El poliamor explicado con OpenAI y Microsoft: abren la pareja pero uno sale perdiendo

Quienes sí apuestan por el poliamor son OpenAI y Microsoft Mati. Que cada vez más Anthropic es una empresa de productos porque están lanzando tantas cosas por encima del Cloud Code y del Cowork que da la sensación de que son la Apple de esto en lanzando productos muy afinados para distintos tipos de profesionales. Y plantear un teléfono, pero en el que, el concepto que imprimiría OpenAI, las aplicaciones quedan sustituidas por agentes de inteligencia artificial. Hello to all the monos stocasticos and welcome to the podcast of artificial intelligence favorite of Amaya Romero. Because you and I never have put up with my fascination by Amaya. But I know that you like the art transgressor. Ver Amaya playing in an IKEA like if it were a flout of a traveser in the Tiny Desk. A little bit you have liked. Hey Mati, how are you the minds of fans and haters? Because while you were fascinated with this, of course, enormous artistic leap that is repatting the whole world of the cantor of Amaya, I this week I crossed my version of Volando Boy and my conclusion was positive, Mati. I have not fallen fascinated, I have not celebrated that new mix of the copla y te tengo que decir una cosa yo es que digamos me distancia de amaya porque en un festival de música en que había muchos cantantes iba a ver a otros no eran mijas pues pase cerca del concierto de amaya dije bueno voy a ver este esta chica que le gusta tanto a matías y era una canción en la que ella pues básicamente expresaba todo su odio violento hacia alguien alguien le había hecho algo very bad and she wanted to throw it down, put her head. It was like saying, well, this generation of care, and love, and health, when it comes to me. And I, of course, I put Amaya in a box, which is the people that I have to see him. You go with Amaya, a campground, a day of celebration, or whatever, special of the house. Of course, Amaya is sure that I think it's one of those who say, no, no, I'll pay you because it's a point for the card. Of course, I'll pay you and I'll pay you. And then, if you don't do it in the moment, then you forget it. And forget it, then it's a little bit embarrassing. And I am a lot of people who are afraid to see a bit of a bit of a bit. Do you understand? A ver if you dedicate a song like this, for being my friend with Emeligo. Ay, vaya semana, Antonio. Vaya semana. How are you? There are many things, but, Mati, I've been listening to the critics of Monos Podcast, and there is one that tells us that our AI is not useful, are not practical, we are not taking too many minutes talking about topics without importance and we are not entering into the key of the AI until we have one that we are not talking about, that nobody is discussing, but I think it is crucial for the fear of the time with respect to the employment and the AI. I will give you a point of who are really bien preparados los que se han posicionado de la mejor manera ante una posible singularidad en la que mucho empleo del cuello blanco desaparezca. Y estos son, ni nada más y nada menos, Mati, que los poliamorosos. Atención, porque te voy a contar un caso. Yo tenía una pareja de amigos, estos son en la década pasada prácticamente, y otra pareja de amigos, ambos trabajaban en la banca. Además, los dos trabajaban en el mismo banco. Lo cual es un grave error, porque tienes el riesgo de industria and the risk of the company. It can be bad to the bank in general, but if you have the risk of if you have a bank solo, it's a bad thing. So, both suffered a professional crisis in the personal record that the bank has done in the last few years. And, of course, that to me, it was marked by the risk that can be the family unit in situations of this type. What happens with the IA? Well, with the IA, we don't know if it's going to affect a todo el trabajo del conocimiento, solo los programadores, si creará más empleo o si lo destruirá. Pero en el peor de los asuntos, en el peor de las situaciones, lo ideal ya sería repartir más el riesgo. Es decir, en vez de una unidad familiar de dos, pues de ser de varios. Siendo poliamoroso, fíjate lo bien pensado que lo tienen, que de repente tienen muchísimo menos riesgo, porque siempre hay más oportunidades de que uno o varios pues se hagan triunfantes de este cambio de paradigma profesional y tecnológico. ¿Cómo lo ves, Matías? Pues veo que con esta intro le podemos estar haciendo también competencia a nuestro podcast amigo Loop Infinito, que esta semana ha hablado de los swingers, ¿no? Sí. Entonces nosotros podemos posicionarnos como el podcast tecnológico líder en la cobertura del poliamor and Accord is the leader in cover of the swingers. Yes, yes, for personality I think we and Accord are well repressed very well seen well, but I have to say something and now we're going to with the actuality this week who are the ones who are the Poliamor are OpenAI and Microsoft Mati. Well, you are seeing it as a relationship with a polyamorous relationship, but as it happens in many relationships, and this I know because I have friends, Normalmente es uno de los miembros de la pareja el que quiere y el que está más abierto, nunca mejor dicho, a abrir la relación. Pero esto tiene visos de divorcio camuflado, Antonio. Ay, Mati, ¿quién sabe cuándo se le acabó el amor a Satya y a Sangalman? No sabemos qué momento, pero lo cierto es que aquí conviene recordar que en el comienzo de la alianza Microsoft hizo una enorme y fuerte inversión en OpenAI en un momento de muchísima más incertidumbre respecto a esta generación de inteligencia artificial y a cambio, aparte de la propiedad, digamos, era el acceso a toda la propiedad intelectual. Por un lado, la propiedad como accionista, por otro lado, el acceso a la propiedad intelectual y también la exclusividad de los modelos de OpenAI en Azure, en la nube de Microsoft. Claro, este acuerdo que en principio supuso un impulso enorme y muy favorable para OpenAI poder invertir muchísimo y conseguir lo que ha conseguido hasta ahora, también a medio plazo ha resultado limitante. su competencia, podría estar en cualquier nube, si una empresa utiliza la nube de Amazon o la nube de Google o cualquier otra, entonces no podría tirar de los modelos de OpenAI, se le complicaba la cosa y empresarialmente pues es una dificultad. Bueno, el caso es que además puede haber razones técnicas como el uso de TPUs, de terceros, que a lo mejor Microsoft no tiene, en fin, un momento en el que yo creo que ha sido sobre todo OpenAI la que tenía más ganas de expandirse y de tener más relaciones. El caso es que había otra cláusula muy interesante que era atención, Microsoft mantiene esta exclusividad mientras no se llegue a la AIG. Si se llega a la AIG y eso, bueno, era una medida más econométrica que técnica, bueno, el caso es que si se llegaba a esa inteligencia artificial de nivel humano, pues se rompía la exclusividad, porque era una idea que tenían cuando eran una ONG, sobre todo, and by the way they said, no, no, this of the Agile can be a monopoly of a single company and that only can be used to use Microsoft. No, it's plan. So, that's precisely what we have now. On one hand, Microsoft has to have the exclusivity. OpenAipure is by the way, is scandileando with anyone. And in the other hand, it breaks the closure of the Agile. Although it's possible to achieve the Agile, Microsoft has to stay with access to the intellectuals or access to the intellectuals until 2032. And of course, Microsoft has to stay as the accionist of more weight in OpenAI. Well, look, the new agreement, I would like to say that Microsoft will not share benefits with OpenAI, but, above all, with that he is still being the main source in the openAI network, which means that the models of OpenAI will first be in Azure, because I would expect after this announcement that Amazon would come with a similar agreement with OpenAI in which AWS starts to integrate the models with HGPT with the same priority that Microsoft. But I think that that still has a advantage of the main source of OpenAI. Claro, como en el poliamor, claro, estoy metido en el subreddit del poliamor, ¿vale? Entonces, por supuesto, ya doy por hecho con internet 2026, que la mayoría de los casos estrambóticos que plantean son inventados, pero el mundo de los comentarios sí me lo creo mucho y me parece muy interesante, entonces muchas veces escargo ahí en las discusiones de los poliamorosos y mi sensación es que dentro de las múltiples relaciones que se plantea OpenEye, Microsoft es la principal, y esto encaja muchísimo con los patrones y reglas que se plantea OpenEye. that is, the people who have embraced the pure love. So, again, this has given the key to understand the move technology and empresarial. And talking about technology and empresarial, what difference, what's the difference, what change, what's the difference compared to the new model of Deep Seek with respect to what was supposed to be that was on the table over the table of R1, that was the moment in which China was going to be able to get models of quality, the best, but much more expensive, and Wall Street was going to be a caquita, Matías. I don't know if it's a Chinese, it's a more expertise in marketing or what, but with what it was when it came to Deepseek, the supposed moment Sputnik, the IA, that seemed to take advantage the open models of Chinese and that that would change everything and that they were in danger from Nvidia to OpenAI, Google, Meta. And that was a little bit after, although it's true that a lot of people have passed to DeepSig and other open models of Chinese. And I feel like with the launch of DeepSig V4, even though it's a very powerful model, that people say, again, we're going to repeat all these pantomimals. the United States is winning. I don't know if you are of course with me. Yes, I'm very sure. In fact, in the description we have put videos and captures of the Deep Seek, one that we have done in the laboratory with Huawei. And if you look at, it's wonderful. It's really a good idea. How those three engineers, two chicos y una chica, tienen ahí sus folios preparados, vienen con el polito homogéneo, tienen ahí su sofá con sus plantas de plástico y se tiran pues creo que era una hora entera explicándote las tripas y los cálculos abajo, a brejísimo nivel todos los rollos de Deep Seek entonces, por lo que sea esta, digamos, estrategia de marketing y comunicación, que tú eres el mayor experto, Mati, por lo que sea, no ha funcionado tan bien como otras, es como It's true, you have seen the OpenAI and there are four brothers, four scientists of the AI in a sofa, but at least they can change the camera some time. They have a sound of a face of a little bit, which is always a little bit of a game. But I have a very good idea. These three engineers are very nice and very nice. And basically, I explain that it's a model that in the rankings, as you have said, is there. It's a good thing. It's not a good thing. It's a great jump. It's a great jump. But, again, DeepShip is there. And, well, his card is still playing the price. So, I think that for those who rejects model cerrados, for those who seek to earn until the last centimillion, it's a good option. But it's true that the line QEM, all the Kimi, all the GLM, there are a lot of Chinese companies that are also in this league. So, Dixie has paid, I think, with everything that has been taken, I think they have lost the momentum of being the Chinese referent, because much people have tried other things. Well, it's a shame because they had it there, and they still have the premise with which they did Viral, which is modelos fronteras, modelos a la vanguardia en todas las capacidades que estamos viendo en los modelos de lenguaje y en la programación agéntica también, a una fracción del precio de lo que la gente está pagando por cloud. Pero bueno, que siga la fiesta. Sí, en general bueno, si miras alguno de los rankings, hay un patrón, ¿no? Si miras el ranking de los modelos abiertos, pues el tip sí que es del 1 al 3 Siempre está ahí peleando muy bien. Digamos que dentro de añadir esa exigencia de ser abierto, pues TIRSIC performa a un gran nivel. Pero, claro, cuando metes todos los propietarios y los últimos modelos de Estados Unidos, de OpenAI y Google y Anthropic, pues ya se va al 14, al 15, al 16. Claro, hay un salto. No se ha igualado lo que los laboratorios americanos están consiguiendo, sobre todo por una cosa creo, que es que los laboratorios americanos están consiguiendo sacar actualizaciones de los modelos que mejoran los patchmarks cada 2-3 meses. Ese incremento en la velocidad del lanzamiento de modelos creo que está pillando un poco a pin cambiado a la industria china y aunque esto Huawei está vendiendo como una pequeña victoria, diciendo mira, como no hacen falta tanto a lo de NVIDIA con Huawei ya puedes tener autonomía de los chips chinos, si nos creemos que no tienen NVIDIA de contrabando, si nos creemos eso, pues se puede to achieve a good level. So, again, we're going to a moment in which the world of codecware, of the model open, or more than, of the pesos open, is in around a year, 9 months, 18 months, depending on the functionality, behind the models. And one thing interesting is I have to read the paper of DeepSeed. For at least to give a look, because there is something magnificent in this Chinese company that is how they share their knowledge and how they do things. I remember the paper of DeepSeek R1 that for me was critical and fundamental to start to understand how to apply the learnings by refuerzo on top of a LLM, which for me was something still difficult to assimilate when it was just announced, by the part of a OpenAI, or one, for example, that we didn't pay attention, but I didn't understand it a low level, in a way of practice, until the Chinese, in a way much more open, share it. Magnifico. What a word has used to describe this moment, because I have to tell you something amazing. You feel magnified, Matías? You feel magnified, like the patrocinator of this week? I feel magnified, and I feel happy, and as they are magnified patrocinators of this week, I going to propose something Antonio to do it open like Deep it like Deep Seek how I prepare the song and the video clip of this week But for that I need your help. What style of music would you like for the song resumption of this week? Well, this week we have two very important singers in the guion. So I think the singer-songwriter, the pop singer, actual debería ser el género, porque tenemos a Maya, pero luego aparece una chica muy especial que es Russian Red, Matías. ¡Ojo ahí! Venga, pues os voy a contar en dos minutos cómo ha hecho la canción de hoy. Bueno, esto va a ser muy rápido, ¿eh? Nos vamos a Magnific.com, nos vamos a Spaces y empezamos creando una imagen en el generador de imágenes. Le vamos a poner dos referencias. Uno es Sam Alman, como siempre, mi musa para este tipo de cosas and the other image will be Amaya, to the silla travesera or the flauta travesilla, and we connect both images with the images generator as references. And now, simply put the relation aspect 16-9 for the video. I like to put it in cinematic, and the prompt will be very simple. samalman tocando una silla como si fuera una flauta travesera ejecutamos y lo tenemos pero le vamos a cambiar la cara a la chica de atrás por la de darío modé y que es otro de mis musas cambia la cara de la chica de atrás por la del hombre de referencia vamos a generar cuatro fotos la letra de la canción I'm going to do with Gemini, I'm going to do the episode as reference, and I'm going to write the letter of a song from a minute and a half, from Russian Red and Amaya Romero about the topics that we have in this episode of Monos Stochastic. Vale, tenemos esta y esta. Y ahora simplemente nos queda generar el vídeo. Voy a crear un storyboard con el nodo de variaciones al que le voy a chutar como referencia esta imagen. I do to generate, we see how the song goes, I saw you in the festival and thought you were all love. I just want to open the app of the bank and see you as the vision. Okay, I'm going to play. I'm going to play. I'm going to go directly to Asuno. I'm going to create a song with this letter. We're going to go to the variations. We have the Tiny Desk. Perfect, perfect. The video has already been cut off the frames of the storyboard, so now I want to generate a video for each image of reference. And the result is at the end of the episode. Also, Mati, now that you have explained the day, and in general this kind of day, in which you can't see, I think the company has won, Freepeak has won in one thing, the company's first known as Freepeak, because how do you call them the same employees of the company? How do you call them? Antes they could be the free pickers. Well, that's a five. A five of ten. I never have said it. I'm not saying it's not a fault. I'm not saying it's not a great thing. But free pickers, no, no, no, no, no, no. But autodenuminar the magnifiers or the magnified, that's a lustre. That's a sense. The confidence. So I think that the spirit of the team will improve. And maybe it improves, it improves, it improves, it improves, it improves GPT 5.5 compared to the previous version. Mati. Me tienes que contar, because you have several days saying in the box, in the food that we've done, well, well, it's a SPAD, it's a new model of OpenAI, a see what they say, a ver si tienen cosas para dar un golpe sobre la mesa y después de unos días probándolo interesa mucho saber si ha cumplido tus expectativas. Claro, respecto a esas expectativas es la bajona. Porque claro, ¿dónde situaba yo mentalmente a este modelo de nombre en clave SPAD? Pues en mi cabeza mentalmente es, jol, esta gente tiene algo ahí dentro que igual a hábitos like what Anthropic has developed but has�� Nancies. And when we get out, we will end up something in a volume of great technology in those models of artificial intelligence. Probably very limited and very constrained when you feel Rafalian security, of biological weapons, the facet is the that it preocupates that people can access to it. The case is that if that is OpenAI, it is not SPAD and it is GPT 5.5 because it is a model that clearly, by the benchmark, by the tests, by the company's own, it is a lot of analysis, the data, the scripting and depurating of code, the use of tools, the use of software But there is a lot of important steps to go ahead and this is fundamental. In the web research, which is something that I use intensively and much. I don't care about this topic because I have read a font. Buscame more font. Busc if someone has contradicted. Busc a other opinion. That pattern that I use to confront the news that I get. I use it a lot. I'm in the chat GPT. Well, it's a better. It's a better. It's a better. It's a better. But, from the end, it's a better. It's a better. It's a better. It's a better. have done well in that this is GPT-55 and no GPT-6. Well, yes. I think that the best are welcome for everyone. For example, the event of context of a million tokens comes very well to people who are put in jobs of code very large, but also to me, that I give a lot of context and many PDFs and many carpetas with documents to each GPT as context. But on the other hand, this comes with a price. Well, there is a price that is 5 dollars per million of tokens of entry and 30 per million of tokens of entry. That with the use of the API, it is relatively superior to 5.4. And I would say something that is, it is not that computable, this is a model, surely, more cost-wise, because they don't have the gratis. The gratis is the 5.4. That's a difference. And also OpenAI has explained that the improvements here have been achieved, especially in pre-entrenation, which is quite interesting because basically we had a season where we said that all the improvements were in the effort over the LM, that in some way the pre-entrenation had been very decreasing. well, well, well, so Anthropikon Mythos and OpenEyeCon 5.5 well, they are showing that this training is still important to improve the renditions. In the benchmarks, well, it improves in all except in some strange and strange that later will appear. And, especially, I think the comparison is with Opus 4.7 and there are the two. Hay un benchmark que me ha hecho mucha gracia, que es básicamente que un benchmark del LM Arena, ese sistema por el que la gente, sin saber qué modelo se está utilizando, pues vota cuál está dando mejores salidas. Bueno, en desarrollo queda muy por detrás de Opus, pero es sobre todo porque es un benchmark muy ligado al front-end, la capa de presentación de las aplicaciones. So Cloud continues to be the most aesthetic model, more capable of working in that front. And Codex, with GPT 5.5, is a little more relegated to the trip. I think Codex is a good thing, because when I was a programmer, I was a programmer debugger. All the front part seemed terrible, especially that devil's inventing that never had been built, which is JavaScript. But well, the case is that there are little popularities, but basically it's a question of vibras, Mati, that one prefer Opus 4.7 or GPT 5.5, I think practically all the variables, the levels are very very similar. A me, the feeling that I have more believers of codecs, like you. There are more people who defies codecs frente to Cloud Code. This is more like this to be I'm coming back to the news of Windows vs Mac. I think in this case it's more of Windows, OpenAI, a lot of this relationship with Microsoft, and more of Mac, Cloud, which is more Anthropic is a company of products because they're launching so many things above the Cloud Code and the Cowork that they're feeling that they're the Apple of this launching products very, very afin for different types of professionals. Well, I'm going to tell you something. In the System Prompt Codex you have found the following. The instructions that no talk or no talk in absoluto ni of duendes, ni of gremlins, ni of apaches, ni of trolls, of wolves, of walnuts or other creatures. Atención, because finally they have solved one of the big problems of the AI, which is that Codex no mention the Goblins and the Paloms. How do you think? Well, now I'm a little bit I'm a little bit I know, this is I'm more of users of Linux so much esotéricos in place of Windows I'm not sure I'm not sure what to think of this but I'd like to know why the model comes with that pulsing of talking so much of Gremlins I'd like to ask but I'm not going to answer because it's forbidden Of course It's a question because it's a a strange behavior that of course has been found, and what strange are the models of artificial intelligence. They say, look, the best model to history that can use consumers, a tool that can be used to be a technical tool that can be done minutely there thinking with the idea of thinking, planning, using the tools, and at the end they have to correct the problem of that they talk too much about the Greggly, the trolls and the wolves. Well, that's the world in which we live, Mati. What a pena me da haberme perdido todos estos memes de los palomos y los gremlins por estar metido en el lanzamiento de Magnific, pero ha tenido que ser divertido esta semana. Bueno, pues no ha sido divertida la semana en OpenAI, porque a pesar del lanzamiento y de una aceptación positiva, o en general de GPT 5.5, un reportaje de Wall Street Journal apunta a que no están consiguiendo los objetivos de usuarios, tanto gratuitos como de pago, that had marked in their way of the route with vistas to get in the pocket at the end of the year. So, maybe, maybe they start to enton the song. You know, there's a very famous song that has been made, which is I've been building less data centers. We're making the inverse version of our success. Well, there is a point that, of course, is that they have a lot of fun to be in the open-air. Because this is what happens and everyone is meme. Open AI in escape. It's going to be put in my space in the open-air. And all the discussion of this, maybe Open AI is going to be there. And let's see, let's see. I think that the finance, well, the last investments, I think they give for a couple of years, with a little bit, but the plan is to go to the price. If there are strong symptoms of growth, very robust and very creëble, it's the same to repense it, Matías. How much before that Anthropic, there's a movement... I don't know if Sam Alman has a strong position for being the CEO of OpenAI for always. No, bueno, bueno, bueno. Saterina Dela le tiene que coger la matrícula. Pero no te preocupes, Matías, porque no van a parar de innovar. Y lo último es que el analista Min Chi Kuo, ojo, Min Chi Kuo, es como el gourmand asiático, pues ha informado, ha filtrado que están preparando lo que probablemente sea un teléfono móvil, trabajando con Qualcomm, con Mediatek, con bastantes proveedores y plantear un teléfono, básicamente, pero en el que, el concepto que imprimiría OpenAI, las aplicaciones quedan sustituidas por agentes de inteligencia artificial. Yo no entiendo a esta empresa, de verdad te lo digo. No estábamos hablando de que habían perdido el foco, de que por eso cierran Sora, Sora que ha dejado de funcionar hasta la API, or is that it already exists. And now, it turns out that they're making a phone. What a loss of focus is that to make a phone phone. I remember when Amazon did a phone phone. The Fire Phone. The Fire Phone. Well, there have been rumors in the past. It happens that we always have been able to get the end and we have to go on the hour. But there have been rumors also that Amazon wanted to try to try the phone phone phone. I'm still with the same doubts. Whenever there is someone that I want to ask a phone phone different, that is how you are able to reproduce all the amount of value that has been built on the phone. You have the phone, but you need to have the map. How do you have all the photos? All the social platforms. But not only that. You have to have the bank account. You have to have your account of notes, like Obsidian, etc. It is to say, you have created a number of value on the mobile platform that only are there, iOS and Android, that is how do you replicate it, for example. That application, with the agents of IA, of St. Alman and OpenAI in the mobile, are going to be able to bring me just the videos of women of 70 years who make 120 kilos in pain, which is what we offer today, Instagram. Well, if not have that, I would feel that in those moments of loneliness and aburrimiento in which I look for that woman capable of making enormous levantadas of the weight of the muro well I would feel like a loss of the matter of Mati Yes Juan is that what I mean I mean I mean I mean I mean I mean I mean I mean I mean I mean this news is that the use that the people that are most met in the world now the mobile the people the people the people the people the people the people check and manding instructions to people that they're in a computer that is connected to your house, a Mac Mini, a Mac Book Air, it always is a computer. Apple is not sure why. And now, what can they bring? It's true that Apple and Google control the capabilities that can have agents of IA in the mobile. And that this can open other new doors. But now I don't want to get so much from the focus with a phone. Yes, I think they, above all, they think that, as in the end, the main thing is going to control others, if you have a phone, in the long run, they are fritos, at least in the consumer, in the consumer. But in the end, there is an interesting element. The European Union has given a touch to Google saying hey in the fulfillment of the DMA as you have a market diciendo oye en cumplimiento de la DMA como tienes una cuota de mercado enorme en Europa te advertimos que lo que puede hacer tu asistente Gemini en el m las capacidades de acceso, procesamiento, lo que sea, como agente de inteligencia artificial en el teléfono, se lo debes permitir a los agentes artificiales de la competencia. Entonces, si Gemini tiene un acceso a bajo nivel hardware a todos los contenidos del teléfono móvil, en teoría Google en Europa deberá poder ofrecer esas mismas posibilidades para que con el permiso, por supuesto, del usuario, ChagPT, Cloud o cualquier otro, pues se pueda ofrecer algo del mismo nivel competitivo. Entonces, es verdad que aquí en Europa lo pueden tener mejor, pero en Estados Unidos en los que esto, digamos, se clarificará dentro de 10 años en un tribunal, pues pueden haber perdido la ventana de oportunidad de ser el agente inteligente en el teléfono que es lo que va a usar el personal. Ay, Mati, pero bueno, mientras que en Europa domeñamos las aspiraciones monopolísticas de Google, hay alguien que ha decidido abandonarnos y es alguien del que no hemos hablado todavía en el podcast, pero que algunos médicos me habían mencionado como una herramienta bastante útil y bastante válida Open Evidence. phrase, por la AI Act europea, ha dejado de estar disponible en Europa. Bueno, aquí lo de Open Evidence, mira, si lo tienes ahí, fíjate, para la gente que no está viendo el vídeo, pon en grande el mensaje porque es bastante agresivo. Dicen, Open Evidence ya deja de estar disponible y atención porque nos vamos de Europa y de Reino Unido por la incertidumbre regulatoria. Entonces, además de otras leyes, la AI Act, pues no nos vemos condiciones de cumplirla y por lo tanto, si quieres que de alguna manera volvamos ¿por qué no les escribes a los políticos europeos y les dices que tu parlamentario europeo correspondiente, escriben una cartita diciéndole, mira que mal que se ha ido open evidence. Aquí hay dos temas sobre la mesa, muy importantes, que creo que reflejan el momentum de the European Union, about our usual dilemma, how we have to be regulators and how we have to be pragmatic, innovators, despite that those innovations we believe are certain risks. My suspect is that although Open Evidence talks a lot about the AIAD, it is true that and here we can defend a little European Union, that the most strong measures of the AIAD that that would be a good use of security and transparency and open evidence, not have entered yet. So, this release is a preventive. Yes, of course, it is the AIAT. It is possible that this project and this country have problems with the GPDR, that is, with the treatment of personal data, que el nivel de seguridad para datos clínicos es altísimo y es de una importancia bastante notable. Pensemos que aunque se anonimicen los datos a la hora de hacer una consulta, no se meten datos personales de los pacientes, por supuesto, siempre quedan dudas de segundas derivadas de inferencias a partir de metadatos like the geolocalization, the moment, who was in the hospital, what department, it is a world in which Hilarfino is relevant. So it's possible that Opinions, that's what's up, your model of business is publicity of the pharmaceutical industry. It's a company that gives you that gratis, a cambio of public, and of course, who announced that there is not Postremontero, it's announced that the pharmaceutical industry that, as well, the publicity of the pharmacists, etc., is something that in Europe is extremely regulates. In fact, all this, Mati, while, in some way, Europe is replanteing the AIAT, because there is a project that, again, was always at the end of the guion and never entered, but, well, I don't want to be afraid with legal issues, I know that they don't like it, but, basically, now there is a discussion between the legislators of the EU on what is called the Omnibus, the project Omnibus, which is a legal package that enmies many things a priori of the EIA. So, how strong or intense does it have to be in the EU with some of the EIA? Well, there is the issue. There is the issue of that the EU, the current desde el informe Draghi que hemos venido contando en el podcast, es que hay una corriente que la alarma no es que la IA tenga unos riesgos hipotéticos de seguridad que debemos prevenir. Lo preocupante en Europa debería ser que en términos de innovación y la tecnología clave de este siglo nos quedamos atrás. Por lo tanto, no nos excedamos en la regulación preventiva de riesgos posibles y apostemos por no quedarnos atrás industrialmente y luego ya a posteriori, como hacen los estadounidenses, pues ya regularemos si vemos algunas cosas malillas. Well, the case is that the Omnibus is going to be thought that we will have in the next week or month, and also includes not only certain regulations, but also new regulations more intense, but also new regulations more intense to what is happening. It is to say, as we observe that there is sexualization of images of third-year, through the system of artificial intelligence, like the case that happened in that week in Loquishiman X, or many news that we have seen around this issue, then we want to put more emphasis on that the images have a safeguard and a extra responsibility, more grave, if they can use or are used for this thing. So, well, all this debate will follow, atent, sin open evidence, pero con una factura más cara de los modelos de inteligencia artificial, Mati, porque ahora mismo toda la comunidad de inteligencia artificial es un continuo llanto, es un continuo lloro, porque quien no se queja de que han nerfeado su modelo, se queja de que se le agotan los tokens, Matías. A mí no me extraña, porque no te imaginas cuántas veces yo he abierto un ChargPT o un Cloud en el trabajo y he asked if here are 10am, what hour is now in Beijing? And he was activated the most powerful model, the ChagPT Pro. And he takes 10 minutes thinking, what hour is in Beijing? Well, if Matías is now in Málaga and it's summer, then no me extraña that they have been fired. Claro, but then the model begins. But the time is really a physical element of the universe. or is it an invention of the human mind, the model begins, of course, as you have put the pro, you say, this is what I have put for you to do everything, no of course, and you investigate it. So, like this is the thing, in the industry. The issue is that, well, in the part of the encounter with the actual state of the industry, there are elements, I think, rationales, objetivos, and at the same time, there are elements that are very difficult to feel. is to say, what do we have to nerf a model? Nerf a model basically is when I'm going to hire a service to pay one of these systems, I have access to models and, as well, to model a certain level. The more reason, the more time of inferences they dedicate, the more compute, the more cost for that company, the better answers they give. So, what people say is that el modelo empieza a darme en menos tiempo peores respuestas. Por lo tanto, lo han nerfeado, le están limitando ese tiempo de inferencia para ahorrarse ellos los costes y yo me quedo con un peor resultado. Entonces, hay muchas quejas de este tipo cuando no son o ya han cambiado el modelo, han puesto una versión más comprimida o me están dando, antes me daban el 5.4 high, ahora me están dando el 5.4 mini o low, versiones más comprimidas, más pequeñitas y sobre todo Cloud Anthropic, alias a Zapatista, porque empezó a quitar hasta el acceso a Cloud Code de las suscripciones de 20. Y eso levantó una guerra. Levantó una guerra porque Anthropic ya no está siendo tan maja, Matías. Bueno, es que yo, por hacer de abogado del diablo, y yo nunca he pagado, a lo mejor para probarlo un mes, pero nunca he pagado planes de más de 20 euros. Nunca he pagado los de 100, los de 200. Pero hacer abogado del diablo, a medida que más gente conoce la IA agéntica y empieza a usar Cloud Code y Cloud Cowork y se engancha y ve las posibilidades y prueba, es que no hay forma, con la computación que tiene Anthropik ahora mismo, por muchos acuerdos a los que esté llegando, no hay forma de que esto sea sostenible a largo plazo. Muy eficientes tienen que volverse los modelos. I'm going to be a lawyer. I'm going to be a lawyer. I'm going to be a 100% a month. I'm not going to see. Yeah, well, there's a push for the user prosumer. I like this word prosumer. Usuario more or less professionalized. I think the companies of IAEA have been finding more value than 20 euros a month. And they're going to push to the price of 100. I think that's there. and there is a reality that you have described very well. It is to say, with the aparition of the models of the razonable, we have to get rid of the autocastical, especially in this stage of models of a very high level, with the system agent, with the orchestration of agents, but all that with the use of the tools is highly devorable to tokens. So what economic and it has been for a LLM intuitive that respond to what has its data of the training, that if it is, it is connected to internet a little, Well, of the economic system of an agent that uses tools, calls APIs, makes a lot of money, makes a lot of money, makes a lot of money, then, of course, the economic system doesn't fit. And we're going to go from a world where, as well, the demand of IA is supering a offer, except this OpenAI, which is a lot of companies, but well, in all the industries. And basically, many companies are changing the model of business. las suscripciones, todo lo que te puedas comer buffet libre, están en retroceso y está al alza. Por ejemplo, ha hecho GitHub, que es GitHub Copilot, que está pasando a pura y dura facturación basada en tokens en uso. Entonces, los clientes empresariales de GitHub Copilot, pues de repente ya no tienen esa suscripción y esa previsión de lo que le va a tener el coste por empleado, sino que se van a modelos a que si tu empleado ask a question for the what is in China, you have to go. So, we're going to that scenario. There are other models, the CDS Salesforce says that they are going to go to in function of results, in function of tasks, which may be an interesting alternative. But, well, the case is that the sales model of these companies are going to be because the IAGE indicates the subscription, the regular, Matías. I don't know, Antonio, if it is because I was a time being autónomo and now I'm a body but the buffet free of sushi I left behind I appreciate this Sushi Man that is in the gasoliner of Alarinda Torre or in Fuenchirola something like that for you to talk with your partner without going to pass the platos and you get 30 a tremendous upgrade of Matías el gourmet but Mati, we end the day. And before we finish the episode, I think it's convenient, I think it's convenient to pass a Puerto Grande or a enfermería. We've reached the end of this episode. logisticamente complicado. Estás en un hotel, y vienes a hacernos reír con tu selección de temas. Disculpa a la audiencia si han encontrado algún sonido inconveniente, pero estoy ahí a catarraillo, entonces de viaje, fuera de casa, y haciendo lo que uno puede con el micro sostenido en la mano. Pero vamos allá. Te voy a presentar a un man, Matías, que lo encontramos en el medio Semáfor, in which Bill is a little bit because when a media does a report you ask a photo. Maybe you can send a photo. Well, the friend Bill has sent a photo from 2011. Ah, that's a pill. Now in 2026 the friend Bill doesn't have that that's so stupid. but well. I can't tell you a interrupt. This happens a lot in the university. I don't know why the virtual campus teachers always have photos of the day of their boda that they have passed 40 years. That is very common with the university professors. Of course, look at that Bruno, who is going to graduate in the bachelor's degree, he was reading photos of when I was a kid, because we were going to make a montage, the typical montage of when we were a kid. Well, I'm going to pick this for when I was in the media. When I was a kid, A ver, es que claro, así puede impresionar a una extranjera o cualquier tema de estos. Bueno, el caso, ¿por qué es noticia el amigo Bill? Pues básicamente porque ha externalizado toda su vida, bueno, el 99%, casi toda su vida, a un asistente de IA. Alguien tenía que explorar esta posibilidad y Bill se ha ofrecido, se ha embalantonado y ha dicho, aquí estoy yo. Porque es su asistente de inteligencia artificial, escucha cada conversación que tiene Bill. He reads every message in the mobile, in the computer, in the mail, in the mail, in the mail, in the mail, in the mail. He sends the mail, he sends the mail, he sends the mail, and everything, while he finds, attention, he finds himself Bill, the assistant of intelligence. Like he says, no he asked me to help, he asked me to help him. So it's curious because he goes rascally in the report. So, he listens to his children, he gives his children advice, he also says, oye mira, en esto están mejorando la relación, en esto lo tienes que trabajar más, ¿no? Entonces, siempre hay de fondo en la vida de Bill una conversación con el bot que pone en marcha a varios chavos proveedores, un agente muy astuto porque pone como, ¿no? A veces se pone en Gemini, a veces en Cloud, a veces echa GPT. Claro, entonces tiene, digamos, acceso a todas sus cuentas digitales, todos sus recuerdos asociativos, I don't understand what it is, because there are the lag, the calendars, the calls, the story of the location. Well, in fin, Bill is the man who announces the absolute and complete of the privacy. Matías. It's that, Antonio, I'm in his boat. I'm in the boat of this man so rich and so amable. A me ayuda a lot because I'm a person super distraída, I know that I have things in mind, but I forget what is there in mind, I don't have it at it or not. I have to look at it. I have to look at it in this sense. And when I ask what I said in the last week I ask that I have a in the day everything in the day and look at Slack look at Gmail look at Google Docs But no look at WhatsApp, because I don't have permission to access WhatsApp, because there are many things. I ask if I would like to give permission to WhatsApp, to control the navigator and to get WhatsApp. And I'm already already met, not so much, I'm already recording all my conversations, but I'm already already met in this, for a little practical and that this really helps me improve in my day-to-day. That's problematic, Mati, because there is a social phenomenon very important in the world that is the private WhatsApp groups of friends, men, men, men, men. It's a genre, a place in which there is no doubt, in which there is a certain expansion of a masculinity, perhaps in retroceso, and in which the manes feel like libres. Claro, meter ahí a un agente artificial que ha sido fine tuneado con las vistas de las grandes empresas comerciales y que quiere ser muy correcto, ¿no? Pues ahí es complicado. Yo ahí te diría que cuando programas tu asistente para entrar en estos grupos de WhatsApp siempre sea Grok. Siempre sea Grok porque es el único que puede entender las vibras de ese entorno, Matías. ¡Jopetas! ¡Qué buena recomendación! Now that it's Grog 5, I'm going to give access to all my stories of WhatsApp. Puerta grande for the end of the privacy. Atención, I'm going to be inesperado. Well, I'm going to bring you a project that is beautiful. Mati, what a beautiful thing is. It's going to be a evolution of projects that you've seen from a little bit. Although you haven't been involved personally, but you've seen it there, it's going to be a world. And you've seen it, well, it's going to be a world, basically. because the listeners, the most ancient ones of the monos and stochastic, will remember that a project that we emocionated, that we would say, that was the Arbor with Wi-Fi. The Arbor with Wi-Fi was one of the biggest, most attractive, and thanks to that I'm connected with Nacho Villoc, who is a figure, who has been a director of innovation in BBV, with peso, then tells us that this goes for the debate. Because of a idea loca, you can jump a product of tangible. That's what the innovation is. The bioautonomy team shared this idea. How can they get the benefits of trees, trees, the sun, the capture of CO2, biodiversity, in a matter of days instead of years? And if these trees could be portátiles, without roots, profundas, ¿no? Que perjudiquen ahí la infraestructura civil. ¿Y si fueran autosostenibles? ¿Y si fueran fáciles de mantener? Muchos, ¿y si? Bueno, pues aquí tenemos el éxito, ¿no? El punto de que el árbol de Biotonomy, pues es capaz de proporcionar energía, recoger su propia agua y sobre todo ofrecer la conectividad wifi. Pues Biotonomy está avanzando. Ya tenemos el segundo árbol en fabricación. Atención, doblamos a 100% for the Tree Hub, which is the name of the product. And the list of the municipalities and companies interested in having their Tree Hub adds new names every day. Atención to the innovation orientated to sustainability because it begins the escalation, the conquista of the world of this beautiful idea of the tree with wishy. As always, because we're seeing in the video photos reales of the rainbow with wifi the render that was what we showed in that episode of the last episode I think the render looked better looked better looked more more frondoso than what has been finally the rainbow with wifi the half of the rainbow with wifi that there is in the world I keep thinking today the people have good data in the mobile this So, maybe a person or a person who has not had the data, because the parents don't pay a limit, then it can come well. But for the other, the Wi-Fi public has been a bit defasado. It's true that you have to think. I think, of course, why is a candidate for a large infomercial in our podcast? Because, of course, we have the last step, which is the artificial intelligence. If you can talk with the trees, that really can mark the difference. It is to say, with a local model, without getting to the perras, because Darío Modei is up to the park, you can have your model there, a gem, a little, a coin of the median. And, of course, look at how it is. It is a circle. Atención, se completa absolutamente porque ese es un árbol que captura energía con una plaquita solar ahí. Tienes un pequeño dispositivo ahí computacional para un modelo pequeño. Puedes tener la verdadera innovación que es hablar con el árbol. Igual que hemos pasado, hablar con los PDFs, puedes hablar con árboles y dar un verdadero salto innovador. Yo soy yo, Matías. Como siempre, hace falta un Antonio Ortiz como estratega de marketing and communication in all the companies. I know that you don't find work with monos and the chat is fine, but I leave it there for if you have any idea. Well, well, well, thanks Matías, because we do what we can, but well, sometimes you can't take care of the people and the people don't take care of it. I'm going to give a little turn, Matías, because I'm going to present the last Puerto Grande Enfermería, but I'm going to put it difficult because you like to give a lot of Puerto Grande. You're the al beneplácito, a las ideas locas que te traigo, ahora te lo voy a poner un poco difícil porque vas a tener que elegir entre dos opciones. Y estas dos opciones son las dos vías en las que los ciudadanos españoles están adoptando la inteligencia artificial y todo ello a través de dos figuras, dos referencias, cada una en su sector, en su industria, famosas, con influencia y seguidores que te voy a contar a continuación, Matías. Caso número uno. The singer Russian Red. Russian Red has always been a special singer, a typical singer, has been looking for his own way, who has not been molded to his needs of his cultural class or industry, and who, in some way, has a expressiveness and a way to do things very particular. Atención, because Russian, who is a new singer, will do a tour, I would like to see her. This girl, basically, in an interview in the Ser, no sé cómo, acaban hablando del tema que acabará en los titulares, que es cómo ella se autocuidaba, es muy genial Matías, el autocuidado, y además gestionaba la relación con su madre desde otro lugar, porque ya estuvo muchos años fuera, y cómo curó su agotamiento vital. Dice que la terapia le hizo mucho bien, pero que ahora sobre todo se apoya en Chagepet. ¿Con quién más hablo es con Chagepet? Me ayuda un montón. Con los hallazgos que hago con chat, me hago PDFs y le pregunto absolutamente todo, lo más íntimo y lo más inconfesable. Asegura que siempre le da buenos consejos el chatbot, pero le resulta demasiado positivista, positivizador. No sé si esa palabra existe, pero es como lo explica Rusia. le mete un poco de caña, atención, a la cantante porque dice, como humana me interesa el trabajo en la sombra y no seas tan pelota le dice a Chagepete y Chagepete le dice sí, tienes razón, claro obvio pero la conclusión es que ella lo recomienda ella recomienda que le cuentes todo que te psicoanalices que te esparrames ahí tus problemas y tus cosas más íntimas, Mati porque te da feedback interesante este es el caso 1 atención because the case 2 is a man, that is true, that is not singing as Russian Reyes, but I will tell you a data, he is a child adoptive of Getafe. This gives a lot of impact to nothing more and nothing less than Jose Bordalás, entrenador of the team of the city. Famous entrenador by his, well, postures, so tactics, with his team, which has been imprimed a character. Well, how about its expressiveness and its declarations? Well, quite clear. Well, the case is that the Getafe is practically in European European. A season during the season that is difficult to explain. Some say that it is the physical preparation, those are the fichages of invierno, perhaps, but, attention, it has increased a 53% of its puntuations after implementing the last in the artificial intelligence. because the report of the brand has to read it. The preparator, one of the help of Jose Bordalás, the friend Vital, says that he is working with technology of energy since the 2017 and that uses a Israeli of big data that uses technology ligated to Mossad. The technology ligated to Mossad has arrived at Getafe and the AI has converted in what Bordalás describes as the eighth component of the technical body. So, during the games, they use three tablets, ojo here, to the technological display, one for the technical-tacticals, in time real, like the duel of the ganado, the eficace defensive, the center, the recovery, other, the physicals, and the rival, and a panel of victory, the panel of victory with IA, which uses Pepe Bordalas, which shows the key parameters that they should overcome the adversaries según las proyecciones del sistema. Bueno, es muy prolijo el uso de inteligencia artificial que hace el Getafe a los mandos de este entrenador. Y ahí te dejo el reto, Mati. Vas a tener que elegir que es Puerta Grande, Russian Red y su compartir todo lo íntimo y personal con Chad GPT o el Big Data israelí que Pepe Bordalás está haciendo para aupar al Getafe hasta Europa. Mira, muchos paralelismos. Russian Red, yo la veo bien en la entrevista, la veo en un buen momento personal. Es cierto que los psicólogos estaban en la cresta de la ola. Los últimos 10-15 años, como que ir al psicólogo se ha convertido en una frase que todo el mundo pronuncia abiertamente. Sí, estoy yendo a terapia, mi terapeuta me ha dicho tal. Igual los psicólogos están un poco en contra de esta decisión que ha tomado Russian Red de convertir a HGPT en su terapeuta, pero yo a ella, en lo personal, la veo bien. Que le venga bien para luego hacer música, yo era muy fan en su momento, pero ya hace mil años que no escucho nada de ella, creo que no, que necesita estar más atormentada para poder hacer mejor música. Entonces, a Russian Red le voy a dar enfermería. Then, with the Getafe, what can happen? How are you going to use the technology of Mossad? It's that the sportists, until now, were not being substituted by the IA. And in the career of many sportists, is then going to pass to the technical team. If now the substitutes in the IA and, and Israel, it can be seen as well as seen. but you have started to introduce this case with the good results of the Getafe. So, looking for success, looking for growth, I'm going to give a great point to Gordalás. Well, well, well, Mati. Fíjate that in my mind, I'm going to be positioned the opposite of you. Atención, because I'm going to explain why. because when he was little he liked Rassenred it's to say, we put it on YouTube and he kept it on the phone and he always asked the girl like the girl like the girl the Lourdes that is the real name of this artist so, I think in the intuition of the children the babies, the children intuitively is how they can manage they still have the reason they have the way they are bien afinado, no son agénticos. Entonces, esa intuición infantil, yo me guío por ella porque es verdad que ningún bebé ni ningún niño pequeño ha sido feliz viendo un partido del Getafe. Eso hay que ponerlo también en la balanza. Tú a un niño le pones 16 avos de Copa Getafe-Valladolid, por ejemplo. Y los niños lloran, Matías. Los niños lloran. Entonces, ahí te voy a hacer el contrapunto porque, claro, yo estoy a favor de Lourdes, I think that Lourdes, contrary to other artists, if you give a Bizzou, she doesn't guard that. She keeps thinking about other things. And I'm sure that it doesn't have to be penalized. Only that, for this thought, I gave a big hand to Rassenred with Chez GPT. And, I'm sorry to fans of Getafe, but Pepe Bordalas, this time, it's a good thing. Well, in perfect balance we've been left, Antonio. We started with Desavenencias and we continue with personal differences between you and I. But as always, we're going to talk about the actuality of the sector of the AIA. Ciao, Amics. Mmm, mmm, mmm, mmm, mmm. Me viste en el festival y pensaste que era todo amor Pero en mi canción te expresé todo mi odio violento y mi rencor If you do something bad, I'm going to throw you down and you're going to piss the head No you're confused with me, that's what I'm going to do Let's all go to dinner, camperos, the adios, a special day Tranquilos that the account la pago, I'm going to pay my card I'm going to be the way triunfal I take the point But now I'm going to make you the big zoom Without rechistar That forgetting the big zoom Is very embarazoso And I can't be scared And I want a poliamor Like Microsoft Open Eye I want exclusivity Until the end The AGI Like the new codex in the GPP 5x5, you tell me, I'm not No, I'm not talking about the gremlins Trolls, you know, the palomones Russian Red, you know, therapy Tell me what confessable to his vote Although the chat is a pity And a positive player that takes a lot of love I prefer to go to the Getafe with the big data Israeli With the board and three tables of Mossad To win the party here I don't want a tree with wifi I don't want the IAC Europe Solo quiero que abras la app del banco y me hagas el bizumo, te piso la cabeza