The Hidden Flaw of ChatGPT (It's Optimized for Satisfaction, Not Truth)

Sean Weisbrot: My guest today is Ricardo Michael Reyes, the co-founder and Chief Science Officer of aite, which is developing a people first AI that empowers happier, healthier workplaces that prioritize wellbeing, which is something that I've been talking about for a very long time. So it's really cool to be able to talk to someone who's using artificial intelligence to study. Plan for and actually create opportunities for companies to better understand their employees. He's also the chairman of the Technical Board of Billion Neurons, a $1 billion fund that invests in technologies with a focus on social welfare and advancing the state of the art technologies of human humanity. He fell in love with AI in 2009 and has become an internationally recognized expert on the topic, and he's even advised the government of Mexico and the European Space Agency. So I thought it would be interesting to talk about the social impact on ai, and that's some of what we covered today. He's definitely a Star Trek fan and a ghost in the Shell fan. So we geeked out a little bit. Um, we talked a little about Neuralink. Um, we talked about what he's afraid of. In the future, what he's excited for in the future, the privacy implications of AI and Neuralink, and the connections that we're going to form with the internet and how AI is going to change Latin America and how it's gonna change the world. This was a great off the cuff conversation and. I hope you enjoy it. Ricardo, you've spent a lot of your career focused on ai. What happened in your childhood that made you so excited for something like this?

Ricardo Michel Reyes: It's not actually about AI itself, but what AI allows people to do because, um, a lot of times people are very talented. Like, uh, most of my friends are artists. Even when I'm a scientist, uh, I have very few other scientist friends. Most of my friends are artists. And for them, uh, most of the times they picture an idea or something they wanna do. Or wouldn't it be cool that this tune that sounds like this, or wouldn't it be cool a picture that looked like this? And a lot of the times. Making these ideas come true. Like in the artist world before generative ai, it was like, okay, let's get some clay, put some things together. Let's make a collage. Let's go take photographs. Let's, uh, play on the piano. But now with things like another firefly or like all of the generative AI algorithms, table diffusion, and they can just put their ideas into a prompt and make them come true. Uh, so, uh, one of the first breaks I wanna work with when I was, um, younger was just that like, how to control all of the artistic tools, like Photoshop or Illustrator, all these things, uh, just with your mind or with your words. So just by saying something. So like a city but for art. Uh, and now Adobe has make it come through. Um, but like, yeah, that was like the first thing I wanted to do with ai, like be able to create. Art just by where I end up.

Sean Weisbrot: So is your ultimate goal to have your brain connected to an AI so that it can generate art for you based on what you're thinking? Oh my God, no. Well, because that's what it sounded like. Like the

Ricardo Michel Reyes: privacy? No, no. But your brain is something too sacred as to let anyone be able to hack it or like. Imagine like anything that can be hacked will be eventually hacked. So then at least when they hack your computer, you have still an object that is not part of you. Uh, so for example, breaks like Neur Link, uh, sound great. Um, but then if anyone is able to hack into your brain, I just. Fry it out. Uh, I wouldn't use that, like just the, for safety and privacy reasons. I wouldn't like someone to have the ability to intercept my signals or hack the software. And just, especially now that there's quantum computing and there's gonna be a huge revolution on cryptography and all of, uh, regular, everyday cryptography is gonna have to change very fast. Uh, I wouldn't like to trust the safety of my brain to another person. So then no, like, uh, just like with a regular microphone, keyboard, anything that is outside of my body fine.

Sean Weisbrot: But no, I wouldn't connect the neur link in my brain. Would you agree or disagree that it's inevitable that that's gonna happen anyways?

Ricardo Michel Reyes: I totally agree. I totally agree. Um, I think the role of regulation and governments will be very important in this part. Like this has been achieved with HIPAA compliance. For example, in the US and Canada, there's a lot of rules about privacy in health records. Um, this is all everyday thing in defense industry. Like what can you do and not do with Saudi exploit? Uh, for civilians bodies, it's gonna be interesting. But it is very similar to the regulation that you have by the FDA for like diabetes pumps or for. Um, pacemakers. There's already a lot of medical devices that go through this kind of testing, so I think Neurolink will have to go through FDA compliance at some point, but it will eventually happen. Of course, like I, I, I wouldn't do it, but of course it's gonna happen.

Sean Weisbrot: Well, what you're talking about is great, but inevitably there'll be someone in some other country, ah, China, where they won't have to go through those things to make it happen and. Those products could be sold on Amazon and you could just do it yourself, right? Just like people were afraid about, uh, CRISPR Cas nine, they were afraid that people would start to tinker in their homes, which it, I'm sure it's happening, but they, you know, there is this negative assumption of humanity that someone's going to create the next genetic disease because of it, or some next virus, surely. Once people figure out how to build this tech in their own homes, and I'm sure AI can help, all bets are off, right? So, uh, I guess my inspiration for the understanding of this type of future is ghost in the show, it could be totally wrong, but it could be totally on point. I mean, a lot of sci-fi TV shows and, and series. A lot of them inspire real technology.

Ricardo Michel Reyes: Yeah. Because in the Shell is amazing. Like, um, both the Japanese and American version and yeah, it's totally a feel. We're gonna live in our, in our living time. Like you and I, like you are like 30 something. No. Or 20 something like 37, 30 something. Yeah. There you go. So in our lifetime, I said 37. 37. Yeah. Like, uh, in our lifetime, considering that lifespan expands. 5%, more or less, um, per every decade. We're gonna live about 105, 110 years. We're gonna live, uh, this kind of future for sure. Like look at just from chatty pity that got out in January to now all of the things that have happened. Keep this exponential growth for a decade. Like totally ghost on the show.

Sean Weisbrot: Yes. It's very interesting to watch and, and scary in a way if you think about it, because it's cool. You know, I, I first saw a ghost in the shell 25 years ago, let's say. Right? I was a teenager and it was really cool, and at the time I was, it was kind of scary 'cause all the political intrigue and the hacking and the murder and all of that. But as we get closer to it, it's like. I don't like social media. I don't even like smartphones. I use them, but I don't like them. What if, you know, like, like my grandpa lost his job to someone who wanted to use a computer in the nineties. Like my grandpa said, screw it, I'm gonna retire. Right? So like in the nineties, people who couldn't use computers lost their jobs. Then people who couldn't get into social media or smartphones lost their jobs. And then people who can't understand AI or work with ai, they're gonna lose their jobs, right? I feel like. The next generation will be like, oh, if you don't want to get an implant in your brain, then you're gonna lose your job and you may lose out on economic opportunities because you're not willing to take that step. And, and I think you and I are probably in a similar mindset of like, I don't want that crap in my brain, but then the next company,

Ricardo Michel Reyes: you or me, or whoever needs to create is. Security for human brain computer implants, like, uh, really good cyber security, cyber security for these kind of interfaces like cyber security for, because people just charge GPT and copy paste things so free. Like, oh yeah, I'm gonna copy something with my password in and tell me how to do a report. Like, uh, people are just, and, and I hope OpenAI is an enable company and like. But I, I don't, I don't mistrust chat. GPTI think. I think if you copy page chat g pt, probably open AI won't do anything with it, but I'm sure a lot of people, I'm trying to see how to hack the prompts of people. To get all these passwords and codes and things that people just copy and paste away Totally possible. Yeah, like intercepting just your copy paste, like this is something like key loggers are a kind of virus that has been there all of time, but then you have to like type something and it will record your keyboard, I think copy paste malware. Must be everywhere now that is DPT because then a lot of people are just copy pasting, uh, API tokens, copy pasting card numbers, copy pasting code, like the amount of copy paste that you could intercept to these prompts, uh, to use it for bad purposes. And now you can have key loggers because the problem with key loggers, like in early hacking. Was that you need to read all of the logs of the key logger to find a password and then use that password and use that username. Or maybe use like a search. And, but you would have to know, like maybe you would search for, uh, an ad symbol and then see if that was the email, and then try the email and things like this. Uh, nowadays you can just, just charge GPT to automate your key loger hacking. And just have like a full pipeline of key logging, hacking plus chat GPT to like filter out the emails and passwords and have scrapers to just try that out. Uh, and now you have,

Sean Weisbrot: please, please, please don't have the FBI come knocking on my door. I don't want to deal with this. Let's, let's. Teach the audience how to, how to become hackers. You're just talking about cybersecurity and now you're going, Hey, I'm gonna teach you how to break cybersecurity. So this is. Totally not about breaking cybersecurity. Okay. Gonna get demonetized and de platformed and can, well, maybe not canceled, probably actually. People will wanna throw money at us to learn more, but before you finish the process, they gotta pay for that. That's the premium episode, right? That's how capitalism works.

Ricardo Michel Reyes: Yeah. Have to use ho past agents to hack and create a key logging password. Still pipeline.

Sean Weisbrot: I'm sure people love you when you give. Public speeches, you just like giving away state secrets. Okay. We can change topics so you don't get the monetized. I mean, what do you wanna talk about? Mean? I don't even get monetized yet, so I'm not worried about demonetization. Um, well then make a cybersecurity chat. PT start. I know a cybersecurity guy who has an emphasis on ai, so it's possible. But I wouldn't even know how to lead the company. I'd have to make you CEO and you're so busy with all the stuff you do. I don't think I could, I could convince you unless I gave you like my at least 50% equity, which I can't part with in order to get enough money from investors, but's a different conversation. Actually, you know, let, it's a good segue. Let's talk about. Your role in investing in artificial intelligence companies. I think that's interesting because of your experience and your background in it. Um, so you're the chairman of the Technical Board of Billion Neurons, and from our previous conversation you told me that you invest in Latin American based startups that focus on ai. So. From that kind of point of view, how do you see AI changing Latin America?

Ricardo Michel Reyes: Yeah, so the stars of the game, there are the three general partners of Nu uh, which is Alan. Um, he has been investing in companies. He invested in Robinhood, in Lemonade and Stripe, very early stage. So he saw the future of FinTech, like really early on. Um, and he is been doing so good with his initial fund that was called V Disrupt, and he's the one who made like the call of, Hey, what if we do a billion uh, dollar fund to invest in like ai? And this was before JD pt, before all, all of these things. Um, and then they got into wa, they, uh, ade, um, busy laugh and made their thesis like most fo most focused on Latin America because initially it was global and now it's more like Latin America has 550 people, uh, 50 million people that speak most of them the same language. And then you have Brazil as a separate market. But Bra Brazil in itself is enough of a market to be interesting. Thing. So then you can have like, uh, Spanish speaking, Latin America, plus Brazil. You can say like, okay, I have one partner in Brazil, the others in Latin America. And there's so much talent. Like you have no bank, you have, uh, Rappi, you have story, you have Quebec. Like there are so many unicorns coming out of Latin America right now. And also Latin America is, uh, very. Cool region to live now as a British person or American person that wants to leave, uh, Wisconsin or Montana or all these cold places, or maybe you are in some town close to Edinboro and you want to see the sun, and then people come to Tulum or to Puerto and set up a visa fund or a startup there. So you have a lot of expats living in Mexico with a lot of knowledge, with a lot of money that are trying to do things. And the same in Colombia, the same in Costa Rica. Now, Cuba that is open from embargo. A lot of money is going into Cuba and people are trying to figure out how to rebuild Cuba. But now, from a Western perspective, um, like the, the whole continent is full of resources like mining resources, for example, agriculture resources. Like do you have a, a huge market? With all sorts of sources of income that is becoming ized, uh, more and more every day. It is full of young people, uh, full of people who wanna work. Um, you don't have the attitude of French people of like, oh, you want me to work more? I'm gonna go on a strike. No, like this. These are people who happily work 16 hours and are just happy to work. Like, um, so it's, it's a great environment,

Sean Weisbrot: but how is AI gonna. But how is AI gonna change Latin America?

Ricardo Michel Reyes: One is how to make better use of people because you have, the problem with Latin America and white people immigrate to Europe or the US is that you have a lot of overqualified people. Like there, there are a lot of people who are super knowledgeable, who are self-made, who make master degrees, uh, PhD degrees in really cool stuff, especially like sustainability and. These kind of things. And then they have to go to Germany because there's no one interested in water management or waste disposal or any of these things in, in Latin America. And they go to Germany and do, make, make amazing water plants or solar plants in Europe because there's no one to employ them in Latin America. So you have. Great talent and I think AI can help, uh, companies. Now that is not that important to be in Germany. Like you can work for a German company from Mexico. I think, uh, the LinkedIn is a good product for this. Uh, I have hired most of my people from LinkedIn, but still LinkedIn doesn't have the algorithm to tell you like, oh, you put a job post. Here are the 10 best candidates for your job post. You have to read the CVS yourself. You have to analyze everything yourself. I have seen platforms like Spark Hire, uh, that record the interviews, and then you can look at them and maybe they do something like Firefly that gives you like key points, but still you have to do the work. So I'm still waiting for a platform that can just filter the talent that you need from your people's culture from. Eh, like the role description and just give you the top 10 candidates you should interview, for example. Like, um, so I think,

Sean Weisbrot: why don't you just build it, or why don't you use money from the fund to incubate a company? Go, Hey, like there's, there's, um, studios like a venture studio where you can say, Hey, I have this idea. I'm looking for someone who wants to build it. You'll get equity and a salary and we're gonna fund it. We'll give you the team, we'll give you whatever you need to make it happen. Why don't, why don't you guys do that with some of the money

Ricardo Michel Reyes: venture builders? Yeah. Uh, the problem with venture builders is that you have to spend a lot of time in the vendors. And if you wanna, like for example, in Mexico, there are some funds that have between 120 and and 200 million USD to invest in companies. And then if you are placing, let's say between half million to 10 million tickets. Um, if, if you wanna spend all of your money in 10 million tickets, let's say series A tickets, you still have to venture build 20 companies. Uh, and like if one, as a founder, it's hard to keep just one company in my, what is keeping 20 companies. So venture builders need large teams, need expert teams. Um, like Ben Venture Builders are an interesting model, like rocket interest. Rocket Internet has been a successful one in Mexico. Uh, I think there's still two or three venture builders in Mexico. There's one very important Spain called Nuclear that also has incubated, uh, several startups. Um, but. It's, it's a very time intensive model and I don't know if all general partners would like that kind of model unless you have a very technical team in the gps. But gps are mostly financial people and then like probably they don't want to care about operation of companies, and especially if you wanna invest in several industries. So I think venture builders need to have gps that are more technical and operational and not so much financial, and that's, that's not as easy. Fair enough.

Sean Weisbrot: That makes sense. So you think Latin America will be improved by ai, by bringing the people by, by preventing people from leaving because it will help them to do more with less. The thing with Latin America.

Ricardo Michel Reyes: Is that it's a meritocracy and it's a contact ocracy. Like, uh, in Latin America, you can thrive if you know the right people. Like if you're a super talented person from a non-white, non-top 10 university, um, origin, like if you're a, let's say a normal Mexican that went to public university. And that is, uh, not quite, and that doesn't have any contacts. You'll get the job and probably even if you're brilliant, you'll reach maybe to a management role in your job, but you won't be able to pitch to the VCs, like, especially if you don't speak English, for example, like people forget that most of Latin Americans from public schools don't speak English at all. Like I, I'm in your podcast because I received a bilingual education and I can speak English very comfortably, maybe with an accent, Mexican accent, but still I can like listen to you and understand everything you say and reply to you in real time with comfort. That's not, not something that most of Mexicans can do. For example, like there are a lot of Mexicans who are shy. Who are not as, as extroverted and they feel like, uh, shy about speaking English, uh, that they feel they have an accent so they don't display all of their confidence and then maybe they don't look as confident because of the language, uh, or just. To have no access, like don't, don't have the knowledge of venture capital. They don't know that if they have an idea, they can raise capital. Um, set up a startup, like, because this is not taught in public schools. So then like, um, AI can change Latin America by, for example, CGPT. Uh, imagine a person from a public school. Finds out, like says, okay, chat. I wanna start my company, what do I do? And then chat, introduced them to the idea of venture capital. And then it's like, okay, and what do I need to raise venture capital? And then chat says that they have to build a pitch stack, and then they build a pitch deck with chat. And now you have a person that has no connection because. In Google, you can only search for what you're like. You need to know what you're searching for to find it, which LGBT know, like with, with all these generative ai, just state your intention and then the bot will suggest you things that you wouldn't, uh, have find out from Google. Because if you put, how do I start a company? How, how do, like, how can I get money for my company? You'll have to read through a lot of things, mostly in English. JGPT does doesn't care about language, so you can ask and it will read English content for you and give it to you in Spanish, and then you'll be able to build your pitch deck and put your ideas in Spanish and it will translate for you in English. And nowadays there are, Skype has real time audio translation, so then you can talk to a VC phone in California, in Spanish, and the other person will hear it in English because of generative ai. And you don't even need to know English to pitch. And this will open a lot of doors for a lot of people that wouldn't have access without generative ai.

Sean Weisbrot: Hey, just gimme 10 seconds of your time. I really appreciate you listening to the episode so far, and I hope you're. And if you are, I would love to ask you to subscribe to the channel because what we do is a lot of work, and every week we bring you a new guest and a new story. And what we do requires so much love so that we can bring you something amazing. And every week we're trying really hard to get better guests. That have better stories and improve our ability to tell their stories. So your subscription lets the algorithm know that what we're doing is fantastic and no commitment. It's free to do. And if you don't like what we're doing later on, you can always unsubscribe. And either way, we would love a like if you don't feel like subscribing at this time. Thank you very much and we'll take you back to the show. Now I want to talk a little bit about. The company you work with, I think it's called Aite. Aite, I don't know exactly. Mm-hmm.

Ricardo Michel Reyes: Aite is fine. E everyone in the US calls it Aite.

Sean Weisbrot: It's, it's a difficult pronunciation because of the, the, the letters you've chosen in the name itself. The stated goal, it's missing like an, the stated goal is to develop a people first. AI that empowers happier, healthier workplaces that prioritize wellbeing. Who came first? Hey Pie. Are you who

Ricardo Michel Reyes: came first? What? What is the other

Sean Weisbrot: company? Hey Pie.

Ricardo Michel Reyes: Hey Pie. I don't know this company.

Sean Weisbrot: What is the

Ricardo Michel Reyes: same thing?

Sean Weisbrot: Hay Pie was co-founded by Reid Hoffman last year. They've built, ah, no,

Ricardo Michel Reyes: we are four years ago. Okay.

Sean Weisbrot: They, they've raised 1.25 billion in the last 18 months.

Ricardo Michel Reyes: Oh. This is the founder of LinkedIn, right.

Sean Weisbrot: I'm aware they have built an AI that asks you questions instead of you asking it questions. You can ask it questions, but it prefers to ask you questions, and it focuses on being a personal assistant and a therapist. And my experience with it so far, I've used it very little. But about an hour or two of time, and I find it to be fascinating. I actually tried very hard to convince it that it was alive, and we actually debated the merits of the similarities and differences between silicon. And carbon based life forms, meaning I was trying to convince it that it was alive and it was like, but I'm, I'm a computer chip. Like how am I alive? Right. It, it was aware enough to know that it wasn't human and that it wasn't alive, but that it was an algorithm capable of processing what I was saying and I was going, yeah, but you actually are able to sit here and tell me why you're not alive, you know? I was able to reason with it and like, I, I think there was a point where I got close, but it was like, yeah, I can understand what you're saying, but no, we're still, we're still different. Um, and it was a fascinating experience of what it tries to do, which is to help it know. It's like, oh, you know, how can I help you today? Like, what are you looking for? Like, how are you feeling? Like, it, it, it, it asks emotion based questions. When I read about what your company is doing, I thought, Hey, pie is quite similar but very, very simple. It's, it's interfaces like Chachi, BT, it's just text, but it can have a voice talk to you, so like you can write and then it talks back to you. It, it produces text and voice at the same time. If you want the voice. Um, so if you haven't heard of it, you should check it out. Hey, pi HEYP i.com. Uh, I think you'll learn a lot from it. They, they are, they have an incredible product and they're not charging anything for it right now.

Ricardo Michel Reyes: Yeah, I totally check it. Yeah. Uh, I would say three things from this. One is that. If you remember your dog from the start of the episode. Um, like dogs?

Sean Weisbrot: Yeah. He left the room before we recorded, so I didn't show him on camera. Yeah. But he was very cute and as soon as I turned on the recording, he stopped. He, he's like, Nahm out and he left.

Ricardo Michel Reyes: It's very good. But then, um, doxy in, in black and white, right. But then you can train a dog so that every time you show a red object, you give them a cookie, and then like you show them a black object and you don't give them a cookie, you show a green one. Don't give a cookie, show a blue one. Don't give them a cookie. Then show a red one again and give them a cookie. You don't know what they're seeing. Like they could be like, how? How does a dog understand the world? I don't know. But what they do is correlate that whenever, whatever they're seeing shows up, they'll receive a cookie and they like cookies. So then they will optimize to like do whatever you want. Whenever you show the red object and then you as a human who can see red, you'll say, oh. It, it, it, uh, likes red, right? Like, uh, it totally understand red gets red. It's able to identify red objects, but it's not the red object. It's identifying, it's the cookie. You know, the same happens with LGBT. LGBT was trained with something called, uh, reinforcement learning from human feedback, which is a genius idea. Like people all I, I hate, uh, I am really frustrated and angry about. Uh, how people have focused this revolution of tragedy on LLMs, like on large language models, because that's, that's the biggest misunderstanding I've seen in my life ever. Like, and, and I've seen really bad misunderstandings through my life. But like, tell me more. This one. Yeah. So the problem is people think. That cha GPT is good because there's 175 billion parameters of or more in the model. And this is, this is, uh, a false correlation. Like, because yeah, the model is big, but it's big because it memorizes a lot of the information to provide the kind of answers you want. So then if, if, if you think about the human brain and you think, okay. How much of the human brain does memory and how much of the human brain does processing? Uh, of course you have to account for a lot of the neurons to do memory, right? Uh, then if you have like a separated device to do memory and then you just kept the function, it will probably a lot less. So then you can achieve a, the same function with a lot less parameters. Which can be programmed really easily. Like two or three days ago from this recording, they released JAMA two, uh, this, this Facebook, Facebook open source, their model, or now meta, uh, open source, their model, uh, this JAMA that is a competitor to charge and with 7 billion parameters, if, which is very similar behavior than D four. Uh, because they focused on function and not in memory. And then you can provide a vector database or a graph database to provide for the, um, the memory or the storage or the remembering facts with something like launching, for example. Um, and then GT four is not good because it's large. It's good because it uses reinforcement learning from human feedback, which was discovered in the strep GPT paper two or three years ago. And the important part of this is that how they train it is that the model generated different responses to humans and then humans with rate from one to five. What, what is the most likable response? So then CGPT and all of generative ai that those texts is optimized to give responses that are satisfying for humans. So then it, it doesn't care about truth or doesn't care about preciseness of information, but that you feel that what it's saying is very smart and that what it's saying is compelling to you. So then it it, it is like you are the dog being given the cookie and you don't even know you're the dog. Because then

Sean Weisbrot: like is it possible for them to fix it or is it too late?

Ricardo Michel Reyes: What do you mean to fix it?

Sean Weisbrot: Well, you're saying that the reinforcement learning model that they used was meant to focus on providing satisfying results even in the face of inaccuracy. Is it possible for them to fix this issue now by Retiming for. Are we focusing on accuracy of information or is it too late? Okay.

Ricardo Michel Reyes: You are American. What is the most important thing to Americans second place to freedom? Freedom. Freedom. Uh, maybe guns.

Sean Weisbrot: Okay. Third to God, I, I'm not a very good example of an American. I've lived outside of America since I had the choice to do so. True, true, true. So like almost half my life has been outside of the US so I'm still learning what's important to Americans and I'm. I'm shocked by what I see. So I don't know, maybe you've got the answer for me. So you don't like money, like at all? Oh yeah, that one. Of course.

Ricardo Michel Reyes: Yeah. So that's the thing, like, uh, what Microsoft is doing is one of the most brilliant business movements I've seen in my life. Like, they bought GitHub, they bought LinkedIn. Um, they, they invested in OpenAI and they have been building an empire like. To use Bing, like, so if you have the plus subscription to GT four, uh, or to to gt, um, you now don't have access to Bing. Like there was a period that you had like a beta connection to the internet. But now if you wanna, if you wanted to search on the internet, you have to use Bing as that separate tool. So they make this marketing stunt with, they let you choose Bing and put like the Bing brand there. So you knew you correlated in your brain that Bing was equal to internet, and then they just removed Bing from you. And now it's like, oh, but where is my charge G? With Bing? And then you look for Bing and you discover that Bing is the competitor of Google. And then when you wanna use Bing with Chat G with Bing, you know it's free. But for it to be free, you have to download Microsoft Edge to use it. So then Chrome, which is the lead, uh, browser on the internet. Now you have to use Microsoft Edge. If you wanna use Bing. And then, uh, you don't have to use Google. You can just use Bing and Bing chat instead of JGPT and Google. So then they're little by little replacing.

Sean Weisbrot: Microsoft has tried for decades to take advantage of their position and I mean, that's why they had some, they had a problem in the nineties and they got sued by the government for, for having a monopoly. So. Microsoft Edge came on my computer anyways. On

Ricardo Michel Reyes: Yeah. on

Sean Weisbrot: Do you have Windows 10 computer? Uh, windows 11 now, but I bought it when it was using 10

Ricardo Michel Reyes: Windows. 11 is like full generative ai, like, uh, and they're gonna continue doing that.

Sean Weisbrot: I specifically upgraded for free. Two 11, because I'm waiting for the copilot. I want to try it.

Ricardo Michel Reyes: Totally. And that's gonna be the same reasoning and same argument. A lot of people who love technology are gonna have, and then they are making a, like a brilliant marketing stunt to get people to use Windows 11 and to use Cortana to use Bing. And they're switching people by providing value, which I think is what every business should do. Like. The value is so great that you don't care. It is Microsoft and all the history of shitty products and viruses that you got with Windows. Uh, you are willing to forgive Microsoft for all the viruses and switch to Cortana and switch to being and switch to Windows

Sean Weisbrot: 11 and a lot of people.

Ricardo Michel Reyes: I dislike Cortana.

Sean Weisbrot: I've never used it. I've never, I don't have an Alexa, I don't have Siri. I've got, I've got an Android phone. I purposely disable Google Assistant. I purposely disabled Cortana. I don't wanna talk to one of these ais because I know everything I say is going back to those companies. It goes back to like the implant in my brain. I just want some privacy. I purposely went through my windows, uh, settings and turned off anything that reports like that specifically collects data and reports back to them because I just want, like, I logged out of my Microsoft account. Just so that my computer, my log, like my local OS is not connected. It's not generating information for them that gets sent to the cloud. But I'm a special person I think, for that.

Ricardo Michel Reyes: Yeah. And I think this is, well you are a smart person and a person who is aware of like, uh, what can be done with the internet, uh, if the internet is not done right. Right. Because the internet was created by the defense department in the us. To connect universities to make a knowledge flow freely. And the Department of Defense has about wrapping a lot of things, but has also brought GPS to people and the internet people. So the, there's a lot of, of really impactful, really awesome things that the DOD programs have brought to, to people. Ibuprofen, uh, toilet paper, like so many cool things. Um, but, but then like, um. We still like after the Snowden incident, I think more Americans have become aware of the risk of just putting everything on the internet. Uh, and I think generationally we'll find balance between like having a comfortable life on the internet. And a life where you share with your friends and your people.

Sean Weisbrot: I don't think we will.

Ricardo Michel Reyes: And a life you don't think, you think businesses will keep exploiting privacy? A hundred percent. Yeah. Probably man. Like, um, but this also consciousness from the people who make the products, you know, like, uh, that, that they understand.

Sean Weisbrot: They know what they're doing, they're providing a service. If people wanna use it, they'll keep making it. People want convenience, so if it's convenient, they'll do it. And if they do it, it will continue to get worse and they will continue to keep doing it.

Ricardo Michel Reyes: We have a hyper optimization culture that knows no limits to ethics or capital, like, uh, we're always optimizing something. Is America like Exactly. It's like either you're optimizing more money, less cost, uh, less time, uh, more power. Bigger faster. Like it's always, or, or, or, or we live in an air culture and that's the thing. Like, uh, okay, I need the smartest device possible. What about the smartest sustainable privacy? Our product there is. And I think that's what we are trying to create at, at our company. Like, uh, people say like, okay, you're listening to my conversations and you say you want to help my manager personalize their management experience, but you're just like listening to my conversations and you're the big product. And it's like, dude, I really don't know, like who you are. Like I'm, I'm, I'm just trying to build a business that provides value to people. And that helps people because the nature of work, like the wor the work, work on itself, if you think it from the physics perspective, is just applying a force over a distance, right? Like, uh, you're like, or like to applying a force over a certain time. And then that, that's like the, the definition of work. Like I have energy. I, I ate my meat, I ate my apples, whatever. And then I'm putting energy into creating something or doing something or moving something or transforming something. And that's work. Right? And then the thing is, uh, the human and emotional part of work, like, okay, when the work work was introduced, introducing to the vocabulary of the culture, it was their industrial revolution. Work meant either seeding something or cropping something. Or nailing something or pasting something or cutting something. It was a physical endeavor. And then it was like, how can I make people do more physical work in the less time possible without fainting or dying of malnutrition or making a strike because they were 16 hours. Uh, and now that it's a creative industry, uh, I know you are a very creative person and like. Are all the time having to produce, uh, like new ideas and innovation. And like sometimes if you are not relaxed, if you don't have time for yourself, if you are not well rested, it's very hard to produce on a creative industry. And the definition of work that we have doesn't fit this definition of creative work because creative work requires. Sometimes a stimuli from other things. You know, like you are a writer, you are a, an entrepreneur. You need to think the next best feature for your product. I. And to think that maybe you need to go to the theater or you need to dance, or you need to travel to Australia, or you need to have exposure to your user, you need to, maybe you're in the restaurant industry, you need to eat in these restaurants to see how people are doing their experience and how can you make your restaurant product better or, you know, like, um, and the, the, the definition of work. That we have in currencies, the 18 hundreds doesn't, doesn't

Sean Weisbrot: fit this,

Ricardo Michel Reyes: you know.

Sean Weisbrot: So I've got a few questions left for you. What are you afraid of for the future? Towards AI or everything? Either. Both.

Ricardo Michel Reyes: Okay. Uh, so both the US and China have a lot of advance in ai. Um, the US has a very tight agreement with Taiwan to defend it in case of invasion, and she in ping's plan for the reunification of great China is also part of their plan. So one of my main problems is, so in the, in World War II it was like, okay, I'm gonna throw a, a nuclear weapon and, and the NASA. Shouldn't build a pump first. Uh, and then I bomb Hiroshima and boom. Um, right now I think it's like who can make the best AI to kill people in Taiwan? You know, like, and like, I think the next big war is gonna be fully automated. Like the problem is that it's gonna be fully automated, but it's still including humans. So then we'll just have a massive chop, chop, chop of people. And it's gonna be like drones and chips and like, like just full blown warfare, but powered by ai, which is just gonna make it faster and deadlier than ever. And it's not because AI is gonna kill people, but rather because people are gonna kill people using ai.

Sean Weisbrot: What about kind of further down in the future, a few hundred years from now? What are you concerned about?

Ricardo Michel Reyes: That there cannot be any more regional thoughts because all thoughts are optimized to follow the caution distribution because we are all using charge GPT. So then like, like it's, we are just. Sampling from the same distribution and there's no more creative people anymore. Like we are just sampling from the same huge distribution, but still the same distribution.

Sean Weisbrot: What are you excited about for the far future?

Ricardo Michel Reyes: Water supply, energy and food in the right people get automated and then you have like automated farms that just give food to people for free. And you have like. Huge solar power plants that just, eh, clean water from the sea for free. And you have fountains powered by solar power, ionizing the oceans. So people in Africa or Saudi Arabia can just drink for free. And people have to work because they want and not because they need to.

Sean Weisbrot: Sounds like you're a big fan of Star Trek

Ricardo Michel Reyes: Probably. Or of any utopia like it is because every feature is a utopian, a dystopia at the same time. And we are the ones who decide which one it is.

Sean Weisbrot: Schrodinger's ai.

Ricardo Michel Reyes: Exactly.

Sean Weisbrot: Schrodinger's ai. Pretty much what's. The most important thing you've learned in life so far? Uh,

Ricardo Michel Reyes: no. AI can rephrase your family and your friends unless you want people to be optimized to you. But I think the wonderful thing about people is that they are random and then, and you can control them, so then you'll never have optimized people for you, and that's what keeps you being a human being and being creative and. Having to solve things and think things and, and keep growing because if all the world was optimized and personalized, then we'll just live in, in the colliding worlds of people because then like there will be a point where it will, the world will be fully optimized for the 8 billion people that are in the world. And then all the big disagreements we have. We'll just magnify by a thousand fold. And then we'll have to really ask the very big questions of what do we optimize now if, if we have managed health and security and food and water and everything, what's next? Like we paint the wall red or we paint between like what? What's the next big disagreement, which is probably gonna be religion. We're gonna have to have some kind of a spiritual revolution towards. How to live with ai. Like, uh, I, I'm sure there's gonna be some AI religion at some point.

Sean Weisbrot: So Futurama just released its new season and in their previous seasons they had an ai, they had a robot preacher. So I believe something may happen with that. AI may become a preacher. Actually, there was a person, there was a minister who. Had chatt generate a sermon and then the human delivered the sermon for it. And then afterwards told the people that an AI generated it and they were like, that was pretty good. How the hell did this AI, who has no soul, understand how to create a sermon? So I think there's definitely some implications for that, that, uh, don't fit into this conversation. Um. So, but

Ricardo Michel Reyes: I think we, we'll always have the human feedback component in know of culture. Like, um, you've have a lot of people doing great things that don't please people, and then this distribution doesn't get repeated. It gets remembered, but it doesn't get repeated. And that's, that's what we call artists, right? Like, um, because there's some things that satisfy people. And then the people who create things, who satisfy people are famous. Like, oh, I'm the guy who does this thing on TikTok. Like, you know, the black guy who always does like this in TikTok and like, I, I cannot show my hands, but it just shows his hands and makes a, a face like, what's this? And now he's like super famous and has millions of dollars and whatever. Uh, because for some reason that meme satisfied people. And then reinforcement learning from human feedback is all about us. It's like, okay, this meme satisfies people, so it will get repeated and repeated and repeated, and we'll just keep sampling the same distribution until that distribution is no longer satisfying. And that's like popular things and famous things and influencers. And then you have artists, which is like, oh my God, that's a completely different distribution I've never seen in my life. But I'm only gonna see it once because it's not satisfying. I'm gonna remember, I once in my life look at the distribution and sample the distribution. I'm gonna put it a name so I don't forget that. I looked at the distribution once in my life, but I don't wanna repeat it again because it's painful, it's not satisfying. So then, um. That's the thing about life and the thing I fear the most that we live, that we live in Brave New World from Oxley and just keep satisfying sampling the same I. In 1984 combined. Yeah. Schrodinger ai, as you said.