AI: Friend or Foe? w/ Raoul Pal & David Mattin

0:00If you've been considering futures trading, now might be the time to take a closer look. The futures market has seen increased activity recently, and Plus 500 Futures offers a straightforward entry point. The platform provides access to major instruments, including the S &P 500, NASDAQ, Bitcoin, natural gas, and other key markets across equity indices, energy, metals, forex, and crypto. Their interface is designed for accessibility. You can monitor and execute trades from your phone with a$100 minimum deposit. Once your account is open, potential trades can be executed in two clicks. For those who prefer to practice first, Plus500 offers an unlimited demo account with full charting and analytical tools.

0:45No risk involves while you familiarize yourself with the platform. The company has been operating in the trading space for over 20 years. Download the Plus500 app. Trading in futures involves the risk of loss. It is not suitable for everyone. Not all applicants will qualify. today's episode of the real vision daily briefing is sponsored by calci we've talked about a few key events on this show recently notably the debt ceiling negotiation going on right now and we've come across a platform that allows you to trade directly on its outcome calci is a federally regulated exchange backed by sequoia capital y combinator charles schwab and other top investors that have invented a new asset class event contracts event contracts allow investors to take yes or no positions on events such as whether the U.S.

1:32will default on its debt obligations, whether the Fed will hike rates in June, what Biden's approval rating will be next month, and much more. Calci's event contracts allow you a chance to profit from being right about where the world is heading in targeted ways and is offering Real Vision listeners$15 when they sign up using the link calci.com slash realvision right now. Again, that's calci.com slash real Vision to claim your$15 credit now.

2:07AI, friend or foe? Hi, everyone. Welcome to this special Festival of Learning AI edition of the Real Vision Daily Briefing. With me today, RV co-founder and CEO, Raoul Pal, and David Matten, founder of New World Same Humans. Hi, gents. I'm so excited to see you. Hello there. It's great to be here, Maggie. Good to be here. So we are only in day two of our Festival of Learning, and we've already had some seriously mind-blowing conversations. And I'm not exaggerating when I say that. I mean, unbelievable. We're going to be covering a lot of ground over these two weeks, but we wanted to bring you in a little early in it so that we could get your view on some of the themes that we're going to be touching on and take some questions because we get so many about AI.

2:53But before we get to that, just a quick programming note. If you are not already a Real Vision member, you need to register in order to access the festival. It's free, but you do need to head over to the link that is pinned to the top of the chats. And we'll be showing it at the end as well. So, David, we spoke to Peter. We kicked it off with Peter Diamandis and Salim Ismail on Monday. And they're not even publishing their book, Exponential Companies, because things are moving so quickly. They just have to have it live online because they can't update it quickly enough. What do you make about what's going on?

3:33Are you surprised at the rate of change and how many applications seem to be rolling out around AI? Yeah, it's a tough time to be in the business of writing a book about AI because the publishing business is, to put it generously, linear. And AI is exponential. People in publishing take about two weeks to reply to an email. AI is eventually advanced, and it's incredibly hard to keep track of. It's a blistering pace of change. I can't say I'm shocked because, like Raoul, like you, I've been watching it for a while. And when you have pace of change this fast, what's happening is the convergence of hugely powerful forces that have been built.

4:17And this was a brilliant conversation, and this was talked about in the conversation. hugely powerful forces that have been building for a long time. And the change has come slowly and then all at once. This change has been building for years, for decades, if you want to look at the history of AI. But the last few years and the last year and the last few months and the last few weeks, it's just exponentially built. The compute power, the complexity of the transformer models, putting chat GPT into the hands of 100 million people in two months, then you get a pace of change that feels staggering and impossible.

4:56But if you look at the history, and it's not, it makes sense, but it's still, it's every time you look at it, it still feels impossible. You have to believe it all over again. And that's the thing with human beings and exponentials, and Raoul talks about this all the time, we'll never be comfortable with them. We just can't quite get our head around exponential change. And that change is only going to accelerate. So today we spoke with Mike Green and Nardo Menolato about what happens when AI meets quantum. I mean, it's sort of, as you say, has been already behind the scenes, but we're also kind of at a tipping point there.

5:34And we got a lot of questions about that too from very smart viewers who were wondering about the same thing. Amazing, amazing conversation. But I wanted to get your reaction to something interesting that Mike Green said during that conversation. Let's have a listen to a clip from that. One of the things that he brought up is this idea of automation in the home, which is to me, one of the primary mechanisms that we're ultimately going to see this play through, whether it's Alexa type applications or self-monitoring thermometers or temperature gauges, et cetera. When you think about this type of dynamic, one of the things that jumps out at me is the regulatory framework that many of the leaders in the AI space are actually pushing for very rapidly.

6:19And I'm known for the expression, why are you reading this now? My immediate reaction to what I'm hearing coming out of the open AIs, et cetera, is a reaction to the Google memo that points out that they're going to lose, right? Google has self-assessed and said, we're going to lose this battle. Open AI is going to lose this battle because open source AI is where it's all at. They can move faster. They can break things faster. They can play around faster. I'm seeing this, you know, I'm seeing this push for regulation as effectively a plea to protect the early entrance more than I am seeing it as a plea to protect the public.

7:01By the way, that comment was right out of the gate of what was almost an hour-long conversation. It was incredibly profound. When it was over, our great colleague Laura said it should be mandatory viewing for every human. I mean, that's the kind of conversation it was. If you weren't able to join us live, make the time to go watch it. If you are not a member, go to the link so that you can register for the festival so you can watch it. It's that important. Raoul, I'm laughing because only Mike could reach back into history and pull up a Thomas Edison reference to fill out that point he was making.

7:36But it's interesting. It raises some serious questions about the valuation of these companies. We see the NASDAQ up again. GitLab is up 33 % because of an AI announcement. I mean, these are big questions about what those business models are going to look like. I don't know. What's your reaction to that? Look, we don't know. And I think, as Mike alludes to, the scaling of truly open source AI means it's going to be ubiquitous. My view on all of this is it's down to the data sets. It's like the internet, right? The internet is the tool. In the end, it's what you get out of it. So yes, open AI might be the Google of the new world, but there's going to be a lot of people who hold data sets that are extraordinarily valuable.

8:22and those private data sets, I think this is what Emad's been talking about as well, is there's a lot of people who's going to build private data sets like Bloomberg's building one. In fact, Real Vision is looking at building ours about what can we do with private data sets with, you know, 7 ,000 hours of transcripts that we've got and all the other things, you know. So I'm not sure if it becomes a utility, maybe the infrastructure becomes a utility. So David, you talked a lot about this in the articles you've been writing for GMI, that you actually changed my framework and put AI as kind of a foundational layer.

9:02Talk through what you meant by that. Yeah, I think that this is a fundamental point. And I think one of the, if we're going to make sense of the exponential age and the shift that's happening now, one of the key characteristics of it, one of the key fundamental shifts is the abundance of intelligence. Intelligence historically has been rare. Human beings within reasonably recent history were pretty rare. There weren't that many of us. And part of the huge value of human beings was the intelligence they could bring. And then we could sort of put that intelligence together, network it in primitive ways.

9:42And creativity. Right, exactly. Yeah. And have forms of collective intelligence. And we started companies and everything. It's very hard to get your head around a world where intelligence itself is abundant. Intelligence is on tap. But that is the world we are, that is emerging around us now. You can hit a button now and access intelligence, right? It's not human level intelligence yet. It's not as flexible as that yet. But that is the world we're emerging into. And that to me is perhaps one of the foundational shifts that defines the exponential age, the big shift we're talking about into a new civilizational system.

10:21So I think, you know, and I reached for analogies in the GMI essay I wrote, and one of them I think that works is essentially intelligence becomes something akin to a new kind of fuel, a new kind of electricity that powers everything. Part of what the exponential age is about at heart is intelligence flowing through everything, objects, devices, tools, the physical world. That is a shift that is just profoundly transformational. And then, yeah, perhaps, you know, open AI, perhaps the big leaders now essentially do become something akin to utilities. And I do buy the argument that this, you know, their calls for regulation is at least in part a play to consolidate their position amongst a sort of whole raft of open source insurgents.

11:16You know, they want to be they think that one likely fruitful future for them is as highly regulated, you know, controlled providers of the most sophisticated advanced models. and then you'll have a whole ton of open source insurgents providing all kinds of other specialist models and I think you're absolutely right about data you know this is a this is going to be an incredible time for people and organizations that can bring proprietary content and frameworks and data and marry that with AI and fine-tune their own model and give that to the world. I mean, one way of thinking about it is we grew up in a world where we would worship something like Albert Einstein.

12:04And now we all worship an algorithm, an AI, because somebody will have something so powerful that's so clever. And the closer we approach AGI, the more we all worship these things as opposed to humans, because humans can't compete. As you said, a world where knowledge and intelligence and creativity is infinitely scalable, reduces so many of the frictions of what is possible. And we go into the realm of what we thought to be impossible, which is an extraordinary thing. Now, yesterday, we had the big Apple announcement as well. How does that all fit with AI? Because as you've said, AI is a foundational layer for everything of the exponential age now, as much as compute power is and energy is that without the, these are the three things that generate everything.

12:59And now we're starting to build stuff like Apple. Talk us through the Apple thing, how it all fits into AI. Yeah. I mean, this, you know, it, this is an incredible development and it's, although people don't think about it necessarily in these terms, it is, it is hugely implicated in the bigger picture we're talking about. You have this foundational layer of the exponential age, exactly as you say, you know, energy is foundational for civilization. Energy is necessary for anything, any meaningful work. Intelligence, compute power to run your intelligence, and monumental amounts of data, of information that you're moving around and that you're doing work on with that intelligence.

13:38Then above that, you have what we're calling in these essays, what you first of all called the productivity layer. How does it change tools, what we can do, how we generate knowledge, the economy. And I think what's fundamental there about the exponential age is that we're going to see this digital, physical, unified field. And I talked about this just a moment ago. We're going to see information flows and intelligence flows through everything and through all physical objects, through the world around us. I don't think people get their heads around what you're saying when you're saying through physical objects, but almost every modern object has a computer in it.

14:23And as you've talked about a few times, and beforehand we were swapping some notes, and you said, well, don't forget, you can put a two gig large language model on a mobile phone that doesn't access the internet. Anybody and any machine, any machine can have a large language model. Yeah, I mean, any machine with enough compute power inside it can. And this is what's crucial about Apple. So yeah, Apple Vision Pro yesterday, you know, this, and it's been, you know, the worst kept secret for a year or two, you know, finally, their augmented reality headset, and they've gone with augmented reality, not virtual reality.

15:04So this is about dropping a digital layer over the physical world around you. There was a whole ton of metaverse hype back in 21. You know, we know that this is perhaps going to be the kind of minimum viable product that gets more people interested in the metaverse. Though I think that word is kind of fading away in the power of virtual worlds. And what it's signaling is exactly what I'm talking about, the beginnings, the emergence of this unified digital physical field where information can become part of the physical world around us. And you can drop physical objects and you can drop digital layers of information over your physical environment.

15:52Now, how does that connect to AI? And you hinted at this. It's because Apple have essentially, you know, in plain sight, but in secret, been building a network of localized AI supercomputers. every MacBook Pro these days has an M2 chip in it capable on its own, on device of running a pretty sophisticated large language model. So anyone now can have their own large language model running on their own Apple laptop, and they'll be able to connect that to now their Apple augmented reality headset and literally, you know, speak virtual objects into existence. I mean, we see these incredible text to image models, text to virtual world models are next.

16:39I mean, we're seeing them now on the internet, but now we're going to see them in real virtual space. Create me a beautiful landscape and there it will be. And people don't realize is why you have it on your phone. A, it's not connected to the open AI network and maybe Apple will respect privacy. Maybe they won't. Let's assume that they do. The thing is, it becomes your AI for you. It learns everything that you do, who you speak to, how you do it. Now, that data is private, but it understands your life. Yeah, exactly. It understands how to reply to your mum when she texts you saying, why haven't you called me?

17:17It understands all of these things. It understands how to navigate your Microsoft systems or your Apple stuff or what you're doing in every way because it's with you. I mean, again, people don't understand this, but embedding AI into machines makes everything even more exponential. Yeah, it's so powerful. And exactly that. I mean, you're going to have, you know, lots of people were hoping that Apple would talk about their AI plans yesterday too, and they didn't. What they've given us essentially is half of the picture in my view. You know, Take a step back and look at the big puzzle that they're building.

17:57A major piece of it was put in place yesterday, virtual worlds. This is their play inside the metaverse, inside virtual worlds. But that is going to be married with their play inside AI, which they've secretly, somewhat secretly been laying out all these years. And now they have M2, highly competent artificial intelligence chips in their latest laptops. and you'll be able to run your own personalized large language model, your own personalized essentially virtual companion who knows everything about you, everything you've ever typed into the laptop, everything you've ever searched for, your tastes, your preferences, and speaks to you constantly.

18:37You run your life through that virtual companion AI and is now synced to this headset that presents, you know, real virtual objects across your physical environment. I mean, this is part of huge. What I think of the mega trend, if you step back from all of this and away from the hype, it's just the digitization of everything. You know, we're seeing the money being digitized. We've seen, you know, film being digitized, music being digitized, And now we've got intelligence being digitized. We've got, I mean, literally, and now we've got images being digitized. So the world is being digitized. Everything is being digitized.

19:22Maps have been digitized. So one after the other, the digitization of everything. So there's going to be a bunch of people in the comments section going, yeah, well, this is all bullshit because natural language models, they're just parrots. Stochastic parrots is the current trendy phase. Like people say about blockchain, it's just a database. and that this is not intelligence. My view on that, and I'd like to hear yours, is when you watch that film about Deep Go Mind, the Deep Mind Go, and you saw that it lost its game to the human, they were like, oh, this is never going to win. And then it completely changed how it played and played games never been played before by any human in any way.

20:10and they didn't understand it when he was doing it. It was like, this is stupid. It's lost its mind, but it hadn't. It had learnt. When some Altman was interviewed, he actually asked somebody, I can't remember who it was, Lex Fridman. He's like, do you think this is AGI? And Lex Fridman's like, don't think so. And why he did that is they don't know how this learns. I mean, that's so fundamental that it's not a stochastic parrot. It actually learns in ways that we don't yet understand. So I'm thinking that if I speak to somebody in Spanish in the street, my limited Spanish vocabulary makes everybody seem dumb to me because I can't understand them.

20:55So my limited vocabulary with how to speak to the AI and understand what it can do and what it knows means that it always seems more stupid than it potentially is. How do you think about this? Because there's a lot of pushback. Like, this is nothing. Siri's crap. Can't you see? I saw that comment earlier somewhere as well. Siri doesn't work. I'm like, I don't think we get it yet. But anyway, what's your cynicism? Yeah, I mean, Siri's like about to be or is now deeply irrelevant or, you know, the last iteration of Siri is deeply, deeply irrelevant. Yeah. What is my take on this? Look, part of the problem is that 50 years ago, 60, 70 years ago, whatever, you know, the pioneers of this field chose to call it artificial intelligence.

21:43Intelligence is an incredibly vexed word. It's something that we feel we as human beings have. We don't have any understanding of how the human brain, this physical object in our skulls, generates the experience that we all walk around with inside our heads. We don't have any idea about the cognitive processes that underpin language, understanding or generation. It is a miracle in plain sight that I'm sort of making sounds now that conjure understanding in your brain. We simply don't understand how that works really at all. So look, as I've said before in my last daily briefing, I'm always open to a bit of scepticism.

22:30Scepticism is good. We need to keep... You're English after all. Huh? You're English after all. So you're going to be exactly I mean you know I'm in London if I was not if I was not somewhat skeptical I'd be I'd be I'd have been yeah sort of sent away by now but when people come out so confidently and say you know this is not intelligence and this is just a stochastic parrot and this is all bullshit essentially it's like hold on a second you know we don't we don't understand how the human brain does these things the AI is doing things we don't understand too You know, this is pure speculation.

23:05My personal belief is that probably some of our language capabilities are essentially statistical, stochastic processes of relation finding anyway. So we are slightly stochastic parrots, you know, and articulate people are just, I mean, are likely to be or perhaps are people who simply are better at building a large language model inside their brain. You know what I mean? So, yes. And on a practical level, I think a lot of these questions around is this AGI or not will somewhat fade away because we are going to start to encounter. And I think we already are with GPT-4 intelligence that is flexible enough and nuanced enough to be surprising.

23:58We can't understand it. It constantly gives us new perspectives. And it just becomes an encounter with an intelligence that essentially feels somewhat human in nature. And also what I find interesting, because people always look at today, assess everything on today. when you can't when something's going exponential. So people say, well, I asked it a question about what was my grandmother's first name. It didn't know it, or it gave me a wrong quote. You don't understand that this is still a child. It's at the child stage of learning. This is not a finished knowledge model. What it is, every question that's being asked by the 100 million people is building its intelligence.

24:43It's not a fixed state thing. So the longer it goes on and the more all of us use it, the smarter it gets, the better it gets in refining the answers. Right, right. And some of the reasoning it's capable of doing now is just absolutely astonishing. You know, I mean, if you're not surprised by that, then you're just not paying attention or you're lying. I mean, it's just incredible. Some of the, you know, I mean, GPT-4 doesn't know the answer to every factual question, but it can conduct fairly high level linguistic reasoning across a range of domains, often better than most people can. That is incredible.

25:22And don't forget, it's the worst AI we're ever going to have now. It's only going to get better. I want to jump in on that wild thought, because that's absolutely true. And we talked about the pace it's going to change, right? So it's not like, oh, in a few years, we'll get an update, And then it'll. So Jim is asking, is AI just information processing on steroids? Isn't everything? Yeah, I mean, this this goes back to this goes back to the sort of philosophical stuff about what human intelligence is. Yes, essentially, it's algorithmic intelligence processing. When you get algorithmic intelligence processing that's complex enough, you start to get outputs like the outputs we're seeing.

26:11It's a highly contentious question as to whether the human brain is simply information processing, simply only information processing, or if there are other kind of dimensions going on inside the brain that we don't understand. But if there are, we don't have any understanding or model of what they would be, because all we can see is physical information processing happening inside the brain. So if something else is happening, we don't know what it is yet. Isn't the human body? I mean, the human genome is just information processing, right? It's just a biological form of computation. I know that's sacrilegious.

26:48I like David's answer about us not being exactly sure, though, leaving the door open that there's something more. But it doesn't mean that this won't also mimic that. since we built it and figure out what that is. But that's going to be maybe one of the things that unfolds. By the way, in the conversation with Nardo today, one of the things, I mean, there were many times where my chin hit the desk, but one of them was when he said, solutions will not be within the human realm soon when quantum computing comes. That's a really wild prospect. Well, quantum computing changes the multiplier of the speed of intelligence and information learning.

27:30And it becomes well out of the realms of human possibility to compete at that level of computational power and speed. But if you think about the implications of that, it's huge. Okay, so Samantha asked, and I think she had to drop, but she's going to come back and watch it in full. Would love to know what AI development Dave is most excited about. I mean, I'm hugely excited about the marriage of really interesting proprietary data sets with powerful large language models. I mean, Raoul mentioned Bloomberg GPT, perhaps empowering Bloomberg and the world of finance to make even more money is not deeply exciting for everyone out there.

28:18But as an intellectual exercise, it is hugely powerful to marry decades of deep proprietary content and knowledge and expertise with a large language model. Or look at Google MedPalm, you know, absolutely incredible. You know, Google's large language model married with deep medical data and knowledge to create an AI that can essentially pass medical exams, that can answer questions as well as qualified doctors. I mean, it can't actually doctor to you. It can't walk into a room and look into your eyes, but it can, in terms of answering questions, making diagnoses, an incredibly powerful tool in the hands of doctors.