Ep. 579 The Process of Developing Tomorrow’s Web3 Applications Today
Ep. 579 The Process of Developing Tomorrow’s Web3 Applications TodayPowered by RedCircle
In this episode of Crypto 101 we have Naveen Agnihotri who is the CEO and Co-Founder of Cumberland Labs and TrustApp. Naveen has had first hand experience building future applications as well as seeing the progress of different projects inside incubators and really has a pulse on the emerging technologies like A.I coming to web3 that are primed to hit the market and bring value during this next bull run.
— TRANSCRIPT —
SPEAKERS
Bryce Paul, Brendan Veihman
Bryce Paul 00:09
All right, ladies and gentlemen, welcome back. We got another high powered high caliber episode here the crypto 101 Podcast. I am your co host Bryce Paul joined as always by my trusty compadres Mr. Brendan Beeman. Brendon. How’re you doing today, man?
Brendan Veihman 00:24
You know, it’s hard to be doing bad when the crypto markets are doing so good. A man I am extremely excited. And man, do we have a really excited exciting podcast in store for everyone? So tune in, turn your phones to silence mode because you’re not gonna want to miss this one.
Bryce Paul 00:41
Yeah, we want all your attention every last drop of it. We know attention is a very scarce resource these days. But please give your undivided attention ladies and gentlemen to the front of the stage. Naveen Agnihotri is joining us today who is the CEO of Kava labs. He’s building a ton of cool things, notably trust app, which we want to talk about. But most importantly, we just want to get to know who Naveen is and why he’s here in the space. So Naveen, welcome to crypto one on one podcast. Thank you for joining us.
01:10
Thank you guys. So happy to be here.
Bryce Paul 01:12
Yeah, no, we love it. We love your energy already from from chat and a little bit before the cameras rolling. I know this is gonna be a great one. You know, you’re very passionate about what you do. You’re passionate about why you’re here. So we want to learn a little bit about your background, even kind of like before crypto, I mean, you know, you were probably around right, you know, had your career unfolding in some direction before crypto pulled you off into another direction. So we want to hear about that kind of that story, if you will. Absolutely.
01:41
So I am an engineer by training, I am probably older than a lot of your viewers. I got started in the early 90s. doing engineering. And then I got sucked into AI, which actually is happening to neuroscience. So I did my PhD in computational neuroscience from Columbia. I was a scientist at MIT for a few years again, to studying how does the brain do things that the brain can find so easy doing things that AI and computers find so difficult. So I spent some time building interesting neural networks. And I left it in 2005, to create a startup that was going to apply AI and neural networks to real world problems. And that became a series of startups, all of whom were aimed at finding interesting uses of AI and broadly interesting uses of technology over the years and name a type of AI and I have probably done it. I’ve tried it like and tell you what works and what doesn’t.
Bryce Paul 02:49
Wow. Real quick, just on the subject of AI. There was just really hot in the news right now, obviously, and I’m in open AI is building this whole chat up T model. And they they recently like fired their their CEO, Sam Altman like the board apparent and then they rehired him or something that was crazy. It was in the news. We don’t need to go into it. But I want to kind of get your your take on? Did they discover something maybe dangerous? And so they decided, Hey, we got to pull the plug on it. We got to fire this guy who’s who’s leading this dangerous researcher? What do you kind of think happened there? If you could entertain us?
03:27
Look, I’m not privy to any information that you guys are not privy to i All I know is whatever you have probably also read. But I can give you a little bit of color that I think you will not find in a lot of places. And that’s what does it really take to build AI? Right? These types of LM models, they’re fairly new, they got started when Google published the original transformer paper in 2017. And since then, the future of field of AI has really exploded in complexity. And these large complex models, it really takes a a huge effort to train one of these models with hundreds of billions of parameters, right. And the thing is that before, before opening, I really nobody was doing it. There was no effort in 2018 2019 to really train very, very large. LLM models, mainly because you couldn’t justify the cost. It’s hard to estimate but standard industry estimates are somewhere between 100 million and a billion dollars combined in terms of the amount of computer takes and the amount of effort human effort that it takes to actually train one model and CPD for right And the thing is, if you’re going to spend a billion dollars, can you and the fact about AI models that a lot of non AI people don’t realize is that the majority of them really don’t work. They don’t, the technical word is they don’t converge, meaning you take a data, you take a very, very complicated machine, and it doesn’t go anywhere. It’s so you have to imagine a sort of a very complex AI model as, like the cockpit of an airplane. Right? Assuming that you the two of you guys aren’t pilots, if I put you in the cockpit of an airplane, what are the odds that the that with all the gadgetry and all the dials and all the complexity that in an airplane cockpit? What are the odds that the thing that actually takes off, I would say, very, very low, right? Even when very qualified AI professionals are at the helm of an AI model, the majority of AI models that are tried actually don’t converge. So now, there’s this thing about this, somebody is taking a billion dollar bet, on a model that we don’t know is gonna go anywhere. And even the best case scenario, even imagine that it converges, imagine that we can train this model, imagine that. It’s amazing. How much money will you make with it? Will you really make whatever four or $5 billion on it to make it worth our while? Do you have a business model that supports that? All of those are unanswered questions. So that really speaks to by somebody like Google or Microsoft, or, like companies like that don’t like we’re doing this before open AI was because it’s not a business case, right? This is not something that Goldman Sachs would do. This is not like, it’s not a cut and dried, sort of cut and dried, where you just say, Okay, we’re gonna spend a billion dollars, we’re gonna make 5 billion that we don’t know, we’re gonna. Yeah, you don’t know if it’s going to train or not, you don’t know if it’s going to make your money or not, you don’t know how much money you’re gonna make, you know how good it’s gonna be. There’s so many unknown questions. So it really takes like a very startup II type of die, to really lead an effort to make it happen, I started to beat up a company. And that’s really what opening I was used to head by combinator before he went to open yet he looked classic startup guy, you need people like that you need a team like that to really lead effort like that. And now, so that was a very long winded way of saying that this is this is a special company in the world of AI. And opening I was especially accompany especial, from several points of view, including the funky way in which he got started as a nonprofit, and then it moved to being a prop for profit. And then, you know, so there’s, there’s that sort of strange things happening in the governance of it at that level. And then there’s the bets that they’re taking, they’re taking these huge bets, with very unclear commercial outcomes. So I’m not sure what the dynamic is. On the board, I’m not sure what’s going on. And that sounds like a outsize sort of figure in that whole ecosystem. And I don’t know what the dynamic is, and what the board was thinking. Sorry, I’m not giving you the answer. I’m just gonna give you some insight into something interesting that I don’t think is commonly understood. No,
Bryce Paul 08:28
that’s exactly what we want. It’s just such an interesting kind of level of insight. And yeah, you know, your background in AI, is so fascinating. And I’m curious if it kind of informs any of the investing that you do at Cumberland labs, and sort of the incubation and any of the thought there are you looking for, you know, crypto, adjacent and AI adjacent projects to invest in an incubator, or what do you look for?
08:56
Absolutely, absolutely. The answer to your last question is yes, we always look for interesting projects, or personally, I am actually quite interested in the way that technology broadly, is applied to solve real world problems. I think I mentioned this earlier, when I was talking about my own companies, all my own companies were all ones where AI was used or blockchain was used or technology was used to really solve a problem that people have in the real world. And that I think, is where things become interesting for, for example, Blockchain. There’s a lot of effort right now, in blockchain to say, what are the use cases? How will we use blockchain to solve problems that people have today? And because of so blockchain has an interesting conundrum. And i i From if you take the narrow viewpoint that blockchain should solve people’s problems then the fact that there are these tokens that are created either fungible or non fungible. The fact that there are these tokens almost looks like a like an interesting happenstance. Because that’s not what sort of solves most of your problems. Because if you think about it, the tokens fungible, non fungible, whatever they are. The majority of people actually the majority, there’s 8 billion people in the world, 7 billion of them actually can’t afford the tokens at a fungible or non fungible right. So how can we solve the problem of a lot that a lot of people have, using this amazing technology? Blockchain, I believe, represents probably some of the best technology of the last 30 years, right. And I say that as an AI, it is an amazing technology, it represents an amazing leap forward in what it can do. I’m happy to talk about that, again, in the context of trust app. But this is an amazing technology, how will we use it to solve real world problems that real world people. That is something that I’m passionate about? That is something that Cumberland Labs is passionate about? And if there are people out there who are working on problems, who have solutions, who are trying to figure these things out, we would love to work with them, we would love to support them in any way that we
Brendan Veihman 11:16
can. Was there like an aha moment for you, where you maybe you saw something, whether it was like a deep fake or AI being used in a negative way? That was like, I need to be the one to combat these issues?
11:28
Definitely, definitely. So we’ll talk about trust. After that, can we get I’m having to come back to camera labs if you want. But talking about trust. Yeah, so talking about trust that the problem is actually misinformation, misinformation or fake news, right? This has always been a problem from the very earliest days, right? 30 years ago, when I first got started with the internet in 19, maybe in the early 90s, it was always an issue, you just didn’t know what was what you didn’t know what to believe. And, as humans, we bring sort of a trust that we have in institutions in the printed word, you know, in the in the 70s, or 80s, you just believe whatever you read, because it was largely produced by trusted parties, by large entities that you believe, you know, it was it was my Dan Rather are by, you know, New York Times, or whatever it is, these big institutions you believe them. And when people came to the internet, they, they they drop their sort of belief system with them. So the instinct that most people have is this believe whatever it is that they read, but very clear from the early days of the Internet, that you just, that was just wasn’t the case. And I think social media really brought that sort of this fact out that you really shouldn’t believe everything that you read, although most people still just, you know, they’ll pass on and they’ll post, whatever it is that they find, without making any effort to see if it’s true or not, if it’s actually even remotely possible, that that is true that, you know, whatever the earth is flat, or whatever it is that people want to believe in. So for me, the ha moment came sort of early this year when it became clear that AI was really part of the problem, right? So this fun fact for you guys. So as a, as a scientist, I care deeply about facts. And I think very deeply that facts are established in a scientific way. So the scientific way is to set up a hypothesis, and then figure out if it’s true or not, that sort of the standards of scientific way. So you can take any theory that you have, and you could, it could be something as simple or something complex. And then you break it down and you say how it will be addressed scientifically. That’s something that I care deeply about. And that’s something that I think the world does not do enough for. Too much, too, too much of the information out there, it’s just sort of is what I would describe as sort of pablum or maybe just people saying very bland things that don’t make much sense with any amount of rigor. So a given that I care deeply about science. And I’ve worked on AI for 30 years, here comes AI and now people are applying it to make misinformation worse. Right. Now people are using AI people are using my thing I consider AI to be my thing. People are using the thing that I put in the last 30 years to make the thing worse that I love. So I’m like, Okay, this is this is not right. This is not I can’t just stand by and let this happen. So I really began thinking about what Technology, what are the best technologies that we have today that we can bring to bear to really address misinformation or fake news? How can we help people stop posting things stop saying things that just aren’t true? Maybe we can stop them, but at least we can discourage them, at least we can give them a little indicator, look, this thing that you’re about to say, is not true, you know, not true, according to experts on both sides of whatever spectrum that you believe in, and not true according to common sense things that just make sense. So that’s really the genesis of trust. And that’s really a very sad, let’s go after misinformation and fake news. Let’s really make a concerted effort to see what is the best technology out there, and how can we apply it?
Bryce Paul 15:44
That’s fascinating. And it just, you know, like you said, I mean, with AI, it’s, it’s a double edged sword. I mean, with great power comes great responsibility. And some people are using this technology for great ends, and some are using it for bad ends. And so trust app is really trying to tip the weight or tip the scales in the favor of the good guys. And it’s utilizing blockchain technology to do so. And so we kind of want to unpack that, you know, at you know, you’re a scientist 30 years experience. I am a Amir plebe, and my listeners are probably not as well versed in AI, as, as you are. So try and dumb it down for us, us mere mortals. We want to understand how this works and and how you’re solving such a significant issue.
16:35
Absolutely. Happy to do that. So the two main headers in Vista technology that we work on for trust app resides our AI and blockchain. Okay. And I’m happy to go into each one of these individually. On the AI front first, it’s very clear that what is true and what is not true, can really only be established by the best AIS of what is out there right now. So for example, you say, Guys, New York Times is reporting that Israel did bomb that hospital, and there was a hospital bombing that happened a month ago that I’m referring to which some of you your readers are familiar with? Well, very controversial, the way it was very controversial. Right. And so some someone posts on Twitter, guys, the New York Times reporting that Israel did bother. Now, that’s a that’s a fact. Or that’s an assertion. That is true or not true? Yes. As undebatable, right, because because either the New York Times is saying that or they’re not doing. So the the, the the approach that we’re taking is using AI, can we accurately determine what are the assertions and what is being said? That you are looking at in your feed? And can we look at the sources of news out there? Can we look at the sources of information out there to actually extract assertions from that? And can we match? Right? And can we match them from all different sources? So there are 18 sources that are saying that what you’re saying is true. Okay. We think that it’s more likely than not that what you’re saying is true. 16 of the 18 sources are saying that we are disagreeing with what you’re claiming in your tweet. So we don’t think it’s true. So what we want to do is if as a consumer as a reader, we, as you’re browsing your feed, we want to give you little indicators that just say, green tech or red question mark, a little green tick that says, Yes, we think that this is likely to be accurate, given that the sentiment here is an agreement with the vast majority of what is being said they’re set out there and or little red checkmark, right in the in your feed that says, okay, and I don’t think so, you know, they just go check your sources like, and by the way, the little checkmarks that I’m talking about, if you hover over our checkmarks, we open up a little window that shows you what our analysis is. So we give you a link to the to the original source, we tell you this is why we think this is questionable, or this is why we think it’s accurate. So
Bryce Paul 19:30
this is you could just plug this into your browser, right? I mean, that’s exactly.
19:34
Absolutely. So the first release of trust app is a browser plugin, sort of like meta mask or the that other browser plugins that enables users to just quickly assess the likelihood that their Facebook although Facebook posts what they’re seeing, or the Twitter posts that they’re seeing is actually true or not, or likely to be true.
Brendan Veihman 19:55
Wow. You know, I’m curious to hear some of your experiences. As with everything that’s happening, there’s like AI and deep fake space, because one of the popular new scams that I’ve seen pop up is this one, especially for people who have a public presence on the Internet or on social media, it kind of listens to their voice, and then recreates it. And it’ll like call their loved ones or people of interest, and call them an act like this person. So for example, you know, me and Bryce, were all over the internet, we now have all these podcasts and YouTube videos, and etc. So what they do is that they’d send an AI model to listen and analyze our voice. And then it would call maybe our parents saying, hey, we need you to send us money, or, Hey, I need my social security number, I lost it, can you give it to me? And then all of a sudden, they go, Oh, this sounds just like my son. Let me go ahead and just give them that, you know, come to find out it’s actually an AI or a deep fake that is recreating our voices. So I’m curious from your end, like, what are the popular new? I guess, scams or ways that people are at risk here that they might not realize yet? And then how can they protect themselves from stuff like that? So
21:08
I can answer the second question, I think, much easier. Which is the, at the end? The quick answer is stop believing things that you read, or that you’ve messages that you receive, stop believing most things on the internet is the honest answer. Because really, what has happened is that AI and other types of technologies have really caught up to where technology should be, right. This goes into the point I was making earlier, we as humans, we can talk about this from a neuroscience perspective, from an evolutionary perspective. As a brain researcher, I can tell you why this makes sense. But we tend to be trusting a trusting species, we believe most things because from a long, long term survival from a longevity perspective, and it makes sense that we should believe most things that we encounter, we should believe most human, most of what most humans say, unfortunately, that’s just not a world that we live in. Given a human intentions, given human plus AI, together the combined potential for harm, it is really important for everyone, including our parents, and everyone around us to just first of all, take everything that you receive. And the default needs to be your natural, they’re not buying it, you know, even if it’s you, unless it’s you in person, and they have good reason to believe that in person, they should really shouldn’t be believing anything that they didn’t see. Let’s just start with that. Let’s just start with when. And by the way, this is also the default scientific view. Yeah. So science begins with, with the sort of a baseline of Yeah, we don’t believe anything that is being said, Okay, now, what evidence will it take to make us believe? Yes, that’s really, the world needs to be more scientific, the world needs to behave more like our science scientists. A true scientists behave, which is when you know, when you hear Newton’s third law, the first thing that everybody says is, Yeah, nah, nah. You know, okay, what evidence is there for this? How will we collect evidence for this? And that really, going back to your point about deep fakes and all of these things about AI? The question is, what systems can be put in place? For as a consumer to say, yeah, that really is price? Or no, not really price, you know? Can we set up set up a challenge, right says, can you just send me your social security number? Okay, Bryce? What’s my middle name? Or Okay, Bryce? What’s something that only you or Bryce would know, you know, like something? A little thing? A little? The the, it’s a personal version of a two factor authentication, you know? Yeah, exactly. Let’s say for something that only Bryce would know. Only you and Rachel. So that’s really, again, it’s a question of, how will we prove something’s true? Or how will we prove that the views being expressed are the views of this person? Absent them? Just the person thing? Yeah, I’m saying so, you know, Neil deGrasse Tyson has this big thing about how eyewitness testimony is the least reliable thing in all of science, right? Somebody said, I saw an alien the whole size again, you know, meanwhile, if in in the court of law, if somebody says, Oh, yeah, I saw him kill that person. And we’re like, Yeah, okay. We believe you and not buying it. You know, I think that we’re telling the world where Sooner or later, courts are going to come to that point where we’re going to stop believing in eyewitness testimony, because it really is what one person sees or says believes that their song is not reliable. So let’s build systems that actually are able to test assumptions and test this hypothesis. And I thought this is this is right. How will we believe this?
Bryce Paul 25:22
I think AI will will absolutely in some form or fashion change, or, you know, yet change law, right? Because like you said, I mean, you will no longer you can’t, you can’t, you’ll be able to prove that, you know, you can’t trust a piece of media about you. Because you’ll be able to prove that hey, like, look, I could create that same thing with AI or something very similar. So, yeah, I mean, it’s crazy to think about what what deep fakes are doing. And, and I see it now, you know, Brendon actually asked, like, what are some of the other scams that you know, viewers could watch out for? I’ve seen these things all over the place, that fake Elon Musk, fake Bill Gates, you know, their words in their mouth are saying, hey, send money to this bitcoin wallet, and we’ll send you back double, it’s a new promotion we’re doing, you know, we’re giving away new all this stuff, right? People on the internet, they see it, they think, Well, you know, how can this be fake? Right? It’s Elon Musk talking to insane send this, send bitcoin to this address, I’ll get double back or whatever. And it’s just crazy how scammy and how good the scams are getting these days. So hopefully, you know, trust app is something that we could recommend to our listeners. Because hopefully trust app does something to maybe mitigate against those fake giveaways or those, you know, spammy sort of YouTube links, you might get to, you know, would trust app kind of flag that and say like, this is not the real Elon Musk? Or is there a way to kind of, you know, have trust app alert people to where there might be suspicious activity with crypto related media?
26:55
Absolutely. I think crypto is a great use case for trust app, as well, because crypto as you are well aware. The the the the advantage and sort of double edged sword of crypto is that any transfers that are done are immediately final. So I like the like, sort of the US banking system where if you do a transfer, and you find that the person was was fraudulent, you have some time to take it back. In the world of crypto you ain’t taking, you know, it’s gone. Right? So crypto is is a much better candidate for fraudsters from for that reason, because I can really get good get away with a lot more like a bank, where people have more recourse to getting things back, people have more recourse to what they can do with the law and so on. And whether whatever the crypto, well, it’s gone. It’s a good thing and a bad thing. It’s a good thing that I have full custody, I take full responsibility. So when I give it my money to Brendan, when I give it when I give it to whoever Sam Altman, it’s gone now. I have no recourse of getting it back if I made a mistake. Oops. So for that reason, crypto I think is a great first use case and crypto people should be doubly aware of everything that can go wrong. And crypto should have better use cases of figuring out cancer. So I can easily imagine something like trust app or a sister app, the trust app being built specifically to solve crypto problems. So trust that is we can talk about the roadmap for trust app and where it goes from here and how we are imagining that it goes. But the use cases that you spoke about, which is sort of one to one interactions and figuring out that really is Bryce and figuring out this problem. I think a it’s a great problem to solve. I trust that doesn’t solve it right now. But I think it’s a great problem for someone to solve. And I think the world of blockchain can actually be used in an interesting way to solve the problem. Because blockchain can be used to recognize what is true and to mark what is true in interesting ways that you really cannot use any other conventional technology.
Brendan Veihman 29:29
I mean, it’s going back to Bryce’s point here. It’s funny because I will have people that reach out to me and say, Hey, I’m Bryce. Paul, have you ever listened to the crypto 101 podcast? And I’m like, No way. I’ve never heard of something like that. Like, tell me more.
Bryce Paul 29:45
It makes my stomach turn. Yeah, I
Brendan Veihman 29:48
mean, all jokes aside, I think that we are stepping into this direction, kind of like what you’re working towards over at trust that Naveen where it’s not exactly what you do now, but I can see someone kind of start to create that whether it’s you or whether it’s someone else, we’re starting to see some really useful applications to combat everything that we’re seeing in the AI space. But I think the big thing that I want to kind of hit home on here is that for every 10, things that happen that are negative, we’ve talked about the negative a little bit in this podcast, there are 100, things that are positive. And so you know, I think what we can’t do is overlook all the positive things that are happening, because there are so many great things happening in the world of AI and everything. And we can’t focus too much on the negative. But there’s a lot of stuff. I’m excited.
30:36
Absolutely. And let me take that as a segue to talk about why blockchain is important to trust. And, okay, this is talking about the positive impact that blockchain has. So this gets a little bit philosophical. So just give me a little bit of rope. I’ll try to be sort of short about that. So in the form of web one, the foundation was a set of protocols that were built largely by academics and engineers, who had no expectation of profitability. So for example, TCP IP was written by people at MIT who were funded by DARPA acdp was written by Tim Berners. Lee at CERN. Again, they were just researchers. SMTP was written by, it’s not even clear who but again, so that’s just a protocol. Now on top of us MTP sets, Gmail, or Hotmail, or Yahoo mail or Outlook, right, so the application was separate from the protocol. And that’s a very important distinction, that you use the same protocol, which is every all the internet is the same protocol, it’s SMTP. But you can pick any provider you want, which is fine. Now, you move into web two, but 10 years later, and you find that actually, that distinction is gone. The platform is the protocol. So Twitter is a platform, but Twitter is also a protocol. So if you have been sending out tweets, and you’ve sent out 10,000 tweets over the past eight years, all those tweets exist only on that capital. If the government of Iran doesn’t like what you what you say, because they find it to be anti Iran or whatever. And they reach out to Twitter. And the management team of Twitter agrees that, yeah, this is not right, they can just remove whatever it is that you’ve ever said, right? The, let’s say Elon leaves tomorrow, a new CEO comes in town, they implement a new policy, all of a sudden, a bunch of Twitter accounts are just gone. That content has been created. It’s just gone. It can be de platformed at any point. Right? So and same thing is true of Facebook, right? The same thing is true of Instagram. It’s true of any of these things. It’s a very centralized operation. A few people are directly in control of every single thing that you say. Everything government you
Bryce Paul 33:16
do Vinit, the government’s proven in some of these investigations right now. Absolutely. Right.
33:21
So the we have moved from a world where the platforms and protocols were independent. And you could just pick any platform that you want the protocol remain constant, to now where the platform is the protocol. And it’s quite bad. It’s quite bad from the point of view of information provenance. It’s quite, it’s quite bad from the point of view of who will believe anything that you say, because it can just go away? Right? Did did Brendan have rights really say this? Well, Twitter’s have decided to delete your account. Now it’s gone. So in my mind, map three, Among its many, many other features. One of them is it’s a public good, and remains, it’s certified by all of us, and it remains forever. Nobody, if we put it out there that dry said this. Nobody can ever say he didn’t. There’s no entity, there’s no company. There’s no individual, there’s no committee that can come in and remove what dry said if we put it on chain. Right. That, to me is a very important point. And a very important facet of vektory of what blockchain technology allows us to do. It allows us to really get in front of this idea of platform, a blank on protocol, sort of juncture and pull it apart and say no, ain’t doing it. The protocol is here. For the protocol. This was set, which cloud Comment percent on is irrelevant. It doesn’t matter. You said it on Twitter, you said it on Facebook, you said it on your personal website, it doesn’t matter, you put it on a phone record, you put it in a telegram chat does not matter. We have recognized that this was said, and nobody can ever deny, you know,
Bryce Paul 35:17
kind of like, kind of like a Bitcoin transaction, right? Like if it’s got all those six confirmations, you’ve broadcast that transaction, everybody attests to that truth. If you run your own node, I mean, you as well, you have a copy of that ledger of that canon of that consensus of that way, that singular version of truth that everybody else is agreeing on. That’s reality, right? Whereas like, in, in the banking system, or whatever, you could send transactions back and forth, and they could get refuted, they could get censored. They’re centralized actors in between, you know, they they might, you know, lose the wire or whatever, you know, in kind of blockchain. I love that, you know, there’s a consensus mechanism, it’s kind of reminds me of what you’re talking about, as well.
36:04
Exactly, exactly the point, which is, let’s use blockchain. So in the world of trust, that we use blockchain as a way, as a technology to establish consensus that something occurred, we are using all of us as a way to say, this is not this is content that cannot be on the platform. I think that’s the word I want, because it cannot ever be deep. Yes. You can never remove it, you can ever cancel. And it’s been said, so So in the world of trust app, we are doing two things. First, we are taking pieces of content. And we’re actually putting them on cheap. This is true for everything that we’re capturing right now. But we want to keep doing it into the future. And it becomes ever more important as the facts that we are collecting interest become sparser and more and more difficult to mobile difficult to represent in centralized places over the apple, if a telegram chat, or if it’s something that the citizen journalist has pointed out, we will, we will create a system where it can be minted, and it can be put in stone and say, Okay, this was this was recognized, this is true, this person was there, they saw it, they posted a video, whatever it is, and we were established by whatever means we have that that actually happened. That wasn’t just AI saying this, it was like nine independent people who took in different videos. And we have ways of establishing that they were in the right place at that time. They also whatever are we it was empty consensus is we can establish consensus about something that happened. And we are putting it on chain. And now it’s done. It’s kind of like
Bryce Paul 37:56
it’s kind of like a blockchain version of Twitter’s community notes. Like they just rolled out this new feature called Community notes where, you know, posts, you know, could have other people kind of check them and basically say, like, this is actually not confirmed, this is confirmed, and all that kind of stuff, which I think is pretty interesting. And so yeah, trust app takes that next level further, but I kind of wanted to ask one other question, kind of, on the subject of AI. But you know, kind of like how we were talking about running your own nodes, always having a copy of the ledger always been able to prove, you know, some people I hear are starting to run their own, like, AI models, like their own private like, Well, I’m not going to rely on Google barred, I’m not going to rely on chat GPT in a GPT for whatever, I’m going to run my own. What is that? Like? Is it how does somebody even start doing that? And do you think that that’s like, you know, maybe they’re a little paranoid? Or do you think it’s a legitimate cause?
38:55
Not only do I think is illegitimate, cause I think it’s actually a requirement. So So, here’s here’s, so here’s, here’s the way, again, this is something coming from an AI guy. This is something that isn’t talked about that much. Okay. So let’s, let’s take, let’s take two steps back from this actual question. And let’s look at these models. So chassis, PT board, all of these models, and you take the entire set of internet data, right? What is the data that dharma has been trained on? It’s everything that’s out there, it does everything without them. And you train a model that can do a number of things. Okay. So it’s a very general listing. Right? It’s a very general listing. You have to think of any of these AI models like activity or bar as like a general human being, right. It’s a general human being that’s good at most things, but it’s not particularly excellent at any given thing. Right. So for example, To any general human being, will likely not be very good at playing table tennis or at running. Why? Because to be good at running good run a lot, right? You need to actually be a train runner. Right? You will be running many hours a day for a long period of time. That’s a bit of good. Right? So going back to AI, if you want an AI, that’s good at x, whatever x is, let’s say x is playing chess, right, or x is what will pick it up, whatever it is right? You will actually train an AI model that is specifically trained for that specific use case. Okay, so and the way you do that is you take one of these more general purpose models like GPT for or actually, it has to be one of these open source models like Allameh, to or lot of these, like a black girl named the new 7 billion model that that just came out, there’s a lot of these open more open source models that are being released. So a younger Cohen is releasing a lot of these models at meta, entirely open source, you release the model, open source, now it’s available, right, so now all 7 billion parameters are available. And now you can use your custom data, to fine tune it, you don’t need to fine tune all 7 billion parameters, you can find you maybe 1%, maybe 7 million or 70 million, depending on how much data you have custom data you have. And that’s the equivalent of taking a human and letting them run a lot. So that they good, they become good at running. So for example, in our use case, if we want a model that is actually good at, let’s say, picking out if someone’s really talking about X, whatever x is, let’s say someone’s really good at talking about or someone’s implying that this content is from CNN, some specific use case, no existing AI chatbot model is actually going to be excellent, they will all be okay, well, maybe they’ll all be 80% accurate, or 70%. Accurate. And that’s enough to impress most people. But that’s actually not enough for any specific use case, right? That’s like being able to run a five hour marathon is probably impressive to a lot of people. But that’s not gonna be impressive, you really want to be good at running. Right? So now you need to take your data, you take all the open source model, and you start fine tuning it, that is neither easy nor simple. Because if you go back to the point I was making earlier, that’s really like taking a sitting in a cockpit and saying, This plane will take off now. You need to be really good at this. And you need to really be able to take that, you know, the 7 billion parameters, and figuring out how many of them you need to fine tune, how will you do it? What’s your strategy? How will any of this work? So you have to fine tune the model. And now you have a model that’s good at what you want to do. Right? At this point. Now, that’s what I’m describing is the basic requirement if you want to be sort of a kick ass AI company today, or have a kick ass AI model that is specific to a use case that you’re solving. Right? So you take one of these open source models, you fine tune it. Now you’re good to start. Well,
Brendan Veihman 43:33
Naveen, we really appreciate you coming on here and enlightening the masses. Truly, I think that this will be something and we’re going to start seeing verification sources like trust app, being integrated into all sorts of daily websites, wallets, crypto, wallets, exchanges, news outlets. And it’s sad to say that, you know, it’s sad to kind of admit that we can see a world where we need to have these third party sources verifying everything that we do, from websites to wallets to exchanges to news to digital self defense, right? I mean, that’s kind of
44:09
it’s absolutely so I’ll just in a couple of minutes I’ll describe to you a vision of where trust app goes a vision of where trust assets are companion which we are now imagining here in this podcast that doesn’t exist right now. Right so album for people here who want to start their next company. So we are imagining trust app to be sort of like Grammarly. Right so you guys are familiar with Grammarly? It’s a company that puts a little squiggles below your text and says, fix your grammar here. We can write not to spell but also grammar. And we are imagining a grammar impacts. You’re about to do something. You’re about to put some content out there. You’re about to write a blog post you’re about to publish something and you’re saying a number of things in there. And we can just go in and check and say yeah, true, true or not true. So everything squiggle below it saying, you know, the correct thing is this. In a Singapore did not become a dependent 1968. It actually became independent and 67, or whatever it is, whatever your factoid is, you know, Clinton was the 41st. President, okay, for the second, most accurately, not controversial, but people this okay, there’s putting whatever content out there, as you can imagine, and then you can become specific at this thing. The New York Times is it not reporting that Israel bombed that hospital? So you put a little squiggle balloon? Now, you can easily imagine a version of this for crypto, you can easily imagine a version of this that actually plugs into a wallet. And you’re about to send, you received a request from Bryce saying, dude, crisis, send me point five E or whatever it is, right? How about we make an app put a little squiggle below it. It says actually, that’s not Rice’s address. I did not address that we know that he uses, you know, the assembly department that is it. I even like if that’s Rice’s address. Where have you seen the last few times we send money to Bryce, who has these addresses? What about this? Are you sure you, we will stop you, but we’ll put it in the schedule below and put it in the morning that
Bryce Paul 46:27
just will be where, you know, it’d be it’d be great to have something like that. Something that just to have our back, a little digital shield in the world of, you know, everybody’s after everybody, everybody’s coming after your money. Everybody’s coming after you for one way or another. Which, you know, that’s just the way the world works. But unfortunately, Naveen we are are coming up on time here. We couldn’t thank you enough for spending the afternoon here with us, and for giving us so much incredible knowledge on the intersection of AI and blockchain, talking to us a lot about Cumberland labs, the work in the investing that you guys are doing there, along with the incredible system that you build a trust app. So with that being said, Where can people kind of follow you stay in touch, maybe even download the app when it becomes ready to rock and roll. So
47:17
the app, you can just search for this trust app. To PS it’s coming online in the next couple of weeks. It’s gonna be available in the Google Play Store as a Chrome extension. You can follow me on Twitter. I’m not terribly active, but that is where I’m begin starting to become more active now, on topics around AI and around blockchain. It’s nav AGI. So the first three letters and the last one and
Bryce Paul 47:46
wonderful and AGI artificial generalized intelligence, also kind of a nice little play there. Absolutely. Naveen, thank you so much. We hope to have you back again soon to talk about some more exciting things.
48:01
Happy to happen, guys. Thank you so much.
Please Support Our Sponsors
Check out Factor 75 and use my code crypto10150 for a great deal:
NetSuite:
Get your FREE copy of “Crypto Revolution” and start making big profits from buying, selling, and trading cryptocurrency today:
https://www.cryptorevolution.com/free
Subscribe to YouTube for Exclusive Content:
https://www.youtube.com/@crypto101podcast
Follow us on social media for leading-edge crypto updates and trade alerts:
https://twitter.com/Crypto101Pod
https://instagram.com/crypto_101
Guest Links:
https://www.linkedin.com/in/navagni/?originalSubdomain=sg
*This is NOT financial, tax, or legal advice*
Boardwalk Flock LLC. All Rights Reserved 2023.
▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬
Fog by DIZARO https://soundcloud.com/dizarofr
Creative Commons — Attribution-NoDerivs 3.0 Unported — CC BY-ND 3.0
Free Download / Stream: http://bit.ly/Fog-DIZARO
Music promoted by Audio Library https://youtu.be/lAfbjt_rmE8
▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬
In this episode of Crypto 101 we have Naveen Agnihotri who is the CEO and Co-Founder of Cumberland Labs and TrustApp. Naveen has had first hand experience building future applications as well as seeing the progress of different projects inside incubators and really has a pulse on the emerging technologies like A.I coming to web3 that are primed to hit the market and bring value during this next bull run.
— TRANSCRIPT —
SPEAKERS
Bryce Paul, Brendan Veihman
Bryce Paul 00:09
All right, ladies and gentlemen, welcome back. We got another high powered high caliber episode here the crypto 101 Podcast. I am your co host Bryce Paul joined as always by my trusty compadres Mr. Brendan Beeman. Brendon. How’re you doing today, man?
Brendan Veihman 00:24
You know, it’s hard to be doing bad when the crypto markets are doing so good. A man I am extremely excited. And man, do we have a really excited exciting podcast in store for everyone? So tune in, turn your phones to silence mode because you’re not gonna want to miss this one.
Bryce Paul 00:41
Yeah, we want all your attention every last drop of it. We know attention is a very scarce resource these days. But please give your undivided attention ladies and gentlemen to the front of the stage. Naveen Agnihotri is joining us today who is the CEO of Kava labs. He’s building a ton of cool things, notably trust app, which we want to talk about. But most importantly, we just want to get to know who Naveen is and why he’s here in the space. So Naveen, welcome to crypto one on one podcast. Thank you for joining us.
01:10
Thank you guys. So happy to be here.
Bryce Paul 01:12
Yeah, no, we love it. We love your energy already from from chat and a little bit before the cameras rolling. I know this is gonna be a great one. You know, you’re very passionate about what you do. You’re passionate about why you’re here. So we want to learn a little bit about your background, even kind of like before crypto, I mean, you know, you were probably around right, you know, had your career unfolding in some direction before crypto pulled you off into another direction. So we want to hear about that kind of that story, if you will. Absolutely.
01:41
So I am an engineer by training, I am probably older than a lot of your viewers. I got started in the early 90s. doing engineering. And then I got sucked into AI, which actually is happening to neuroscience. So I did my PhD in computational neuroscience from Columbia. I was a scientist at MIT for a few years again, to studying how does the brain do things that the brain can find so easy doing things that AI and computers find so difficult. So I spent some time building interesting neural networks. And I left it in 2005, to create a startup that was going to apply AI and neural networks to real world problems. And that became a series of startups, all of whom were aimed at finding interesting uses of AI and broadly interesting uses of technology over the years and name a type of AI and I have probably done it. I’ve tried it like and tell you what works and what doesn’t.
Bryce Paul 02:49
Wow. Real quick, just on the subject of AI. There was just really hot in the news right now, obviously, and I’m in open AI is building this whole chat up T model. And they they recently like fired their their CEO, Sam Altman like the board apparent and then they rehired him or something that was crazy. It was in the news. We don’t need to go into it. But I want to kind of get your your take on? Did they discover something maybe dangerous? And so they decided, Hey, we got to pull the plug on it. We got to fire this guy who’s who’s leading this dangerous researcher? What do you kind of think happened there? If you could entertain us?
03:27
Look, I’m not privy to any information that you guys are not privy to i All I know is whatever you have probably also read. But I can give you a little bit of color that I think you will not find in a lot of places. And that’s what does it really take to build AI? Right? These types of LM models, they’re fairly new, they got started when Google published the original transformer paper in 2017. And since then, the future of field of AI has really exploded in complexity. And these large complex models, it really takes a a huge effort to train one of these models with hundreds of billions of parameters, right. And the thing is that before, before opening, I really nobody was doing it. There was no effort in 2018 2019 to really train very, very large. LLM models, mainly because you couldn’t justify the cost. It’s hard to estimate but standard industry estimates are somewhere between 100 million and a billion dollars combined in terms of the amount of computer takes and the amount of effort human effort that it takes to actually train one model and CPD for right And the thing is, if you’re going to spend a billion dollars, can you and the fact about AI models that a lot of non AI people don’t realize is that the majority of them really don’t work. They don’t, the technical word is they don’t converge, meaning you take a data, you take a very, very complicated machine, and it doesn’t go anywhere. It’s so you have to imagine a sort of a very complex AI model as, like the cockpit of an airplane. Right? Assuming that you the two of you guys aren’t pilots, if I put you in the cockpit of an airplane, what are the odds that the that with all the gadgetry and all the dials and all the complexity that in an airplane cockpit? What are the odds that the thing that actually takes off, I would say, very, very low, right? Even when very qualified AI professionals are at the helm of an AI model, the majority of AI models that are tried actually don’t converge. So now, there’s this thing about this, somebody is taking a billion dollar bet, on a model that we don’t know is gonna go anywhere. And even the best case scenario, even imagine that it converges, imagine that we can train this model, imagine that. It’s amazing. How much money will you make with it? Will you really make whatever four or $5 billion on it to make it worth our while? Do you have a business model that supports that? All of those are unanswered questions. So that really speaks to by somebody like Google or Microsoft, or, like companies like that don’t like we’re doing this before open AI was because it’s not a business case, right? This is not something that Goldman Sachs would do. This is not like, it’s not a cut and dried, sort of cut and dried, where you just say, Okay, we’re gonna spend a billion dollars, we’re gonna make 5 billion that we don’t know, we’re gonna. Yeah, you don’t know if it’s going to train or not, you don’t know if it’s going to make your money or not, you don’t know how much money you’re gonna make, you know how good it’s gonna be. There’s so many unknown questions. So it really takes like a very startup II type of die, to really lead an effort to make it happen, I started to beat up a company. And that’s really what opening I was used to head by combinator before he went to open yet he looked classic startup guy, you need people like that you need a team like that to really lead effort like that. And now, so that was a very long winded way of saying that this is this is a special company in the world of AI. And opening I was especially accompany especial, from several points of view, including the funky way in which he got started as a nonprofit, and then it moved to being a prop for profit. And then, you know, so there’s, there’s that sort of strange things happening in the governance of it at that level. And then there’s the bets that they’re taking, they’re taking these huge bets, with very unclear commercial outcomes. So I’m not sure what the dynamic is. On the board, I’m not sure what’s going on. And that sounds like a outsize sort of figure in that whole ecosystem. And I don’t know what the dynamic is, and what the board was thinking. Sorry, I’m not giving you the answer. I’m just gonna give you some insight into something interesting that I don’t think is commonly understood. No,
Bryce Paul 08:28
that’s exactly what we want. It’s just such an interesting kind of level of insight. And yeah, you know, your background in AI, is so fascinating. And I’m curious if it kind of informs any of the investing that you do at Cumberland labs, and sort of the incubation and any of the thought there are you looking for, you know, crypto, adjacent and AI adjacent projects to invest in an incubator, or what do you look for?
08:56
Absolutely, absolutely. The answer to your last question is yes, we always look for interesting projects, or personally, I am actually quite interested in the way that technology broadly, is applied to solve real world problems. I think I mentioned this earlier, when I was talking about my own companies, all my own companies were all ones where AI was used or blockchain was used or technology was used to really solve a problem that people have in the real world. And that I think, is where things become interesting for, for example, Blockchain. There’s a lot of effort right now, in blockchain to say, what are the use cases? How will we use blockchain to solve problems that people have today? And because of so blockchain has an interesting conundrum. And i i From if you take the narrow viewpoint that blockchain should solve people’s problems then the fact that there are these tokens that are created either fungible or non fungible. The fact that there are these tokens almost looks like a like an interesting happenstance. Because that’s not what sort of solves most of your problems. Because if you think about it, the tokens fungible, non fungible, whatever they are. The majority of people actually the majority, there’s 8 billion people in the world, 7 billion of them actually can’t afford the tokens at a fungible or non fungible right. So how can we solve the problem of a lot that a lot of people have, using this amazing technology? Blockchain, I believe, represents probably some of the best technology of the last 30 years, right. And I say that as an AI, it is an amazing technology, it represents an amazing leap forward in what it can do. I’m happy to talk about that, again, in the context of trust app. But this is an amazing technology, how will we use it to solve real world problems that real world people. That is something that I’m passionate about? That is something that Cumberland Labs is passionate about? And if there are people out there who are working on problems, who have solutions, who are trying to figure these things out, we would love to work with them, we would love to support them in any way that we
Brendan Veihman 11:16
can. Was there like an aha moment for you, where you maybe you saw something, whether it was like a deep fake or AI being used in a negative way? That was like, I need to be the one to combat these issues?
11:28
Definitely, definitely. So we’ll talk about trust. After that, can we get I’m having to come back to camera labs if you want. But talking about trust. Yeah, so talking about trust that the problem is actually misinformation, misinformation or fake news, right? This has always been a problem from the very earliest days, right? 30 years ago, when I first got started with the internet in 19, maybe in the early 90s, it was always an issue, you just didn’t know what was what you didn’t know what to believe. And, as humans, we bring sort of a trust that we have in institutions in the printed word, you know, in the in the 70s, or 80s, you just believe whatever you read, because it was largely produced by trusted parties, by large entities that you believe, you know, it was it was my Dan Rather are by, you know, New York Times, or whatever it is, these big institutions you believe them. And when people came to the internet, they, they they drop their sort of belief system with them. So the instinct that most people have is this believe whatever it is that they read, but very clear from the early days of the Internet, that you just, that was just wasn’t the case. And I think social media really brought that sort of this fact out that you really shouldn’t believe everything that you read, although most people still just, you know, they’ll pass on and they’ll post, whatever it is that they find, without making any effort to see if it’s true or not, if it’s actually even remotely possible, that that is true that, you know, whatever the earth is flat, or whatever it is that people want to believe in. So for me, the ha moment came sort of early this year when it became clear that AI was really part of the problem, right? So this fun fact for you guys. So as a, as a scientist, I care deeply about facts. And I think very deeply that facts are established in a scientific way. So the scientific way is to set up a hypothesis, and then figure out if it’s true or not, that sort of the standards of scientific way. So you can take any theory that you have, and you could, it could be something as simple or something complex. And then you break it down and you say how it will be addressed scientifically. That’s something that I care deeply about. And that’s something that I think the world does not do enough for. Too much, too, too much of the information out there, it’s just sort of is what I would describe as sort of pablum or maybe just people saying very bland things that don’t make much sense with any amount of rigor. So a given that I care deeply about science. And I’ve worked on AI for 30 years, here comes AI and now people are applying it to make misinformation worse. Right. Now people are using AI people are using my thing I consider AI to be my thing. People are using the thing that I put in the last 30 years to make the thing worse that I love. So I’m like, Okay, this is this is not right. This is not I can’t just stand by and let this happen. So I really began thinking about what Technology, what are the best technologies that we have today that we can bring to bear to really address misinformation or fake news? How can we help people stop posting things stop saying things that just aren’t true? Maybe we can stop them, but at least we can discourage them, at least we can give them a little indicator, look, this thing that you’re about to say, is not true, you know, not true, according to experts on both sides of whatever spectrum that you believe in, and not true according to common sense things that just make sense. So that’s really the genesis of trust. And that’s really a very sad, let’s go after misinformation and fake news. Let’s really make a concerted effort to see what is the best technology out there, and how can we apply it?
Bryce Paul 15:44
That’s fascinating. And it just, you know, like you said, I mean, with AI, it’s, it’s a double edged sword. I mean, with great power comes great responsibility. And some people are using this technology for great ends, and some are using it for bad ends. And so trust app is really trying to tip the weight or tip the scales in the favor of the good guys. And it’s utilizing blockchain technology to do so. And so we kind of want to unpack that, you know, at you know, you’re a scientist 30 years experience. I am a Amir plebe, and my listeners are probably not as well versed in AI, as, as you are. So try and dumb it down for us, us mere mortals. We want to understand how this works and and how you’re solving such a significant issue.
16:35
Absolutely. Happy to do that. So the two main headers in Vista technology that we work on for trust app resides our AI and blockchain. Okay. And I’m happy to go into each one of these individually. On the AI front first, it’s very clear that what is true and what is not true, can really only be established by the best AIS of what is out there right now. So for example, you say, Guys, New York Times is reporting that Israel did bomb that hospital, and there was a hospital bombing that happened a month ago that I’m referring to which some of you your readers are familiar with? Well, very controversial, the way it was very controversial. Right. And so some someone posts on Twitter, guys, the New York Times reporting that Israel did bother. Now, that’s a that’s a fact. Or that’s an assertion. That is true or not true? Yes. As undebatable, right, because because either the New York Times is saying that or they’re not doing. So the the, the the approach that we’re taking is using AI, can we accurately determine what are the assertions and what is being said? That you are looking at in your feed? And can we look at the sources of news out there? Can we look at the sources of information out there to actually extract assertions from that? And can we match? Right? And can we match them from all different sources? So there are 18 sources that are saying that what you’re saying is true. Okay. We think that it’s more likely than not that what you’re saying is true. 16 of the 18 sources are saying that we are disagreeing with what you’re claiming in your tweet. So we don’t think it’s true. So what we want to do is if as a consumer as a reader, we, as you’re browsing your feed, we want to give you little indicators that just say, green tech or red question mark, a little green tick that says, Yes, we think that this is likely to be accurate, given that the sentiment here is an agreement with the vast majority of what is being said they’re set out there and or little red checkmark, right in the in your feed that says, okay, and I don’t think so, you know, they just go check your sources like, and by the way, the little checkmarks that I’m talking about, if you hover over our checkmarks, we open up a little window that shows you what our analysis is. So we give you a link to the to the original source, we tell you this is why we think this is questionable, or this is why we think it’s accurate. So
Bryce Paul 19:30
this is you could just plug this into your browser, right? I mean, that’s exactly.
19:34
Absolutely. So the first release of trust app is a browser plugin, sort of like meta mask or the that other browser plugins that enables users to just quickly assess the likelihood that their Facebook although Facebook posts what they’re seeing, or the Twitter posts that they’re seeing is actually true or not, or likely to be true.
Brendan Veihman 19:55
Wow. You know, I’m curious to hear some of your experiences. As with everything that’s happening, there’s like AI and deep fake space, because one of the popular new scams that I’ve seen pop up is this one, especially for people who have a public presence on the Internet or on social media, it kind of listens to their voice, and then recreates it. And it’ll like call their loved ones or people of interest, and call them an act like this person. So for example, you know, me and Bryce, were all over the internet, we now have all these podcasts and YouTube videos, and etc. So what they do is that they’d send an AI model to listen and analyze our voice. And then it would call maybe our parents saying, hey, we need you to send us money, or, Hey, I need my social security number, I lost it, can you give it to me? And then all of a sudden, they go, Oh, this sounds just like my son. Let me go ahead and just give them that, you know, come to find out it’s actually an AI or a deep fake that is recreating our voices. So I’m curious from your end, like, what are the popular new? I guess, scams or ways that people are at risk here that they might not realize yet? And then how can they protect themselves from stuff like that? So
21:08
I can answer the second question, I think, much easier. Which is the, at the end? The quick answer is stop believing things that you read, or that you’ve messages that you receive, stop believing most things on the internet is the honest answer. Because really, what has happened is that AI and other types of technologies have really caught up to where technology should be, right. This goes into the point I was making earlier, we as humans, we can talk about this from a neuroscience perspective, from an evolutionary perspective. As a brain researcher, I can tell you why this makes sense. But we tend to be trusting a trusting species, we believe most things because from a long, long term survival from a longevity perspective, and it makes sense that we should believe most things that we encounter, we should believe most human, most of what most humans say, unfortunately, that’s just not a world that we live in. Given a human intentions, given human plus AI, together the combined potential for harm, it is really important for everyone, including our parents, and everyone around us to just first of all, take everything that you receive. And the default needs to be your natural, they’re not buying it, you know, even if it’s you, unless it’s you in person, and they have good reason to believe that in person, they should really shouldn’t be believing anything that they didn’t see. Let’s just start with that. Let’s just start with when. And by the way, this is also the default scientific view. Yeah. So science begins with, with the sort of a baseline of Yeah, we don’t believe anything that is being said, Okay, now, what evidence will it take to make us believe? Yes, that’s really, the world needs to be more scientific, the world needs to behave more like our science scientists. A true scientists behave, which is when you know, when you hear Newton’s third law, the first thing that everybody says is, Yeah, nah, nah. You know, okay, what evidence is there for this? How will we collect evidence for this? And that really, going back to your point about deep fakes and all of these things about AI? The question is, what systems can be put in place? For as a consumer to say, yeah, that really is price? Or no, not really price, you know? Can we set up set up a challenge, right says, can you just send me your social security number? Okay, Bryce? What’s my middle name? Or Okay, Bryce? What’s something that only you or Bryce would know, you know, like something? A little thing? A little? The the, it’s a personal version of a two factor authentication, you know? Yeah, exactly. Let’s say for something that only Bryce would know. Only you and Rachel. So that’s really, again, it’s a question of, how will we prove something’s true? Or how will we prove that the views being expressed are the views of this person? Absent them? Just the person thing? Yeah, I’m saying so, you know, Neil deGrasse Tyson has this big thing about how eyewitness testimony is the least reliable thing in all of science, right? Somebody said, I saw an alien the whole size again, you know, meanwhile, if in in the court of law, if somebody says, Oh, yeah, I saw him kill that person. And we’re like, Yeah, okay. We believe you and not buying it. You know, I think that we’re telling the world where Sooner or later, courts are going to come to that point where we’re going to stop believing in eyewitness testimony, because it really is what one person sees or says believes that their song is not reliable. So let’s build systems that actually are able to test assumptions and test this hypothesis. And I thought this is this is right. How will we believe this?
Bryce Paul 25:22
I think AI will will absolutely in some form or fashion change, or, you know, yet change law, right? Because like you said, I mean, you will no longer you can’t, you can’t, you’ll be able to prove that, you know, you can’t trust a piece of media about you. Because you’ll be able to prove that hey, like, look, I could create that same thing with AI or something very similar. So, yeah, I mean, it’s crazy to think about what what deep fakes are doing. And, and I see it now, you know, Brendon actually asked, like, what are some of the other scams that you know, viewers could watch out for? I’ve seen these things all over the place, that fake Elon Musk, fake Bill Gates, you know, their words in their mouth are saying, hey, send money to this bitcoin wallet, and we’ll send you back double, it’s a new promotion we’re doing, you know, we’re giving away new all this stuff, right? People on the internet, they see it, they think, Well, you know, how can this be fake? Right? It’s Elon Musk talking to insane send this, send bitcoin to this address, I’ll get double back or whatever. And it’s just crazy how scammy and how good the scams are getting these days. So hopefully, you know, trust app is something that we could recommend to our listeners. Because hopefully trust app does something to maybe mitigate against those fake giveaways or those, you know, spammy sort of YouTube links, you might get to, you know, would trust app kind of flag that and say like, this is not the real Elon Musk? Or is there a way to kind of, you know, have trust app alert people to where there might be suspicious activity with crypto related media?
26:55
Absolutely. I think crypto is a great use case for trust app, as well, because crypto as you are well aware. The the the the advantage and sort of double edged sword of crypto is that any transfers that are done are immediately final. So I like the like, sort of the US banking system where if you do a transfer, and you find that the person was was fraudulent, you have some time to take it back. In the world of crypto you ain’t taking, you know, it’s gone. Right? So crypto is is a much better candidate for fraudsters from for that reason, because I can really get good get away with a lot more like a bank, where people have more recourse to getting things back, people have more recourse to what they can do with the law and so on. And whether whatever the crypto, well, it’s gone. It’s a good thing and a bad thing. It’s a good thing that I have full custody, I take full responsibility. So when I give it my money to Brendan, when I give it when I give it to whoever Sam Altman, it’s gone now. I have no recourse of getting it back if I made a mistake. Oops. So for that reason, crypto I think is a great first use case and crypto people should be doubly aware of everything that can go wrong. And crypto should have better use cases of figuring out cancer. So I can easily imagine something like trust app or a sister app, the trust app being built specifically to solve crypto problems. So trust that is we can talk about the roadmap for trust app and where it goes from here and how we are imagining that it goes. But the use cases that you spoke about, which is sort of one to one interactions and figuring out that really is Bryce and figuring out this problem. I think a it’s a great problem to solve. I trust that doesn’t solve it right now. But I think it’s a great problem for someone to solve. And I think the world of blockchain can actually be used in an interesting way to solve the problem. Because blockchain can be used to recognize what is true and to mark what is true in interesting ways that you really cannot use any other conventional technology.
Brendan Veihman 29:29
I mean, it’s going back to Bryce’s point here. It’s funny because I will have people that reach out to me and say, Hey, I’m Bryce. Paul, have you ever listened to the crypto 101 podcast? And I’m like, No way. I’ve never heard of something like that. Like, tell me more.
Bryce Paul 29:45
It makes my stomach turn. Yeah, I
Brendan Veihman 29:48
mean, all jokes aside, I think that we are stepping into this direction, kind of like what you’re working towards over at trust that Naveen where it’s not exactly what you do now, but I can see someone kind of start to create that whether it’s you or whether it’s someone else, we’re starting to see some really useful applications to combat everything that we’re seeing in the AI space. But I think the big thing that I want to kind of hit home on here is that for every 10, things that happen that are negative, we’ve talked about the negative a little bit in this podcast, there are 100, things that are positive. And so you know, I think what we can’t do is overlook all the positive things that are happening, because there are so many great things happening in the world of AI and everything. And we can’t focus too much on the negative. But there’s a lot of stuff. I’m excited.
30:36
Absolutely. And let me take that as a segue to talk about why blockchain is important to trust. And, okay, this is talking about the positive impact that blockchain has. So this gets a little bit philosophical. So just give me a little bit of rope. I’ll try to be sort of short about that. So in the form of web one, the foundation was a set of protocols that were built largely by academics and engineers, who had no expectation of profitability. So for example, TCP IP was written by people at MIT who were funded by DARPA acdp was written by Tim Berners. Lee at CERN. Again, they were just researchers. SMTP was written by, it’s not even clear who but again, so that’s just a protocol. Now on top of us MTP sets, Gmail, or Hotmail, or Yahoo mail or Outlook, right, so the application was separate from the protocol. And that’s a very important distinction, that you use the same protocol, which is every all the internet is the same protocol, it’s SMTP. But you can pick any provider you want, which is fine. Now, you move into web two, but 10 years later, and you find that actually, that distinction is gone. The platform is the protocol. So Twitter is a platform, but Twitter is also a protocol. So if you have been sending out tweets, and you’ve sent out 10,000 tweets over the past eight years, all those tweets exist only on that capital. If the government of Iran doesn’t like what you what you say, because they find it to be anti Iran or whatever. And they reach out to Twitter. And the management team of Twitter agrees that, yeah, this is not right, they can just remove whatever it is that you’ve ever said, right? The, let’s say Elon leaves tomorrow, a new CEO comes in town, they implement a new policy, all of a sudden, a bunch of Twitter accounts are just gone. That content has been created. It’s just gone. It can be de platformed at any point. Right? So and same thing is true of Facebook, right? The same thing is true of Instagram. It’s true of any of these things. It’s a very centralized operation. A few people are directly in control of every single thing that you say. Everything government you
Bryce Paul 33:16
do Vinit, the government’s proven in some of these investigations right now. Absolutely. Right.
33:21
So the we have moved from a world where the platforms and protocols were independent. And you could just pick any platform that you want the protocol remain constant, to now where the platform is the protocol. And it’s quite bad. It’s quite bad from the point of view of information provenance. It’s quite, it’s quite bad from the point of view of who will believe anything that you say, because it can just go away? Right? Did did Brendan have rights really say this? Well, Twitter’s have decided to delete your account. Now it’s gone. So in my mind, map three, Among its many, many other features. One of them is it’s a public good, and remains, it’s certified by all of us, and it remains forever. Nobody, if we put it out there that dry said this. Nobody can ever say he didn’t. There’s no entity, there’s no company. There’s no individual, there’s no committee that can come in and remove what dry said if we put it on chain. Right. That, to me is a very important point. And a very important facet of vektory of what blockchain technology allows us to do. It allows us to really get in front of this idea of platform, a blank on protocol, sort of juncture and pull it apart and say no, ain’t doing it. The protocol is here. For the protocol. This was set, which cloud Comment percent on is irrelevant. It doesn’t matter. You said it on Twitter, you said it on Facebook, you said it on your personal website, it doesn’t matter, you put it on a phone record, you put it in a telegram chat does not matter. We have recognized that this was said, and nobody can ever deny, you know,
Bryce Paul 35:17
kind of like, kind of like a Bitcoin transaction, right? Like if it’s got all those six confirmations, you’ve broadcast that transaction, everybody attests to that truth. If you run your own node, I mean, you as well, you have a copy of that ledger of that canon of that consensus of that way, that singular version of truth that everybody else is agreeing on. That’s reality, right? Whereas like, in, in the banking system, or whatever, you could send transactions back and forth, and they could get refuted, they could get censored. They’re centralized actors in between, you know, they they might, you know, lose the wire or whatever, you know, in kind of blockchain. I love that, you know, there’s a consensus mechanism, it’s kind of reminds me of what you’re talking about, as well.
36:04
Exactly, exactly the point, which is, let’s use blockchain. So in the world of trust, that we use blockchain as a way, as a technology to establish consensus that something occurred, we are using all of us as a way to say, this is not this is content that cannot be on the platform. I think that’s the word I want, because it cannot ever be deep. Yes. You can never remove it, you can ever cancel. And it’s been said, so So in the world of trust app, we are doing two things. First, we are taking pieces of content. And we’re actually putting them on cheap. This is true for everything that we’re capturing right now. But we want to keep doing it into the future. And it becomes ever more important as the facts that we are collecting interest become sparser and more and more difficult to mobile difficult to represent in centralized places over the apple, if a telegram chat, or if it’s something that the citizen journalist has pointed out, we will, we will create a system where it can be minted, and it can be put in stone and say, Okay, this was this was recognized, this is true, this person was there, they saw it, they posted a video, whatever it is, and we were established by whatever means we have that that actually happened. That wasn’t just AI saying this, it was like nine independent people who took in different videos. And we have ways of establishing that they were in the right place at that time. They also whatever are we it was empty consensus is we can establish consensus about something that happened. And we are putting it on chain. And now it’s done. It’s kind of like
Bryce Paul 37:56
it’s kind of like a blockchain version of Twitter’s community notes. Like they just rolled out this new feature called Community notes where, you know, posts, you know, could have other people kind of check them and basically say, like, this is actually not confirmed, this is confirmed, and all that kind of stuff, which I think is pretty interesting. And so yeah, trust app takes that next level further, but I kind of wanted to ask one other question, kind of, on the subject of AI. But you know, kind of like how we were talking about running your own nodes, always having a copy of the ledger always been able to prove, you know, some people I hear are starting to run their own, like, AI models, like their own private like, Well, I’m not going to rely on Google barred, I’m not going to rely on chat GPT in a GPT for whatever, I’m going to run my own. What is that? Like? Is it how does somebody even start doing that? And do you think that that’s like, you know, maybe they’re a little paranoid? Or do you think it’s a legitimate cause?
38:55
Not only do I think is illegitimate, cause I think it’s actually a requirement. So So, here’s here’s, so here’s, here’s the way, again, this is something coming from an AI guy. This is something that isn’t talked about that much. Okay. So let’s, let’s take, let’s take two steps back from this actual question. And let’s look at these models. So chassis, PT board, all of these models, and you take the entire set of internet data, right? What is the data that dharma has been trained on? It’s everything that’s out there, it does everything without them. And you train a model that can do a number of things. Okay. So it’s a very general listing. Right? It’s a very general listing. You have to think of any of these AI models like activity or bar as like a general human being, right. It’s a general human being that’s good at most things, but it’s not particularly excellent at any given thing. Right. So for example, To any general human being, will likely not be very good at playing table tennis or at running. Why? Because to be good at running good run a lot, right? You need to actually be a train runner. Right? You will be running many hours a day for a long period of time. That’s a bit of good. Right? So going back to AI, if you want an AI, that’s good at x, whatever x is, let’s say x is playing chess, right, or x is what will pick it up, whatever it is right? You will actually train an AI model that is specifically trained for that specific use case. Okay, so and the way you do that is you take one of these more general purpose models like GPT for or actually, it has to be one of these open source models like Allameh, to or lot of these, like a black girl named the new 7 billion model that that just came out, there’s a lot of these open more open source models that are being released. So a younger Cohen is releasing a lot of these models at meta, entirely open source, you release the model, open source, now it’s available, right, so now all 7 billion parameters are available. And now you can use your custom data, to fine tune it, you don’t need to fine tune all 7 billion parameters, you can find you maybe 1%, maybe 7 million or 70 million, depending on how much data you have custom data you have. And that’s the equivalent of taking a human and letting them run a lot. So that they good, they become good at running. So for example, in our use case, if we want a model that is actually good at, let’s say, picking out if someone’s really talking about X, whatever x is, let’s say someone’s really good at talking about or someone’s implying that this content is from CNN, some specific use case, no existing AI chatbot model is actually going to be excellent, they will all be okay, well, maybe they’ll all be 80% accurate, or 70%. Accurate. And that’s enough to impress most people. But that’s actually not enough for any specific use case, right? That’s like being able to run a five hour marathon is probably impressive to a lot of people. But that’s not gonna be impressive, you really want to be good at running. Right? So now you need to take your data, you take all the open source model, and you start fine tuning it, that is neither easy nor simple. Because if you go back to the point I was making earlier, that’s really like taking a sitting in a cockpit and saying, This plane will take off now. You need to be really good at this. And you need to really be able to take that, you know, the 7 billion parameters, and figuring out how many of them you need to fine tune, how will you do it? What’s your strategy? How will any of this work? So you have to fine tune the model. And now you have a model that’s good at what you want to do. Right? At this point. Now, that’s what I’m describing is the basic requirement if you want to be sort of a kick ass AI company today, or have a kick ass AI model that is specific to a use case that you’re solving. Right? So you take one of these open source models, you fine tune it. Now you’re good to start. Well,
Brendan Veihman 43:33
Naveen, we really appreciate you coming on here and enlightening the masses. Truly, I think that this will be something and we’re going to start seeing verification sources like trust app, being integrated into all sorts of daily websites, wallets, crypto, wallets, exchanges, news outlets. And it’s sad to say that, you know, it’s sad to kind of admit that we can see a world where we need to have these third party sources verifying everything that we do, from websites to wallets to exchanges to news to digital self defense, right? I mean, that’s kind of
44:09
it’s absolutely so I’ll just in a couple of minutes I’ll describe to you a vision of where trust app goes a vision of where trust assets are companion which we are now imagining here in this podcast that doesn’t exist right now. Right so album for people here who want to start their next company. So we are imagining trust app to be sort of like Grammarly. Right so you guys are familiar with Grammarly? It’s a company that puts a little squiggles below your text and says, fix your grammar here. We can write not to spell but also grammar. And we are imagining a grammar impacts. You’re about to do something. You’re about to put some content out there. You’re about to write a blog post you’re about to publish something and you’re saying a number of things in there. And we can just go in and check and say yeah, true, true or not true. So everything squiggle below it saying, you know, the correct thing is this. In a Singapore did not become a dependent 1968. It actually became independent and 67, or whatever it is, whatever your factoid is, you know, Clinton was the 41st. President, okay, for the second, most accurately, not controversial, but people this okay, there’s putting whatever content out there, as you can imagine, and then you can become specific at this thing. The New York Times is it not reporting that Israel bombed that hospital? So you put a little squiggle balloon? Now, you can easily imagine a version of this for crypto, you can easily imagine a version of this that actually plugs into a wallet. And you’re about to send, you received a request from Bryce saying, dude, crisis, send me point five E or whatever it is, right? How about we make an app put a little squiggle below it. It says actually, that’s not Rice’s address. I did not address that we know that he uses, you know, the assembly department that is it. I even like if that’s Rice’s address. Where have you seen the last few times we send money to Bryce, who has these addresses? What about this? Are you sure you, we will stop you, but we’ll put it in the schedule below and put it in the morning that
Bryce Paul 46:27
just will be where, you know, it’d be it’d be great to have something like that. Something that just to have our back, a little digital shield in the world of, you know, everybody’s after everybody, everybody’s coming after your money. Everybody’s coming after you for one way or another. Which, you know, that’s just the way the world works. But unfortunately, Naveen we are are coming up on time here. We couldn’t thank you enough for spending the afternoon here with us, and for giving us so much incredible knowledge on the intersection of AI and blockchain, talking to us a lot about Cumberland labs, the work in the investing that you guys are doing there, along with the incredible system that you build a trust app. So with that being said, Where can people kind of follow you stay in touch, maybe even download the app when it becomes ready to rock and roll. So
47:17
the app, you can just search for this trust app. To PS it’s coming online in the next couple of weeks. It’s gonna be available in the Google Play Store as a Chrome extension. You can follow me on Twitter. I’m not terribly active, but that is where I’m begin starting to become more active now, on topics around AI and around blockchain. It’s nav AGI. So the first three letters and the last one and
Bryce Paul 47:46
wonderful and AGI artificial generalized intelligence, also kind of a nice little play there. Absolutely. Naveen, thank you so much. We hope to have you back again soon to talk about some more exciting things.
48:01
Happy to happen, guys. Thank you so much.
Please Support Our Sponsors
Check out Factor 75 and use my code crypto10150 for a great deal:
NetSuite:
Get your FREE copy of “Crypto Revolution” and start making big profits from buying, selling, and trading cryptocurrency today:
https://www.cryptorevolution.com/free
Subscribe to YouTube for Exclusive Content:
https://www.youtube.com/@crypto101podcast
Follow us on social media for leading-edge crypto updates and trade alerts:
https://twitter.com/Crypto101Pod
https://instagram.com/crypto_101
Guest Links:
https://www.linkedin.com/in/navagni/?originalSubdomain=sg
*This is NOT financial, tax, or legal advice*
Boardwalk Flock LLC. All Rights Reserved 2023.
▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬
Fog by DIZARO https://soundcloud.com/dizarofr
Creative Commons — Attribution-NoDerivs 3.0 Unported — CC BY-ND 3.0
Free Download / Stream: http://bit.ly/Fog-DIZARO
Music promoted by Audio Library https://youtu.be/lAfbjt_rmE8
▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬