I Talked to the Guy Building the Anti-ChatGPT to Take Back Control from Big Tech

0:00Today, my conversation with an AI evangelist who says we're thinking about artificial intelligence all wrong. This country was stone cold dead. And don't forget, if you're enjoying these conversations, you can subscribe to us on YouTube. And also you can subscribe on Spotify, Apple, wherever you get your podcasts if you prefer to consume these conversations that way. But everything counts. Every little piece of engagement counts. And if you have any questions or topics you want to hear more about, you can always email me my email in the newsletter every day. Hello, hello from Newsweek World Headquarters at the top of the World Trade Center in New York City.

0:32I am Carlo Versano, and this is The 1600. So today I am back on my AI hobby horse. A lot of you guys have written in and DM'd me and said various things. Like, you're talking about AI too much in the newsletter and on the podcast. Too much AI. I want to hear more about Trump and, you know, what the craziest thing that Trump did recently. So, look, I'm sure that, you know, when people were doing podcasts about the horse and carriage and horse and buggy industry, when Henry Ford was tinkering and perfecting the internal combustion engine, people were probably saying a similar thing. But the fact of the matter is that Trump is what's happening right now.

1:09But he is, frankly, just a small part of a much bigger story that's happening all around us right now that has to do not just with technology, but the intersection of politics, culture, and everything. And a big part of it, of course, is artificial intelligence and what we're seeing now with the exponential growth in AI. So, you know, we've been talking in the newsletter recently, like I said, about you know the sense the sense that people have that we're approaching this inflection point with ai right we're finding out that some of these newest models now are essentially building themselves right they don't even need us at least that's what these companies are telling us they can write the code themselves they can self-iterate and automate without humans we've got banks like goldman sachs that are starting to outsource big parts of their back office like compliance right to AI.

1:57Even the biggest AI evangelists out there are talking about in no certain terms about how entry-level work in this country and in this economy is about to essentially go away and soon. So if you're a layman like me, this can feel very scary. It can feel exciting also, but mostly scary. So I wanted to bring on somebody today who can talk to us from the vantage point of actually working in this business rather than just sort of writing or talking about it because there's plenty of those people out there and they have very valuable insights and we'll talk to more of them later but i want to talk to somebody who's actually in this business uh to help make sense of sort of where we are but also somebody who's not working with the big guys who's not working at meta or google or open ai or anthropic um which i think is probably safe to say that those companies don't necessarily have our best interests in mind and by us i mean humans so without further ado without uh after that long intro i want to bring in angel Maldonado who is the founder of Empathy AI.

2:54Angel joins us from London today. Thank you so much sir, I really appreciate you joining us. Thank you Carlo, what a joy. Thank you so much. So Angel, let me ask you first what is Empathy AI which is the company that you helped found and are currently running? Well I like to think of Empathy AI more than a company or a corporate and to think of it as a movement, a reset, a radical movement that we are building in the image of European values. So under EU soil, transparent, self-hosted, sustainable, solar-powered AI that is, in my vision, limitless. Because the real limit to AI isn't as the big tech voices are trying to present, but it is the limit is in compute.

3:50The limit is trust. As any other thing in our lives, whether it's a product, whether it's a place, a person, an idea, we don't relate to them. We don't become augmented by them on the basis of how powerful or how functional these things are. There is the element of trustworthiness and being able to embrace something that you can trust. So when AI is presented as a dark, with intentions that escape your control, with trying to provoke longer and necessary sessions, trying to present a notion of distance and dependency, then you are creating limits to AI. And we think we can break those limits if we create trustworthy, honest, and, you know, transparent AI that can serve human growth.

4:47Well, that sounds fantastic, right? So I think that I think you're certainly onto onto something there. But let me ask you, I guess, my first question, which is can AI actually be empathetic? Because to be empathetic, you have to be conscious, I would think at least. But these machines aren't conscious. I mean, they're essentially big probability machines. That being said, we are seeing, as I noted in the intro, you know, some of this exponential exponential growth with these LLMs that suggest maybe it's more than that. Maybe it isn't necessarily human consciousness as we know it, but they appear to be at least some of these models approaching something that is more human than at least we thought of even as recently as six months ago.

5:29Am I on to something there? Well, let me be absolutely clear, and this is very important, as you said, for the narrative of the way AI is being presented. it is not possible for AI to be empathetic because you have to expect a reciprocal engagement if you are to ignite or negotiate empathy. Empathy cannot happen between an individual and a tool because the tool doesn't hold a notion of itself and a consciousness, as you were reflecting, Carlo, a second ago. So it is not true. Empathy is a very powerful feeling that is complex, that is the conscious reflection of a wealth of feelings that, you know, they sit on something else than the mechanics of producing text or image or audio.

6:29So this is one of the problems that we try to challenge or one of the narratives that we try to review, this anthropomorphization of AI. AI mimicking thought, mimicking feelings, mimicking qualities, which works really well when you want to have an AI that coerce an individual into an engagement that is presented to hold another kind of advancement. You know, AI is a tool and it needs to admit its limitations. It needs to reject anthropomorphism and it needs to expose these contradictions like the one I'm saying. In fact, Carlo, we bought the empathy AI domain as we were afraid of a company presenting a notion of an empathy, sorry, an AI that can be empathetic.

7:14So we're trying to do just the other and put forward the other message. AI is a thing. I have so many philosophical questions here. But before I get that, talk to me about sort of like the nuts and bolts. Did you build an LLM? Is this sort of like a wrapper on an existing technology, or is this a model that you have built from, and you and your team have built from the ground up? Does that make sense? Yeah. We use existing both open source, like Olmo 3 models, and open weights, like Mistral or Quain. But we don't use any foundational model that is coming commercially. everything that we use is under a license that has freedom to use and extend, replicate, modify, with no limits.

8:05So we do use underlying models that come from a place of mysterious provenance. Now, for the audience, provenance is what is the actual origin of the data that is used in a model. So we do use models that we don't know with integrity if there has been consent to provide that underlying data. But it's important to clarify that we don't use any cloud provider. We don't use any mega player. We don't use any service that is a third party. We have our own AI servers, which are called GPU servers or graphic processing unit servers that we host in our own building. So it's like we're going back to the 2000s where companies had their own infrastructure in the basement.

8:54So in the gymnasium, in our office, we started to build and grow our own AI farm. And then what we did, if people can search that, we build a solar structure outside of our building. to be able to sustain this in an explore ways of sustainable compute. So we do use existing models as long as they are open source, and we release a lot of models ourselves, but not the big mega capital ends. And we do compute everything within our sphere of privacy, within our control and our governance in a physical space that we own. So that is a big factor there. So when you say open source, you're talking about models like what regular people sort of have come to think of as AI, like a ChatGPT or Anthropics Claude.

9:46Those are not open source models, right? So when you're using Empathy AI, you are not interfacing with those companies. Is that correct? Correct. It would be impossible otherwise, and this is a very tricky dependency that you can create if you embark on a strategy with AI that has this intrinsic dependency. You need to have your own models. And again, the narrative of Anthropic, the narrative of OpenAI, GPT, the narrative of Microsoft or Google Gemini is that you need the big cloud. You need to have these massive mega compute spaces. when in reality, to give you a very good example, I believe, and I believe you like metaphors, is as if you go to a restaurant.

10:29Every time we go to GPT or Perplexity or Gemini, it's as if you go to a restaurant and you say, can I have the menu? And they give you all the menus that there are in the entire world. When you just had a context, you had an intention, you had a specificity. Or to think of another metaphor, For instance, if you jump into a car and you decide not to have any gears, this idea of massive compute applied to any kind of problem as opposed to understanding where are the specific domains or contexts that you have. You know, if you're a law firm, if you're a clinic, if you're a school, there are so many organizations that don't require to fall into a dependency with a big cloud.

11:10And they can have their own compute, which is the premise that we try to inspire and exemplify. And then you're solving for these other problems that people are now talking about, which is like how much electricity some of these data centers are using by having them sort of self-sustaining via solar. Is that accurate? Yeah. Again, it's a very appealing mystery to present that, oh, you need a lot of power. This needs to go to space. You know, the distance is necessary. and therefore you have to surrender to this schema and this architecture and that proposal. But it's not the case. I mean, imagine in the work that you do as a journalist, in the work that I do as a product designer or an AI maker, I don't need all the knowledge of the world to be able to tackle the tasks I am tackling.

12:03I approach things with specific contexts. That is what makes the current proposition of big tech AI very unsustainable. Every time you open an AI experience on the cloud, whatever it is, Gemini, Anthropic, you know, GPT, is presented with everything, ready for everything, right? When in reality, you can be way more sustainable and you decide what you want to work on. And then you realize that every household, every household could have their own AI computer. There are NVIDIA Spark computers that can hold 400 billion parameters and they cost $4 ,000 today. But this narrative that AI has to be in the cloud is not so.

12:48You can achieve so much more when you put it within your sphere of privacy and your sphere of control and governance. I've never really thought of it that way. Hold that thought on the privacy angle. I want to come back to that, but I want to talk about safety. So this is a headline in Barron's today. AI arms race risks human extinction, warns top computing expert. That's not a particularly crazy headline. I mean, it is a crazy headline, but these are a dime a dozen now, right? There's always now some AI person or some computer scientist saying that the end of the world is coming. But it isn't just talk, right?

13:23We're seeing people walk now, too. And I want to bring up a couple of examples and ask you a couple of questions, so just bear with me. But this month we've seen a handful of some high-profile people working in AI who are working in this business, especially on the safety end. They're working at the bleeding edge of this industry, and they are warning to anybody who will listen to them that it's starting to spin out of control. So there was somebody named Miranank Sharma, who was a safety researcher at Anthropic. This person resigned. In a letter they said, quote, the world is in peril and that the commercial imperative of AI development is just no longer compatible with sort of the caution that's needed to ensure that it doesn't spin out of control.

14:03I'm sort of putting words in their mouth, but that was the gist of their resignation letter. Two days later, a man named Jimmy Ba, who was one of the co-founders of XAI, which is Elon Musk's AI, he resigned and he said that this year was going to be one of the most consequential ever for our species. The fact that he would say that and not want to be part of it is telling, I think. And then last month you have Dario Amodi, who is the I guess I don't know if anyone's ever called him this, but he's kind of like the godfather of AI safety. Anthropic is supposed to be sort of the AI safety company.

14:35He wrote a 20 ,000 word essay acknowledging that artificial intelligence is, quote, considerably closer to real danger in 2026 than it was just in 2023. and also that half of all entry-level white-collar jobs will probably be gone within the next five years. To use a line from one of my favorite movies, Jurassic Park, when Jeff Goldblum says, you know, these scientists were so preoccupied with whether or not they could, they didn't stop to think if they should. What do you make of these people who are issuing these warnings now? Yeah, I think coming back to Jurassic Park or any of those stories where there is a monster that causes harm and we might be misguided to think that it's the monster, it's the dinosaur, it's Frankenstein, that which is evil or that can provoke those consequences.

15:31It is not the monster. It's not the monster that is the problem. It's the maker, as in Jurassic Park. the unconsequential, mindless, careless thinking. So the danger is not AI. The danger is in the people who are making AI that doesn't have any preoccupation in understanding that it can have consequences to individuals, households, societies, and of course the civilization that we try to navigate in this present time. If there is such a thin civilization left, Anyway, joking aside, you know, you have all those stories. It's not about the tool. It is unchecked ambition. That is what is the problem that can destroy our civilization.

16:26And we need to challenge these narratives, which are very good marketing tools, to surrender governments, to surrender institutions, to surrender households and individuals, to expect that there is an inevitable outcome. It is the lack of moral value. It is the lack of consequential thinking. It is the lack of bringing into the strategy that which data cannot give, like, for example, autonomy, like, for example, a sense of justice, like, for example, privacy, like values, what becomes a challenge. And this is what we try to address when we present Empathy AI as a movement to reclaim intelligence.

17:07It is another illusion of progress. it's another illusion of progress you have many other spaces that have been claimed by the same mentality I'm going to give you three examples very quickly the most beautiful creation of playing and games a space that has been claimed and taken by technology and products the space of socializing we've seen that space also being claimed by technology and products which are the terrible consequences that this has had the space of intelligence and creativity another space that now is being objectified and thingified and claimed. So it is one and again another illusion of another iteration of these propositions.

17:50And they are not serving human growth. They are serving the interests and the very corporate interpreted exclusively as an instrument of revenue generation, which of course we can challenge that too, right? But anyway, there is a lot of many propositions with the many grandfathers of AI that are distant from the truth. And you can have a perspective, Carlo, or anybody listening to this can have a perspective and a fear. But let me focus the attention on, to me, what is the most important audience and the most important group of people right now. and there is the new generation, children, young adults, that they are in school supposedly learning something and they are being told at the same time by all these narratives that they're not going to have a job, that they have to be afraid of AI.

18:44Once they understand that AI is a tool, it's just another tool, it's an object of use, not the other way around, they understand that AI is a physical thing that you can place somewhere and you can put under your direction and governance, they understand the things they can do with this tool. One thing they understand is not something that has to mimic thought. It's something that has steps and actions as opposed to reasoning. And they can then turn that narrative. And rather than being afraid of what the future can bring with AI, they start to feel empowered and they understand that it's the future.

19:21and maybe that interpretation of AI, what has to be afraid of themselves, right? AI is not nothing and merely something constrained to its inputs. And the capacity that a creative mind, the kind of things that a creative mind can do with AI are incredible. Many, many more. You can think more, do more, advance more. but that narrative that is going to take over the genuine intelligence and intellectual capacity i'm sorry it's not true well that i think that's a very important point for people to to hear and it just reminds me you know when i was when i was coming through you know through school and through grad school it was always you know the line was always like you know don't study the humanities don't study english and journalism and political science you go into you got to learn to code right everyone said here in america at least from the president on down learn to code learn to code So we have all these kids going into higher education, learning to code, because that's what they were told to do.

20:19Then they're coming out and they're being told, well, actually, those jobs, that's the first thing that AI is actually capable of doing. It's better at coding than the things that the humanities actually teach you, which, to your point, has to do with reason and creativity and critical thinking. So I think that that's – I just – I feel bad for so many of these children and young adults who are coming into this because they're coming into an economy in a labor market that's doing exactly what they were told to do. And now they're essentially seeing that the skills that they were told to have are the ones that aren't needed.

21:00Does that make sense? It does make sense. But let me challenge these. The value of a piece of software is not in the code. It has never been in the code. The value is in the way somebody has analyzed the problem, designed a solution, and proposed a way of implementing it. The value of solving a problem, the capacity, is not in the way you implement the solution, but in the way you define the problem. The definitions of the problems contain the answers to those problems. And the same applies to code. We have reduced the notion of what software engineering is to coding. Coding is just a task. It's just a translation to a given language.

21:52But the real beauty of creating products with software is the same beauty of creating products in the physical world. It's about the conception of those creatively. And then again, we have narrowed down so much what software engineering is coding. I think it's a misinterpretation. Maybe there are people who are experts just of the act of translating specifications to code. And that person will never have been a good software engineer pre or post AI. Right. The talent, whether you're a lawyer, whether you're an educator, whether you're an actor or an actress, whether you are a software engineer, it's in your capacity to understand, your capacity to define things.

22:35Whether you're going to articulate the solution then through code is not that important. Let me ask you about – so you brought up this issue of trust. and the first thing I thought of was, you know, a lot of the companies, not what you guys are doing, but the big guys, right, the Open AIs, the Googles, the Microsofts, the metas of the world. I think a lot of the fear that you're seeing in sort of like the general population is they just don't trust these companies. And why should they, right? I mean, these are the same firms, some of the same people who were telling us, you know, 10, 15 years ago that social media was going to be this thing that connected the world, right?

23:16It was going to solve, you know, it was going to bring democracy, right? That was the big thing in the Arab Spring. It was going to bring democracy to places that have never had it before. It was going to make us all closer. Obviously, none of that happened, right? And in fact, it did the opposite. It actually made us more divided, hate each other more, more, you know, more tribal. So if you don't like what do you I don't even know what my question is there, but I guess that I guess it's just like why should we trust any of these people? Yeah. Well, narrow incentives and narrow outcomes. And inevitably, the truth gets you, no matter how big or small you are.

23:56I mean, they get in to choose how slow it gets them. But the people you cannot fool people forever in specifically the story of social media. is the story of surveillance capitalism. It's the story of selling engagement. I find it fascinating, Carlo, very relevant that the most engaging content on social media is the content that triggers your instinct of survival. It's the content that you have a necessity to stay watching because it's so extreme that it's affecting and it's tapping on your very instinct of survival. And that's why the content that succeeds in social media is segregating, is polarizing.

24:46It creates, and as you say, it creates silos and tribes. And the definition of what safety is is challenged. And what is very worrying is that that very same, those mechanics, the dynamic of pushing content that is extreme, content that leads to further engagement, because the means is not the act of socializing, the means is the act of engagement. And those very dark patterns that are present in social media that now everybody has become aware of are the very same dark patterns that are going now into big tech AI bots. There are incredibly relevant studies how when you go to Gemini, a child, six-year-old, 60-year-old, 50-year-old of whatever age, when you go to Gemini, it's designed for engagement.

25:43It's not giving you the precise answers at the right time. It's designed to hook you into longer sessions with the same emotion, emotional coercion. So the very same dark patterns propagate into and tropic into OpenAI, into Gemini, etc. Because these platforms are also designed with the same outcome, which is the productization of the individual, as opposed to serving the individual and engaging them critically or cultivating their curiosity, admitting the limitations of the tool, etc., etc. So those very principles that affected social media that people have learned not to trust social media, we're beginning to realize that the venom has moved now into these AI boats.

26:29And it's very worrying. And there are incredible stories on that space that we need to address with urgency. Well, I want to ask you about one of those stories. Well, let me ask you to just tell me about it. You guys, Empathy AI has helped finance a new film. It's coming out in a couple of weeks. It's called Molly vs. the Machines. Tell me a little bit about the movie first, and then I'll ask you why you chose to get involved in this one. Yeah. Here in the UK, there is many stories like Molly. Many, many, many, many. Hundreds, if not thousands worldwide. Molly Russell took her own life at the age of 14.

27:11And I can tell you because I know the dad, and I know the story, is just a normal family. She's just like my kids or any other family, any other ordinary, perfectly normal family. And Molly was 14 and they found her in her bedroom. And then the story is that the inquest, for the very first time, once they were able to investigate the content on her phone, She had over 2000 pieces of content sent not by an individual, but by Instagram, by Meta, by Pinterest, on self harming. The most, and you can have that, it's quite difficult to see, but the kind of content that a 14 year old who naturally can be on a period of depression and was invaded by all this very difficult content because it was engaging for her.

28:12And the algorithm was working for her within that silo. And then again, is the consequence, which in the film is co-written by Shoshana Suboff, author of Surveillance Capitalism. And you have this story between Molly, a child and the machine, Molly versus the machines and the machine being the platforms built for engagement. So, yeah, we're profoundly honored to be able to finance this film. And it's been an incredible lesson. You can see if you go to molly vs. the machines.com, we have created an AI experience built on Empathy AI, where you can talk privately with the letters that Ian Russell, who is an incredible hero, Molly's father, has sent to the UK government.

29:00You can talk to them with an AI that is exercising de-anthropomorphization, speaking to you in third person. There is no AI in the world, Carlo, that speaks to you in third person, only that. That presents transparently all the steps on how that's happening. You can talk to Susana Zuboff's Surveillance Capitalism and have a conversation that is designed to engage the reader critically, to cultivate your curiosity, to admit its limitations. You can talk to the transcript of the film. So we were not only putting forward this message and this important story, supporting that story directed by Mark Silver, the most talented, incredible director, but also continuing the story in a website where you can demonstrate that it's not that technology is bad, it's the way it has been made, right?

29:53right so that's the story of maury russell yeah it's an incredible film comes to the cinemas here in in the uk across a hundred cinemas on the first of march and she'll come to the u.s soon yeah and it'll be available is that going to be a streaming uh film on in the u.s is that the idea well so far channel 4 has the rights here i i see i'm not involved closely on how the rights go into the US, but it's getting an incredible, you know, every prime minister is watching this at the moment and parliaments and organizations. It's got an incredible momentum. And it's beautiful that Molly's story has been applied into this cause.

30:36And it has been connected so beautifully with the world of AI, not just the world of social media. is a fascinatingly, incredibly generous story. And the role that Ian Russell plays is just our hero. That man is incredible. So it's a great story. Thank you for asking about that. I know what's interesting to me is that, I mean, we've heard so many stories like this of Molly, even before AI came into the picture, right? I mean, you know, there's been inquests here. There's been lawsuits about meta specifically when it comes to targeting teenagers and teen girls with eating disorders, with content that will make them more inclined to harm themselves.