Claude G. Theoret: "We rushed into AI without Historical Context."

I think one of the tragedies of humanity is our intelligence supersedes our wisdom. Science is intelligence, and let's say philosophy and ethics is wisdom. And we always do things, we rush towards new technologies before we've understood how those things are going to have impact. And we're doing that with AI right now. So yeah, I'm not a big fan of the, "Oh, let's stop all AI research" stuff.

Because there's a lot of AI scaries and what could go wrong. So where do you stand in this continuum?

I think like all scientific revolutions, we have to look at it from a historical perspective. And I think this is not happening today. So I worked with a guy, his name is Bruno Latour. He invented something called Actor Network Theory. I was still a physicist back then. I was introduced to his group. We started looking at patents. How do scientific or technical revolutions propagate? And so it was basically something called the impact factor, mostly was what Google search algorithm is based on, citation networks. I started doing the math for citation networks and that kind of graph theory. And I saw a lot of founders get divorced. It was just Sheryl Sandberg said, like, there are 5 things: fitness, friends, family. If you're an entrepreneur, you can pick 3 of the 5. And obviously I sacrificed fitness.

And you have to choose. You have to choose fitness for later.

Fitness for later. I lost a lot of weight, but man, I gained something like 60 pounds during that startup.

Yeah, it's very stressful. I mean, I had my own company and like I broke all my front teeth. Yeah, and now they're all back.

How'd you break your teeth?

I think the stress. So when I was running Keto Media, I was not even aware that I was grinding.

That journey we talked about, like grinding your teeth?

Yeah.

Like, I see you.

It's real. Like waking up in sweats at night. Eating your emotions.

Oh, eating my emotions. Like, I have nightmares of Excel. Excel spreadsheet nightmares.

Why?

Cash flow.

Oh yeah? Burn rate.

Burn rate. Okay.

Welcome to Driving Impact. Welcome to Driving Impact, which is a podcast about all the leaders, or also game changers in the world, who want to drive massive impact in the industry. And I was visiting Montreal and I told my team, I said, we have to interview Claude. And they said, why? And I think today we're going to discover So tell us a little bit more about you. You're an astrophysicist. You're also an AI entrepreneur. You're a venture capitalist. So tell us a little bit more about you as a kid. Where did it start?

Oh, boys.

And also, who was your biggest influence, like teachers, friends, for the making of Claude Théorêt?

Wow, that was a great question. So I'm a Franco-Ontarian, a French Canadian from outside of Quebec. And I was raised in, I would say, like almost poverty. There were a lot of laws that prevented francophones from going to school in English Canada. My parents, my mom had a grade 7 education, my dad had grade 5. Both of them hadn't even finished grade school.

So you really started from the bottom.

Really from the bottom. So I was trailer in a 10-acre field with unclean water. The water I grew up with. So now the things I'm grateful for is like, clean water every day.

Yeah, the little things that we take for granted.

Rule of law.

Yeah. Wow.

I grew up always curious. My biological father was a hunting guide and really instilled in me a love of nature that to this day.

Yeah, I still see you like going fishing with your son.

Fishing was one of the most important things he transmitted to me.

And what's special about fishing for you?

Oh, well, fishing is an analogy for life. Right? You do, you know, from a sales perspective, as a good, like a really senior salesperson, there's so many good sales analogies for fishing. Right? Like, you know, if you always be closing, if your hook's not in the water, then you're not gonna catch fish.

Nothing's gonna happen. Oh yeah, I'm gonna tell that to my team. And you have, in sales, you also have to stay zen. Right? You can't be like frantic, the fish is not gonna come. You can't, exactly.

You can't hunt fish, that's why we call it fishing.

Yeah, I love it.

Hunting is a different thing. Hunting is going after something and like chasing it. Fishing, you just gotta be in— you got— what I love about it is that you have to learn the ecosystem.

Mm-hmm.

Okay, and every lake, every ecosystem is different. And I think that's the thing I love the most about fishing is that initial exploration, right? And like when I— my wife is Australian, so when we first went to Australia, I decided to snorkel. Yeah. Okay, parts of Sydney Harbour, just to figure out what kind of fish there were. And I nearly got eaten by a shark, by the way. Oh my God. So it was like, all the locals were saying, "There are sharks here." And I was like, "Oh, I'm like 2 feet away from the side. Nothing's going to happen to me." And then the one guy on the paddleboard paddled up to me. I'm overweight. And the guy goes, "You know, if a shark bites me on the paddleboard, there'll be no problem. But if it bites you, you're like a seal." It's like, it's a jackpot. It's a jackpot. He's not going to do one bite and let go. It's filet mignon.

The best quality of meat.

Yeah, so for me, like the things that, you know, love of nature and then I'm a little bit dyslexic, dyslexic, I think as we'd say, so I wasn't doing well in school. And I went to Catholic school and I learned the Bible just to spite my teacher.

Wow.

Like I really hated her. And so I was like, I'm going to read this and I'm just going to make you look bad.

Oh my gosh. That was one of the pivotal moments in your career.

It was like, it was, I read just like she was, you know, we didn't get along. I was, like grade 5, but my priest saw this kid who like read the entire Bible in grade 5, so he thought I was touched by God. And so my priest took me in and like I became, you know, a reader in church and, and, you know, he taught me a lot of philosophy. He like taught me how to play chess. That really set me on the right path. There's something called Reach for the Top, which is like a quiz show, and we won the nationals when I was 17.

Wow.

Yeah.

So one of your pivotal mentors in your career was a priest?

Was a priest.

Saw the potential?

Yeah, yeah. Who advised me not to become a priest.

Wow. I'm sure there's a whole story about that one.

Well, you heard my confessions.

Okay. The confessions of Claude G. Théorêt.

He was literally like, I'm not sure, maybe I should become a priest. He was like, Claude, I hear your confessions. You don't have it in ya.

So then what brought you to the world, the Claude that you are today, which is AI, but really astrophysics?

That love of nature, that always wanting— like I said, like fishing is about understanding the ecosystem. And so when you really want to understand nature, you want to converse with or commune with nature, so you got to understand nature more and more and more and more. And you know, when you go deeper and deeper down, you get down to subatomic physics, and that's like the lowest level. And so even though I was an astrophysicist, I started a new field of astrophysics, and you want that combined particle physics, because I did my master's at CERN, which is particle physics. So it's electrons and anti-electrons you accelerate close to the speed of light.

And tell us a little bit more about CERN for the audience who doesn't know what it is.

CERN is the biggest machine ever built by mankind on the border of Switzerland and France. It's where the Higgs particle was discovered recently. The Centre Européen de Recherche Nucléaire. I don't know how to say it in English.

Nuclear Research Center. European Center.

Yeah. And then I worked at Fermilab in Chicago.

So you've been like, you've been in Europe, you've been— that's amazing.

Well, that's the thing, you see, I think if you give in, if you find what you're good at, what you love, and you really pursue it until you fail at it or you excel at it, well, you're gonna do one of the two. And that was a great thing about physics, is that it was something that, well, okay, I was passionate about it because of that love of nature, and you get to see it in action all the time. I just went as far as I could in physics, and that took me to Chicago for the beginning of my PhD. I did my PhD at Sandia National Labs in Albuquerque.

Which is Breaking Bad's territory.

It's the home of Breaking Bad.

I'm discovering so many things about you now. I have a whole new perspective.

I loved Albuquerque. I loved, actually, I've done a lot of work with the military when I was doing my PhD, but also in my first startup, so my work with the military was entirely fairly positive. I— and in terms of like where I stand on AI, I think that there's an AI arms race happening. We talked about it last night. I have no qualms whatsoever with working with the West, right? My biggest fear is, you know, some, uh, some dictatorships in the world winning the AI race.

So AI dictatorship. So just to take a step back, so what brought you from astrophysics to the world of AI? Because it's a jump. Like, I was a kid, I loved like black holes. I was reading Flatland. I was like, oh, this is so cool. It's 5 dimensions. But it's not like— like, you actually did it. You studied. And on top of that, you also founded the astronomy department of McGill University, which is not a small feat. So can you tell us a bit more about that period and then what brought you?

Yeah. So I was doing particle physics at Fermilab and there was a great— my supervisor. I've always been lucky to have one or two people who see talent, raw talent, and like take me under their wings. Like my priest. I had a great British physics professor in high school, right? When you see people trying to help you, you have to take the hand. A lot of people don't know and they reject it or whatever, but I've always been really happy to have various mentorships, right? So, uh, David Hannah, I'll say his name, is great physicist. But what happened is that when you do astronomy and you do particle physics, the amount of data you take is truly astronomical. And that's the leap to go from astrophysics to AI. So I have two patents in unsupervised learning. We didn't call it unsupervised learning back then. I have two patents. I had algorithms that automatically identified groups or signals and things in textual data. And so the patents were reclassified, I think, as unsupervised learning. Learning. Yeah, and they're used by IBM, they're used by— I forgot who the list is, but like huge, huge AI companies are still citing my patents. But at the time, I'm just a physicist. I, I adapted— okay, this is the truth— I adapted to— for those patents, I adapted an algorithm I developed when I was 24 working at CERN.

Wow.

And though the algorithms were to spot weak signals of supersymmetry in particle collisions.

Wow.

And I was like, well, if this can identify these weak signals through this algorithm, I adapted it to textual data. And that's what I got two patents in.

And that's— that was where things evolved.

Because you work with so much data.

Yeah.

When you work with particle physics and astronomy. So McGill— while I was at McGill, you know, switching from particle physics to astronomy, my supervisor, my unofficial supervisor, Diva Hanna, was like, hey, I want to start an astronomy department. I go, "There is none." Wow, that's so cool. He goes, "You can do—" and I'll never forget this, he says, "You can do your PhD and get out of here," because at the time I was kind of in— just wanted to do a particle physics degree and get out and be a quantitative guy, a trader or something. That's where they hire all the physicists anyway. And he goes, "You could do that, and that's completely respectable, and you're talented, you get through it very quickly." Or he says, "We could change the history of physics in Canada." Which one are you going to choose? Which one are you going to choose? That's literally how he stated it. —Okay. —That's a good sales pitch, by the way. He is a smart guy. I said, "Let's do it." We took this risk. My real supervisor was on sabbatical. When he got back from sabbatical, the 3 of us just got this thing going. We worked with the University of Chicago, Barnard University, and UCLA, and UC Santa Cruz. Some of the best universities. Sounds next to me. Exactly. UCLA, hello. I spent some time down there. Renee Ong, who was the lead scientist, is still a prof at UCLA. And I learned immensely from these people, right? And then traveling to the US, seeing like just the sheer amount of money they had for science compared to what we have in Canada, like working on Sandia National Labs, just being exposed to that at such a young age, right? Yeah. 24 at CERN. I didn't travel to Geneva, but like that's— that was— that's a— I stayed in Montreal for that, but going to the US Seeing Fermilab, meeting my first Nobel laureate while I was at Fermilab, all this stuff, it was like a huge opening. And what happened is that you became very good at handling colossal amounts of data.

Which is incredible, and that's how I also heard more about you when I was in Montreal. So let's talk a little bit about the zeitgeist of AI, because you were in AI before it was formally called AI. Yeah. Artificial intelligence before in the '90s versus today?

So AI, I worked, you know, in 1991, I worked at Northern Telecom, Bell Northern Research, we used to call it Big Nerd Ranch, right here in Montreal where we did, I did a work term where we were trying to do speech recognition. Okay. Today speech recognition, nobody calls that AI. But in 1991, that was revolutionary. That a computer can understand a human speaking. So AI has gone through these different manifestations where as soon as AI does something, or as soon as computers do something that humans used to do, we call it AI.

It's interesting because, I mean, in Montreal we talk about Nuance. Yes. And all the voice, uh, text-to-text recognition.

So Nuance grew out of that group. When Northern Telecom went bust, that same group became Nuance.

And then a lot of these engineers were hired by big American companies, even in Montreal.

Nuance was acquired by Microsoft, and Microsoft has a huge office here. So some of the Nuance people started another company. This guy is called Andy Morrow, who started a chatbot company, Automatt AI, right? And that's all from that Nuance lab. But also, you know, Nick Cappados, who's another big Montreal AI guy, was at that lab as well. So. That's what is AI is always when computers start doing something that humans used to do. Strictly speaking now though, Montreal is one of the biggest AI hubs in the world thanks to Yoshua Bengio, who's one of the three fathers who won the Turing Prize for deep neural networks. There's Yann LeCun, who works for Facebook in New York. There's Yoshua Bengio, and there's Geoff Hinton in Toronto. Literally what we call modern AI, the papers that founded that are published by Yoshua Bengio, who's in Montreal. That's incredible. We're a little city punching way above its weight.

Because of the brains? Because of the power? What is it?

There's a lot of things that came out of Montreal. Ernest Rutherford, who worked at McGill way back when, discovered the nucleus essentially here. He won the Nobel Prize for that. Then the Geiger counter, Geiger was Ernest Rutherford's student. So that goes way back. But I think Montreal goes historically, the Quebec government invested a lot in education to make education cheap and free for everyone. Accessible. Accessible, right? And that's the consequence of the colonialism it suffered. But those cheap tuition rates make Montreal the second biggest university city in North America after Boston. It's kind of a big deal. It is a big deal. I'm very proud of it. We have great universities, and so people come in all over the world to go to those universities. The federal and provincial government put a lot of money into a very risky proposition, right? Thinking that computers could start thinking and doing these tasks was not clear. But they took the risk and it paid off, and so did the University of Toronto. And, you know, Toronto has a great AI community. As well.

I think it's incredible. And then in layman's terms, could you tell us the difference between AI and deep learning?

Deep learning is part of AI. So what, what the term AI now is, I think people abuse the term. Yeah, yeah. Machine learning is part of AI. What I used to do, which was like people call unsupervised learning, I just called it clustering algorithms. Like, I didn't call myself an AI person. I thought like I would lose respect because I had too much respect for the people who did deep learning.

Yeah, yeah. And now everybody's talking about AI, everybody and their mother.

Yeah, decision science, there are all these things that are just being lopped into AI. Yeah. So yeah, deep learning is basically a form of— it's a bunch of algorithms that accelerate neural networks and do a pre-form of learning so that the neural networks learn on their own. Okay, and machine learning is anything from like statistical fitting and doing that kind of stuff. So there's a lot of different branches of machine learning, then there's reinforcement learning, and all these things now are all being lumped into AI.

Yeah, yeah, it's like the big tagline. Yeah. Now let's like zoom in a little bit more about you. So you founded Nexology and you sold this company, which was all about big data. Yeah. And now you're the founder of Ex Machina. Yeah. I want to just go do a little bit flashback about Nixology, because that's the time when I met you initially, and I was part of like social— I was doing social media at the time. Yeah. Flickr, and I was like, oh, this guy is smart. It's super interesting what he's doing, and you were analyzing all this data about behaviors and predictive types of analysis.

So Nixology was straight from, like I said, from my work when I was at CERN. I took— while I was in Paris, so I lived in Paris for 6 years. Paris is its own amazing place. But one of the things about Paris that's incredible is its sheer density. There are more Nobel laureates on one street in Paris than all of Canadian history. Incredible. I worked with a guy, his name is Bruno Latour. He just died recently, but he invented something called actor network theory. I was still a physicist back then. I was introduced to his group. and we started looking at patents and how, like, what prop— how do scientific or technical revolutions propagate? And so it was basically something called the impact factor, mostly was what Google search algorithm is based on, citation networks, and I started doing the math for citation networks and that kind of graph theory, and I actually transitioned out of physics into sociology briefly. So I published twice in sociology, and I worked as a postdoc And I really loved it. I loved it. Every year McGill has a Nobel laureate from the previous year who comes and gives a talk. And I remember once, I forget which Nobel laureate said, we asked a question like, as a physicist, how can I drive the biggest impact? And that Nobel laureate, he said, get out of the physics. Get into the world. Get into the world. Take what you've learned as a physicist and apply it to the real world. And I think that's very pertinent to your podcast. Yeah, it, you know, and I think physicists do that over and over again. They get out of physics, you know, and you see them in all kinds of fields that are too new, right?

And do you think that's what brought you to create Nexoology?

Yes. So you got into the world, I got outside of physics into sociology, and I started seeing, oh my God, like, we can actually see how scientists work together and which, which kind of scientific ideas will propagate why. And I just said I was like, "Hmm, there's something called a blog." This was like 2006. And the data is not clean, but it's free. And none of the people I was working with in the sociology—

Department.

—wanted to do anything with blogs. They thought it was crappy data. And it is. It's massive crappy data. And I just said, "Hey, I think I could do this on a much bigger scale than just patents and scientific literature." That was the birth of Nixology. Wow. And then of course, like, as soon as a new social network come along, like, we jumped on Twitter. Twitter had amazing terms of service in the early days. They shared their data with everyone, and like, we, we built basically an engine. It's real-time data.

It's massive amounts of data. Yeah.

So the data processing, everything else is very challenging. And yeah, I really enjoyed— I, we, you know, we did a lot of work. We tried to work with agencies, so Eventually got Droga5 as one of my clients, big agency in New York. I had a, one of my first software clients was in San Francisco, right off the bat. So working in Montreal, we didn't have any clients at all from the city, but eventually we did a lot of work. Of course, you know, we did some work for the government, various political parties, always worked on the progressive side.

So you were not the Cambridge Analytica?

No, we did meet with them 2, 3 times. They were looking to buy data and they ended up going with one of my competitors, so I was super happy about that. Yeah. But we always drew the line at Accelogy to not do ad buy. And Cambridge Analytica, we can go deep on that if you want, but like what Cambridge Analytica did very well was psychometrics, right? So they were able to really understand psychometric targeting and then ad bias. So if you look at the Brexit controversy, like one of the biggest spends for the leave side was Cambridge Analytica, an associated company in Victoria, in Canada, right? They did a whole bunch of psychometric ad targeting. And that, you know, it's controversial. Pollsters will tell you it doesn't work. To this day, people say Cambridge Analytica was all 'Cause people who do traditional ads and polling don't like the psychometric kind of data, but— So Nixology didn't do psychometrics. We didn't do it. What we did is that we spotted networks, and we spotted, if you want, proto-ideas before they would become actual networks of people knowing each other. So when you spoke about zeitgeist, we were able to identify zeitgeist.

Can you give us some examples? I wanna bring back for everybody of like, what did Nexology do and what—

Yeah, so some of the things that we're able to do is gauge the temperature around public policy. Okay. Right, so just before an election, you know, what are actually the people talking about? Is it what you think it is?

What's top of mind? Yeah, yeah.

So we did one where people— I can't talk about specifics— but people were extremely— the political party thought that one social issue that was on the media all the time was with the elections going to be about. And when we actually polled right across— we didn't poll, we just sucked in as much as we could— turns out health was the biggest thing.

Wow. Right? And what did the people think? What did the politicians think that was top of mind versus health?

Well, politicians, you know, one of the things is that they often don't listen.

So you help them listen to what— get a pulse of the population.

Correct. Correct. My main investors were polling firm, in Veronix. Okay. And so they're great angel investors. I'm thankful to this day, and I'm still very good friends with both the CEO of Veronix, Barry Watson. And so they saw it as a way of polling and doing real-time focus groups. Which is super exciting.

And then now, tell us a bit more about— so you sold Nexology. Yeah. It's another chapter. Yeah. And you started X Machina. Yeah.

So X Machina We talked about it briefly. So what I've seen is that, so the first wave of AI funding, there's a really seminal paper that Andrej Horowitz wrote that says AI is not like software. The initial round of VC investment in AI was done with the idea that it would be SaaS and highly profitable and high margins, and then you do the model once and you can just go. It turns out it's not like that. No. No. It turns out there's a lot of services that need to be done. Corporate America is not quite ready. Now we're seeing that people are really starting to uptake AI. But to this day—

If you go on LinkedIn, that's all people are talking about. I know, but— You can't just—

So if you look at, let's say, big data, you know, Wall Street Journal reported that 85% of the Fortune 500 big data projects failed. Right? Gartner stopped doing the hype curve on big data in 2017. Right? Why? There's not enough people who understand data in the US economy.

And then now there's all this hype around AI, which is based on data. Yeah. So where do we, I mean—

We're in another hype cycle, that's for sure. Okay. Without a doubt. But the thing is that I've always believed in AI way back from 2006, okay? So even though we didn't call it AI, we called it big data or whatever you want to call it back then. Because it drives efficiencies, right? And it's auto— it's automated insights. And so one of the interesting things right now about AI is that it's taking over spheres of human activity. Like I said before, like voice recognition was something that only humans did. Yeah. And in the '90s, we've got to have computers to do it. So that every time we do one of these things that automates a part of what we think is purely human, it's a massive, massive amount of efficiency and change happens, right? So one of the things that, you know, deep learning does very well is visual recognition, right? So that visual cognition, that is being automated. Yeah, that is huge.

And what applications do you think are more interesting? Because of course we can talk about the social networks, but there are other applications that must be more—

well, for social networks, you know, like You know, one of the things there is that what we're doing now, let's— so we'll split it up into, you know, what people used to call computer vision, and people didn't call that AI either, right? In the old days, computer vision was computer vision, and now that's part of AI. So if you think of the ability of a computer to do something that only humans could do before, all the boring tasks— like, there's one example in Quebec that's, that's really hilarious, but it's very practical— is potatoes. Inspecting the potatoes. That's probably one of the most boring jobs in the world. Could you spend 8 hours a day on a factory line like kicking out bad potatoes? Those are jobs we don't want.

Yeah, or even analyzing cucumbers. Yeah, well— Classifying the cucumbers on the chain.

There's a company in Quebec called Vouban, and they built a potato inspector so that you don't get rotten potatoes in your poutine. Right?

Because you want to have the best potatoes in your poutine. Oh my. Me, I'm like, I'm very— Picky about my cheese.

Yeah, yeah, it has to be curds. So that's a job that like— so people are afraid of the jobs that are going to be automated, but I'm actually like, well, a lot of those jobs nobody wants today, okay? And we can actually— I am a big believer in human potential. I believe we could retrain people to do better things. So visual cognition is something now that's being done in the realm of computers, and that's going to be just, just visual cognition.

So everybody's leveraging like ChatGPT, Bard, all these technology is a lot for text, but even more like images and creation of videos.

And well, this is, you know, with generative AI now, these are the things is that we think— I mean, so let's just talk about text and text generation as one. It's, you know, it's a new thing that like, yeah, it's not going to write maybe a Pulitzer Prize-winning novel, but 99% of what's written is crap anyway. And so everybody says, "Oh, it's not that great." Well, it's better than— it's good enough. Like YouTube, there are dozens of companies that tried to do video streaming before YouTube. Dozens and dozens. The key thing about YouTube is that the visuals weren't that great. It was just good enough. Yeah, and it worked. And it worked good enough. It's up. And this is what— we've hit the good enough level. Where actually, what's actually I find interesting by generative, by like the large language models is that we've turned, it's actually a user interface. We created a conversational user interface to everything.

And then everybody, most people who want to access can access. Yeah. So then it accelerates some of the tasks that maybe are not as interesting.

Oh my God, like writing emails. You know, writing is a difficult task. For a human, you know, we're hunter-gatherers.

We're not— Yeah, I mean, to be thoughtfully write something. Yeah. Not just to write.

Well, I'm dyslexic. So for me, writing is a pain. Like, I've sometimes spent like half an hour trying to get an email right. Do you use ChatGPT?

Oh, I used it.

I jumped on it right away and I used it. I use it when I have to compose something from an angle that I don't normally think about. Like, write this from an HR perspective. Those are the prompts that I like, right? I may not have the best language for this. So when I have to take it outside of my own context, I have used it. That's what I find tricky.

Because you want to make sure it's politically correct. Not politically correct.

I really believe that every context has its place. You know, like, I know I'm an extremely limited person. So one of the things about being a physicist is that you know what you don't know.

Which is good because some people don't know what they don't know. I know. You founded xMachina.

So xMachina, what we saw is that there are going to be problems in the way that AI companies are funded. Mm-hmm. 2, 3, we, there are a lot of, and so this paper, Anderson Horowitz was like, AI is not like software. The models have to have to be retrained. There's a services component when you have to interact with and implement it into enterprise. Mm-hmm. There's always a service component to this stuff. and if you try and do it straight up like a SaaS software, you're going to screw up in that part. And you know, there's a really good capitalist, uh, venture capitalist called, uh, Schuster. He has a blog called Both Sides of the Table, and he's always been a big advocate of having services in enterprise startups. Because I, I lived that experience myself with Mixology, so I took a lot of the failures, a lot of the learnings from Mixology, about how to grow and finally get right. And I, I'd say I never scaled Mixology to what it could be. I sold it early on, um, but I— we spotted enough of the problems, and I lived enough of those problems. I say, oh, I think I can actually help a lot of AI companies. And the way we do it is that we acquire the AI company in a public structure. Okay, so I worked in the public markets. I was acquired by a publicly traded company. Company, and then I helped turn around another publicly traded company. And I realized that, like, the sheer amount— certainly in Canada too— like, the sheer amount of money there is in the public markets compared to VC, right? And so I said, I think I can convince public markets people to invest in AI. And so what we're doing is, so we're, you know, we created a public structure to acquire AI companies, fund them, and help them grow.

Okay. And basically, this is the umbrella where, with your background as an astrophysicist and an AI, before it was called AI, you can vet the companies that you think have a high potential. Yes.

And ultimately, you know, we've tracked— so I've teamed up with Daniel Drouet, who we talked about last night, you know, one of the founders of, you know, Real Ventures, one of the first seed round venture capital firms in Canada, a very smart guy called Vaughn DeMarco, who who was an investment director at InvestAI, the provincial government AI fund. And so, you know, Vaughn had done 200 due diligences in AI companies. So, you know— Oh, that's a high volume. I always get people who are smarter than I am, right? It's like, go get somebody who's done, you know, 10 times more due diligence than you and have them come join your team. And the way I see my job as CEO is I, it's kind of like to put in, put together the initial team, raise the money, give them— try and give them the tools they need to do their job. Along this vision. And we tracked 1,124 companies. We met, we wrote to 340. If you're a young profitable AI company, you probably don't want to be bought.

Yeah, you want to keep scaling and growing.

If you are, yeah, but it's really hard to scale these. And we met 97 companies in one year. Most of the time when people decide to sell, it's highly psychological. M&A is extremely psychological. It's not about dollars and cents. It's about where people are in their lives. When I sold Nixology, my father had just died, my stepfather, and I was in the part of my life where I was like, "You know what?" I was struggling to a certain extent, and we hit one thing that worked very well, and so I was raising money when I decided to sell it. It wasn't, you know, we just got an incoming offer, but I said, okay, let's do it.

And then it created space for you to process. Yeah, yeah. What was going on in your personal life?

Well, yeah, my father's death. And I think that's when the decision to sell— often before, it's purely an economic decision, you know, like when most people want to go to IPO and then, you know, it's not selling, you still stay. Yeah, but that M&A decision comes like, okay, I've done enough of this in my life, or I need to move on to something else.

Yeah, it's time to pivot.

Yeah, I just started a family. Yeah, you have a son. I had a son. My son was born in 2013. I saw a lot of founders get divorced. It was just— Sheryl Sandberg said like, there are 5 things: fitness, friends, family. And if you're an entrepreneur, you could pick 3 of the 5. Mm-hmm. And obviously I sacrificed fitness.

And you have, you have to choose, you have to choose fitness for later.

Fitness for later. I lost a lot of weight, but man, I gained something like 60 pounds during that startup. Yeah, it's very stressful.

I mean, I had my own company and like I broke all my front teeth. Whoa. Yeah. And now they're all back.

How'd you break your teeth? I think the stress.

So when I was running Keto Media, I was not even aware that I was grinding. And I mean, the life of an entrepreneur, you know what it is. It's, yeah. You don't know who to turn to. Some people give a lot of advice, but at the time I was very young and I didn't have the net, the support network. Yeah. And, uh, my, my older sister actually said, what happened to your teeth? You always had perfect teeth. And I was like, what do you mean? And then I looked, I was like, oh my gosh, I had done this TV interview and I was, and then I went and I fixed it. Well, um, but that's like, I understand I had gained weight as well. So it's just health is important as an entrepreneur, as a, as a human being, but it's It's not like you're healthy, it's like you just have to keep it going.

I developed diabetes, my health was not good when I decided to sell. My dad died, uh, you know, that's a lot, heart failure. And I was like, that's it. Like, it was a big sign, like there was no way I want to go out that way. Yeah.

And then you want to— your, like, your body is your vehicle to what you want to accomplish. So now with Ex Machina—

so Ex Machina, now, you know, of course we struggled, uh, a little bit. We raised initial round very quickly. It was a great idea. It's based on a company called Constellation Software.

Mm-hmm, which is publicly traded as well.

Publicly traded, listed at $18. I think it's trading at $2,700 today. You know, nice factor of 144. Yeah. So again— Over how many years? 20 years. 20 years. So, um, we, we, we have an amazing, um, former VP of Constellation Software who's on our advisory board, Nick Nardi. He's like, he's a great mentor to me. He's giving us so much insights, avoiding us so many mistakes. Which is important, right?

Because you need to have like a lighthouse. You need to have the advisors to support you in your journey. So how do you think xMachina is going to drive impact in the world, especially in the world of AI where there's a lot of noise?

Well, first off, cutting through the noise. We'll be able to like, the idea is that, you know, if you want to invest in AI, well, it's very hard for a retail investor to pick which company. Mm-hmm. Even I, in my own portfolio, I got wiped out in 2022. Yeah. Like when the markets went down. And I'm a pretty seasoned tech guy. I thought I knew what I was doing, right? Yeah. So a mom-and-pop person— so what we're trying to do is create a vehicle where we want to acquire maybe 20 companies in the next 3 years. So if one company doesn't do well, then we want to be a safe bet. For normal investors on AI. And there's a democratically— the idea is that VCs in private markets get the massive growth, right, from seed to IPO. That's like, that's when you're getting the factor of 20, 50 on your investment. And so the idea is like, maybe we can acquire companies and we can't do a factor of 50 like a VC, but, you know, if we do the Constellation model, normal people who don't know a lot about AI will be able to invest in AI and get a pretty big growth factor. Yeah, because then they can buy shares and, and, and participate in the AI revolution.

Yeah. Who's your team in Extra Machines?

Well, Daniel Drouet, who founded, uh, founded Real Ventures a long, long time ago. Van DeMarco. Yeah, who, uh, was at Invest AI. And then I have two public markets guys, Scott Munro and Patrick Brown, and they know everything about public markets, right? So Scott lived, uh, Patrick lived in Houston, did some US, uh, public markets So I know what I don't know, like I said. And so, like, you have a solid team. Well, the public markets aspect, man, it is complicated. Securities law, so it's in its own world. I, I just got people who are better than I am. Yeah, you know, and they're the— I love working with them. I'm very, very fortunate every day. And then the same thing as we're growing, you know, we have to get Nick as an advisory on the advisory board. But the other part though that's more personal, that's not just like the way we're driving impact is that that journey we talked about, like grinding your teeth. Yeah.

Like, I see— it's real. Like waking up in sweats at night, eating your emotions. Oh, eating my emotions.

Like, I have nightmares of Excel. Excel spreadsheet nightmares. Why? Cash flow.

Oh yeah, burn rate, burn rate.

Okay, it's like, so we, you know, Daniel has invested in, like, when he was at Real, and we help entrepreneurs. That's what Daniel does today. So, you know, when you're in that tough position, maybe you need a team around you. Okay, a big part of this is that these, I would say, struggling entrepreneurs, but we help in the entrepreneurial journey. Yeah, right. And so you're not going to get the massive exit you would if you went all the way and you get IPO, but you know what, your family will be all right, you have a good quality of living, and as a public structure, we're going to attract better talent. Yeah. Because we can give options, right? It's not like giving shares in a private company where maybe one day those shares will be worth something. We're giving you options, a publicly traded company, to where we can attract better talent.

So I think that's super interesting. Well, congratulations on your new venture.

Knock on wood. We're listing in a few months, and so the raise is going great this summer. Yeah, it's very exciting.

I'm going to follow your journey very closely.

Hey, ticker symbols XMAI. You can buy up the float.

So let's talk a bit more about AI for the people, because I think that's an important topic as we talk about AI for writing, for video, for images, for voice. But how do you think AI can empower the average person?

So this is a very controversial topic, I feel, because I really do believe we are going to need to rejiggle capitalism. Yeah, well, actually, Sam Altman actually also said something recently about, you know, AI may be the end of capitalism. And like, I'm not a communist or anything like that. I'm a social democrat, maybe because I live in Quebec and I've lived in Europe. Well, so listen, the ethics— when you look at human creation, the process of human creation is really like taking all these things in your subconscious, and this is what AI is doing. So it's taken a whole bunch of images And it's— so if you look at like Stable Diffusion, all these things, they do random iterations around, let's say, an image, uh, your prompt, and they'll take it— they'll take images from that they've stored, and that's what it does for the randomness, right? And it minimizes the randomness until it gets something— by reinforcement learning— until it gets something that's like close to what you wanted. I don't think we've automated this part's this— so many parts of human activity so quickly at any time in history. Okay, and we've seen kind of like the sociological ramifications of technological revolutions in the past, right? When we did the steam engine, we didn't automate people, we automated horses. Okay, but that, that already was a huge change.

It was transformative. Oh my God, it gave—

provided access But also massive people moved from the country to the cities, right? Massive unemployment, all the exploitation we saw there. So there's— this revolution that we're doing is probably one of the biggest technological revolutions in human history. Let's let that sink in. And that's going to have sociological ramifications, like, right, probably massive unemployment in certain industries first, right? And what's interesting is that people always thought that it would be the blue-collar jobs that would get automated first. But those blue-collar jobs need 3-dimensional recognition of objects, and that's hard to do, right?

So they're going to keep their jobs. So are you saying that the white-collar jobs again are—

white-collar jobs are going first because it's based on structured data. So if we look at what— how you train AI, it's actually, you know, all the algorithms for AI were around, like, you know, neural networks were around 1991. Why weren't we able to do what we did? Because it's compute power that was lacking in the sheer amount of data we need to learn. And when big data combined itself with the neural networks, it created what we call the modern AI we have today. So if you look at where you're gonna get, where are you gonna get all the structured data to train the AI? We talked about this yesterday, this Wall Street Journal article with like millions of people around the world now just labeling data. That data, when you do machine learning—

So basically, you need humans to label. Data. You need humans somewhere to be able to help organize it.

A computer can't tell the difference between a banana and an orange. Our minds come preloaded with tons of stuff that computers don't have. Yeah. So there has to be a whole bunch of data somewhere that's labeled orange for the computer to go, oh, this is— these are the things that resemble an orange. Yeah, this is a chair. This is a banana. Yeah. So all that stuff has to be labeled by humans first.

And there's huge farms of human beings across the world. Millions. To do just that. Correct. Right? And that's what nobody talks about.

Well, so nobody talks about it, yes, but one of the things that like, you know, Scale AI does this, and there's 2, 3 other companies that, you know, label data. Now the consequences of this are coming up, but one of the reasons why creative industries are being disrupted first is that later it was all pre-labeled. How many pictures are there on the internet of Tamara de Lempicka or any artist you want? It's accessible. And so we trained it with what was free and what was labeled and accessible, and that hit the creative industries first. It's incredible, right? So there's tons of image data with tons of like, you know, this is a movie. Like, what I found interesting too is doing this in the style of, let's say, Wes Anderson, right? Yeah, it did such a good job because there's so much stuff on the internet to train.

So what do you think is next after the creative industry?

I think structured, uh, areas of structured data, but everything where this data is public, okay. I think where we're going to see a lot of acquisitions now coming in the future are going to be around data moats, where data that the various, you know, OpenAI, whatever, can't train on. If you have some of that proprietary data and you could do your own machine learning on that, um, then you— that may be acquired. Certainly you get to monetize that, right? So the next phases are gonna be all these forms of data that are highly structured, that are pre-labeled, but are sitting maybe in legal firms, or sitting in health. It's like closed doors.

Yes. Like black boxes.

And that's gonna be the next phase. And then once that data, once you get machine learning models around that, those jobs will be automated as well. And if it, because you need, the data needs to be structured and labeled.

But will they give access to that data? Will a law firm just open up the door?

Those are going to be some of the acquisitions we're going to see. And some of— if you're big enough, because the amount of money now that's needed to do any kind of compute is colossal. Yeah. So we're seeing a lot of investment in AI and they're done. Then right now the multiples are a bit out of whack again. We're getting into bubble land. Like latest round into Hugging Face, I think was 100x revenues or something, right? That's yes, but what we're gonna see is that if you're able to do a model, when you have to train these models, the sheer amount of compute that's needed is now only in the realm of a few companies. So it's one of the reasons I sold Nexology. We were analyzing billions of— High volume, yeah. And I knew I needed, much more money than I could raise at the time.

Just to process that, that—

Just do next round of training. So that's where we are, is that, you know, you know, OpenAI raised a billion bucks from Microsoft, I think, in 2018. And that's a billion dollars mostly in Azure credits, right? Most of that was compute. So not many people can afford that.

No, it's companies that have massive cloud.

Correct. —and storage. Yeah. So there may be some places now where it's going to be about data moats. And then even if you have the data moat, do you have the compute? Do you have the expertise to monetize that data? Because I always say, you know, people used to say data was oil, right? And I had a talk way back when that was like on SlideShare, and it was the most shared talk on all of SlideShare for one week. Wow. Yeah. It was one of my biggest accomplishments to get that email like, "Your talk on SlideShare." I was like, "What?" What was the talk? Responding to the "data is oil" thing. And I was like, "Yeah, data is like oil, but not quite.

It's more like New York snow." Your talks are so good. I think you're very good at making the complex simple and also attainable for everyone. Yeah. Which I enjoyed watching your talks online.

You know what? I'm a working-class guy. Okay, keep it real. I, I— it's one of those things where it's hard for me to— you know, my mom had a grade 7 education. I used to explain particle physics and Big Bang Theory to my mom, right? So like, I want to make sure that everybody can understand, and you switch your language. Like I was saying earlier, I use ChatGPT when I need to switch voices. Yeah. And I think if you're a decent communicator, you're going to try and switch voice so that the person in front of you gets a message.

Yeah. And you talked also about your dyslexia, and there's a lot of very successful people who have dyslexia. And when we talk about AI for the people and for health, just hearing from you that you can leverage AI to help you get things done in terms of writing an HR email is very powerful. Oh yeah. Do you see other applications in health that like we should be triple-clicking on?

Well, you know, sadly there's a company in Montreal that just shut down called Imagia. I think health is gonna be a tricky one because, but cancer recognition, anything to do with like, again, like We look at the things that are being— AI is very good at, computer vision. So it's kind of stupid that a human looks at an X-ray, right? Because you could actually analyzing that X-ray data straight from the data. Yeah. You're doing a two-dimensional projection of a three-dimensional thing so that a human can process it. Why don't you just keep it in three dimensions, have a computer analyze it, right? So I think having to do with like diagnostics and anything touches on computer vision and diagnostics, but also all the other forms of data around diagnostics. But the health professionals, you know, this is a thing where like there's no professional liability issues around, let's say, getting rid of graphic designers. They don't have a huge association. They don't have a union. They don't have anything like that. But medical associations have a lot of clout. Yeah. Right. Legal associations have clout. Accounting associations have clout. Engineering associations have clout. Like, so this is where the liability issues around AI, which professions are well organized enough to defend themselves against AI. Yeah. Those are the ones that will be later to automate, right? Doctors, I think, to a lot, and that's changing now, but I think the medical profession fought back against medical, assisted medical diagnosis. Right? And it's just, it's like a guild in that sense.

Do you think AI will be leveraged for everything that we talk about, environment, sustainability, to change the course of our climate boiling?

So there's a great company in Montreal called Brainbox. Okay, they've raised, I think, $60 million now. Okay. And this is a really good example of AI they use exactly for that. So they automate heating systems. Oh, wow. In big buildings. Right? And it's like, if you want to— and I, you know, I've worked with some climate funds. We talked about it briefly yesterday, and I think the intersection of climate tech and AI is very important for the simple reason that most of climate tech is so infrastructure dependent, right? And that infrastructure takes decades to build. Yeah.

Can you give us examples?

Well, let's just say getting rid of coal power plants. Okay. Well, sadly, getting rid of coal power plants, what are you— right now you can't replace it with any form of of power that's as steady as coal, right? Coal or gas or anything else where you need more power.

So it's chicken and the egg.

Well, we built coal as a 150-year-old infrastructure. We've been building coal infrastructure for 150 years. We have to take, just to build, let's say, the oil infrastructure that we have took maybe half a century. So to build a new infrastructure from scratch, it's not done in years. It's done in decades. So if you want to reduce, let's say, carbon emissions, probably one of the easiest things and cheapest things to do is just to make the existing coal plants run cleaner with AI. Data is cheap compared to completely getting rid of an entire power infrastructure. Like, that's in the trillions of dollars. Hundreds of millions for training a new model to do an immediate impact. So I think AI in climate tech can have an amazing impact right away because it's It's the cheapest, fastest thing we can do. But let's say in a 5-year window, not in a 20-year window.

So that's where a potential future—

I see a lot of overlap between climate tech and AI. So let's talk a little bit about astrophysics.

Because I feel like we talked about it. We didn't talk about it. We talked about it. So I want to go in a territory that's like when you look at the world and galaxies, And I was telling the team, I want to talk about the world and the parallel universes and whatnot. I don't know how looped in you still are in everything that's astrophysics, but what are the big inventions or discoveries in the world of astrophysics? Because we live on planet Earth, we know about Mars and all these other planets, but what is groundbreaking that you've heard?

Well, listen, like, I haven't done astrophysics in reality since 2005. Obviously, you know, I was on the paper that helped prove the supermassive black hole at the center of the galaxy. I did a very small part. I did like a lot of data acquisition things. I'm still very close to the hardware. But I think one of the things we've discovered is, you know, dark energy, where, you know, it kind of proves that Einstein was right. This is the thing that, you know, that happened in the late '90s. And so dark energy, are you familiar with that? No. So it's the idea that there's almost like an extra force that pushes against gravity. And what does it do? Well, it keeps the universe from collapsing onto itself.

So we want to keep it there.

It's one of the things that Einstein, when he did the gravitational constant, said it was the biggest fudge he ever did in his life, right? Because like, when you look at the universe, it's kind of like it's just expanding slowly and slowly and slowly, and the expansion seems to be slowing down. And he wanted to have a stable universe, right? And he didn't want one where it collapses back into itself because the force of gravity is too strong. So he made this constant, right? Completely fudge factor. But it turns out it's right. So it's real. It's real. It turns out it's right, and it balances gravity. There's something maybe we call dark energy that seems to be balancing gravity.

And is it at any risk at the moment of losing its balance?

No, it's more like this happens over billions of years, so there's no immediate risk in anything in astrophysics, right? Everything is, and in astrophysics everything is, it's not too precise. It's like, the universe is like, I think we know now, like 13 billion or 14 billion years old, but plus or minus a billion.

Where was I a billion years ago?

Yeah, those are the scale that like, you know, the error bars on that kind of stuff are quite big. And so those are— and then I guess the big one is like a few years back we saw the actual visual picture of a black hole, which is astounding. Like the amount of work that needed to go into do that was astronomy. That's not so much groundbreaking in terms of like actual pushing the boundaries of human knowledge of astrophysics, physics. But in terms of like for everyone to understand, black holes are real. Right? There's tons of them, they're all over the place. So those are the things like I think those are the latest big developments that happened.

But tell me more about the fact that black holes are real. Like how does it— what does it give access to? And in science fiction, they talk about like you go in a— you just go in time wrap and then you're like 50 years in the future. But what is the impact of black holes on us? Like, why does it matter? I mean, I love black holes. I just think the concept is interesting, but it feels so far away.

When I was an astrophysicist, you know, like I said, I did it because I love understanding. But ultimately, one of the things that I realized that from a sociological perspective is that every time we expand the knowledge of humanity, and we, you know, we have a new way of understanding things with the science, you're actually like displacing other crazy myths, right? And I think black holes for me, I don't like— so I never— but I'm not actually a big reader of science fiction, right? I never, you know, didn't like Star Trek, all that stuff. I think what's important about black holes is actually that they're the end. The end of the end of matter or anything else like this. It's like black holes, there's nothing like— from what I— everything in literature, there's always crazy theories you can read somewhere, but for the most part, they're neutron stars, and they have neutron data, which is extremely dense. And the black hole just gets bigger. So it's a creative process. It gets bigger and bigger and bigger and bigger and bigger, and there's no coming out of it. There's no thing on the other side. When you fall into a black hole, you will become part of the black hole. So you get— you become part of that thing that's causing that singularity.

Okay, can you tell us a bit more about it? I mean, we're not going to fall into a black hole, but like, let's say the Earth would fall into a black hole.

Yeah, we'd last microseconds. They would last— I think there are simulations that are done that it would last like 2-3 seconds. Like, as you're approaching a black hole, the force of gravity is so strong that just over the distance between your feet and your legs, the tidal forces would tear you apart.

Okay, so it's annihilation, which is— yeah, which is vanish.

It's just that force of gravity, as you get closer to something, gets stronger. So when we talk about tidal forces, so if let's say a solar system would fall into a black hole, the sun would be ripped apart like even before it'd be close to the black hole because the force of gravity is so strong. And then you probably likely become a plasma before you actually go in.

Okay, but none of these things are gonna happen. No, no, no, no.

But I always found that, like, the thing that's important about black holes is that, yeah, the sources of wonder and the other— but what I found is that it's the end, right? But it's the end and the beginning, because they're probably universe of a new galaxy. No, it's where galaxies prob— they're, they're, they're black holes, probably the center of every galaxy. And if you do simulations of how the universe evolved, unless you had some initial variations in density, the universe would— there wouldn't have been enough time to form galaxies. The force of gravity didn't have time to squish everything into a galaxy. So it's probably primordial black holes that were around right in the beginning that started having enough attractive force to draw in all the matter. So that's why black holes are a lot of ways, it's like the beginning of the universe, right? Where there's, you know, you started forming galaxies and they're important in galaxy formation. But also at the end, like when a star explodes, if it's heavy enough, it has something called the Chandrasekhar mass limit. And I think it has to be about 10 times the mass of our sun. I'm rusty on this. 9 or 10 times the mass of our sun. If you have that mass limit, then you will have like a supernova and what's at the core of a supernova is a neutron star. And that neutron data, that neutron matter is so dense that one spoonful of neutron data, of neutron matter weighs as much as the entire planet Earth. Wow. And that's what the center of, let's say, a black hole. As the order of magnitude of its importance. Of the sheer mass. So when you have supernova, supernova is an interesting process, right? It's a It's like stars are like onions and you see, you know, we are made from stardust, right? You've heard that. Yeah. It's actually true. From Hubert Reeves. Hubert Reeves. Proud Montrealer who lives in Paris, right? And who I interviewed when I was doing radio a long time ago. Really?

Yeah. I got the book in French.

Well, it's so, you know, stars are element-creating ovens formed like onions and the outer elements of the stars are very light, like hydrogen, helium, you know. You take 2 hydrogens, you put them together with 2 extra neutrons, make a helium. You put an extra hydrogen, you get a lithium. You put an extra hydrogen, you get a boron. Or you could take 2 heliums, put them together and get a boron. So all the elements are made together in stars because the force of gravity squishes these atoms together that don't usually want to be close together with each other. Okay, and the force of gravity gets so strong that it overcomes the electromagnetic repulsion and forces the, the nuclei to combine. And the difference in mass between the nuclei as they are apart and when they're together gets emitted in light. That's why the fusion process generates light, and that's why our sun generates light. It's that difference in mass from fusing 2 different atoms together. Okay. And so you keep this process going until you get to iron. When you start doing fusion with iron, it takes more energy than it gives out. Endothermic process. So the, the core of the sun is a huge iron ball, right? And what happens is that at one point, the force of gravity gets so strong that there's something called the weak force. And the weak force is like a sex change force. What does that mean? It takes quarks and makes them change flavors of quarks. Okay. So you'll go from an up quark to a down quark. Okay, that's what the weak force does. But when it does this, it changes a proton to a neutron. Okay. And so protons don't want to get close to each other, right? They have a huge positive electromagnetic force that keeps them apart. So when I hit this table, it's mostly empty space. Okay, what's pushing back on me is protons that don't want to get closer to each other. Hmm. Right? But I know as a physicist that this wooden table has a lot of empty space. When the weak force comes into play and changes an up quark or down quark to flavors, you can go from up, up, down, which is a proton, to down, down, up, which is a neutron. Okay, whoa, when everything changes to a neutron, what happens? You no longer have that force pushing back. So all the nuclei can squish together immensely. That's why you get that neutron matter that's so dense, right? That spoonful of matter that's as dense as the entire planet Earth. That's what happens when you get rid of the electromagnetic repulsion of protons.

Interesting.

So that iron core collapses, right? The entire sun would collapse to something that's like 50 kilometers wide.

Imagine that density. It's like an impact collision.

It's an implosion. A supernova is an implosion. And so the iron core collapses into this super dense neutron data— neutron matter. And then all the other elements that are lighter then iron, and there are onion— they're like an onion layer around the, the iron core— comes smashing down onto the neutron star that's now being born. When it bounces off, the collisions of the different levels of matter, as some are coming down and some are bouncing off, that's when all the elements that we are made of— lead, uranium— they're only created in like the 5 or 6 seconds said a star blows up. Wow. During the collapse of a supernova. So we are literally made from our stardust. Okay, except for the lighter elements in you like hydrogen, lithium, which, you know, we're from the Big Bang.

There's enough hydrogen in the Big Bang and that's how everything heavier—

lithium— only made—

we're made of a massive impact, which I'm going to bring back to the podcast. Yeah, that's only made—

there's another type of impact. Absolutely. Absolutely. Yeah. It's beautiful. Yeah, but that's another form of why I love astrophysics. There's so many analogies. And do you see any analogies?

I mean, it's going to be a bit of a stretch. You see any analogies in astrophysics, not the one that you've leveraged in data and where AI is going in the future? Oof. Like, I know it's a bit of a stretch, but I like to look into the future of potential scenarios of things that No, I don't see any, none that come to mind.

I think that the thing that, the only analogy I would have is that what's interesting about astrophysics is that it makes you, you have to have immense humility. Like I keep going back to, like it seems I don't have any humility, but I do. I feel like you're very humble. Yeah, like I really don't know a lot of stuff, and like I'm very open about not knowing, but that's the basis of science.

We know what you don't know. Yeah.

And so— But so I think in AI, we have to have that attitude. We are facing—

Being humble with AI of what we don't know. We don't know what the future holds. Absolutely. And what can we do to drive impact in AI in a way that's sustainable for us, for the planet, for—

Well, Bertrand Russell, I was a huge fan of Bertrand Russell when I was a kid. And he said that— I can't quote it exactly, but that I think one of the tragedies of humanity is our intelligence intelligence supersedes our wisdom. Science is intelligence, and let's say philosophy and ethics is wisdom. We always do things, we rush towards new technologies before we've understood how those things are going to have impact. We're doing that with AI right now. I'm not a big fan of the, "Oh, let's stop all AI research" stuff because I think—

You're not a big fan of like— because there's a lot of AI scaries and what could go wrong. Where do you stand in this continuum?

I think like all scientific revolutions, we have to look at it from a historical perspective, and I think this is not happening today, where we are like, every time we have a massive technological revolution, how has society been impacted? Okay, and so we have to look at the structure of scientific revolutions and technological revolutions, and we really have to like understand how that's going to change humanity. Yeah, so looking at the lens of human rights Human rights, geopolitics, you know, my— like I said before, I, uh, capitalism. How are we going to take care of each other when there— we've automated 90% of the jobs? Kurt Vonnegut wrote an excellent novel called The Player Piano that looks about this. Yeah, I'm not sure if you've heard about this. No, I'm going to read it. Yeah, so Kurt Vonnegut actually, actually has a very good novel about, uh, so it's the idea that he— there were machines that just record Human behavior. And he looks at what's— what's— what's— if you want to read a book now, the fiction, of course, and the implications of AI, I'd recommend The Player Piano. Okay. Because it's basically— he takes the last craftsman on Earth and he records— this machine they had would record the craftsman's movements and then automate that craftsman's job. So what are we gonna do with all these people who have no work?

Are they gonna do like AI prompting or There's not enough. It's not enough.

Not enough jobs. Not enough jobs. And I think this is one of the things where, like, we are going to automate some of the things we thought were core, like creativity. Yeah, like, this is the thing that, like, I think is surprising a lot of people. Uh, all these things we thought were core to humanity, man, there are a lot of them are being automated fast. And, you know, like I said earlier, the, the jobs where there's no protection are often the jobs where there's more people already, like, who— that exist. Nobody wants to do, you know, 5 years of differential equations. PhD. Nobody wants to— you have to have a really screwed up weird brain, right, to do, oh yeah, I'm gonna do differential equations. I didn't see a movie for 2 years, I think, like when I was doing my undergrad. Oh wow. Right?

Like, it's intense. It's a commitment.

I went to Waterloo, a very, you know, a very tough university. And, um, yeah. And so that stuff, like, OpenAI can't do math. We don't, you know, that kind of stuff is going to stay human longer than we think. That kind of creativity. Wow. Right? Because it's rule-based, it's based on logic, and it's not just based on, on reproducing what already exists. What already exists. So like what we used to think with the creative industries, they may get wiped out very quickly.

So what would be your recommendation for somebody who wants to evolve in their career?

Learn prompt engineering as fast as possible. And then what else? Uh, learn to code, but maybe even not even that anymore. Yeah, because GitHub is owned by Microsoft.

And yeah, and ChatGPT can also help.

Yo, it's structured data. Yeah, it's you know, this worked, this didn't work. GitHub, we just created the biggest repository of code in the world. Yeah. And it's owned by Microsoft. So like, without a doubt, now that structured data, there's gonna be machine learning on that, and that, that's going to be reproduced as well. So this is why I really think we got to think, what are we going to do with all these humans? How are we going to change capitalism? Yeah. Okay, I, I believe in the creative destruction theory. Okay, I do believe there'll be new jobs that'll be created, but I don't think the speed at which things are happening is the issue, right?

What do you think is the issue?

Well, when the, the idea of creative destruction is that when you have a new technology destroys a previous industry, that there's new things that come up, like video games. Yeah, who would have thought that like we'd be playing video games as much as we thought we'd be playing video games and employ as many people? Right? So we may have destroyed, let's say, traditional cinema by doing that. You know, I'm a— I, I've never been to a cinema now, like, certainly since the pandemic, right? So what do we do? We watch Netflix at home. That's been disrupted. And I, I love playing video games. Video games are like, you know, a form of movie now. They, they're so intricate, right? If you look at The Last of Us, right? A great series, but it was a video game before it was a series.

Yeah, it's interesting how things evolve. I mean, but there's some classic that's— I mean, I don't know, I just— Claude Lelouch movies, I was a big fan when I was 17. Yeah, yeah. So it's like, I don't know, maybe AI is going to be good in all these different languages and create movies.

I think AI is going to really move fast in the movie industry just because, again, there's so much outlying data and there's so much like, this is a good movie, this is not movie. It's all rated, so it's easy for machine learning to go, oh, this works, this doesn't work. If you look at prompt engineering, okay, when you look at movie financing, it's all done by like going, oh, we're gonna do, uh, you know, uh, Star Wars, Vampire Star Wars, right? Well, that's prompt engineering.

That's incredible, right?

So we're just— there's so much classifications. We can say Vampire Star Wars because there's There's so many vampire movies and so many Star Wars movies.

You can mix and mash them.

That a human can understand the classification. That means there's a lot of data that a lot of movies are to this, and that means we could train AI on it. Mm-hmm. Right? So as soon as you have these classifications where there's enough labeled data, that's where AI is gonna go.

Yeah, it's super interesting. So parting words. So thank you so much, Claude, for all the time and knowledge. Super excited about this episode. A lot of people would love to drive impact in their career, right? Just like you did and you're doing, 'cause it's always evolving. What would be your top 2 or 3 tips? How do you drive impact in your career, whichever field you're in? From Claude G. Thierry. Oh my God.

Get out of your comfort zone. Okay, number 1. Okay, I actually think, you know, this is a tricky one because It's two things. So focus. So find out what you do, do it as much as you can, and nail it. Like really, like in those— in your 20s when you have all that energy, you know, you should be wasting it trying to become the best in something. Okay, and doing that as much as that work as possible. And then, you know, and that means getting out of your comfort zone.

So basically focus Right? Find your superpower and triple down on it to become the best at it. And then thirdly is get out of your comfort zone. Yeah, like getting out of physics. So when you become very good at something, just go—

Well, yeah, sometimes inside that field you have much bigger impact outside of that.

Mm-hmm. Right? Because you're leveraging what you've learned to import it into a different domain, and then there's—

And that's really out of your comfort zone because I was a decent physicist, like I wasn't great, but at least I had, going to startups, I started from nothing. Zero, like, and then on top of that, when you're a professor, no one wants to fund your startup. Right? Yeah.

'Cause they don't know if you're gonna be real applicable in the real world, or if you're just gonna, somebody's gonna, theories.

Well, I finished, so there was something in Canada called the CIX Exchange, which is a competition of the 20 best startups. Third way back when, when I started Mixology. And I remember I had a spy kind of like on the admission committee helping me out, and like he said the biggest thing they had against me was that I was a professor, right? There's this whole like, oh, he's not gonna be an entrepreneur, and none of that. But nobody knew I grew up in poverty. I was a cattle dealer before I was a professor, right?

So it's like you've had this history of really like overcoming adversity. Yeah, reinventing yourself.

And but getting out of your own field, starting from zero, is is colossal. No, and most people never want to do that. They always want to be, you know, on top of the game, on top of their game, bigger fish in a smaller pond.

It's amazing. Well, thank you so much, Claude, and super excited about this episode and, and what the future holds for you and Ex Machina and all the wonderful things you're going to do.

Katheline, it's fantastic to see you again. Thank you. I'm so glad you're back in Montreal just for this, and it was really nice.

Thank you. Thank you.