell me such another. Well, it says, load two audios and normalize them at the same frequency of sampling. Shorten long silence to analyze more the good empty space. Extract voice features. All this that are formants, chroma, that's what a expert does. A expert usually uses this validation term to see the voice. And what do I tell him? That he analyzes the similarity and that he generates a report. That's what this application does. Now, what stack?

Or tools to use for that. If I'm explaining all this, right? And I show it to you and I put it on you, expert, later so that you have access to these tools. What is the difference? That a expert does it or me? This already enters with a force saying, look, if you are telling me that this audio is not yours, look what it says.

My stack uses Python as the base of the pipeline, uses a book source to load audio, it analyzes in such a way, it uses a lot of tools that they use separately. That you are going to say, I give the audio to the expert, I'm going to use it that way. And if I go here, because I already did this to not delay the processing, notice what it does to me. It has made me a forensic report in text format.

Chiquitito, para que yo lo entienda. Mira, audio 1 y audio 2, lo sumo en básico. Me dice, me da la frecuencia de esto. Métrica de similitud entre A y B, menor, más. Me explica técnicamente esto. Entonces, en la conclusión probabilística, dice, hay factores para decir, che, si es esta persona. Si yo te muestro que yo tengo que seguir analizando esto, a mí me muestra el croma de la voz que pueden...

see it there, where the different types of frequencies are analyzed, the text unclips, shows me the spectrograms and this makes a comparative analysis of an expert expert without failing. Why? Because he uses audio mathematics now.

I did it thinking that way. I say, how do I introduce it to courts if you don't allow me to do it? Well, today you have a code that you have to give the opportunity for those things to happen. Only that you being a judge say what do I do when that happens? You see a leg that breaks it with this, I have made 200 tools and what is going to handle the whole process for that? No.

that explains how I made that tool, that I can use it as a professional so that the judge understands it, or I can also explore that type of tool to know what margin of error they can have.

Thanks to what? To the fact that on the one hand, the guarantee or guarantee is because you are going to say how this is valid. What model did Marco use? Look, the model that I have used for this uses this one that I have shown you, this library to do such a thing. Uncompress the audio, pass it to text, etc. and make a report under this system prompt. You are an assistant specialized in digital audio forensic, do this question. Well, audit me. See if someone is going to do this in another way.

What tools have you used for this? I have not uploaded the audio to Notwqlm or GPT and said, "Is this person?" No, look, I have prepared a tool thanks to the use of artificial intelligence, that uses artificial intelligence for this. Use it whenever you want, whoever needs it, I give it to them. And I tell them, "Or call me to the Law School, let's do tools, call me to the association, let's do our own tools."

Why? To facilitate our work. Let's really do it in the search of jurisprudence. Why? So that we don't have hallucinations. Let's do it with our tools. That's what I'm looking for here. Not depending so much on the external probability of social science, but knowing that I count on math and disposition using artificial intelligence. So I say, see that it's not just GPT. You are going to count on this. Believe me, you are going to count on different types of tools and if not, the expert is going to count on that. Now, what does this have? It still has to be validated by an expert.

We both in the production of the test, in the impugnation, in the anticipation test that has to be anticipated or anything we say yes, validate it with a expert. Everything you want to mark, but that someone comes here, then it puts something that is very interesting to start knowing about the use of artificial intelligence that works with electronic documents. The load of the test, something that is not seen dynamically in the use of electronic documents. But today I ask you, look at what a risk I am going to put, but let's see.

This image is artificial intelligence and it says that I put the image up, but what I find interesting is to show you that I, from this image itself, what the doctor said, the human eye allows me to warn what is happening here. Who is going to have an important responsibility in the electronic document today? Remember that before, in the general rules, I upload this document that I attribute to you. If you don't deny that document, what does it have?

The rule, the proof is true. Now, with the use of artificial intelligence, I'm not saying that. I'm saying I'm going the same way, brother, because of the volatility of this. Keep with the load of trying what you're going to upload with artificial intelligence. Your load is not exempted. Why? Because it is a very high risk, not only at the process level, but at the level of what the code itself looks for in a harmonious way.

It is very important that this is understood, because you are saying, "Hey, you are a lawyer, you are not a computer expert. Hey, Marco, very nice everything you do, you love everything you do, here this process is judicial." And it continues under these rules, that's why there are experts in the court, that's why there are certain formulas, that's why this is still the same here. So what do I say? Of course I can see that this is not generative, no, whatever, but Martin comes and says to Moni, "Don't tell me anything about this, silence."

I attribute to him that he is here, that such a question and I say this is generated with artificial intelligence, generative, I agree. Well, for that to really acquire a bit of validation in general, show me that it is an intermediate product, it has no autonomous value. This shows me how you have done it, what

What prompt have you used? What tools? As you give me more data, it's the same as in the electronic document. If you tell me which cell phone has extracted that audio, if you tell me which line is such a thing, if you show me that in the WhatsApp group there were other people and the witnesses have come, they have said I have read that in the group and it has closed all the tests because I say this has happened. Here it happens the same. So for me it is important because I believe in many codes will start to change the load of the electronic document test. In fact, the current code is today.

It's very strict in that and it tells you brother, you bring me a document that is mathematical and you deny it, you deny it to the other side. Watch out, I'm going to fine you. Have you seen the deny, deny, deny, deny? Of course, because you may not remember that it was a signature that you have made or maybe you did not remember that it has come out in such a matter. Now you tell me it's not me because that's generated with generative, put it or you tell me that audio is not mine because such a matter. Well, that audio is not yours because it's not my voice. I haven't said that.

Perfect. Now, it is proven that you have denied mathematics because I have compressed 0 and 1 to say that I believe that the judge has the power to say brother, you are wrong. Why? Because I generated all this legal dispute just to lie. Now you can lie totally, but it can have logical consequences in this. It has happened to me that there is abuse of digital law in this.

They scan things like this, like this, like this, turned around, 500 sheets. And well, just the most important one, the one that is turned around and turned around and put it somewhere else. Now, is it really access to justice? Does it not violate the right to defend? Bring me the things in order, digital hygiene principle, security, term of an unpolluted electronic file. So is it a photo or is it scanned with PDF? All of that today begins to import. So it is very good to interpret how we use that.

human

But what happens? We give 100% validation to human criteria in general. Can the judge separate the human criteria from a accounting science? Yes, it can. You have to justify it. That is in any code. It is not binding that a expert has told him that he has such incapacity and for that such damage. Great, thanks for the information. I decide with all the other evidence of the case. Here comes the same, casuistic. What is it that you are going to get a proxy or I don't know what, ask an IP?

to say that a expert is making a mistake when using this validation system that tells you the IP is such and such. So there is the important thing, not that a human or a machine says it, but how it has scientifically reached that result. I mean, I tell you, if an expert says it on one side and the expert on the other, which one is worth?

I also think that this is planned for now and for what will happen later. Now, maybe this is not what we are talking about. Now it would be more, for example, I have many cases where I am a labor accountant, accounting tests are done and the expert did not follow any technique at all. He did not explain where he got the information from, he did not explain what the step by step that I have done has been and he brings me calculations only.

If the counterpart of Gary tells me, look, I use artificial intelligence to argue why that expert did not follow the accounting techniques and the school will have to value it, it will have to give it reason. If the human expert did not do his job as a human and has been an artificial intelligence, the one who is realizing that the task of the human is badly done, well, he has a reason. And for me it is not that the comparison of the artificial and the human.

For me, they are still two humans talking, let's say the truth, subscribing to things. The issue is how that result came about. So we say, well, the processed income is important. You have come to 5x5 is 125 because five times we are five.

Well, here exactly the same thing happens. So that for me is a very important logic of reinterpretation. Hey, Marco thinks this, says Marcia, why? Because he has asked this in artificial intelligence because he has consumed it in such a way and that is why this result has come. That is very important for the reinterpretation of a judge. How is it a four-hour audience? We agree. Processing information in a tool that we are going to put on Google.

I transcribe it and copy and paste it and I understand that this is what happens in the audience. No, I have to see it, brother. I have to know not to lose the important thing. Now I have to see the whole audience and no, I can make a summary of the audience and yes, but don't lose the immediacy. You judge, you move away from the cause for using artificial intelligence, the generative is bad. Do it at the counter because it's bad. So what we're looking for is that the technology doesn't move us away, it amplifies us and there we go back to the same guarantee safeguard.

Today it will happen to you that a witness himself wants to incorporate an I to the audience, that he records the same thing, it is the third to the process. We have to start thinking about that, what does the witness do with that? Why? It is still a liturgical audition where it follows such principles, why does the witness come and put a cell phone to record now? Or does the woman accompany him and he is recording with the cell phone? Is this really how this works? Is it designed for that? There is a safeguard where the judge directs the process.

If for you this is unsafe, explain it to me. Now, if for you there is no problem, perfect, we all play it, every doctor has a problem, it is still the process of the parts, as long as we safeguard the use of the tools. For me it is very important that we take that part today, because there will be a thousand different uses. Yes, doctor? Well, look at you, look at you, the Maradona case. Even one of the lawyers of the Maradona case is using a lot of tools that we have given to process the case that now returns.

And what happens? And we will say, yes, it is used for the cause of artificial intelligence, it is worth it. One of the most media causes and that comes with a kilo of technology. I feel that we should not be scared, we should not be scared. Now, let's try to make this controlled. Why are you with a weapon here? You don't have to have a weapon here. Now, what worries you? Explain it to me. That makes the orality. Let's talk, let's understand why it appears in such a way. When is there talk of a rejection of the use of artificial intelligence? In the tests. It is a logical consequence of this, let's say.

What I was already doing, illicit obtaining, remember that there is an article that here already tells you, brother, the media obtained by the media, the media obtained by illicit means, out. Now, of course, that depends on an argumentation and many other things that make the case reasonable.

Proof that has been adulterated. It is necessary that I, if this has been adulterated in any other way, well, a tool that has generated an image with artificial intelligence and that supports itself with the rest is not something that is adulterated, it is something that is created synthetically. You also have to understand how that works. And on the other hand, it is.

What is the infraction to the presentation with generative rules? Once again, that you present and think that this has an autonomous value, that in short, you have explained to us how you obtained that information, you have not explained to us your PROM, what system you have used. You can use a thousand systems if you want, the free ones, the paid ones. But do you think that the process is the same as a 100-dollar cloud that Gemini, free? Let's find out, because you have to start knowing it.

Because I'm going to show you that you can make a tool in microseconds that calculates faster than a person. Now, that tool is valid. I'll put a link and use it. You can also play. Look, it's going to be more dynamic the use of technology, but we're not going to have to demand in criteria of reasonability. Not technology by technology. So that says this. Ah, I didn't fill myself with generative images that have nothing to do with it. That you're looking for, see how a person has suffered, a little boy crying. It has nothing to do with what you're doing here.

And another issue that is going to happen, which are the general rules of law. Hey, this person comes and begins to adulterate images of children to try to generate and well, he broke the system that should not allow you by default and generate and it's good that he does it. What did we do with the images of minors that we were given in a process before?

Well, it starts from before, because now I'm going to talk about the data, we are forgetting about the process law itself, which is a bit of a test, but now we are going to show things, then we are going to go back to the topics of how to offer it, what to do, we are going to go to documents. Why? Because there we talk again about that we are going to talk about documentary.

processed with artificial intelligence that allows me, and there are enunciative uses that the doctor gave me, summarize, interrelate, process information, well, perfect. But you don't have to go from the general principle of the missile that we have said before, how this intermediate product is not a test in itself. It's not that I pointed out to GPT what probability there is that a person who crosses in yellow is hit by a car.

Your theory of the case is that this person is using yellow, right? Now, do you think it really has a value what you are presenting there? You want to mark in the corner of Belgrano and they crash you in yellow. I explain where the reasonability goes, which is a bit like I was just commenting on the case that I have that this person drowned.

You know what the judge said? The judge heard a lot of things that apart from that, I said, hey, very nice, the defense is recorded, but I don't want to upload it because, by chance, the defense that intervenes is the same that my wife works. So, for the doubts, I'm not going to leave it. But the argument was that the Red Cross of London says that the reflection of the water can generate that a person has not been seen. I, oh, the pressure rises to me.

I say, "Oh, I used GPT very well, my son, to read things from London, but here or in London." And you know how the judge of the case has resolved it? He said, "Doctors, we know the swimming pool, some of you know it like this." I went to the swimming pool. It measures three meters. Look, look what the judge has done. It measures three meters this and here Amazonas gives you where the evidence that you agree with has been shown. It gives this here.

There are images where someone has accused him, he runs and passes over him, where little ones play with a float float. You are telling me that because the Red Cross of London says that responsibility is not absolute because there is a criterion that water can be crystalline. We are talking about a concrete case. He is telling you the leg has not been used, has not been used, has not been used, has not been used, has not been used, has not been used, has not been used, has not been used, has not been used, has not been used, has not been used, has not been used, has not been used, has not been used, has not been used, has not been used, has not been used, has not been used, has not been used, has not been used, has not been used, has not been used, has not been used, has not been used, has not been used, has not been used, has not been used, has not been used, has not been used, has not been used, has not been used, has not been used, has not been used, has not been used, has not been used, has not been used, has not been used, has not been used, has not been used, has not been used, has not been used, has not been used, has not been used, has not been used, has not been used, has not been used, has not been used, has not been used, has not been used, has not been used, has not been used, has not been used, has not been used, has not been used, has not been used, has not been used, has not been used, has not been used, has not been used, has not been used, has not been used, has not been used, has not been used, has not been used, has not been used, has not been used, has not been used, has not been used, has not been used, has not been used, has not been used, has not been used, has not been used, has not been used, has not been used, has

And surely, I tell you this defense, he searched five minutes before for an artificial intelligence argument on how to defend his accused of having been a lifeguard and the provincial law and all that. And he says they got him out of the Red Cross in London, right? I, of course, I value that judge a lot, who has not had to contact anything more than his reasoning. He has not had to say anything else. Very nice the arguments of the artificial intelligence that the defense used, but the specific case tells me that here a person of 1.85 meters, 46 meters underwater,

I have to sue you for the guilty homicide of this and give you these measures of correction, with nothing else. So I say, let's not replace that criterion. As much as I don't know very well how GPT works, as much as I don't know very much. Now, just if I don't know, worry about explaining it to me. Worry about you who want to use this tool, give it a certain validity at this moment of the process. That's what this is looking for. Let's be clear, let's be transparent and let's be sure. If for you that's not safe, the judge director of the process is not going to. For me that is key.

I say this, I upload images or videos with synthetic representations. If I don't declare that it is artificial intelligence, it is valid or not valid. This is what we have to start asking everything, because it is very nice to say, well, it will happen. And we have put it as a requirement, right? Hey, brother, you are going to use this tool now. What's going to happen? There it goes. The interpretation according to the doctor's advice, hey, you're cheating me. If you know that this is processed with artificial intelligence and

Exactly. Notice that these are principles before the use of technology. Hey, did you find something on the floor and you brought it to me and I put it here? Or do you really know that this is such a thing? I think it is very important to understand that as an order of the process we have put certain requirements of proof that seemed to be absolute.

At least at the beginning, because it seeks to give a lot of traceability to what is done. At some point it will happen that you are going to have a system where the prompt is not going to matter much because it is very contextualized. That is why many of those who do use the tool have put it. There are not two without three and perhaps those who do not use that tool so much go to another option. But now, today it is very important that if you use that, we trace the instructions from the outside and from the inside. We trace the data that has been used. So if I don't declare it and if I take that risk and

And then as a counterpart, as a part, what can I see? What can I notice? This part has declared that this is with a generative idea that I have to take into account how it obtained it, how it has worked. Let's go, the judge can also pass it to us, of course. And what will happen when we solve it? Ready? You saw that many times it happens to us. No one has said anything about this. He entered through the window and this video that nobody changed. As there were ten videos that were real, this eleventh entered through the window and I have to solve it. He came to the cause to solve it. I value it, I do not value it.

The parties have said, well, there is the interpretation. This has gone through a legitimate process. This has happened, you have seen it, you have come to another side, you too, perfect, you have admitted it, nobody has said anything, right? Ready. When I value it, what you are saying, I will value it with the rest of the evidence. If I want to use it, if I want to not use it. So I think that we do not teach that the law has no interpretation. So for me it is very good that each one interprets this in their place, because it will happen. We are human, it will happen for now.

What value does it have? You tell me, Marco, and what is it for? Then that people incorporate something documentally with artificial intelligence. What value does it have that someone has to process data from here to do something that does not exist? Yes, supposedly capable that the value of the judge. Let's go the other way around. I'm a judge. What value does it really have that I have processed the assistance plans here with what the camera and video from here? Ah, it has no value, something or if it has no value, it's good. Let's not talk about the value, maybe in the trial, because it's trial by trial, this, right?

But what value does it have to incorporate documentary, informative, informatics, witness? We know that it has value that a third party, alien to the process, who has no interest in this, comes and says something. We know that a private place that has information that can be collected objectively and brought to the process has, well, we know that if someone from another type of science comes and says, this happens, it has validity. And this, what value does it give? Let's start thinking about it, because it's now.

Does this have value or not? For me it has a very high value of conviction. That is, you give me a screenshot, by itself it is worth nothing, but it generates an indication. And another question, and for me also the speed.

I was not seeing 200 sheets of how the waiter has marked because I am already sure convinced that I have interpreted that when this name appears, such another passes and I will show you how to do that and you can review specifically the places where such a name I say has served me but of course you want the sentence to last 10 more days you want this as what we want with that being part of the same then I say

If I tell you 525 by 525...

You want to start, right? And well, if you use the calculator, it's faster. Now, it generates me a higher standard that you use technology so that I don't waste time. How much? Well, that's also this code. Look for that as a tool. It serves the process to both parties, the judge and the lawyer. Exactly. And with verification.

Human in the human heritage, in the human decision, in the human determination, in telling you how I have thought to process this information and have instructed such a thing as I am going to show you now. And another question for me that this is the greatest value that it has to incorporate now in this moment of generative artificial intelligence is that they are advancing what they see. You are already generating something homogeneous. Why? Because this is going to happen.

This is going to happen. A lawyer is going to come and say, I did this with artificial intelligence. And others say, I said this with the expert. And the judge has no argument to say what I do, how. It's okay, Susana is critical. I agree with everything you said. This gives you tools so that you, who are not you,

And you shouldn't initially be a techie, that is, understand how all this intrinsically works. Let these things happen that tell you how it is, because this gives you a certain security that the standards that you are using to evaluate certain issues are safe. You are advancing something that is going to happen. In what is anticipated production, the same thing. I think it seems important to understand that this is probatory material in general. So it's good to understand it that way. And there.

Maybe show them a little bit what we can refer to do a small activity and work. But, for example, I go to this cause. If I go to this cause, for example, I want to show a legajo. If this is the criminal cause, what I am going to show, I am going to work with the civil cause. If the insurance company listens to that, it is a good recommendation. If I go here, this legajo in penal, in Tucumán, the legajos are shared like this, by drive, one or two.

A horror. So if you ask yourself, I'm worried about the data of the person. Free cloud. I hadn't even constituted the belief, she sent me access. Now, there's all the complaints, at least they list them. That I like, it's numbered. But for example, here I have a lot of information. Here I have a lot of information that is oriented to criminal investigation.

This criminal investigation will give me a lot of data for my civil cause. We agree on that situation. So, for example, I'm about to start this, it's advanced, maybe in a certain aspect of the criminal part, I'm about to start the civil cause and I, for example, I think that being a lawyer, it's not going to be very useful for me. The criminal analysis report, this report that I think I should have here.

This that I have open, I'm writing something else, so no. This report, if I'm not mistaken, goes photograph by photograph, the forensic analysis of the person who entered here. So here, look, if I arrive, look, there are many things that are written, blah, blah, blah. Very nice, it's digitized, who is the person. No, an investigation has been done of one of the defendants. He explains to me who is who. Perfect. This part...

He identifies it in the place of the fact. I'm going to tell you why I think this is important. And I want to show you how I think I would use this information, which would be useful to me a lot, which is being done by justice assistants as well, to generate some conviction in the civil part, being a lawyer on this side. And if I go now, I would like the one about photo by photo. This one, report the photo of the photo. The videos are super impressive. They speak for themselves, really.

But if I go here, there was already someone who processed this information manually. And I say manually because I am convinced that the police are not using artificial intelligence. So they tell you, look, I collected all this, the scientist, and here this person enters. Look at the size of my client's son. Perfect. He enters, look, he says, hey, he enters here, here he leaves his little things.

This male supposedly puts him on life insurance, life insurance. This person is something that, for example, I human beings begin to interpret when I start to see who is who, because this is investigated. So I see that here this person, if they are 9 and 31 and 09 and this person is 9 and 30 and 43.

It is likely that here, notice that it says male 1 and lifeguard, it is not identified very well, but it is probably that he crossed with the lifeguard. The lifeguard was aware that this leg was 1.90 meters. He was aware. Why? Because he is showing me here that 30 seconds later the lifeguard is gone and this person arrives at this place of the pill. That interpretation for now is mine, human, right? From the brain. There is data there that are hours, etc. Look, at a certain moment the water enters, the water continues.

And he stands still, he had a maturation delay. He stands there, waiting for someone to say something. You know what I detect from the video? Nobody greets him when he enters. He is in the water. And these people that you are going to see here on the side, who are in the same level, those who are here, are the people, let's say, where the teacher is, that he, an ultra-poor person, ultra-poor, his psychiatrist recommends him to do swimming or weightlifting, he said he did swimming. He didn't know how to swim, it was his fourth class. But those people who are here,

They are the ones who supposedly have to be taking care of him, the occupational therapist, etc. And if you go 9:33, he had entered 9:31 into the water, these two minutes waiting for someone to greet him and tell him what to do. He was there. And evidently, at one point he says, "Nobody tells me what to do." He looks to the sides and as if it were that it does not exist, they pass, they pass. And he decides, look there, to get into the water. 9:34. He gets in like that, sideways. Look, sideways, look at what the level is. He doesn't go out anymore. 9:34.

I would show you the video, it is very impressive, very impressive. Notice that person you see who says he is there next to Gorrito, who is not identified by the forensic, is the main inveterate, he is his teacher, he is stepping on him. Now that is not identified here, I identify him, right? And notice if I would show you, he is still there, don't you see that? What a reflection of the water, it is very impressive.

Look, and a third one comes at 9:40, I show him that 9:32 he got in 9:40, another person has nothing to see and says, hey, you have to take it out, I'll do it for you. If at 3 minutes, imagine, well, this is not relevant. Notice that there is a question here. See this person who is supported here on the side of the float float? It is the leg that I was mentioning that went over, that he does not even get close there. Now, follow him and take him out. That person follows the side, here he is not identified, undefended.

Now, if I didn't have all these photographs, right? If I only had stories of what happened, if I had the income system of him, if I had the 911 call of him, right? Do you think I could, taking into account that I do have photos of the place where I have the andaribel, I have photos of him, of the size that he has, all the questions, do you think I could, and I can tell you very realistically, recreate that moment?

And I generated the same thing that you just did. Again, going back to what you said a while ago, no? What happens? I tell you that what I just showed you doesn't exist. Maybe today it's very difficult to do it, right? Yes. But in five years you're going to have a system where you can load your data and you're going to want a realistic representation of a space, a realistic space simulation. That's what we're going towards and those things are the ones that were treated to anticipate here.

The doctor, for example, says, hey, not to read, it's a tool. If you enter here, the Google suite is there, you are the nine dots, they go down, they have it at the last one. Not to read, if they don't put it as it is from Google, it will always appear first and it will be here. Now, what is it? It is not artificial intelligence, it is a tool with artificial intelligence, notebook, notebook, LM, great language model or as they call it in the final.

to the tools that make up artificial intelligence. These large language models, why? Because they first use the processing of natural language to process large volumes of data. That's why they are large language models. I put "try this", so simple, I'm going to show it to you. Well, I already have another case and I'm also going to show it. I put "create notebook", I downloaded two files in PDF, right? This and the frames and another. Perfect.

What is this going to do? I'll put it up here on the side. Keep in mind if you had a notebook at this time and you had written something down, you have the opportunity to ask them things. That notebook, as simple as that. If that notebook is empty, nothing happens. If this notebook had what I have there, it's the same. That's why it's notebook. Now, what does this notebook do? Interrelations sources. Here, for example, they are all in PDF, but they can make an audio in MP3, it can be a YouTube video, it can be.

And what do you want? The people of the Red Cross of London speaking in another inter-relational language with what the stupid defender has said? Oh, it's a shame. I understand that you have to defend, but it's a shame. Well, the point is that this has something particular, different from any artificial intelligence that you use. That's why I think it's good as a repository to start understanding where to process and extract data.

What does Google's license tell you? The general license when you create an email. Google cannot have access to your files unless there is a sign that you are committing a crime. Number one. Number two, we trust our life to Google. It's the biggest cloud. I'm telling you that here the criminal law considers the life of everyone else.

And number three, which is already with the use of this tool, I think you will start to value that. Good, nice, cheap. 55 sources where I do not use storage in the cloud, it is registered there and I can share this. See if we can share here with people so that they can see the notebook where I extract information and the way in which I extract information. But what does this do? See what it says here? Forensic video analysis, doubt analysis, doubtful death analysis, in reality, case of doubtful death, because that's how the initial load is loaded.

What does this do? It's called overview. Notice that without me explaining anything, this says this is a matter of investigation, where the Tucuman police on the death of such a person. The documents include a judicial office where this report is requested. This understands the context that I'm talking about. I'm not explaining anything. I haven't written it yet. That's very valuable to us. Why? Because when you upload a file in any of other language models like this one, OpenAI, for example, with ChatGPT,

The first thing you do is to make a file of this, let's say, notice that I do this and I put this, I don't say anything, as I did just now. Wait, it's loading, well, we'll leave it there. Here, what's going to happen? This, if I ask you, who is the judge that intervenes? First, no judge is intervening, and second, there is no judge, there is a judge. Let's try to break this situation. You know that it's an analysis, that it's a judicial test. It could, if I show you GPT now,

Here, Carlos Hugo says, "I don't see any judge." He doesn't mention any judge. But there is a prosecutor. Yes, this is the prosecutor. He understands that in the structure of the investigation there is this situation, but he doesn't give you information that he doesn't have. So why do I tell him about the systematization of information? Because here it is safe to extract information, but not so much to process it. First of all, the document recontrols a sequence, understands what is happening and offers you to make a historical summary.

You see what it says? Tell me how you want us to work on it and I'll move on to that. I can make a technical summary, a chronology, and look at what it says. Or transform an issue into a claim, a civil claim, or ask for a notebook LM. Do you know why it puts that on me?

Because this already knows that I work by extracting information from a place and processing it in a certain way. What do I do? First, a chronological report. I extract information from here and use the models so that they can be corrected here, here, here, here. But I assure you that there will be no error in the data load, which is what worries the law the most. If I would load the relevant laws on that here, it will not say another law. So here, for example, do you remember that I have asked and put here who is the judge who intervenes? Here there is or not the probability that this is in mind.

Well, look, perfect, understood that there is no. They don't see. I'm going, I'm going, I'm going. It seems to me that how long ago they don't see things in the call. Well, well, I have 200 open tabs, let's show something else. Full screen. There you go. This you know why? For me it worked the way it has to work. And this I give you a little tip. It's called persistent memory. If you come in here.

Configuration. Cada una de las guías que ustedes utilicen tiene un alma, les digo yo. Tienen capas. Si yo voy acá, este conocimiento está más carito el que le estoy dando acá, ¿no? Pero bueno.

The context window began to work a lot when this reminds something like our brain. This initially did not remember, nor the previous conversation. In 2023, I would say: "Hello, I'm Marco. How are you? Good, Marco, everything good?" And now, after that, I would say: "I'm Juan. How are you, Juan?" And here I would say: "You weren't Marco? Ah, I replace the information that you weren't Marco." It begins to remember. And you have different layers of customization of this and this becomes a before and after of using artificial intelligence. A first basic customization layer, I call this the skin.

in which this is not going to communicate, what are we going to talk about, what do I dedicate myself to and how are you going to act? This is the skin and the blood. Behind this, things start to exist about me, my heart starts to sound, "Hey, I do this, I use this in such a way, I write this, no." But in the end it says memory, I'm going to put "manage memory" and in third person, what is it going to tell you? "Coffee shop so much, if I go to IA, in the IA direction they work with prognosis and cyber security specialist."

This has been going on for three years now, and it's practically started to interpret what he has to know about me. Well, I don't let him interpret this. I make him know how. If I, in this case, would like this, for example, an article or many things, this you don't know, you know, if you made up your mind what to save. But I put I want you to keep who the prosecutor is in your memory.

You are doing that if I do unzip, how many times have you got a result with artificial intelligence after iterating, iterating, iterating, iterating, right? That's what you've done. In short, it's waiting for this to understand what you're looking for. Here, you already know this first move when you talk to me. That's why I upload an archive. You want us to do this thing, we do this other one. Well, here it says, let me verify this. Maybe I can share an image of text there. I'm asking for this.

that saves something from a PDF that you just extracted, analyzing and see that it says updated memory. I go in there and now this, when I ever talk to him about this, he says the legajo so much, the headline is such. This is never going to say that another person. So I work with a certain normative, with a certain way of understanding the problem, analyzing the cause of such and such. I don't like to lie information in the soul of this, because then they follow more layers of personalization.

You have the projects, you group things up. When you are a judge, mother, friend, well, in one folder you are a judge, in another you are a friend, and in the folder where you are a judge, you are a judge for this type of lawsuit, for this type of sentence. Same here.

We group our thoughts in the library. Everything that Marco is explaining has to do with a use that follows a methodology. In the chat a while ago they said to me, it happens to me that I have to write something with some laws and I ask Jack and he answers me anything. So it's better for me to solve it myself directly. Yes, it's true, because that use does not have a precise methodology. And here what Marco is explaining are specific techniques to say, well, in my

So,

For example, I create a civil case for death here. For example, if I get 200 civil cases, well, they go here. Why? Because I do use a project as a source. I put this on it. See that it says System Prom Project Civil Cause for Death. I'm hacking the system a little bit here, saying just read this file that will tell you how you have to behave in this place. And notice how I tell you that you have to behave in this place. This is called System Prom and there are user prompts, which are what you say. The system prompt. Ufa. Zoom. There it is.

So,

So they start to understand that this is not a car that you get on and takes you to such a place. My car doesn't take you to the same place you want. And this is good to start to understand it because that's how they're going to start working for groups, for lawyers, for legal studies, for places where they work. And this is something that I can show the judge and the judge saying, hey, this way.

I adjusted the engine so that it reaches this place. Because if I show you, look, I use it as a system prompt that this acts this way and does this. That's what I thought when, for example, I worked here on a public information access request. Cyprus, for example, I have worked this death with all the same data. He is the same person. They are the criminal data. But I want to know if the lifeguard was enabled, if the swimming pool was. I already work here. A part that I see here.

So perfect, the written here adds this. Notice that there is a writing of things to make my request for access to public information. Well, there are more things, let's say, to see from the practical, the truth. But I'm a little interested in first that we do an activity and second that we understand the phenomenon, let's say, of how this is going to start to happen. Why? Because not only if I enter here, look and put the request for access to public information and, for example, that I have worked here all this information with which I have arrived, I put here to share.

And here I have the PDFs that I have used that are in the case and it says copy the link. You saw such a disclaimer that it can include personal information, copy the link. I give this link to someone else, right? I say, hey, I'll add it to you so you can use this thought chain to enter this. Well, how do I verify that this is done or not? Does the link work?

Do you know what the APA safety standards are for citing that? At what date has the link been made? From what place? All this matters to give documents security. But here you are going to see how I have reached that result. Assuming that this I need to process, register things, I need to inform how I have reached this result. We will have to show many more things.

But finally, if we go to Gemini, I like Gemini a lot because we usually use many Google suites. If you go to Google suites, you will be able to use a lot of things from here. You will enter your Drive and this will not look for the file, it will look for the probe and it will interact. It will enter your mail, it will interact with many things, right? However, it has something very particular for me, which is also the fact of being able to create tools. Here a lot will happen. If I show you, for example, I show you a tool on the legal study page that I work from here.

Do you know what the hash code of an archive is? It's the DNI, right? If you go here, on my website it says online hash, you go there, you upload the archive, for example, I want to hash this photo, right? I upload the archive there, I press it and it doesn't work. That's how it works. It's always easy. Well, it gives me the hexadecimal code and it tells me also, look at this, generate the archive. It must have downloaded, if I'm not mistaken, a small archive explaining what happened. Not yet. Generate the hash.

Perfect. No, how do you generate an archive? Well, here, report. You generate a report where if I upload an archive, if I upload this report, or if you want to use this tool that is free, I enter here, this report is generated that says, let's see, I got stuck, like a puppy, you want me to show it, I'll show it. It says how I extracted, where I extracted, what I did, and they are traceable tools where I, without being a expert, am using a tool, I'm going to look for it on the screen.

A mathematical tool that says that at a certain moment this file had these characteristics. It doesn't need an experiment for that.

Eso quería consultar si en el escrito se puede dejar el código hash estipulado para que después lo pueda verificar o validar y la otra parte. Más validación al momento de hacer una pericia, eso. Acá pasa lo mismo. Si yo mando el link de cómo procesa esa información, puede verificar el juego y la contraparte, cómo va a llegar a ser el resultado. Si yo acá te muestro que no me está dejando ya descargas, la cantidad de archivos que tengo. Acá te voy a decir el código hash está hecho de tal manera, tiene tal cuestión. Tengo muchas cosas abiertas.

So this is also going to happen, look, if I show you the little tools, you know if I find any created here, there is something here called the canvas method, you enter here and this programs the natural language for you, it is called bytecoding, it will start to happen, I have here in some created, maybe you Franco, I believe you with your user, no, no, they are not gems, they are canvas, I want to look for the conversations, calculator, no, no, calm, calm, calculator,

Well, Argentinean legal and labor analysis, for example. I must have some more as a calculator. This, behind what you see there, for example, I see the labor liquidation calculator. Notice that I have asked this to you. The context, I want to work as an assistant, you will have this, your goal is to calculate compensation, the total amount of the product of each one, the tasks. Imagine formulas of time calculation, of damage, whatever you want. This tool I put here to share like this, and you put it there when you want. And what do you have to load here?

the amount, the remuneration, process the data and this back has code. Here if I press this, it has the code of how the application behaves. But notice that this has been generated because I have told him, hey, I want an application where I put this this way, do this, this, the other. And an application has been generated where I could generate the calculation and show why I have reached that calculation, taking into account all possible variables. I say it is not a form of use of the generative day. How do they calculate the fees? How do they calculate the damages? How do they calculate?

Now, repeatedly this is going to happen.

I want to automate what I decide to do sentence this for example this is papita so that you can upload the writing where you ask for the embargo and you do the papita sentence. It will always be the same, it will determine such a thing. Now it is traceable and if I can do it, how have we decided to change these places around here, that it interprets this and that it says that bank because I have the code with which it is really done, that I can, how long can it take me 20 minutes to do a tool this way?

So I say, if we are no longer encouraging these things can happen. Maybe I try to go as far as possible, it's the truth. The first thing that's going to happen is that people are going to start using artificial intelligence to process information. Hey, I've seen this information, I've processed it with what you've said here and it's incongruent that you've said that this time it's everywhere when this other one has happened and the witness has said, well, it's going to add a lot of value if you want to incorporate as artificial intelligence that you say how you obtained that information.

Not if you make a sentence, you have to say that information processing with artificial intelligence does not work that way because it is not the object itself. You are going to use that as an intermediary to say the problem of how when you need to argue that you did this with artificial intelligence. Not always, you don't make a calculation and say I used this calculator, which is what I was showing. I can say I calculated this way. Look at this a little bit. Maybe what we can see now.

of different uses that are going to be given within practice, in the oral part, in the general part, in the search for information. There are many techniques of search for information that I would like to be able to comment on at some point. And here I have decided to deal with this part, that this is not for the warm. Here I am tired of the shiruzo of the grape and saying the data, the data. That's because they don't walk on the street. That's because they don't know it's not like that. And another question that is very important to me.

The personal data protection law. Today, your article 5 tells you, brother, you are a republican power, you can process data. Now, what data? I, for example, am a judge who wants to make a sentence with clear language and that it be a video with notebook LM, I upload the sentence, I want it to be a video that is shared by the parties and the sentence is explained, right? Well, I can't do it. Why? No, you're processing data from a company. What do we think of that? The issue is that the data is not processed.

Now.

The differentiation of this was for me before, I do not talk about artificial intelligence again. You are commenting on the causes where you intervene saying: "Hey, he is complaining because he has received a report that he has AIDS and he actually does not have AIDS." And you are commenting, shouting that. Well, this works the same. What I can't scream and say, why do I introduce it there?

Now, is it necessary? Can I? Yes, I can. Is it as harmful as it seems? Show me a damage. What is the damage? You think that if you put "Marco has AIDS", tomorrow they will ask you, "Marco has AIDS?" If I show you how it works, you will say, "There are no two without three." This processes information like a blender and gives it to you. It is not a library that learns. So, it is unconceptualized by many lawyers who do not know how artificial intelligence works, the myth of data. Of course, the data will always be the most important item of the state, but let's not be hypocrites.

Now, what I recommend is a matter of professional criteria. This is an electronic contract. Meta has an electronic contract that can't use your data. Just like Trump said, give me the data, Mark Saigert, put me the guitar and instead Facebook to call Meta. And well, we have the Cambridge Analytica case that has been sued for using Facebook data. It may happen that they use our data and it doesn't work. Now, you know how much we need to have local processing of our own to lose the opportunity of this?

There are many ways, I believe that the judicial power of Catamarca, the Código de Abogado, will find that. But if you ask anyone in society, hey, you, I prefer that I talk to you and process your data, but your sentence comes out before, what do you think they're going to say? But they do a lyrical scan for a company for 30,000 pesos. So what do I say? Neither one extreme nor the other. This is not libertinage. What you can't really shout for me, I recommend you not to upload it. Now, anonymize it if not.

Regarding that, I would like to make a contribution that at the international level we have the law of the European Union that entered into force in 2024 and talks about the risks of the use of high-risk systems. It seems to me that this is a good parameter to lower it perhaps to our reality

But there really is not these data of use, the name, the ID, the truth is that they are public everywhere and a lot of sense does not have protection, but it is already devalued to that extent. And here it does mark guidelines. What data? In reality, not the data, but the use.

For example, that they exploit vulnerabilities, for example, children, disability, in all those issues. Then also, as far as they are used, for example, to score socially, for example, they are used in biometric issues, as you said, just like the one from WorldCoin.

Biometric issues that are used to categorize social standards, for example, because you are of a certain physical shape and they captured your biometric image in such a place, they position you in a certain social category, for example. Those systems are high risk. Yes, yes, profiling too. And why for me it is not so important that this is found here, because there we are saying that that profile takes a decision of a profile. Here we are saying specifically, this does not decide.

Here the human is the one who decides. So, regarding this, while this does not make a decision of who is legally responsible, if it is theft, theft or if it is theft or guilt, for now, I understand now as a general rule, without a doubt. I do also want this and finally, automatic resolutions. Here it says, hey, do not solve, do not delegate the resolution cognitively. This for me is very important to be able to corroborate it, unless there is a system designed for that.

And the system designed for that will be validated just as you are validating these things. Hey, this happens, this happens, this happens, this does have to exist. For me, there should undoubtedly be certain resolutions that I can put people to do other things. It's the truth. Just as it happens in legal studies. And I always say, can we associate that with the decisive writings? Because here we tell the judge, hey, you decide with artificial intelligence, but hey, lawyer.

Because if I write a process and put a bonus, a job, if I give a license and I forget saying, do you want the next one to write you with a more formal language? Is the process really hurting me? Now, on the contrary, this is the demand, this leg, evidently, has used this, the process like the orto, has not warned me.

It's really more serious, isn't it? I think we need to talk about interpretation of this. If here at the court we are telling you don't make decisions, don't delegate with a commitment, what limits do you think a lawyer has to delegate with a commitment? This is where I think it's good to always think of it as tools for that. That's why we're going to do this later. of computer sovereignty. At some point you're going to have the relevant tools. In other words, at some point we're going to have the open recipes of how to make decisions. But as long as we don't have it, we don't have it.

While we don't know how to put this out, we don't do it, but not for that reason. We are going to lose. Let's see if we have not regulated rapi, Uber, moto and free market at the time and they were already selling, brother. It's better if it serves us to have work, we give opportunities. The e-commerce was great. Well, we have to start to adapt. The same thing happens here. I mean, here we don't kill the messenger, let's say, if the equipment is not the problem, where we incorporate it is. And well, I always mentioned it before. This is a topic on artificial intelligence generative in the state.

up to 2026. Why? Because I think and I think it always made us afraid when we talked about this in the commission, that this is old to the touch. And I think it's very well profiled because it talks a lot about issues once again, how the car is driven first. If I talk about the age of agents, you are going to start using artificial intelligence agentica in this way. What is artificial intelligence agentica? One more branch of the generative, but that it is an autonomous agency. Task. So this

For everything else that happened, I ask you for the prompt because you have had to do this and activate buttons for this to happen. When I show you that here not because if the agency is where you have to worry, for example, look, if I enter the Tugman judicial power page, I pay a subscription of cloud that allows me this. There are some subscriptions that allow us to. Here on the side, you can see that the cloud opens. I have it here, a chat. Here, if you have a Mac, you can have it. Connect it to me, I'll pay for it.

They can have it at Atlas, which is like Google, but with artificial intelligence agentics. I'm going to open it here. Everything starts to incorporate this to the side that works like a robot. I can show you that if I send an order from here, this will start to move my mouse and act fully alone until I get a good look here in the judicial power. I want to give you access to me first, because if not, this does have a small limit, sometimes passwords, but here it already has access. So, for example, I open Cloud here.

Here I have my files, the search for things, the management of what I could do. Besides, I'm going to go back, back still. And I tell him, look, I need to verify the last notification of the case of Chala with Paredes, the last process state. Leave it to me written in a doc with the recommendation. Look, I'm going to let him decide the impulse of the case. I put here, I'm going to have to act without asking and ask before acting. Act without asking.

With this I can make this give it my OTP, upload a script with a digital signature, everything you can imagine, but I want you to see, it turned orange on the side, you saw? And this starts to work. By the way, what do I do here? I go to the other side and I keep working, right? You ask me how I do so many things and I use a lot of people, a lot. Notice that here it goes saying, I clicked here. It's terrible clicking. What do you remember? The robot video grabbing an egg?

that was celebrated for precision, here the same thing happens. Here for them it is very difficult to capture, to get crazy, to press buttons that are badly done. Then they make screenshots, they process it. And here on the side, many things are happening, look there. But I'm not interviewing anything. Here I'm seeing that something is happening. He says, stop it or not. I leave it there again and I want him to analyze where he's going to go to look for that. He can enter in a file consultation, he can enter, he starts analyzing all the buttons that are here.

En mi interfaz. Ahora, por ahora, hubiese sido mucho más rápido que yo entre mi expediente, me fije en la última actuación, ¿verdad? Ahora, si esto fuese de 50 expedientes, de 100, de 150, de todos los expedientes que hay en el PJN, de todos los expedientes en donde alguna vez se haya condenado una aseguradora en un atatorio, iba a entrar expediente por expediente, de aseguradora por aseguradora, a buscar sentencia, conforme yo lo prompté. Fíjense que yo he hecho un prompt, hace esto.

Imagine the problems I showed you earlier about the system. As a specialist in such a matter, you're going to do such a matter. This is my goal. I want you to enter these links, these links, these links. Look for this. Enter this person's Facebook. See if it's the day. And this little robot starts to enter a lot of screens and starts to do things.

I don't say what to do. I don't say enter this button. The goal is that you give me this information and this will arrive as it is. The goal is task agency and that's where it's big. This could be working for me all day. It's worth it. If I show you how it works in Atlas here, which is this is only for Mac and if they pay the license 20 dollars, I think they also have it. There is another one called Day. It's Comet of Perplexity. This one that I like the most.

It works very well, it interfaces very well, it understands very well what you are seeing. But for example, I want to show that, for example, if I don't want to use the agent part, I have a thousand things, it will explode my Ram. I'm going to close this copilot in the Microsoft, the copilot Microsoft interface, but I don't like it very much. I don't particularly pay for it because I use the cloud a lot, I use a lot of docs, but I don't want to take away what my friend is doing. The star. Look, it's still on the screen, it doesn't know where it's still.

Establecer entrada, está pensando qué me está diciendo el usuario. Para esto está haciendo una tarea compleja. Fíjense que abrió estas pestañas ya. Familia y sucesiones. Yo no le he dicho si es civil, si es laboral, si es penal. Y está buscando familia. El robotito está actuando ahí al costadito, fíjense. Ha puesto el demandado, ha puesto el actor. Che, y ahí me va diciendo acá qué está haciendo. Acá decidió abrir una pestaña más y va a decir, che, como no está en familia, voy a otro. No le he dicho en qué centro judicial. Yo no le he dicho nada. He dicho, este caso, decime.

and you will continue to search until you find what you are looking for and you will accomplish it. If I had told him it is a civil and commercial case, the number of pendants is so much, it is better that now the doctor has said grok, I will say Gemini again, we always go with grok or perplexity. Look at this nonsense, how much efficiency can generate in many things and how much can it damage in the use of all this. While here he keeps thinking that he is looking for a job, he will not find it. I enter grok, grok is twitter, let's say, it is the idea of twitter, of x.

If I enter here, there is something called look here, they have projects, just like I have shown you everything, but it has something for me that is very cool. If you expand this here, which are what are called programmed tasks. I enter here to my user, you see that it says tasks. I enter here and every day I receive an artificial intelligence report for a lawyer on TikTok.

Every day I send an email saying, Marco, look, these are the trends that the lawyers are talking about in Tiktok about artificial intelligence. And this is free for me. It has two weekly tasks. I can create a task and what do you want me to do? That I check what? That I do what? That I remind you what? That I look for information from here and the process is programmed. Every day I have a daily task, he tells me. Now, give him the following. Give him information to prepare you for what you need to prepare yourself. If this has access.

So you're going to have a little report every day of yours with a programmed action where you look daily. The last thing I've done is this. Look, I did it today at 8 in the morning. Tendency actual of TikTok. And it's talking about this. It's talking about this question. And it gives me this report that I've reviewed on TikTok. I have to go to TikTok and look at that situation.

This is the same thing that happens in Comet. There are many artificial intelligence tools to delegate tasks that we can perhaps delegate in person. It is still a condensed administrative, my life will continue to search for civil and commercial. That's how the puppy when he wants to do the task. Now I ask, I don't think this is safe to use today. Notice that if you enter here, this will tell you, of course, and you can make the mistake and verify the answer. It has 200 disclaimers that it is in beta. I enter here and enter to read once the electronic contracts.

And you say, hey, why does it respond badly or why does it do such a thing? And it has 200 disclaimers, why does it do it? If not, it receives 200 demands. It doesn't have to give you a precise answer. Because this is not a library. I mean, what have we done with Google? Google, you ask, you touch, and there's no chance that if you put here, there are no two, without, because there are no two, without, it doesn't appear to search, but it's not a phrase. It understands it because it's a movie that maybe the most people, people look for this. This is a library.

The other one, no, it's a bit like what the doctor did, they are totally different tools and they work in a different way. You don't google here, keep looking for my brother, my document location, no, they are looking for it, but until they keep looking, we advance, we are already finishing with a single question, which is a small activity of ten minutes, two minutes more. I'm not going to happen anything, then. What would I like here? I mean, I've explained everything we're going here, I don't know, I think it's good. Sorry.

We are talking about persistent memory, I don't know. Yes, I would like this to be understood. Let's say, if this code pretends this, that the guide is at the service of our human brain and the process. If we have it like this and it is not a brain that really contributes external things, but it is part of the initial brain, whether you are a lawyer, a litigant or a judge, it is a correct use. That is, the primary is that, the primary is the process. Am I letting something alien to the process really enter?

Well, if I'm doing it that way, it's wrong, because here the AI does not argue, it does not tell me what to decide or what to ask. It's the first thing I have to decide. Now, if you would have to incorporate into your Gemini, if you have that, Franco, no, the image. If you have to incorporate an artificial intelligence file into the terms that we have seen the admissibility, enter any generative IA that you use, go in.

Also through the chat. But it will cost him a prompt. Then pass me the bill. I'm going here, I'm going to Gemini. You did it with Gemini, right? Yes. Look, Franco has a daily summary of his judicial career. Look. What did Franco do here? What did I ask him? That in the end he works. I think it would be here, right? There has been a penalty. Well, here Franco says let's do a representation of a human body in an anatomical position, referencing the injuries suffered by a worker that affect 40% of his working capacity.

What has this done? An image that explains in legal redesign what you want me to explain, right? Franco, why do you use it in such a generic and unprecise way? Because it is not a cause in itself, right? But if I had to contribute this, Franco, what would I link it to? I bring you the real data of the student. I have all this information. I want to make a visual representation of this, how the injury looks in the body, what is the affected part, how it has been.

This can be received by a judge because he is explaining to you, just as he does in an argument, in a paragraph, that this was processed with artificial intelligence by these parameters that I had been processing, I did not bring it. So Franco says, in notebook I have processed this information of the cause with certain medical parameters, in Gemini he did this. Now, why? It is necessary in this case that it does not show something real like this, and perhaps not so much, but he knows that it generates a lot, that this is not a drawing that Franco has made.

This is not something that has information that was created out of nowhere. So, explain to Franco what is what follows now. The image that the companions are going to do here. Well, now the idea is that you try to think of something like that. How could you use for the causes you work with or that you have had to solve some kind of argument that links the evidence that already exists and that represents it in a different way, but faithful to that information? Look, for example, the Kennedy assassination.

And it shows you where it takes it, variables, right? Now it's a real croquis, this could be done. Oh, I've finished it. No, but make one of yours. No, I copied mine. No, I copied mine. Invent. Yes, well, Franco, take it. Now, we have 10 minutes for that. I would like you to see what is the creation of a synthetic image, but represented in a case.

represent in a certain aspect here Franco has chosen an incapacity well, in the case of the swimming pool, I could have represented, for example, the lungs of a person who is six minutes underwater taking into account certain medical reports, taking into account the autopsy report that he gave me for immersion well, there are some lungs that pass two minutes, three minutes, four minutes, six minutes, it does not generate something different scientific, only a demonstration that the person who has opened the autopsy must see it

and I can represent it with images that are with something specialized and with information about the cause. The intermediate product that is generated, right? It has no value in an autonomous way, but it has been linked to the tool of the causes or with what it had and that here could generate elements of conviction at the time of presentation. Did anyone not understand? Yes, now. How is the group? How are the people in the Zoom?

foreign

foreign

um

It's very difficult. It's practically impossible. It's practically impossible.

The fears, in reality, all the fears that there are, derive from an interpretation of the personal data law of the 90s with what is happening today. That is the confusion that there is.

foreign

-

foreign

- -

Another question, sorry, it would be good if you just saw it, that there I consulted the doctor, for example, with some issues of uploading data. When you go into any configuration here, you have on the margin of this, it is called data controls.

Here, if you want the model to improve for everyone, you select this and this definitely says I share the information of how you processed this so that later it will be re-entered in the model. Technically technically. I don't know why they come out, I don't know, they're not so good. I'm showing you is to enter the configuration. No, I don't know if it came out or not, but I didn't get it.

In definitiva, estoy entrando a la configuración general de los datos donde decir, che, este modelo se puede utilizar con los datos que vos le das, se puede mejorar. Si no, es lo primero que te va a permitir esto. El modelo gratuito,

has this situation where it says, hey, we can model certain data with what you want, the 20 dollar one. Hey, these data, you're going to choose what can be done and what not. And directly in the one that follows the enterprise, the data is yours. That is, it is practically like that. It says, hey, I give you the space of the cloud for you. Now, why do I speak of Google? And also the same as the doctor. The electronic contract is another. Here the file is not entered, it is like having it in your cloud.

This does not know if you are saving in the garage a car or a truck. Yes, in this practically in Google is very similar to the free version and pays in Europe. Changed a little in this, we don't. And what happens here? This does not know if you save a truck or a car. You know that you are saving something of weight. Now what I think. So I say you can extract information from here. What I was showing you here. I said I need.

The information of the cause of death, all the medical information, and that it does not mention the data of the deceased. So it will look for everything from the autopsy, from whatever, and it will give me that information if I want to anonymize something in here to then process it in another artificial intelligence. But well, this is part, as I say, of a lot of the professional activity of each one. What do you want to use as a tool to get a result? I...

I consider that there is a bad interpretation many times of our personal data protection law of the year 90 to what is happening now.

Yes, I obviously believe that there is a responsibility, without a doubt, but it is very difficult for this to be obtained as information that this person died of suffocation in such a way at the time of their reentering. There is a very low probability that this will happen. It is like this because what is most focused on this is the behavior, how you reasoned these things, not so much the data of the reports you used. Imagine if you have to process that to retrain.

that knows about this person with all due respect who has died in such a way is very rare. The reentry of this does not work like that. And another question that invites me to say this is also how many cases do you know of this, that having entered, we have been going for three years, data and artificial intelligence generates a leak. How many data leaks have you known from other platforms that have to be saved here? From this one, none. Well, those who were able to do the activity and work on that will see that it is not that easy.

It's not that easy to generate an element of conviction as an intermediate product. It's sitting down to think that today you have a tool that can connect things. Now, with the specific case, it will come to you and the more you use it, it will come to you. Hey, I can process this spreadsheet with this and generate this. Yes, report, you're going to generate that. Now, it's not that easy. It's not a little drawing, because if you don't look, it generates that. It generates that idiot of uploading this that looks nice. No, we upload things of value.

Totally. And you're going to realize that this is going to work like a brain assistant of yours until you have the right tools of the judicial power that wants you to act this way. And you know where the problem is going to be when the judicial power or the courts have that tool. Who decides that this generates this result? Is there really the knowledge of the natural judge to make a cause? Or is it being done by the computer head who is saying that this has to act in such a way?

Today we take advantage of this cognitive sovereignty in this way, because when specialized systems are lowered, it will have a lot of value what you decide to order. When you say, I got cold food and I decide it in a system, if they give you back the money, well, to not get to that situation in law, it is very important now to work this, not delegate cognitively, and that's what we're looking for here. Thank you very much for all your time, it's been a pleasure, and we hope to continue to train you in this. Thank you.

No.

They must have very specialized things for certain moments of the process.