Sesión 9

Good morning. On behalf of the Digital Transformation Agency, Infotec and the National Technological Agency of Mexico, I welcome you all to session 9 of the digital competencies training course. My name is Thelma Leonor Estevez Durantes and I will be with you in this second class of the artificial intelligence module.

Before we start, I want to remind you that this session is only for the course. It does not replace videos, readings, or activities on the platform. Its function is to help you organize ideas, clarify concepts, and help you recognize real applications.

The group can be very diverse. Some people have already used artificial intelligence tools or even know technical concepts. Other people are just starting. To those who are just starting, I ask for your peace of mind. We will go step by step from the beginning. To those who already have more experience, I ask you to take advantage of the session to organize your basic vocabulary that will be necessary for subsequent qualifications.

Today we will discuss some topics that are studied in more depth later, such as language models, diffusion models, agents, tools and human supervision. We will not go into programming and technical architecture because this is a propedal class that only starts for the subsequent certifications. At the end of this session, I hope we can explain five things

that I am going to tell you about. First, we are going to see what is artificial intelligence and what kind of content it produces. It is important to recognize that we are talking about new content such as texts, images, audio, videos or initial ideas.

Second, what is a PROM? It is a very simple level. The PROM is an instruction or an indication that we give to the artificial intelligence system to request a result. Third, we review examples with tools like ChatGPT, DALI, Copy.AI and Copilot. It is not about memorizing marks, but about relating each tool with the task it can support.

During the entire session, we will be reviewing these QR codes. In this case, this gives you the route so you don't miss out on the sessions that have already passed. And this session, if you couldn't see it in...

live so that you can have the link to each of them. At the end of the session I will be presenting a QR code of a special session that was done on Friday to explain everything related to the exam that you have to present to credit the course. So, well, if you can see this QR code, there we will be seeing all these situations.

As I was saying, the purpose of this session is that we can explain five basic ideas. First, what is artificial intelligence generative and what kind of content can it create? It is important that we talk about new content such as texts, images, audios, videos and main ideas. Second, what is a PROM?

which is a very simple level, a promise, an instruction or an indication that we give to the artificial intelligence system to request a result.

Third, we review tools and remind you that it is not about memorizing brands, but relating them to tools that can support them. The ones I am going to present are not the best or the only ones, they are simply the ones that were chosen for the examples, because otherwise we would never finish, but it is not about them being the best or the worst.

In the fourth point, we will distinguish the artificial intelligence generative from the agentive. The generative produces content while the agentive can organize steps to try to achieve a goal. And finally, in fifth place, we will remember that every generated response must be reviewed, contextualized and evaluated with human criteria.

Without this, artificial intelligence results would not be correct to use. If at the end we can recognize these elements in everyday examples, we will have fulfilled the purpose of the session. This is our learning route that we will follow for the next 90 minutes.

First, it is divided into three blocks. The first block is to review what artificial intelligence is and how it works in a basic way and what role it plays in PROMPT.

In the second level we will talk about tools and applications and there we will see videos of ChatGPT, DALI to identify each of the instructions and how it is that it can generate texts or images. In block 3 we will introduce the angentive guide, which is: how is it differentiated from the generative intelligence and what are the basic components and why does it require human supervision?

Then we will have frequent questions to review and if the public decides, we will answer the questions that you are sending us in the forum of the course, which is our direct communication. Meanwhile, we will wait with questions of toners to recover the main concepts.

Finally, we will close with a cumulative review of the two sessions to prepare you for the exam. And an important recommendation, although the exam is already available, those who are still starting, I suggest that you should review the previous sessions and accompany those that are still missing

We will accompany you in the ones that are still missing before using these attempts. Remember that they have limited attempts and then they must be much safer. This session still needs two sessions of data science, which would be very important and very enriching for those who are starting to wait before presenting their exam.

You will also see the QR codes at the beginning of each session and at the end we have the specific explanation session of the exam platform. So, before we start again, Carlos, if you can help us please with the QR, we would be very grateful.

As I was saying, this QR allows you to access the link where the classes are already shared. This resource is especially useful if you could not continue a live session, if you want to review a topic before the exam or if you need to reinforce a concept that is not yet clear.

These codes do not correspond to the exam and it is not a list pass either. It is just a consultation resource to support your learning. We are going to leave it a few more seconds on the screen so that you can scan and save this link. Ready then.

Let's get into the first block. Generative artificial intelligence is what we will see now, its concept, its basic functioning and what a PROM is. Here we will focus on three basic questions: What is artificial intelligence? What kind of content can it create? And why is the PROM so important?

The idea is not to explain the deep architecture of models, but to understand the general logic. A person gives an instruction and the system generates a new result that must be reviewed later. Let's go now to this key idea. What is artificial generative intelligence?

The generative AI is a branch of AI capable of creating new content from instructions and patterns learned during training.

When we say new content, we can talk about texts, images, audio, videos, codes or initial ideas. For example, you can write a draft of an email, propose titles for a marketing campaign, generate an image from a description or help create a script.

It is important to distinguish it from other artificial intelligence systems. Some systems analyze, others classify or recognize information. Generative artificial intelligence produces a new output. This output may seem very convincing, but this does not mean that it is always correct.

From the beginning we must maintain a rule. This rule is always very important. Generative artificial intelligence can support work, that is the idea, but the result must be reviewed, contextualized and evaluated always with human criteria.

In more advanced levels of subsequent certifications, they will study in much more detail how these models are trained, what data they use, how they are evaluated and what specific limitations they have. So, well, we are left with the part of creating new content and that human review is very, very important. This table will help us avoid confusing, make a frequent confusion.

Not all intelligence is generative. An artificial intelligence, analytical or predictive, focuses on analyzing, classifying, recognizing or predicting information. For example, a system of feelings analysis can review comments, classify them as positive, negative or neutral, but nothing more.

The agentive artificial intelligence, on the other hand, creates new content. It can produce text, an image, audio, video, code from an instruction. The agentive artificial intelligence takes a step further. It can receive a goal, organize the steps and rely on tools to try to achieve a goal. Today we will see it at the introductory level.

This comparison will be important so as not to confuse examples. A video of an analysis of feelings or a health application can show the applied artificial intelligence, but not necessarily the generative artificial intelligence. What difference do you think is clearer? Analyze, generate or act?

Artificial Generative Intelligence can produce different types of content. In text, it can generate answers, emails, reports, ideas, summaries or explanations. In this case,

more visibly in conversation tools such as ChatGPT. In image and audio, we can generate illustrations, design, voices, music or sound effects. Here are some visual or multimedia creation tools.

También podemos apoyar la generación de video como guiones, avatares o código inicial. La idea importante para el examen no es memorizar todas las herramientas posibles, sino reconocer que la inteligencia artificial generativa puede producir diferentes tipos de contenido, no solo texto.

Even so, all production must be reviewed, I insist, it must be reviewed by a human. A text can have errors, an image can contain biases, or a response may not fit into the context, and this is always decided by a human. Let's see how it works in a simple way.

First, the model is trained with many examples. Those examples can be texts, images, sounds or other data depending on the type of system. Second, during that training, it learns patterns, styles, relationships and structures, not like a person who understands the world, adjusts internal relationships from its data. Third, when it receives an instruction,

produces an output and this new output is based on what has been learned previously. Fourth, the quality of the output depends on many factors. It depends on training data, the model, the context, how we put the instruction and the subsequent reviews where adjustments have already been made. In the certifications that will be studied later,

They will study neural networks, parameters, language models, evaluation diffusion and security. Here we will only stay with the general logic to recognize how they are used. So we will avoid giving very technical architecture questions.

As I was saying a while ago, the PROM is the system instruction. Let's review what the PROM concept is. A PROM is an instruction or an indication that the person gives to the artificial intelligence system to generate a result.

It can be a question, it can be an order, a description or a request in specific conditions. A very, very basic PRO would be the one we have here. Make me a text about artificial intelligence. The tool can answer, but probably the result is very general because we are asking for something very general. Make me a text about artificial intelligence. That is very, very basic.

A better PRON can be the one we have here on the screen.

that includes more information. As you can see, this includes the audience, the objective, it has a specific tone with which to speak, the extension we want, the format in which we want it, the context and the criteria. So, it writes a brief explanation about generative artificial intelligence for beginners students, with simple language and two daily examples, it is a very

much more robust than making a text about artificial intelligence. Therefore, the result of this must be much more

to what I need than the previous one, which was very general. This is achieved if we are very clear about what the PROM structure is. This structure is a structure that we are proposing because it is the most extensive structure of how to make a PROM. If you review what different platforms propose,

artificial intelligence, they will suggest what are the techniques or their proposals of techniques to generate a prompt. Some only propose four steps, some propose three steps and some propose up to seven steps. This is the longest we've found, perhaps the most complete, and what the others do is not omit it, they simply group them in fewer steps to make it easier for those who are going to do it.

This is like any skill that you develop in life. Practicing and practicing can make you more familiar with it. So, what I propose is that you make a template with these labels in a column so that you can write in the next column all the PROM. Once you have done it, you put it in

And then you can play with if I don't put the context in, if I don't put the criteria, if I don't put the objectives, what is it that returns me? So that then you can develop the ability to generate these prompts as complete as possible. As it is more complete, the result that it generates back to you will be much better than what you are waiting for.

So I was telling you, the difference is not in using complicated words, but in communicating with clarity. Clear communication is what we need. Let's remember this central idea. The prompt means an instruction and although the prompt is good, the answer always needs to be reviewed. So, of course, the more skill we have in the

in the PROM's writing or in the PROM's dictation, depending on how you like to use artificial intelligence, the result will be better, but I repeat, the answer always needs human review.

With this we close the first block, we have already reviewed what is the artificial intelligence generative, what content can it create, how it works in general and what is a PROM. Now we will make a brief pause and Carlitos if you help me with the QR.

I repeat, this QR gives you access to previous sessions. The QR is not a list pass, it is a consultation link to review classes already given. We will leave it here for a few seconds and we will give five minutes to stretch your legs and have all the adjustments required on your computers. We are ready. Very well, let's start block two.

In this one, we will see tools and applications. Now we will move on to the definition and examples of text generation, images and other content. We will also have two moments to watch video, one from ChatGPT and another from Dalí. So I'm going to ask you...

Mientras observan el video que voy a correr a continuación, les pido fijarse en tres cosas. Uno, la instrucción. ¿Cómo se escribe la instrucción? Dos, la forma en que el sistema genera esta respuesta que van a poder observar en el video. Y tercero,

that the interaction occurs immediately. Text and conversation will appear completely. So let's run the video so you can see these three little things that I'm asking you. It's very, very short, so let's go.

is

that can generate and support text and conversation tasks. It can help to write email, reports, explanations or messages. It can also help to generate initial ideas, organize ideas, summarize a text or reformulate an explanation. This does not mean that we must cut and paste everything that produces.

This answer can be useful as a starting point, but it always needs a review. I insist, it needs a human review to see if it can be applied. We must check this information, if it is true, if the tone is adjusted, take care of the context and make sure that the content corresponds to the purpose we are looking for.

In the context of the course, it is important to recognize that a tool like ChatGPT is mentioned just like this, as a tool that allows you to talk, write texts,

Solve doubts and generate ideas. In more advanced levels that we will have ahead, you will be able to study prompting strategies, verification, data security and professional use of these, as well as conversational assistants. For now, we will only stay, as I was saying, at the propedeutical level, where we are analyzing

simply the concepts. What I could tell you is that take it as a skill that you can develop as you practice. The more you practice, the more you practice, doing a complete PROM will have greater skills for this. Here we are going to present another video, which is the one from Dalí. Again, I ask you to observe

which is a textual description that identifies a generation of image and that relates the prompt with the content. So, we are going to run the video, it is also not very small, but very illustrative of how this tool is used, which is called Dalí. ♪ ♪

Now horses are coming, so you better run. For your mother, for your children, for yourself. I'm loving it with you if you want to sell my dog.

So here we have seen that the key idea is that artificial generative intelligence not only creates text, but it can also produce visual content like the one we were seeing there from a single description. The two images that I showed you at the end

They are a clear example of how you can interpret the same prompt in two different ways, one much more contextualized

al juego y otro más contextualizado hacia las mascotas. Entonces, bueno, pues tenemos aquí los dos videos en los QR que les hemos presentado. Están los enlaces para las sesiones y así ustedes pueden recuperar cuántas veces necesiten el video para verlo a la velocidad que ustedes necesiten.

I recommend you to check it out and remember that DALI is an example of a tool that generates images from textual descriptions.

The person describes what they want to see: an object, a style, a scene, a color or a composition. And the system produces a new image based on that instruction. In practice, the first result is rarely perfect. Many times you have to adjust the prompt, add details, change the style or ask for a new version so that this can

We must also take care of ethical aspects, copyright, use of images of people, visual biases and transparency about the use of artificial intelligence. It is being legislated that every time an image with artificial intelligence is generated,

a little note or a warning that can tell the viewer that what they are seeing is a product of artificial intelligence because the better they put the plum and the better they give the context, these images will be

more and more and closer to reality and then this can cause a lot of confusion in that some people are believing that that image that they are presenting belongs to reality and not

to something developed with the creativity and artificial intelligence of a person and a machine. In subsequent courses, concepts such as diffusion models, image editing, multimodality and watermarks can appear. Here we only need to understand what is the relationship between a textual instruction and visual generation. So, well, we have

artificial intelligence also produces images. In the table that you can see below, some tools that support these tasks are summarized. I repeat, the fact that these brands are here does not mean that they are the only ones or that they are the best. They are the ones that we chose simply for the example, but in the market today there are many and they are not

And remember what one does at a time.

can improve the other maybe in the next two hours or the next day. So that's why you can't do a ranking with them. They are always moving and fluctuating because this technology or these technologies are developing very, very fast. So, well, I told you this table summarizes those tools, for example,

ChattyPT is mainly associated with text, conversation and ideas, and can help explain, write, summarize or propose some alternatives. CopyIA is associated with marketing texts, it can support campaigns, advertising ads or product descriptions.

In a simple case, a person who needs to create texts for advertising campaigns could take advantage of being supported more by copy AI than by chat GPT, but either of the two can give very good results depending on the prompt they are facilitating. DALI is oriented to image generation from text.

So Copilot is related to productivity and a lot with productivity because it has a suite of tools like the office that has documents, presentations, spreadsheet and then like all these tools are very generalized in the market, Copilot has a lot of

.

So, well, the one that is commercially much more used does not mean that it is the only one, the other is free and also gives very good results. So, well, Copilot is more focused on productivity because of its off-limits tools. It is not about saying "back", which is a

to say that one tool is simply better than the other. The important thing is to link the tool with the task and always, always check the result. So,

Enter them, see what results they give you, try with the same different program and then see which one is better for the task you are looking for. So maybe it's not worth marrying one, but learning which one gives me the best result according to the context that I am having.

We see applications nearby, then. In education, artificial intelligence can help explain topics, create examples, prepare materials or suggest exercises. At this point we must be very careful. It must support learning, but in no way replace its reasoning.

En el trabajo puede apoyar la redacción de informes, ideas de campaña, documentación de procesos o ahorro de tiempo en tareas repetitivas de texto. En creatividad puede ayudar con bocetos, guiones, imágenes o contenidos de multimedia. La recomendación es comenzar con usos de bajo riesgo, como las lluvias de ideas, hacer borradores, ejemplos o reformulación.

When the result has consequences, whether academic, legal, medical, financial or institutional, the review must be more rigorous. The review always, without a doubt, but the bigger the consequence, the more rigorous the review of what we do as a result must be. Therefore, the result of an artificial intelligence simply

Es una materia prima con la que ustedes puedan trabajar, pero de ninguna manera pueden dejárselo nada más así. Entonces, pues veamos la otra.

We are going to the end of this session. We have already reviewed examples of ChatGPT, DALI, CopyAI and other tools according to the task they support. Now we will show the QR code again to access the previous sessions so that you can see the videos. Remember that the QR allows you to review classes and in games and better prepare the contents before the exam. It is not in any way a list pass.

So we're going to leave it here for a few minutes and we continue with agentive artificial intelligence. Five minutes to rest your legs a little and we'll be right back. Ready. Let's go to block three. Okay, let's go into block three.

artificial intelligence, its objectives, actions, tools and supervision.

Until now we have talked about generative artificial intelligence, systems that produce new content. Now we will review a related but different idea, it is called agentive artificial intelligence. The agentive artificial intelligence allows us to think about systems that not only respond, but can organize steps to try to achieve a new goal.

The agentive artificial intelligence refers to systems that can act more oriented to specific objectives. At a basic level, an agentive system receives a goal or a general task, can divide it into steps, use tools or learn information and adjust its processes according to the necessary progress.

For example, it is not the same to ask a generative artificial intelligence to write an email than to ask an agent to help me organize an activity, check the necessary steps, prepare an answer and tell me what is missing to confirm.

The main difference is that the generative artificial intelligence produces an answer. The agentive artificial intelligence can organize a sequence of actions

to achieve a goal. This does not mean that you should act without control. The more capacity you have to take steps, the more important it is to define limits and supervise results and maintain human responsibility.

And well, later on we will study some architectures and frameworks of agents that will be studied specifically in certifications. In this slide you can see a summary of the difference. Generative artificial intelligence produces new content, for example, create a text,

an image or a draft of something, while the agentive artificial intelligence plans and executes steps to achieve a new goal. It can look for information, complete alternatives, use tools or propose a sequence of actions.

The third element is always, without a doubt, the person who supervises, decides and above all assumes the responsibility of how these results are used.

In this session, we are not saying that agents should replace people. The instruction is always very clear. The responsibility of the supervisor and decision is that of a human being. We are identifying how an evolution takes place

of systems that respond to systems that can coordinate steps, but supervision and human automation are always indispensable. Not automation, it is authorization. It is revision and human authorization that are always indispensable. Remember that generating content is not the same as executing steps.

A agent needs at least three basic elements. First, an objective. The goal must be clear.

What is needed? With what criteria? What limits exist? And what results are expected? They are always essential. Second, actions. The agent can propose, consult, use a tool and adjust steps. For example, he can organize information, look for data in a document or prepare a summary of alternatives.

Third, human control. People must review, authorize, correct and above all assume the responsibility of using the result. A frequent mistake is to think that an agent knows

What is convenient to do, who knows what is convenient to do without a doubt, is the human being. In reality, it needs instructions, limits and supervision. An agent could not be completely autonomous. In subsequent certifications, you will study architectures of agents, memory, tools, planning and evaluation of performance.

There are different patterns of agents. Today we will only mention them in a very introductory way. A reactive agent responds to an immediate situation. For example, it detects a condition and acts according to an instruction. A planner agent divides a goal into steps. First, it identifies what to do and then organizes the order and then generates an answer.

A tool agent can rely on search engines, files, or apply to complete a task. A human agent in a cycle requests authorization before performing sensitive actions.

In a much more advanced technical certification, these patterns are related to architectures, memories, tools and security evaluations. In this session, we only seek to recognize the general idea as we had already mentioned. So, let's review. A reactive agent responds to an immediate situation. A planner agent divides a goal into steps.

Here we have

a graph of autonomy against control. Autonomy increases the need for control. In AI, we must avoid entering personal or confidential data without authorization, especially when these are not ours. We must also verify facts, sources and coherence.

There are risks related to bias, copyright, privacy, misinformation and lack of context. In agent systems, we must also be super careful that actions are not executed without supervision when it can affect people or institutions.

The principle is simple: the greater the impact of the task, the greater the human supervision. In education, intelligence must support understanding and productivity, not replace reasoning or academic responsibility. In work, decisions must be strengthened, not to hide errors under an appearance that does not convince anyone.

Reviewing, contextualizing and evaluating will always be the most important thing. This is block 3. With this we close it. We already distinguish between generative artificial intelligence and agentive artificial intelligence. And let's remember that both require human supervision. Carlos, can you help me with the QR, please?

The QR, I repeat, gives access to the previous sessions. If we are starting, you can still review a concept that has not been clear, so you can save this link that will allow you to review classes and lessons before the exam. We are going to give you a few minutes so that you can organize, copy the QR and stretch your legs and we'll be back in a few minutes.

Ready, Carlitos? Okay, we open now the space for questions and comments. You can write your questions in the forum of the course. If there are still no questions, I will use some that I have here as detonators, in which we are downloading from the forum the questions that you are sending.

So let me check what we have in the forum or we start with the questions we have here. Okay, here's a question that says: to use applications like Cloud, GPT, etc., you need to know about bases or models

¿Y cómo se construye? No, no necesitan conocerlos. Básicamente lo que necesitan saber hacer es un buen PROM con la estructura que ya habíamos dicho, entre mejor estructurado esté su PROM, mejores resultados les puede dar y ustedes no necesitan saber qué hay detrás de la herramienta para poder usarla.

-

some content, whether it's video or documents, that are already on the internet and that allow you to attend the consultation that you are doing. While when you write a PROM, you are asking for an artificial intelligence that generates text, image or something. It can be based

in the sources that you are asking for, but it is not the same as just searching here, it is generating. Google does not generate it when we are in the Gemini C search engine because it is an artificial intelligence that generates, while the search engine as we normally know it, it will only search in the network universe

or the immensity of the network, what is the closest to you and then from there it will give you so that you can choose which is the resource that you like and you take it. In artificial intelligence it is being generated. Let's see if there are any more. It says what would be the most effective specific methodology to evaluate the output of artificial intelligence?

The most effective is the review of each person. When the person knows well what they are going to use those results

that person can evaluate it. The person is the most effective strategy, right? That he knows what the context is in which it is going to be used, the better he gives the context or shares the context, the result will be better. There is no such strategy as the book, here logic has a lot to do with it,

The part of the human review has been constantly repeated in all the slides that I have been explaining to you in this session. Human review is the one that decides, it is the best strategy that you can use.

to see if it can be used in one context or another. The better I know my material and the results, the better I will see it. Also, the more experience I have doing the prompts, the better results I will have. Here the recommendation would be that

see very well that when in that context they are not putting some privacy data at risk. So if it's my information and I don't have a problem sharing it, that's fine. But if it's the information of my work, then I should think very well about what I'm going to share with an AI and what I can't share with it because then I'm putting

the confidentiality of some sensitive data at risk. So that's what we have.

What is the cost of artificial intelligence? Well, the different artificial intelligence or artificial intelligence platforms have different costs, many of them, the vast majority have a free part and the free part gives some services, some very basic, some advanced and they can give it limited, so you can

start reviewing anyone without making more economic outlay and simply see which one can be adapted more to your needs. If you need to scale it, for example, we were talking about confidentiality. If you want that to remain very private, most platforms, if

offer that privacy, but only in their payment models. So, well, depending on privacy, depending on the power they need for that processing of the information that they are sharing, they will tend to more, right? Some of them are, they range between $20, maybe the most expensive are those of design, but for example, the entire Google platform

has a lot of tools and those tools give some things that are called credits. So every day they give them a free amount and at any time they tell them, "I got here, I can't take this anymore, what you're asking for is not enough with what I'm giving you and then you can scale and they give you the option to pay like

So some of them have like a suite of tools that allow them to use many different tools.

There are some specialized ones, like GPT, which is specialized in generating text for answers, but also generates images. So you can see if the quality of the image for what you need from GPT is enough for you and you are already paying for it, then only with that or maybe the free one. Free ones can be used in two ways.

with or without registration. Without registration, you simply enter from the search engine, you give the name of the artificial intelligence, it can be Google, it can be this GPT, Gemini, Copilot, for example, Cloud, Perplexity, right? Any of these, you can enter and give your full instruction, right? Fill in your prompt and it will give you the answer.

They have a free model where you register. So there you are sharing your email or your phone and also your name. And once you are doing that, it gives you a little more services. If you ask her what she can give you, she can give you up to a table. You can put in the prompt that I make a comparison and I can give you different

Even a single artificial intelligence can make recommendations of which other artificial intelligence could help them more precisely for the task they are doing. So, well, we have those. Let's wait to see if some people have any other questions. For the moment, it seems that we have already answered all those who arrived. Let's go with the detonators.

My students often ask in class things like if all artificial intelligence analyzes data, that it analyzes data is generative, not necessarily. Some systems analyze, classify or predict. Generative artificial intelligence is distinguished because it creates new content. Another question may be why the PROM changes the result.

And the answer is very simple, because the instruction guides the system, it is what gives it light. When we give context, format, tone and objective, it increases the probability of obtaining something that is more useful.

the test that we put in the PROM slide, put something completely general and then go applying it with this, with the context, with the format, the tone, the objective and then you will see how the

The answer is much more concise and much more useful for what we need. Another would be what is the difference between creating an image and analyzing an image?

Creating an image corresponds to visual generative artificial intelligence. Analyzing an image corresponds more to an artificial vision. Remember what we saw in the last class? Making the pattern, putting all the dots. So that's a complete analysis of the artificial vision. Another is, for example, when it is convenient to use

Copy AI or when it is convenient to use ChatGPT for a campaign. When we need ideas, erasers, advertising ads or product descriptions, copy is the best. Even so, the result must always be reviewed. The tone, the speed and especially the context must be taken care of.

Another very common question in class is why does an agentive artificial intelligence need human supervision if we already have the agent done? Well, because it can organize steps and use tools. This increases the responsibility of defining limits, verifying results and authorizing actions. For example,

These cars that go by themselves, one arrives, sits down, puts on the seat belt, turns on, tells him where he is going and the car starts. If the driver falls asleep, the car stops because even if the person is not driving it physically, it always requires human supervision to be alert.

So it's not that humans don't have to do anything. Humans define the limits, verify the result, authorize actions. So don't lose sight of the fact that the human factor is still indispensable. The tool helps a lot, but the human is the one who designs. So let's review the chat again to see if anyone has any other questions.

Remember that if the questions are exceeding the propa-deutical level, we can give a brief answer, but it will be noted that it is addressed in the certifications that will be issued later.

Ok, we don't have any more questions, so let's start reviewing module 5 of artificial intelligence. Let's go with the slide. We are going to do a cumulative review that can help you a lot for the exam. Before closing, then we have this cumulative review.

Let's remember from the previous session that the processing of natural language works with human language, works with texts, with voice, with translations and conversations. Artificial vision will allow us to interpret images or videos.

Remember that the artificial intelligence generates new content. It can generate texts, images, audio or videos. Also remember that a PROM is an instruction given to the system. If the instruction is vague, the result can be generic. If the instruction is clear, the result can always be more useful.

Finally, responsible use implies reviewing, contextualizing and evaluating. Artificial intelligence can support, but it does not replace human criteria in any way. This review seeks to organize concepts so that you can recognize these different examples: artificial vision, generative intelligence, PROM and responsible use.

In this slide, a different QR appears to the one we have been presenting in the previous spaces. This QR takes you to the video that was made on Friday to support the presentation of the exam. It is very, very important

Porque debemos aclarar algunas cosas. Aunque el examen ya está disponible, no creo que todas las personas deberían presentarlo de inmediato. Solo les estamos diciendo que ya está disponible porque hay personas que están avanzadas y que ya se sienten lo suficientemente fuerte.

Those of you who are still at a beginner level, I highly recommend you to see the sessions that you still need to go even harder for this exam. I repeat, if you already have experience and master the concepts, you can probably advance with greater security, but it is never too late to expect what your colleagues have in the following sessions.

If you are starting and you still confuse some areas or if you have not seen all the sessions, my recommendation is that you first review the previous classes and we will accompany you in the following sessions. The following sessions are also very important and will help you organize the material much better than you saw in the exam.

The objective is not only to answer an exam, the objective is to understand the basis that allows you to move forward towards the subsequent certifications with much more strength and much more confidence. The support video can guide you a lot, but remember that it does not replace the materials that are in the course.

Everything that is in the course is very important so that you can, on the one hand, present the exam and, on the other hand, go much stronger to the certifications that come. So, we have there this QR, it is the one that has the exam. Another very, very important recommendation is that everything that we are working on, we do it with the time of Mexico City.

For those who are at other times, this is very important, especially for the hours that they will be explaining to you for the closure, so that they do not have any problem with this. Also take into account that the schedules are limited, so for example, if they are going to give us an hour to present the exam,

The entrance time to the exam will be closing an hour before because they have an hour to present it. They can't be entering the closing time because it won't give them time to do the exam because the platform is set up automatically.

And then if I take them to the middle of the exam and give them the closing time, the platform will automatically close them. So schedule your time.

with enough ease so that you can have in this session, for example, they will explain what happens if the light goes out, what happens if the internet goes out with your exam and how you can solve it and what are the support strategies that you can have to carry out the exam with great success. So, well, to close,

We will recover the central idea of what we have been handling in these sessions of artificial intelligence. Generative artificial intelligence creates new content, text, image, audio, video, codes or initial ideas.

The agent artificial intelligence organizes actions to achieve goals. Both can be useful, but both require human criteria, without a doubt. The PROM is the instruction we give to the system and tools like ChatGPT, which can support text in conversation and ideas. Tools like DALI, which can generate images. Tools like CopyAI, which can support marketing texts or campaigns.

If you are starting, check the previous sessions and join us in the classes that are still missing. I remind you that this propaedautical course prepares you to face the subsequent certifications, where you will find a much deeper technical level on all the topics that we have been reviewing. I thank you very much for your participation. Check the materials of the course, save the support links

present in the exam when they feel prepared to distinguish concepts and recognize applications. Thank you very much and we conclude this session 9 of the course. So we wait for you tomorrow and who are not, good luck in your exam, we are seeing each other. Thank you very much.