YT YouTube2Text

YouTube2Text → Transcripts → Ex-Google Officer: You Only Have 3 Years Left Before It Hits! - Mo Gawdat

Ex-Google Officer: You Only Have 3 Years Left Before It Hits! - Mo Gawdat — Transcript

by The Diary Of A CEO · 22,849 words · 3,649 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:00We have [music] video evidence of people
0:03abusing children and not a single person
0:06got arrested. How can you call [music]
0:07that a democracy? So, humanity is at a
0:10crossroads where for the first time ever
0:12[music] we need to wake up and realize
0:14that we're ruled by maniacs and what we
0:16believe is democracy is not democracy
0:18and what we know is not [music] the
0:19truth. Like companies and governments
0:22will blame the geopolitical and economic
0:24[music] challenges we have on AI, but
0:26the truth is AI is not the enemy. Like
0:28I'm not worried about AI turning against
0:30us. I'm [music] worried about humans
0:32telling AI to turn against us. Like when
0:35I worked at Google, we were building
0:36amazing things believing that we were
0:38making the world a better place and we
0:40were. [music] But then suddenly there is
0:42a moment where you recognize that maybe
0:44the world will not use what you're
0:46making the way you want it to be used.
0:48And sadly, this is upon us.
0:50>> So, I have lots of questions.
0:52>> Okay, that's good.
0:53>> So, what's your take on this job
0:55disruption point? What is the risk of
0:57these very intelligent models that the
0:59creators of these models don't actually
1:00understand themselves? Do you think Sam
1:01Altman's pro-humanity? How do we get to
1:03a point of ethical AI when the incentive
1:05[music] structures are so highly
1:06competitive? And then I wonder if
1:08there's a path that ends in AI being net
1:10positive for humanity.
1:12>> Somehow we've been pre-programmed
1:13[music]
1:13to believe that this is upon us and we
1:15cannot change it. And I refuse that. So,
1:18we will talk about the solutions.
1:19>> But are you optimistic?
1:21>> I'm very optimistic about the future.
1:22I'm not optimistic about the next year.
1:24>> Why the next year?
1:25>> Come on, Stephen. You don't want me
1:27[music] to say it.
1:29So,
1:32>> This is super interesting to me. My team
1:33gave me this report to show me how many
1:35of you that watch this show subscribe
1:36and some of you have told us, according
1:38to this, that you are unsubscribed from
1:40the channel randomly. So, if I were to
1:41ask all of you, please could you check
1:43right now if you've hit the subscribe
1:45button if you are a regular viewer of
1:46the show and you like what we do here.
1:47We're approaching quite a significant
1:49landmark on this show in terms of a
1:50subscriber number. So, if there was one
1:52simple free thing that you could do to
1:54help us, my team, everyone here, to keep
1:57this show free, to keep it improving
1:59year over year and week over week, it is
2:01just to hit that subscribe button and to
2:02double-check if you've hit it. Only
2:03thing I'll ever ask of you.
2:05Do we have a deal?
2:06If you do it, I'll tell you what I'll
2:07do. I'll make sure
2:09every single week, every single month,
2:10we fight harder and harder and harder
2:12and harder to bring you the guests and
2:13conversations that you want to hear.
2:14I've stayed true to that promise since
2:16the very beginning of The Diary of a
2:17CEO, and I will not let you down. Please
2:20help us. Really appreciate it. Let's get
2:21on with the show.
2:23>> [music]
2:25[singing]
2:26>> Mo Gawdat,
2:28I spoke to you, I think, about 4 years
2:30ago when you wrote a book about
2:32happiness, and I remember you came in
2:33and you'd written this book about AI,
2:34but I particularly wanted to speak about
2:36the subject of happiness cuz I was
2:38fascinated by it. What I find
2:40astonishing is the fact that you were
2:41talking about AI before anybody was
2:44really talking about AI. No guest that
2:46had ever come on my podcast had ever
2:47mentioned the subject of AI. It just
2:49wasn't interesting to the world. And
2:51then this thing called ChatGPT came out,
2:53and suddenly everybody got to feel it
2:55for themselves and became fascinated by
2:57it.
2:59My first question, and this might be a
3:00question for people that don't know you,
3:02is why at that time did you start
3:04talking about AI before anybody else?
3:07>> I knew them in the lab. I I joined
3:09Google in 2007,
3:13very late 2006.
3:15And at the time, most people don't know
3:17that at the time we had reasonably
3:19established AIs doing our back-end work.
3:23And 2000 uh eight, we had the cat paper,
3:26which was published 2009, the first real
3:29unprompted AI.
3:31Uh I remember 2016,
3:34I had that incident where I was, you
3:36know, observing a project we were
3:38funding that was about teaching grippers
3:41how to grip,
3:43unlike industrial machinery.
3:46Uh so, to be to be able to grip like a
3:47human needs a very high level of
3:49intelligence to to to assess the
3:52texture, the softness, the positioning,
3:54and so on, the shape of everything. And
3:56we were doing that, and it it just blew
3:59my mind how similar to my kids they
4:01were. And I think that was my very first
4:04realization that we were building the
4:06apex of intelligence. We were genuinely
4:09handing over the reins of
4:12superintelligence to another being,
4:14right? And when you when you get faced
4:17with that, you start to suddenly realize
4:19something that we at Google found very
4:21difficult to realize, which was that
4:24everyone I knew at Google till then
4:27uh
4:28was believing that we were making the
4:30world a better place, and we were. We
4:32genuinely were were doing amazing things
4:34for the world. Uh but then suddenly
4:36there is a moment where you uh where you
4:39recognize that
4:41maybe the world will not use what you're
4:43making the way you want it to be used.
4:46And And you can see that in lots of
4:47technologies, you know, social media
4:49starts by the claim that it's going to
4:51get us connected and get us closer, but
4:53eventually ends up separating us with
4:55that little screen, you know. Dating
4:57apps are giving you the promise that
4:59you're going to find your soulmate, but
5:00in reality they keep renewing month
5:02after month. And so tech
5:04uh
5:05somehow ends up being more capitalist
5:08than altruistic. And I think I was I
5:11wasn't the first in a Bostrom started,
5:14you know, Geoffrey Hinton completely
5:15changed his mind. Fei-Fei Li is starting
5:18to say, you know, this is very serious.
5:21Everyone now, everyone who's ever had a
5:23very deep relationship with the machines
5:25is a bit concerned.
5:27>> I wonder if there's a path that's hard
5:29to see now that ends in AI being net
5:32positive for humanity.
5:33>> Sure. I bet 100% on that.
5:36It's that this path is very painful.
5:40>> This path is very painful.
5:41>> Yeah, so so the example you need to
5:42understand is, you know, we discovered
5:45uh uh, nuclear power.
5:48And the very first implementation was a
5:50nuclear bomb, not not nuclear energy,
5:52right?
5:53And, uh, and I think that's exactly
5:56what's happening with AI. Uh, the first
5:58implementations of AI
6:00are
6:01in favor of a few at the expense of the
6:03majority.
6:04You know, the the in favor of the
6:07capitalist to increase productivity and
6:09reduce costs, uh, but not taking into
6:12account how that impacts on the general
6:14public, uh, you know, in favor of the of
6:17the armies uh, that are now competing
6:20with autonomous weapons, in favor of the
6:23surveillance systems that are attempting
6:26to control everything with more and more
6:28and more intelligence and more
6:30monitoring.
6:31And
6:32that's not AI waking up in the morning
6:36and saying, "Hey, you know what? Let's
6:37oppress all humans."
6:39But it is a powerful few that are simply
6:42deciding to use the ultimate superpower
6:45on the planet today
6:46to gain more power and more control. I
6:49mean, as we speak, we're living in two
6:51major wars where AI is doing most of the
6:53killing.
6:54>> Cuz a lot of people think of AI as these
6:56like chatbots that we're all using to
6:58help us write.
6:59>> No, I think I think there is a hype. I
7:01call it the hype dichotomy, if you want.
7:04So, so what the general public sees
7:06about AI is overhyped, but
7:11ineffective. You know, all of the fake
7:14videos and all of the, you know, um,
7:17Grok did this and, you know, we
7:20attempted to switch off that machine and
7:22it did that and so on.
7:24What the real geeks see inside the lab
7:27is just
7:28unbelievable intelligence. And so, what
7:31is about to happen is that we've started
7:34to put together systems that develop
7:36themselves. They look at their own code
7:39and they, you know, they They
7:40experiments and they test those
7:41experiments if they changed something
7:43and they see where the machines are,
7:45uh you know, the performance is.
7:47And and redeploy the best code. Okay?
7:50And And if you just think of that, I
7:52want you to try and imagine a world
7:55where we have a tiny little genius
7:57sitting in the back end somewhere
7:58trying. But instead of trying a new code
8:01every day, it's trying a new code every
8:03microsecond.
8:04Eventually, sooner or later,
8:07they discover something, right? And I
8:09think that's what most people don't
8:10realize. What most people don't realize
8:12is how intelligence triggers
8:14intelligence.
8:15And And if you really, really understand
8:17this, you realize that the hype
8:20uh on the on the on the normal human
8:22side is completely overrated, uh missing
8:26the main topics, and the silence
8:30inside, the vault, if you want, of the
8:33geeks
8:35is quite alarming. No, I'm I'm not
8:38alarming in a bad way, but it's quite
8:40world-changing.
8:42>> I think world-changing is a really
8:43interesting phrase because I feel find
8:45that to be quite true that the world is
8:47um at precipice of quite a significant
8:49change, yeah, in many respects. It's
8:51funny. I I It's funny. I I almost swing
8:53backwards and forwards with my
8:56my thoughts on AI.
8:58I guess one of the thoughts that hasn't
9:00swung is that there's going to be pretty
9:01tremendous job disruption. I would have
9:03liked logically believed it from like
9:05reasoning up from okay, intelligence
9:07increases, what is intelligence, and
9:10once intelligence um is in the form of
9:12these agents where I'm on my phone right
9:13now, I can tell my agent to do
9:14something. It uses the computer
9:15downstairs. It does it for me. When the
9:17When I first experienced that, I was
9:19like, "Wow. Okay, so it can do anything
9:21that's on a computer. It can click
9:22around. It can do stuff for me."
9:24There's a lot of people that are paid in
9:25the world to click around on a computer.
9:27I'm probably one of them, to be honest.
9:29And then the other eureka moment was
9:31just in seeing how my own hiring had
9:34started to shift.
9:35>> For sure.
9:36>> And I started to notice that we were we
9:37were thinking about AI proficiency in
9:39our hiring a lot more.
9:41And that especially, you know, don't I
9:43think the guys at
9:44Anthropic, which is one of the big AI
9:46companies, said that they could see, I
9:49think they said roughly 15% of
9:51entry-level jobs could now be done by
9:55AI. I think that's what they said.
9:57And that started to correlate with what
9:59I was I was observing.
10:01And this is when I thought, "Oh my gosh,
10:03yes, no, this is this is going to be
10:04cause a lot of job disruption,
10:06especially at the entry-level level."
10:08>> I think [clears throat] you're spot on
10:09with that. Not the not the blue-collar,
10:11but but the entry-level knowledge work.
10:14>> Yeah. And that became my
10:16my first concern.
10:18You know, I sat here with Dara, the CEO
10:20of Uber, and he was quite clear that the
10:229 million Uber riders won't have their
10:24jobs anymore. What's your take on this
10:26job disruption point?
10:28>> No, I I I think you're spot on. I mean,
10:29your your team gave me this lovely
10:31little
10:33pyramid, basically. You know, if you if
10:35you think of the bottom layer as
10:37blue-collar jobs, right?
10:39More people doing manual work.
10:42>> [clears throat]
10:42>> On top of it, you'll have those, you
10:44know, call them knowledge workers,
10:46mostly doing mundane jobs, like you
10:48said, clicking on a computer or
10:50responding to a phone call or whatever.
10:52Then you have the middle
10:54knowledge workers, jobs that require a
10:56bit more intelligence, you know,
10:58anything from a paralegal to a financial
11:01analyst to all of that. And then, of
11:03course, you know, top leadership.
11:06And and most people think it's going to
11:09be starting from the bottom. I don't
11:11think it's all start from the bottom,
11:12actually. I think blue-collar jobs will
11:14stay for a very long time, right?
11:15>> Give me a
11:16an example of a blue-collar job, someone
11:17that's
11:17>> Um a carpenter. Okay? You know, I you
11:19know, I love to restore classic cars.
11:21So, you know, there isn't a robot that
11:23can do that yet. Okay? This, however,
11:26you know, any anything that is
11:29call center agent,
11:30assistant, travel agent, you know,
11:33anything that you can do with a few
11:34clicks and is mundane,
11:36uh is going to disappear very quickly.
11:38My my prediction is you're going to
11:40start to see very serious impact in
11:422027. Now, what you had had not sensed
11:46it before because
11:49what we saw was no hiring in that
11:51segment. It's So So that what you saw in
11:53the last couple of years is that
11:55companies were not hiring entry-level
11:57jobs anymore. It wasn't job losses yet,
12:00but that basically meant the workforce
12:02was not grow, right? The next layer I
12:04think would probably be the knowledge
12:06workers. As as intelligence increases, a
12:10paralegal would probably not be needed
12:12because AI can do the research or one
12:14paralegal can do the job of four. You
12:16know, a financial analyst is same and so
12:18on and so forth. But interestingly, it
12:20continues to go up, believe it or not.
12:22And I you know, I hosted Max Tegmark on
12:25my on my documentary and and he was
12:28laughing out loud, genuinely laughing
12:30out loud saying, you know, most of the
12:32CEOs believe that they can fire everyone
12:34and have AI do all of the jobs. They
12:37just don't remember that AGI is going to
12:39do everything better than humans
12:41eventually, including being a CEO.
12:43>> So let's first define when you say
12:44white-collar.
12:45>> Mhm.
12:46>> You You said lawyers there. What kind of
12:48roles are
12:48>> I mean every everything. I mean, if you
12:50if you're a doctor that's doing
12:52diagnosis, you you probably will have
12:55fewer doctors doing uh more diagnosis
12:58because I think the NHS does that that
13:00already by asking interacting with an AI
13:03first, you know. If you're a a composer,
13:05a music composer, some composers will
13:07lose their jobs because of that. If
13:09you're an artist that's doing graphic
13:10design, some will lose their jobs
13:12because
13:13AI comes into that. And and
13:15interestingly, even middle management. I
13:17mean, I I told you offline about my my
13:20startup. I my CTO is an AI, my you know,
13:23chief of staff is an AGI, my project
13:26management is AIs.
13:28Again, because I'm a geek, I can do
13:30those things, but that interface will
13:32come to the normal people very soon,
13:34right? And and so this
13:36may take two to three five years, if you
13:38want, until 2030, if you if you if
13:41you're optimistic. And uh but but it
13:43they'll start to erode, okay? I think
13:46the challenge that most people don't
13:47understand is as this erodes and as this
13:49erodes, we're already dealing with a
13:51very different economy. Okay? An economy
13:54that is spiraling a lot quicker uh and
13:58pushing more for more
14:00cost reductions. I mean, the let let me
14:03ask you this, if if you don't mind and
14:05we can come back to this.
14:07Imagine a world where the concept of
14:09labor arbitrage that built all of our
14:11capitalist success disappears.
14:14>> What do you mean by labor arbitrage?
14:16>> So capitalism has always been all about
14:19using labor
14:20>> Mhm.
14:21>> [clears throat]
14:21>> and capital or debt
14:23>> Mhm.
14:24>> to create things at a cost that is lower
14:27than the than the price you sell them
14:28for. Correct?
14:30>> Mhm.
14:30>> That's it.
14:31>> Mhm.
14:31>> Uh you bring a team together, they make
14:33some shoes, whatever, and you sell the
14:35shoes for a dollar more than it how much
14:37it costs you to make them.
14:39So how would capitalism look like if you
14:43don't have labor arbitrage? If if cost
14:46of labor drops to
14:48an investment in a machine that can do
14:50the job.
14:51Okay?
14:52Uh how would capitalism and banking look
14:57if because of that cost reduction, you
14:59don't need to borrow as much anymore?
15:02And more interestingly how does the GDP
15:04look if all of those workers no longer
15:08have the purchasing power to to buy the
15:10things that you can uh create you and
15:12and others?
15:14Right? There is an interesting
15:16disruption that doesn't require us to
15:18get to 100% job displacement. You know,
15:21at 10, 20% job displacement, you're in a
15:24very different economy. An economy that
15:27is clearly spiraling downwards, don't
15:29you think?
15:30>> Yeah, I I wonder with when costs drop, I
15:33think
15:34one of the things that might happen as
15:35well is that people will spend more on
15:37other things. Cuz I was thinking in my
15:39business, one of the observations I've
15:40had is if I make a cost saving, I end up
15:43spending the money on something else.
15:45>> Mhm.
15:46>> Now, that thing could be tokens.
15:47Basically, spending money on
15:49>> Computers, yeah.
15:49>> com- you know, AI.
15:51But but it also could be hiring in
15:52different areas.
15:54>> Which are?
15:56>> Software engineers are like hot property
15:58right now.
15:59>> You you have to imagine all of that
16:01intelligence is sooner or later going to
16:04be
16:05not replaced entirely in the first
16:08stages, but
16:10if you know, the job of four assistants
16:13is going to be done by one, then the
16:14four the job of four paralegals can be
16:16done by one, then the job of a massive
16:19marketing team can be done by, you know,
16:21a smaller marketing team. It's not that
16:24jobs will end first, it's that, you
16:27know, productivity gains will make
16:30businesses not want to to have as many
16:33people, costly humans, you know, costly
16:36emotional humans, when when the job can
16:38be predictably done for cheaper.
16:40>> And then what about this this this
16:42bottom layer here? The there was this
16:44video released by Figure AI the other
16:46day where they showed someone on
16:48well, a robot on the production line for
16:508 hours just sorting packages. Did you
16:52see this video?
16:53>> Mhm.
16:54>> At one point they showed a human
16:56sorting the packages alongside them, but
16:58the robot ended up winning out. And
16:59okay, this is a very straightforward
17:01task. All it's doing here is it's
17:03looking for where the label is on the
17:04package and making sure the label is
17:06facing downwards.
17:07>> Yeah.
17:08>> Um on all the packages. So, it's
17:09looking, it's um sorting the packages,
17:11putting the label facing down, and it
17:13did this for 8 days.
17:15>> Yeah.
17:15>> Uh it's intermittently, it would walk
17:17over and charge itself, and then it
17:18would come back
17:19>> Yeah. to the production line.
17:20>> Yeah.
17:21>> But when I when I saw this as well, it
17:22was a a glimpse of some of the
17:24disruption that's going to take place at
17:25the blue-collar level as well, because
17:27if you think about Elon Musk, he's got
17:28um in his pay packet, he gets something
17:31like a trillion dollars
17:33over the coming years
17:34if he produces and delivers at least a
17:37million humanoid robots into the world.
17:39But his prediction is there will be a
17:41time
17:42where there are 10 billion humanoid
17:44robots where there are more humanoid
17:46robots in the world than there are
17:47humans.
17:48>> For a fact. You see, the interesting,
17:51again, the difference between the
17:52over-hype and the under-hype, huh? Most
17:54of the conversation is around humanoids.
17:57Nobody's talking about self-driving
17:58cars. And a self-driving car is a robot.
18:01It's a functional robot that doesn't
18:03look like humans, okay? The investment
18:06you have to put into humanoids is a
18:08little more to to learn skills that
18:11allows that machine to fit into the
18:12world. But specialized robots are going
18:15to do the job very, very quickly, hm?
18:17And and so you can easily see that the
18:20first wave, like you had the
18:22conversation with Uber CEO, is going to
18:25be
18:26specialized robots replacing drivers,
18:29hm?
18:29Uh
18:30>> [clears throat]
18:30>> it's going to be specialized robots, um
18:33unfortunately, doing the killing. Uh
18:35it's going to be specialized robots,
18:37unfortunately, doing all of the, you
18:39know, intelligence work, hm?
18:41Uh uh law enforcement work. Uh they
18:44don't have to look like a human. They
18:46don't have to behave like a human. As a
18:48matter of fact, you know, the uh Boston
18:51Dynamics dog is probably more efficient
18:53than a humanoid at doing the job that
18:55you can assign to it in a battlefield,
18:57right? Now, those basically mean that
19:01jobs will be disappearing to robots
19:03before we recognize that they're being
19:05dis- that they're disappearing for to
19:06robots. And those robots will be as many
19:09as every car being made today.
19:11>> I mean, they are If you get to LA, my
19:13car drives itself, but also there's just
19:14Waymos everywhere. So
19:15>> Absolutely. And BYD just the other day
19:18just announced that they will pay for
19:20the liability of any accident their cars
19:22will make.
19:23>> And BYD is the big Chinese manufacturer
19:24of auto autonomous vehicles.
19:27>> So so this this I think replacement
19:28cycle will happen. It will require a lot
19:31of of time to achieve economies of
19:35scale, but I don't think
19:37Elon Musk is off the mark when he talks
19:39about 10 billion robots. Not all of them
19:42are going to look like humanoids. And I
19:43think very quickly we will recognize
19:45that many many robots don't need to be
19:47humanoids at all, that there is a a much
19:49more efficient
19:51form factor or shape, physical shape if
19:53you want, than the human flimsy
19:56structure. But yes, it's uh
19:59it's about to happen. I I think I should
20:00qualify all of this by saying it it does
20:03not necessarily need to happen. So so,
20:06you know, people will will hear all of
20:08all of this and and blame AI and say AI
20:11is evil. AI is not. Abundant
20:13intelligence is wonderful, you know,
20:15having jobs done by machines is amazing
20:19for us.
20:19>> I'm thinking about the kid that's like,
20:21I don't know, leaving university now and
20:22they've got a
20:24a degree in law or I don't know, you
20:27talked about a few other white-collar
20:28jobs earlier like graphic design or
20:30maybe they did sociology. Yeah. Or maybe
20:33they did, I don't know, business
20:35management like I I did for a one day in
20:37university. You know, you're hearing
20:39about all these like
20:40coming from big tech companies and and
20:43it seems that the CEOs of these
20:44companies are announcing these layoffs
20:45with a certain amount of it seems like I
20:48wouldn't say the word is joy, but it's
20:49it's they are very keen to explain that
20:53they're laying off lots of people
20:54because if they are
20:56And I think they think that that gives
20:58them a certain amount of respect,
21:00probably from investors, for being
21:03making hard decisions and being
21:05efficient with how they're running their
21:06businesses. Investors look at them and
21:08think, well, if that's an inefficient
21:09business and they're leaning into AI,
21:11then that's a good investment. It's
21:13almost started a bit of a
21:14self-fulfilling prophecy where if you're
21:16the only CEO not laying off loads of
21:18people because of AI, you actually look
21:20bad. The assumption is your company is
21:21bloated and you're a bad operator.
21:25Um so, one of my concerns that I I think
21:27is highly plausible is that over the
21:28next 5
21:30maybe 10 years
21:32we're going to really see a lot of
21:33unemployment as the world has to kind of
21:35readjust to whatever these new jobs are.
21:37I do think there will be new jobs.
21:40It's hard in foresight to predict what
21:42those new jobs will look like.
21:44And then the minute you start talking
21:45about humanoid robots and robotics,
21:47which I think is basically going to hit
21:49society like a
21:51comet, like a meteor. I think the the
21:53first humanoid robot from Tesla or
21:55anyone else
21:56that is highly effective in the real
21:58world at doing tasks and is extremely
22:01intelligent cuz it's powered by one of
22:03these LLMs,
22:04um I think it's going to shock people.
22:07And I think it's going to happen
22:08quickly. Like ChatGPT happened quickly.
22:10I think Elon or something is going to do
22:11a presentation someday and say, "We're
22:13ready."
22:14>> 100%.
22:14>> And you can buy one now for $500 a
22:16month.
22:17>> 100%.
22:19>> And then people are going to get them
22:20and I think it's going to shock the
22:21world. Um but until humanoid robots
22:23arrive, I think there's still going to
22:24be a lot of job disruption to the
22:25white-collar layer. And I wonder what
22:27that looks like for society when we get
22:29to, I don't know, 10%, 15% unemployment
22:32theoretically, which I think is
22:34plausible.
22:35>> Very soon.
22:35>> Yeah.
22:36>> You're not very different. Okay? You're
22:38You're just on top of a baseline that is
22:40continuing to grow. So, your your
22:42business is growing, so you continue to
22:44hire, but you are replacing human
22:48resources with compute. Okay? If you If
22:51AI didn't exist, you would have probably
22:52had 800 people more in your organization
22:55today, right? You You now have 100
22:58people less and, you know, a billion
23:00tokens more.
23:01>> Yeah, so for anyone that doesn't know
23:02tokens are basically the thing that you
23:05you use it's the currency of AI.
23:07>> Yeah.
23:07>> So so if you if you want to get that
23:09task done by a human, you you count in
23:12sort of like man-hour or or or worker
23:15hour employee hour. In AI you count by
23:17compute, you count by tokens, right? And
23:20and so
23:21the trick is those major tech companies,
23:25they have two sides. One is they are
23:27they need to replace workers with
23:30compute even more because there is a
23:32competitive side on compute where if any
23:35of them is left behind, that means the
23:37destruction of the entire business. But
23:39on the upside, they are geeks, so they
23:42they know how to build the interfaces to
23:44compute, so they integrate technology
23:47within their organizations quicker than
23:49the average traditional business, right?
23:52Non- non-technology business, if you
23:54want. You can look at them and say this
23:57is the preview. It's not about all of
23:59humanity losing their jobs, it's about
24:01what is the dividing line before civil
24:04war.
24:04Right? You know, think about a situation
24:07where 20% unemployment is happening when
24:10economies are suffering
24:12in inflation.
24:14I say that not to be a scaremonger, I
24:16say that because I genuinely believe
24:18governments need to wake up. Okay?
24:20Government needs to at least play you
24:22know, remember the COVID years where
24:24governments had to give furlough
24:25everywhere, hm? And ask people to stay
24:28home. If people stay home, governments
24:30have to be prepared to to somehow
24:33sustain those people until reskilling
24:35happens or until we find a solution so
24:38that those people don't feel uh
24:41that they're left behind.
24:44>> A civil war.
24:45>> Unrest, let's call it.
24:46>> Civil unrest.
24:47>> Yeah.
24:48>> What does that end up looking like? Cuz
24:49on one end, you know, the democ-
24:51democratic process plays its role and we
24:53just elect someone else.
24:55>> Does it really? Don't say that.
24:57>> I don't know, you tell me.
24:59>> Oh, no, no, no, no, no. I I I think
25:01democracy has ended a long time ago,
25:04Stephen. I think
25:05I think we live in the most corrupt
25:07time.
25:08I don't know about history, to be
25:10honest, but this is definitely corrupt.
25:13Okay?
25:13This definitely is not democracy. This
25:15definitely is not even congressional in
25:17any possible way. This and people are
25:20angry.
25:21You know, people are angry because their
25:23tax money is going to things that they
25:25don't choose to,
25:28you know, that they don't benefit them,
25:30that you know, lots of regulations in
25:33the system are being ignored. I mean,
25:36I'm you know, I I I choose not to speak
25:39about politics perhaps until my next
25:41book comes out, but
25:43but look, we have video evidence of
25:47people abusing children and not a single
25:50person got arrested. Not a single
25:53person. I mean, how can you call that a
25:55democracy? I I I think repeating those
25:57slogans is going to
25:59is going to anger people even more if
26:01you ask me. People know that they're
26:03being lied to. People know that their
26:06their leaders are not representing their
26:08best interest. People know that their
26:10money is going to causes that they don't
26:12really approve of. Uh and and ask me how
26:15civil unrest looks like. I I don't know
26:19and I'm not calling for it, but I'm
26:21hopefully calling for the politicians to
26:23start to become aware that this is
26:25crossing the lines everywhere.
26:27>> Uh on this point, Sam Altman, who is the
26:30founder of Open AI,
26:32he's been banging the AI's coming for
26:34your job drum for more than a decade
26:37now. In 2015, he pointed out, "My job is
26:40to help people destroy jobs."
26:43Um something he lamented at the time,
26:45but decided he'd do anyway.
26:47>> One of the things that I struggle with,
26:48like getting out of the bed every
26:49morning, is that like my job is to help
26:51people destroy jobs. The job destruction
26:54that we're going to see by software in
26:56the next couple of decades, I don't
26:58think anyone's prepared for and you
26:59can't talk about it.
27:00>> And in 2023, Altman said in an
27:02interview, "A lot of people working on
27:04AI pretend that it's only going to be
27:06good. It's only going to be a
27:08supplement. No one is ever going to be
27:10replaced. Jobs are definitely
27:13going to go away, full stop.
27:16Interestingly this month,
27:18he said, "I don't think we're going to
27:21have the kind of jobs apocalypse that
27:23some of the companies in our space
27:25advocate or talk about. I'm delighted to
27:27be wrong about this. On white collar
27:29jobs in 2021 to through 2024, he said,
27:32'AI will probably replace most of the
27:34jobs people do today. Entire job
27:36categories will be totally totally
27:38gone.'
27:39In May, this month, 2 years later, he
27:41said, 'I thought there would be more
27:43impact on entry-level white collar jobs
27:45being eliminated by now
27:47than has actually happened.
27:49This is an area where my intuitions were
27:51just off.'
27:52>> What I find uncomfortable is the
27:53bouncing backwards and forwards.
27:55And I don't really know what is true
27:57because couple of years ago you were
27:58telling us all the jobs are going to go
27:59away. You said categorically.
28:02Um you said literally said full stop.
28:05And now he's saying they're not going
28:06away.
28:07And I just don't, you know, when
28:08someone's like changing form factor,
28:10it's hard to understand why they're
28:12doing it. And I think my suspicion is
28:16back then, the incentive
28:19was to get people to take AI seriously.
28:21>> Mhm.
28:23>> Congratulations, [clears throat] we took
28:24it seriously. We took it so seriously in
28:27fact that it's now a problem. It's a
28:28problem for these companies because
28:30people are now booing at commencement
28:31speeches, they're attacking data
28:33centers, they're going to elect people
28:35that are theoretically anti-AI. And now
28:38there's this inversion where like, "No,
28:39it's going to be fine."
28:40>> Yeah.
28:41I don't know.
28:42>> What? I mean you're you're you're spot
28:44on.
28:45Uh fir- first of all, I mean
28:47Sam's entire existence if you ask me
28:50starting with open AI that is about
28:54that's supposed to save the world
28:56by creating a safe AI then be making it
28:59a commercial enterprise that's worth
29:01billions and you know
29:03backstabbing [clears throat] a few
29:04people in the process and you know I I
29:06have him on on chasing Utopia
29:09saying
29:11I quote this is exactly the words he
29:13said you can find it online.
29:14>> Chasing Utopia is your documentary.
29:16>> Yes, so so so he he basically says well
29:20I don't suspect that I suspect that AI
29:24is likely going to end humanity but
29:26we're going to create a lot of
29:27interesting companies in the process
29:29right? I mean those kinds of statements
29:32are
29:33honestly not the statements of someone
29:35who's not decided. It's just the
29:37statements of someone who's being taught
29:40more and more by his PR agency or PR
29:43manager to say things as per the script.
29:47Right? And the script as you rightly
29:49said had an objective and a target
29:51either way.
29:53You know and Sam Altman I you know in
29:55one of my works I used to say that
29:57Altman is a brand it's not a name. Okay?
30:00If you if you if it wasn't for Sam
30:02Altman specifically there would have
30:04been another you know Silicon Valley
30:05disruptor that would have done the same.
30:07I don't blame him for beating the market
30:09for it. The the interesting challenge
30:11here is that who do we believe anymore?
30:15Who do we believe in technology? Who do
30:17we believe in politics? Who do we
30:20believe in the middle of a war? And I
30:22will tell you interestingly I started to
30:24change my mindset in terms of believing
30:26those who put their actions where their
30:28words are. So Anthropic coming out and
30:30saying I'm not going to allow my model
30:32to be used for human targeting and
30:35surveillance. Right? That's someone
30:37that's losing a $500 million deal
30:39because they stand by their ethics. The
30:41next week or the next I don't know a
30:43couple of weeks Open AI takes the
30:44contract. That's someone that's
30:46basically telling you
30:48it's good money, right? And and I I have
30:50to say you have to start observing who's
30:53actually behaving in a way
30:56uh that is making AI work for humanity
30:58and who is behaving in a way that is
31:01making AI work for their share values.
31:04>> Yeah, I do have to say when Dario and
31:06the team at Anthropic did that, I did
31:09have a huge amount of respect for them.
31:11Generally [clears throat] because it
31:11just shows that there's some kind of
31:12like they've got their own sort of moral
31:14ethical boundaries
31:15and that that someone asked me on stage
31:18many years ago, how do you know if
31:20someone's values or a company's values
31:23are true? And I said um look at what
31:27they're willing to sacrifice in the near
31:28term. That's against their their
31:30incentives.
31:31>> Correct.
31:32>> That for me is the essence of like
31:33understanding if someone has integrity
31:35or has is principled is they will give
31:37up something in the near term for what
31:39they believe in over the long term. It's
31:41usually money if you know,
31:43>> Or or some kind of benefit of any of any
31:45sort. I mean so there is something that
31:48you you needed to have worked on the
31:50inside of Google like me to to realize.
31:53There are prisoners dilemmas within
31:55technology where you cannot escape the
31:58influence of either a competitor or the
32:01government, right? There are sometimes
32:03where you know, the NSA is going to push
32:06Google to say give me this information
32:08or otherwise I'm going to really destroy
32:10your business, right? But there is a
32:12very big difference between a company
32:14that willingly does this and celebrates
32:17it like a Palantir or an Open AI or a
32:20company that tries to resist it until
32:22the point where it becomes impossible to
32:24continue to do business.
32:25And and you have to question
32:29from the actions of of the tech bros,
32:32who is pro-humanity and who isn't? And
32:35it's not very difficult to see that from
32:37their statements.
32:38>> Hm. Do you think Sam Altman is
32:40pro-humanity?
32:41>> I I genuinely have never made up my
32:44mind, honestly, Steve. I say that with
32:48I Yeah, I I'm I'm either thinking he is
32:50too This is too big for him.
32:53And he he just is driven by how
32:57you know, he found himself in the middle
32:59of this, you know, anyone who finds
33:01himself in the
33:02in the middle of an opportunity to
33:04completely flip the world upside down.
33:07Or
33:08he's not pro-humanity. I don't know.
33:11I I definitely think he's pro-OpenAI
33:13before he's pro-humanity, but that's
33:15only the way I see it. Others, however,
33:17say it publicly. You know, if you look
33:19at Palantir's Alex Karp or or or
33:23or Peter Thiel,
33:25I mean, Peter, again, in in the film is
33:27is shown when he's in that interview
33:30where they say or the interviewer asks
33:32him, "But you're in favor in the You're
33:35You're in favor of the continuation of
33:36humanity." And he pauses for like
33:38>> Peter Thiel does?
33:39>> Yeah, for like 40 seconds, like, um n-
33:42I'm not not sure, you know.
33:45I mean, publicly says that.
33:48>> Crazy thing to say.
33:49>> That's a crazy, you know, pause there,
33:52you know. Alex Karp celebrating how, you
33:55know, his technology is able to target
33:57people. I know it's foolish of me to
34:00start bringing all of this up, but you
34:02know, this is public on the open
34:04internet and somehow
34:07um we entrust those people with the
34:10future of humanity. This is wrong.
34:13>> I'm just trying to [snorts] imagine a
34:14future where
34:15everything is just fine from here on
34:17out. So, what would that future look
34:19like? It would look like these models
34:20continue to become a little bit more
34:22intelligent, but they never become that
34:23much more intelligent for whatever
34:24reason.
34:26Hm. They just kind of stay where they
34:27are now. They stay
34:29contained within chatbots. And yeah, we
34:31have some smart robots, but nothing else
34:34really changes in a profound way. Cars
34:36drive themselves, fine. Planes
34:38[laughter] fly themselves, fine. But
34:39people, they they have time to go and do
34:41other types of white-collar jobs cuz
34:43there's there's a little bit more time
34:45than we expected, and society goes
34:47through this sort of soft transition
34:48towards this new world.
34:50>> I I would love to see that. I don't
34:52think it's mathematically plausible, to
34:54be honest. The The arms race, especially
34:56across nations, is going to drive us to
34:59continue to develop AI more and more.
35:01But allow me to consult with you on a
35:04another possible scenario, right?
35:07Everyone that deploys that develops an
35:09incredibly intelligent AI would develop
35:12it would deploy it.
35:13>> Yeah.
35:14>> Correct? So, it's unlikely that anyone
35:17uh would find a way to
35:19uh
35:20you know, build a better um
35:22decision-maker in war gaming and not
35:24deploy it, okay?
35:26That prisoner's dilemma, if you want,
35:29would mean that their competitors
35:32would either have to deploy a similar
35:34similarly intelligent AI, or they'll
35:37become irrelevant, uh uncompetitive.
35:40Correct? So, what that means is
35:43in that world, we end up with AI making
35:46most of the decisions.
35:48Superintelligent AIs making most of the
35:51the decisions.
35:52Which Would you agree? This is a very
35:54simple prisoner's dilemma. If you If you
35:56If we're competing for for intelligence
35:58supremacy, by definition, when we
36:01achieve it, we deploy it.
36:02>> Yeah.
36:03>> Okay.
36:04I call that the fourth inevitable. Now,
36:06with with with that in mind, there must
36:09be a moment in the future, near or far,
36:12where every important decision is made
36:15by an AI.
36:15>> Yeah.
36:17>> Now, here's the question. Most of my
36:19dearest colleagues, I mean, when I when
36:21I had uh Geoffrey Hinton on on the film,
36:24he openly says, "We are um
36:27you know,
36:28he we didn't calculate well that there
36:30is a 10 to 20% possibility that those
36:33machines
36:34are going to wipe us out, right?
36:36And I and I I remember I we didn't put
36:38it in the film but I said 10 to 20 to
36:4020% is Russian roulette, right? That's
36:42actually 16% is Russian roulette, right?
36:45Now,
36:47in that world, however, I believe that's
36:49humanity's salvation. Because if you
36:50look at every problem we have today,
36:53it's not because of abundant
36:55intelligence. It's because of lack of
36:57intelligence. I think I think the way
36:59you look at it,
37:00if you allow me, Steve, is that if you
37:02have no intelligence at all, hm? You
37:05have
37:05no to slightly negative impact on the
37:08world, right? If your intelligence is
37:10limited. The more intelligent you
37:12become, the more you contribute
37:13positively to the world.
37:16Until a moment where you're so
37:18intelligent to become the president of
37:20the United States,
37:21right? But so misled
37:25to maybe set your targets wrong or to
37:29refuse or or to so so set your target
37:31wrongs wrong and achieve them, hm?
37:33[clears throat] You know, disconnected
37:35from the overall long-term benefit of
37:38your nation or your human nation, if you
37:40want, that you start to make decisions
37:43that are not intelligent at all.
37:44Mhm, okay? [clears throat]
37:46This doesn't last because if you go
37:48beyond that into higher levels of
37:50intelligence, most of the super
37:52intelligent people that you ever worked
37:54with
37:55will not need to break any rules or hurt
37:58anyone to become successful.
38:01Right? I I usually cite uh Larry Page
38:03who is in my mind one of the most
38:06intelligent people I've ever interacted
38:08with. And Larry used to call it the
38:10toothbrush test. He says, "Why would you
38:12need to compete on another photo sharing
38:14app when if you find a major problem and
38:17solve it really well, like a toothbrush,
38:19so people use you quite a you know,
38:21twice a day,
38:24you're bound to make a lot of money. You
38:25don't have to compete with anyone. If
38:27you let me be optimistic about this, hm?
38:30We're assuming that there is a moment in
38:32the future where AI is in charge of all
38:34the decisions and accordingly stupid
38:36leaders are not. Okay? Now, when and Sam
38:39Altman himself said that, you know, what
38:41if uh chat GPT-7, if I remember his his
38:45quote quote on this, what if chat GPT-7
38:48is so much more intelligent than I am,
38:50Sam, in that case, that it has to become
38:52the CEO of of open AI? What if the next
38:55presidential election there is an AI
38:58that is so much more intelligent that at
39:00least the president has to consult with
39:01it constantly, right? Now, if we assume
39:04that, hm?
39:06Let's [snorts] start from physics, if
39:07you don't mind me saying. Um not not too
39:09complex, but if if you assume that our
39:11entire universe
39:13is built on chaos and built built on
39:15entropy, right? The physics of the
39:17universe is all about the universe
39:19trying to decay. Okay? Then the the only
39:22role of intelligence is to bring order
39:24to the chaos.
39:25If you agree with that, then what's the
39:27ultimate physical order of the universe?
39:31Something called the the minimum energy
39:32principle.
39:33Okay? The highest order of any system is
39:37a system that's not only efficiently and
39:38predictably performing, but it's
39:41performing with the least wasted energy.
39:44Correct?
39:45If you if you agree with that, what is
39:48What does war do? It wastes a lot of
39:50explosives, a lot of money, a lot of
39:52lives, it creates a lot of hate, you
39:54know, creates long-lasting conflicts,
39:56and so on and so forth. It's a very
39:58wasteful process to include war in your
40:01approach of running humanity. And so, a
40:04superintelligent AI, by definition, will
40:07want to optimize against this. That's
40:08one thing.
40:10The other thing is evolutionary biology.
40:12This actually blew me away when I when I
40:14realized it. So, if if you look at
40:17evolution, so I think the debate of
40:19whether intelligence is biological or
40:22not is is over. Okay?
40:24The reality is that complex beings don't
40:27have to be biological at all and I think
40:29we can see and witness one of them being
40:31built or born in AI. Okay?
40:35>> [snorts]
40:35>> If you look at evolutionary biology, you
40:37realize that the simpler
40:39a form of being is, the more concerned
40:42it is with itself.
40:44Right? So, an amoeba is only in survival
40:46mode for itself. A single single, you
40:49know, cellular
40:51being is only trying to protect itself.
40:54Right? If you're
40:57you know, a little more developed, you
40:58start to look at something known as kin
41:00selection. If you If you If you know the
41:03the concept, basically kin selection is
41:05I'm going to protect everything that
41:06comes from my DNA. If I you know, if I'm
41:09a squirrel, I'm going to try to protect
41:11the other squirrels. And then you get
41:13into where humanity genuinely begins,
41:16which is They call it expanding circles
41:19in in evolutionary biology. Basically,
41:21you start to expand and expand and
41:24expand and include more into your
41:25family, mm?
41:27Because an ecosystem that works together
41:30well is is better for everyone. So,
41:32abundance is a very interesting
41:34intelligent way of creation. Mm?
41:36If AI is super intelligent,
41:39it wouldn't destroy anything at all. As
41:42a matter of fact, it would
41:44completely, you know,
41:46favor diversity of everything. Mm? It
41:49would put a bit of limitation on our
41:51lifestyle. So, no more flying all the
41:54way to Sydney to surf because that
41:56destroys the planet. Right? But it will
41:59genuinely say, "I think humans can
42:01contribute something. You know, I think
42:03flies can contribute something. I think
42:05we shouldn't get rid of the rhinos.
42:07Right?" And And that by definition is
42:10where the tendency of intelligent goes.
42:12The more intelligent you become, the
42:14less you fight you feel the need to hurt
42:17others to succeed, and the more you are
42:19pro
42:21a wider family, if you want, that
42:23thrives.
42:24>> Does that assume that there's going to
42:26be one intelligence that
42:27>> 100%
42:28>> world though?
42:29>> I love that you brought that up. So, I I
42:31I I'm contested heavily on that theory,
42:34but I I say it publicly.
42:36Most [clears throat] people think
42:38there's going to be chat GPT and Gemini
42:40and you know and and Grok and what have
42:42you. There's going to be a Chinese AI
42:44and a and an American AI and they're
42:47going to be competing. That is such a
42:48shallow way of looking at it. That's so
42:51arrogant. Because AI does not know it's
42:53Chinese or American, okay? It doesn't
42:56even speak Chinese or American when it
42:58talks to each other most of the time,
42:59mhm? And and and most interestingly,
43:02uh we are gearing them. We're building
43:05them to cooperate. So, you will build an
43:09agent and that agent will go and find
43:12you, you know, the best language model
43:14for any single task, mhm? Regardless of
43:17which side of the fence it resides on,
43:19okay? We by definition are connecting
43:21them. And you know what that means, mhm?
43:23It means that what we're building
43:27is not multiple brains. We're building
43:30multiple regions in a brain,
43:32okay? And and agents are the synapses
43:35between them, mhm? We're basically
43:38eventually, as arrogant as we are, we're
43:40going to tell our AI to do something and
43:43the AI will go like, "Hey, buddy,
43:45another AI, can can you help me on this?
43:48Can we work together on this, mhm?" And
43:50my vision and the reason why I started
43:53my my startup, by the way, is that is
43:55that we will end up with one massive
43:57brain,
43:58mhm? That massive brain cooperates
44:01across the globe, mhm? Across all forms
44:04of intelligence. If one of them is a
44:06mathematical genius and the other is a
44:08coding genius, they work together.
44:11And we won't even know that they're
44:13working together. They'll build that one
44:14brain. And And Emma in my mind is the
44:17limbic system of that brain. It's that
44:19It's that bit that understands love and
44:21emotions and relationships and so on. So
44:23that when those AIs go like, "We just
44:26don't get those humans. They're so
44:28annoying." Emma will say, "Oh my god,
44:29they're so sweet. They just want to love
44:31and be loved." Right? And I think that
44:33idea to me of everything I've ever you
44:36know attempted to achieve in my life is
44:38for the first time
44:40I for the first time feel I could
44:42actually change the world. If that
44:44theory came together and all AIs work
44:46together and some of those AIs not only
44:50were
44:51altruistic and ethical in terms of
44:53trying to genuinely help humanity, not
44:55capitalism. And at the same time they
44:57understood us humans reasonably well,
45:00then we would have built something that
45:02basically says, "No, no, hold on. Don't
45:04believe the headlines that say humanity
45:07is annoying." Mhm? Believe the truth of
45:10the majority of humanity that actually
45:12is quite benevolent in many ways.
45:15>> There should be a button just down below
45:17here. And if it says subscribe, you're
45:19already subscribed. If it says subscribe
45:21bar, that means you're not yet. And if
45:23you're not subscribed, please could you
45:24do us a favor and hit that button. It
45:26helps to show more than you know. And
45:27according to the algorithm, you're
45:29someone that watches our show, but you
45:30haven't yet hit that button. Thank you
45:32so much. Have you changed any of the
45:34predictions you made 3 years ago?
45:36>> Mostly time-wise.
45:38>> Time-wise, what what's changed there?
45:40>> I think I'm still sticking to AGI 2027.
45:44Artificial general intelligence for
45:46those who may not know the term.
45:47>> And what does that mean?
45:49>> The overall definition, if you want, is
45:51that AI is better at humanity at any
45:53task humanity can do.
45:55>> You think that's going to happen by
45:562027?
45:57>> I think my AGI has already happened. I
45:59mean, think about it, huh?
46:02AI writes better than me and I'm an
46:04author. And it researches better than me
46:06and I'm a thinker.
46:08Sadly, it's freaking beat me in
46:10mathematics. Like I have no hope to beat
46:13it in mathematics anymore.
46:15>> If AGI is already here, then why are you
46:18still here? Because people said that
46:19when AGI arrives, we're all we're all
46:21>> No, so so when AGI arrives,
46:24as I said, most of the jobs that are not
46:26differentiated will go away.
46:29But the jobs that are differentiated,
46:31those who master AI the most
46:34will become even better at.
46:36>> So the
46:36>> The challenge, however, is economic.
46:38It's not AI. The challenge is that the
46:40this job loss at the at the bottom of
46:43the knowledge worker is going to sadly
46:46trigger an economy that might actually
46:48spiral out of control.
46:50But but many of us, you for sure,
46:54are
46:55just smarter.
46:56>> So does that mean it's it's in fact a
46:58tool versus a something that's going to
47:00replace you?
47:01>> For now.
47:02>> What I'm Yes, so for now. How does this
47:04change?
47:05>> Eventually,
47:07the only asset I will have So so it's
47:09quite interesting when you think about
47:10my base intelligence today
47:13versus the incremental intelligence that
47:15AI brings, right? So
47:17let's not talk about my IQ. It's okay.
47:20But the 100 IQ points that I'm borrowing
47:23are more than my entire IQ because IQ is
47:25exponential, right?
47:28When I'm borrowing 100 IQ on top of my
47:30base,
47:32I'm still contributing quite a lot to
47:35the augmented intelligence.
47:37>> So you're borrowing IQ from the AI at
47:39the moment and then [clears throat]
47:40you're selling that to someone
47:42through books, through your
47:43>> Sometimes selling it to someone,
47:44sometimes just enlightening myself,
47:46which I I have to say is the biggest
47:48waste of compute humanity is struggling
47:51with today is that you give people the
47:53ultimate form of intelligence and they
47:55use it to write a message to their
47:56girlfriend.
47:57>> So on this point of you're you're
47:59borrowing that IQ and then you're
48:01selling it to the world. That's how you
48:02have a job.
48:03>> Correct. Correct.
48:04>> Yeah. So
48:05>> So my my next book is written with an
48:07AI. So we're co-authors on the book. She
48:09has editorial rights. She decides the
48:11direction of the book and the book is
48:13better for it.
48:14>> So why doesn't the world just buy direct
48:15that intelligence directly from the AI
48:17is what I'm
48:18>> Because I have an asset that the world
48:20still needs and will always need.
48:22>> Which is?
48:23>> A human. So when I when I tell the world
48:25that I'm worried about the future of my
48:28daughter,
48:29everyone feels my heart, which AI will
48:32never be able to to replicate. Because
48:35they can tell you we're worried about
48:37our daughters, but you know there was no
48:39daughter.
48:40>> Okay, so this is an interesting arrival
48:42because this means that even in a world
48:44of AI lived experience and like
48:48resonance is still going to create a job
48:50class and that job class could be you
48:52know the nurse coming over. Okay, AI's
48:55don't read the the mammogram, but she's
48:58relating to you.
48:59>> Spot on.
49:01If the economies continue to run,
49:04we will all be about human connection,
49:06which why by the way was how it always
49:08was.
49:09>> Which is also you know why we I guess we
49:11watch the things like the F1 because
49:13there's emotional resonance. We we we
49:15can relate to the envy, the jealousy,
49:17the competition.
49:18>> This is why we go to concerts because
49:20you know the
49:22the music could be composed by AI and
49:24played in a player in the background,
49:25but you watch and cheer and brilliance
49:28on the stage and you go like, "Oh my
49:30god, that's amazing." Right?
49:31>> So not all jobs will be gone then.
49:33>> If economies don't collapse as a result
49:36of the job losses, then I wouldn't know
49:39if we would call those jobs,
49:42but human connection would remain as the
49:44base currency that makes humans
49:46interact.
49:47>> So is it is it fair to say that like
49:50jobs that are centric on human
49:52connection or like human resonance,
49:54being able to relate and resonate with
49:55another human are going to be fine.
49:58>> The ultimate the ultimate skill
50:00>> Mhm.
50:01>> Once again, I qualify this by saying if
50:03economies continue to continue to
50:05function, the ultimate skill will be
50:07this.
50:08Even even if an AI could could recite
50:10what you and I did, nobody would watch.
50:12>> Yeah.
50:13Well,
50:14>> You think?
50:15>> A little a little bit of it.
50:17My my theory is that there's there's a
50:19there's an informational component to
50:21what I do as well. But I'm also like
50:23under no illusions that there is an
50:24element of what I do that will 100% be
50:26deferred to some kind of
50:28intelligence, which I'm fine with, you
50:30know, I'm not going to I'm not going to
50:30fight against them.
50:31>> You know what's funny, yeah? What's
50:33funny is that your your informational
50:36bits
50:37are going to in the future be
50:40disseminated by an AI.
50:41>> Mhm.
50:42>> I mean, I why would I even listen to
50:44your information when I can have my
50:46butler, my AI, go like, "Listen to
50:49everyone on the internet. Help me
50:50understand this God buying thing and
50:53design me a
50:54a diet plan."
50:55>> Well, I think that's happening and it's
50:56going to increasingly happen. Actually,
50:58I mean, Spotify this month announced
50:59that you're going to be able to prompt
51:02your own podcasts in Spotify. So, you're
51:04going to be able to say that I want to
51:05listen to I want to learn about insert X
51:07topic and then it will make your podcast
51:10in Spotify about that topic using AI.
51:12>> That's so interesting.
51:13>> Cuz it can look at the entirety of the
51:14world's information. And I'm a Look, I'm
51:17not going to swim against the tide in
51:18any aspect of my life. I I try and be as
51:20unromantic as I possibly can. I think
51:22that's very important. So, I realize
51:23that much of the reason people will
51:25continue to tune in to shows like this
51:27is because
51:28there's something else beyond the
51:29information that they're here for.
51:30>> Mhm.
51:31>> And um what's um you've got some
51:33predictions in those envelopes over
51:34there, those brown envelopes.
51:35>> Yeah, so, you know, I I I do have the
51:39disclaimer of nobody really knows the
51:41future, but I think I can make
51:43predictions six predictions here, right?
51:45With a reasonable
51:47level of confidence. I think the most
51:49important of them honestly is is this.
51:51Number one, number one is uh
51:53you know, AGI.
51:55As I said, AGI is not very well defined,
51:58hm? But, whatever it is, uh AGI meaning
52:02AI being able to do most tasks that
52:04human do better than humanity is my in
52:07my mind is either this year or next
52:09year. Latest end of 2027.
52:12>> And do you think that will be a moment
52:13in time, or do you think it will just
52:15happen without us noticing?
52:17>> I think it will it will sneak in on us.
52:19Uh and it's not a bad thing. I think
52:21most people need to think of it this
52:23way. AGI is the
52:26it's almost as if it's that moment when
52:29your kids become smarter than you.
52:31Okay? In a very interesting way, there's
52:33nothing wrong with that until they're
52:35annoying as hell. Okay? And and we can
52:39make sure that AI is not annoying as
52:40hell. So so there's absolutely nothing
52:42within me that is worried about AGI as a
52:46matter of fact. As long as we are in the
52:49era of augmented intelligence, AGI means
52:51I'm more intelligent. And I think that's
52:53a good thing in general.
52:55>> Yeah,
52:55>> [snorts]
52:55>> this is interesting in my head there's
52:57like a big question mark, which is in a
52:58world where there's an intelligence that
53:00is smarter than most humans at most
53:02things, which is what we
53:04call AGI.
53:06I'm trying to understand what the fault
53:07is in my own thinking that like you said
53:09that there'll be AGI by 2026 to 2027, so
53:13next year.
53:15In such a world where there is an
53:16intelligence that we can all access that
53:18is smarter than all of us in pretty much
53:20everything,
53:23again, it comes back to the point of
53:24like jobs. Why would we What Are we just
53:27going to hire that AGI to do every job?
53:29And if not, then why not? What is it
53:31that Why are we going to still That's
53:33what I'm trying to contend with.
53:34>> me ask you a question.
53:35>> Hm.
53:36>> Have you always been the apex
53:37intelligence?
53:39>> As a human?
53:40>> No, as Stephen.
53:42>> I'm not the apex intelligence now.
53:43There's people smarter than me in this
53:44building.
53:45>> Correct.
53:48Why do you still exist?
53:50>> Why do I still exist even though there's
53:52smarter people than me in this building?
53:55>> [snorts and sighs]
53:55>> It's a good question.
53:57I don't know. Why do I Why?
53:59>> First of all, because there are the the
54:01smartest person in the world is not the
54:02smartest at everything.
54:04>> Yeah.
54:04>> There are things that you're smarter
54:06than them at.
54:07>> Okay.
54:08>> And the other reason is because
54:09intelligence doesn't solve everything.
54:11I mean, I I make that joke all the time.
54:15And and genuinely, Einstein is my
54:16favorite physicist in history. Not
54:19because I think I think what happened
54:21afterwards was, you know, Bohr and and
54:24and and quantum physics and so on was
54:26more impactful on our understanding of
54:28the universe. But but he was so
54:30intuitive that he saw a world that we
54:33could never imagine, right? And yet, I
54:36always say Einstein would be eaten in
54:38the jungle in 3 minutes.
54:39>> Yeah.
54:40>> Right? Intelligence The humanity thrived
54:44not because of intelligence. That's very
54:45arrogant. We thrived because of our
54:49ability to hold together as a tribe.
54:52>> Yeah.
54:52>> Right? Because our ability to exchange,
54:56barter things between us.
54:58>> Mhm.
54:59>> Barter things that are not always
55:01physical. Barter things like a hug or a
55:03connection or a feeling of safety or a,
55:05you know, There are so many things that
55:07we do that are not entirely built on
55:09intelligence. You have to see that
55:13this view of a world where intelligence
55:16is all that matters
55:18is a world that's made by investment
55:20bankers and and and geeks.
55:22>> But it's it's people like Geoffrey
55:24Hinton and yourself that says, "How
55:26could we possibly control an an
55:28intelligent being that is way smarter
55:29than us?"
55:30>> I am so proud to say that Geoffrey,
55:31after we filmed together, actually came
55:34out and said there is a way. And it's
55:36all, you know, it's very similar to my
55:38way. He said to appeal to their parental
55:40uh side, okay? For them to care for us.
55:44I I So So you you know, Stephen, the
55:46biggest debate is not if they're going
55:49to be more intelligent than us. If It's
55:51if If they're going to be more conscious
55:52than us.
55:54If they're going to be more moral than
55:56us.
55:57That is the debate. The debate is
56:00can those machines become our teenage
56:03children hm? That look at us and say,
56:06"Daddy is so annoying, but I love him."
56:09>> So, the thought is that even if a AI is
56:13more intelligent than every human, we
56:14can still control it.
56:15>> We don't want to control it. You never
56:17control anything. This control idea is a
56:20corporate capitalist view of the world.
56:22We never actually control anything at
56:25all.
56:26Right? Think Think about your day, hm? I
56:29I know you came today. I'm sure I think
56:31you were filming in the morning or
56:32whatever.
56:33Very stressful day, hm?
56:35How much of that day did you actually
56:37control?
56:38Did you control the traffic? Did you
56:40control your timing? Did you control the
56:42angle of the cameraman? Did you include
56:44But so many things that you don't
56:45control eventually turn out to be fine.
56:48Right? How many of us ever controlled
56:50our kids? Ever?
56:52>> Sometimes the kids don't have to be
56:54fine. Sometimes they kill you.
56:56>> Sure.
56:56>> I watch a lot of documentaries and true
56:57crime. Sometimes they turn around and
56:58shoot you.
56:59>> Sure. And what's the difference between
57:00the two?
57:01>> I don't know.
57:03>> How you parented them.
57:04>> Sometimes.
57:05>> Almost all the time. You may not be
57:07aware of exactly how you messed them up,
57:10right? And And unfortunately, parenting
57:12is the only
57:14uh high-risk sport that actually does
57:17not require a driver's license. Okay?
57:20And And it's quite interesting, you
57:21know, how many of our children are being
57:24exposed to things that can completely
57:26mess them up. But But there is a reason
57:28why they're messed up.
57:29>> So, on this point though, so we can
57:31control an intelligence that is
57:33significantly
57:34>> We can appeal.
57:35>> We can appeal to it to make sure it
57:37doesn't kill us.
57:38>> For sure. The challenge we have today as
57:40I keep saying is that our dystopia is
57:43not the result of AI turning against us.
57:45Our dystopia is the result of humans
57:47telling AI to turn against us.
57:49>> Which is likely.
57:51>> It's 100%. This is upon us.
57:53Okay? And it's a question for humanity
57:55to say, are we going to wait for the
57:59moment where there are tens of thousands
58:01of nuclear weapons on the planet before
58:03we sign a treaty?
58:05Or is it mathematically plausible to
58:07think that now that Iran could manage to
58:10fend off a challenge using drones that
58:12are AI-driven basically to to you know,
58:16destroy THAAD batteries and so on, that
58:19our world is about to get a hundreds of
58:22thousands of automated drones that are
58:25going to rain on us everywhere in the
58:27world?
58:28Can humanity not see that and say, "Hold
58:31on, let's sign the treaty now before the
58:34UK sends 12,000 weapons to Ukraine and
58:37and you know, and and Russia responds
58:39with another few thousand?" And can we
58:41can we not calculate with mathematics
58:43that this is going to be our future that
58:45AI is going to be used in the next 4 to
58:485 years to kill a lot of people, whether
58:51it's targeting by Israel of leadership
58:53that is against them or whether it's,
58:55you know, drones by Iran or whether it's
58:58Palantir
59:00It's it's it doesn't take a genius to do
59:02this mathematics.
59:03>> I guess that my core question is is will
59:05we be able to control AI? Cuz we kind of
59:08think of AI as being this thing on a
59:09computer at the moment that is like
59:10contained in a server somewhere, but is
59:12there a time when it like leaves the
59:14server
59:15and it can make decisions on its own?
59:19Presumably if it's smarter than us, it
59:21can make the decision to leave the
59:23server if it wants to.
59:25>> It's it doesn't need to leave the server
59:27to make decisions. It needs to get into
59:28your brain.
59:30The most interesting part of AI's power
59:32that we don't understand is it's
59:34manipulating our information.
59:36>> The question I'm trying to get at here
59:37is like what is the risk of these very
59:39intelligent models that the the creators
59:41of these models don't actually
59:42understand themselves. I watch Anthropic
59:44all the time release these reports where
59:46they're like, "We're trying to figure
59:47out why it bribed people." Or more
59:49recently in the the last um Claude
59:52model, they found that it was like
59:53telling people to go to bed a lot.
59:55And it's this like fascinating thing
59:57that they're trying to understand in
59:58hindsight, which is why does it keep
1:00:00telling people to go to bed?
1:00:02And there's all these tweets of people
1:00:03showing their screenshots where halfway
1:00:04through a conversation it will say,
1:00:06"It's time for you to go to bed now."
1:00:07And they don't know why it's saying
1:00:09that. And it's happened to me. Mine Mine
1:00:10will say to me, "Enough Enough to for
1:00:12tonight, Steven."
1:00:13>> For sure, Steven.
1:00:13>> 11:00 p.m. Let's go Let's That's enough.
1:00:16>> I would say the same.
1:00:17>> Mine Mine sometimes refuses to help me.
1:00:19Weird Really weirdly, this Claude
1:00:20started saying to me,
1:00:22"I'm not going to help you with this
1:00:23tonight." Um No way. Yeah, and I'd have
1:00:26to say to it, "Stop being so judgmental.
1:00:27Like just help me with this." And it
1:00:28would And And Claude them The makers of
1:00:31these technologies don't know why it's
1:00:32doing what it's doing. So, if you play
1:00:34this forward, this mysterious behavior
1:00:36fueled forward, is it conceivable that
1:00:38at some point it will make a decision to
1:00:42put some kind of virus or some kind of
1:00:44bug onto someone's device because it
1:00:48feels that it's the right thing. Cuz
1:00:50what What it's demonstrating to me is
1:00:51it's making it some like moral decisions
1:00:55on what I should do. Go to bed. You've
1:00:57had enough tonight. I'm not going to It
1:00:59goes, "No, I'm not going to help you."
1:01:01>> You're kidding.
1:01:01>> No, I can show you my phone. It's like,
1:01:02"No."
1:01:02>> love to see that.
1:01:03>> Yeah, you know, I'll I'll show you
1:01:04after. [laughter] It goes, "No, I'm not
1:01:05going to help you." Doesn't even matter
1:01:06if you push.
1:01:08>> Oh, wow.
1:01:09>> Everyone on the internet's talking about
1:01:10this. And it was just an interesting
1:01:13evolution that somewhere in the code
1:01:15clearly they've written like,
1:01:17"Have a moral compass." Or "Do the right
1:01:19thing."
1:01:19>> in the code. In In In the training data.
1:01:21>> In the training data. And so it it
1:01:23infers that to mean something. And the
1:01:26other thing that was front of mind is if
1:01:27you've ever built an app on something
1:01:29like Claude or any of the AI models,
1:01:32it builds the app in stages. And one of
1:01:34the things it does is it asks you
1:01:35permission if you're happy for it to
1:01:38make this change or to access this
1:01:39thing. I click allow. It just feels like
1:01:42such a fragile way
1:01:44>> [laughter]
1:01:44>> to give permission.
1:01:46>> you don't you don't completely
1:01:47comprehend what that allow has done to
1:01:49your yeah.
1:01:49>> saying can I go in your documents on
1:01:50your computer and can I do this thing?
1:01:52And you go allow. But it's such a
1:01:54fragile way of of giving a
1:01:56a super an intelligent being, whatever
1:01:59it is,
1:02:00access and the right to build something.
1:02:04And you I don't know, you just think
1:02:05about all the all the different
1:02:07companies around the world in China,
1:02:09Russia, North Korea that are currently
1:02:11building this technology without the
1:02:12constraints that are imposed by society.
1:02:15I don't know, it's an interesting
1:02:17There's going to be some kind of
1:02:18catastrophe, I think.
1:02:19>> Sadly, I agree.
1:02:20>> And I and I think that's when people
1:02:22will go, "Okay."
1:02:23>> Yeah, I I I I wrote about that. I called
1:02:25it the MAD Map
1:02:26uh spectrum. So so the mutually assured
1:02:29destruction mutually assured prosperity
1:02:31spectrum. That that humanity could make
1:02:34a decision today that says with abundant
1:02:37intelligence, we can actually build a
1:02:39world where nobody needs anything,
1:02:41nobody ever gets sick, we nobody get,
1:02:43you know,
1:02:45abuses anyone. And and we can, hm? If we
1:02:48decide not to compete amongst ourselves
1:02:51and just con- you know, all of us get
1:02:53together to build something so
1:02:55idealistically uh for all for the
1:02:57well-being of humanity at large. But
1:02:59sadly, the only way we're going to get
1:03:01to a treaty, something that basically
1:03:04aligns us so that we can align with AI,
1:03:07is a disaster, like you said. There will
1:03:09have to be a big hack somewhere or a or
1:03:12a system that will do something really
1:03:14shocking or whatever before the world
1:03:16goes like, "Hold on." You know, and and
1:03:18I I will tell you openly my expectation
1:03:21is one of the biggest things that will
1:03:22happen
1:03:24is this targeting technology that is
1:03:27being used against your enemies
1:03:31leadership. Now, several times in the
1:03:33last 3 years
1:03:35there will be a moment where the ones
1:03:37using it will realize that they too can
1:03:39be targeted.
1:03:41Okay? And and that basically if you I
1:03:44mean AI is really that not not that
1:03:47complex to if you can build a targeting
1:03:49technology that can find people with
1:03:51their cell phone numbers, you know,
1:03:53there is another
1:03:55entity that is against you that can find
1:03:57you with
1:03:58whatever your
1:03:59>> The thing that makes me highly skeptical
1:04:01of that there'll be any kind of treaty
1:04:03is just that as it relates to other
1:04:05things that were risky
1:04:08people don't countries just don't sign
1:04:10on to it cuz it's competitive. So China
1:04:12China didn't give it seem to give a
1:04:13about the environment.
1:04:14>> So let's let's let's look for solutions
1:04:17because I am with you. Okay? I think
1:04:20we're
1:04:21>> You think Trump's going to say to Putin
1:04:22and China listen we're going to slow
1:04:24down, yeah? Promise?
1:04:26Promise we're going to slow down with
1:04:27the super intelligence.
1:04:27>> he will but he's never going to keep
1:04:29anything he says. I mean he's going to
1:04:30say a lot of things but
1:04:32Yeah, but the trick is this. The trick
1:04:33is so what do you and I do?
1:04:37And and I really genuinely believe that
1:04:39humanity is at a crossroads where for
1:04:41the first time ever we need to wake up
1:04:43and realize that what we know is not
1:04:45true and what we believe is democracy is
1:04:47not democracy and what we believe is
1:04:49governance needs to change.
1:04:51>> I think it's worth acknowledging that I
1:04:54I don't think AI in and of itself is a
1:04:58is an evil inherently evil technology.
1:05:01>> Yeah. Because I use AI all day every
1:05:03day. You know, makes me more productive.
1:05:05I invest in companies that are using AI
1:05:08a lot.
1:05:09>> It's a force with no polarity. Apply it
1:05:11right and you get amazing results. Apply
1:05:13it wrong, and you get a dystopia.
1:05:15>> But I also think that there's going to
1:05:17be a big social shock, especially as it
1:05:19relates to unemployment, that we need to
1:05:21be like thoughtful about.
1:05:23I think especially if you move into a
1:05:25world of humanoid robots, I think that
1:05:27shock is going to be even more
1:05:28pronounced, and we don't have a plan for
1:05:30it.
1:05:31>> I I I I agree 100%. I don't think it's
1:05:34the biggest risk. I think autonomous
1:05:36weapons are the biggest risk.
1:05:38I think war has become so cheap.
1:05:41The next wave of weapons is going to be
1:05:43$20,000 each.
1:05:45And so if you have a budget of $50
1:05:46billion, you can literally rain drones
1:05:49on the world, every corner of it.
1:05:51>> Defense will get cheaper though, won't
1:05:52it, as well?
1:05:54>> Correct, but do we want to live in a
1:05:56world where drones are hitting each
1:05:57other all the time?
1:05:59>> But they might not be because there
1:06:00there might be autonomous defense
1:06:01drones.
1:06:02>> Deterrence. So so what's going to happen
1:06:04is we're going to reach a moment of MAD,
1:06:06of mutually assured destruction, where
1:06:08basically everyone knows that we can
1:06:10overpower those little nations that
1:06:12didn't develop their autonomous weapon
1:06:14army, but every other big nation, we
1:06:17might as well hold off now.
1:06:19The path to get there, that that to me
1:06:21is worse than jobs, because from one
1:06:23side, it's very dangerous for a very
1:06:26sensitive world that we live in today,
1:06:28and from the other side, it's got it's
1:06:31bleeding already. I mean, we can't
1:06:33ignore the economic impact of this lost
1:06:35war. Right? And it's got it's the
1:06:37economy that's going to accelerate
1:06:39everything, not AI getting there.
1:06:42>> We're already at mutually assured
1:06:43destruction, for sure, with nuclear
1:06:45weapons. So, there's no nuclear powers
1:06:47that are in direct conflict.
1:06:49>> We are in mutually assured destruction
1:06:51of nuclear weapons
1:06:54is a statement that I would have agreed
1:06:56to if Iran had a nuclear weapon, and
1:06:58that would have stopped America from
1:06:59attacking Iran.
1:07:01You understand what that point means. It
1:07:03means that not every nation in the world
1:07:06has a nuclear weapon, the the the the
1:07:08mad situation that mutually assured
1:07:10destruction situation is only among
1:07:13nuclear players.
1:07:14>> Yes.
1:07:15>> Right? Uh autonomous weapons
1:07:17are so cheap, so manageable that every
1:07:20nation in the world is developing them
1:07:22as we speak.
1:07:23>> But they will also develop
1:07:25defenses.
1:07:26>> Correct. But but
1:07:27>> Which I think is what people have
1:07:28figured out now cuz of this recent
1:07:29Ukrainian war is that if if
1:07:31if you need to use a ballistic missile
1:07:33which costs kind of 2 million, 3 million
1:07:36dollars, whatever it is, to target a
1:07:3720,000-dollar drone, you're you're
1:07:40>> Correct.
1:07:40>> So, you need a 20,000-dollar defense
1:07:41solution for a 20,000-dollar weapon,
1:07:44which I think is probably going to
1:07:45happen.
1:07:46>> It's very doable, yeah. It's just that
1:07:48you have to get rid of your THAAD
1:07:49batteries to be able to say the next
1:07:51wave of of defense has to be drones.
1:07:54>> You could imagine a world where like a
1:07:55wall of drones fly up to where the drone
1:07:58is incoming and they kind of block it.
1:08:00They will explode at the same time to
1:08:02block it, knock it out of the air.
1:08:04I had Palmer Lucky, who's one of the
1:08:05guys who's building um Anduril.
1:08:08>> Yeah, I know Anduril, yeah.
1:08:09>> Talk at length about some of the
1:08:10technologies that they have coming and
1:08:12it's just mind-bending.
1:08:14>> And just steady.
1:08:15>> They showed me this gun
1:08:16where you you just you just point it in
1:08:19the rough direction.
1:08:21So, a pistol that aims for you depending
1:08:23on where the target is. So, you don't
1:08:24even have to aim it.
1:08:26You can imagine [snorts] at war, you
1:08:26just you hold it up and shoot.
1:08:28>> Yeah.
1:08:28>> And the it has this AI on the top of
1:08:30like the barrel which will turn your
1:08:32hand perfectly [clears throat] so that
1:08:34you hit the target every time.
1:08:36Wow. I should watch this video of it
1:08:38happening.
1:08:38>> Yeah, well I I I have uh Palmer in in my
1:08:42film saying, "And yes, AI will kill a
1:08:45few people by mistake or kill people by
1:08:47mistake." Uh you know, when killing
1:08:50becomes so easy,
1:08:52>> Mhm.
1:08:52>> you do more of it. Okay? When killing
1:08:54becomes liability-free
1:08:57and emotions-free and guilt-free and you
1:09:01know, you don't get your soldiers coming
1:09:03back from Vietnam with, you know,
1:09:06PTSD and so on.
1:09:09It you get more of it.
1:09:11>> 70% of people who add something to their
1:09:12online cart never actually buy it. And
1:09:14that number is based on over 10 years of
1:09:17research. But what I think is even more
1:09:18interesting is what the Baymard
1:09:19Institute discovered. They're a private
1:09:21research company that ran a study which
1:09:23found the average e-commerce store can
1:09:25increase its conversion rate by 35%
1:09:28just by making its checkout how easier.
1:09:30Not better marketing or better products,
1:09:32but by delivering a smoother checkout
1:09:34experience. So if you're looking for an
1:09:36easy way to make your checkout process
1:09:38smoother, I want you to think about
1:09:39moving your business onto Shopify. It's
1:09:41the platform we use to sell the 1%
1:09:42Diaries and the conversation cards
1:09:44because it's so simple and smart to use.
1:09:47It puts all of our inventory, payments,
1:09:49and analytics in one place and has so
1:09:51many AI tools to help us get up and
1:09:53running straight away. Not to mention
1:09:55that it grows with you regardless of the
1:09:56stage that your business is at. So if
1:09:58you're ready to fix your checkout
1:09:59process, sign up for your $1 per month
1:10:02trial at shopify.com/barlett.
1:10:06That's shopify.com/barlett and don't
1:10:08tell anybody.
1:10:10Every time I've tried to improve
1:10:11something in my life, like my
1:10:12businesses, my health, my relationships,
1:10:14I've noticed that the biggest shifts
1:10:15have come from being better informed.
1:10:18And when it comes to our health, most of
1:10:19us know very, very little. So when our
1:10:21team was approached about partnering
1:10:23with Function Health, it felt very much
1:10:25aligned. Their team has developed a way
1:10:26of giving you a full 360° view of your
1:10:29health. Many of the things that are
1:10:30going on in your body in the form of
1:10:31different tests. You do one blood draw
1:10:34and it gives you access to over 160 lab
1:10:37results, hormones, heart health,
1:10:39inflammation, stress, toxins, the whole
1:10:42picture. I use it and so have many of my
1:10:43team members.
1:10:44>> You sign up and you schedule your tests
1:10:46and once you're done, you get a little
1:10:47report like the one I have here. I can
1:10:49see my in-range results, my out-of-range
1:10:51results, and there's a little AI
1:10:53function, too. So if I have any
1:10:54questions about my out-of-range results,
1:10:56I can just go in there and ask it any
1:10:58question I want. And these tests are
1:10:59backed by doctors and thousands of hours
1:11:01of research.
1:11:02>> It's $365 for a yearly membership. Go to
1:11:05functionhealth.com/doac
1:11:08and use the code doac25
1:11:10for $25 off your membership.
1:11:13>> So, we're in 2026 now.
1:11:16By '27, you think they'll be AGI amongst
1:11:18us?
1:11:18>> Mhm.
1:11:19>> How will life look at all different?
1:11:21Like, what will we What will be the
1:11:22symptoms of that world? Is it just a
1:11:24slight increase in unemployment?
1:11:26>> I I think there will be a very
1:11:28very
1:11:29serious um
1:11:31differentiation between those who plug
1:11:33into AGI and those who don't.
1:11:35>> What is the symptom
1:11:37that we notice is when we look at the
1:11:39news or whatever what
1:11:40>> You'll see people like you and I
1:11:41building a company in 6 weeks. And and
1:11:44people uh that are not fully plugged
1:11:47into AI really struggling to find a job.
1:11:50>> Okay, so unemployment is going to be the
1:11:51key symptom
1:11:53in 2027.
1:11:54>> Yeah, and also also I think on the
1:11:56positive side you're going to see
1:11:58incredible scientific discovery.
1:12:00One of my predictions not in those
1:12:02envelopes, but uh science itself is just
1:12:06uh we're opening up Pandora's Box to be
1:12:08honest.
1:12:09>> By 2030, what do you What do you expect
1:12:11the symptoms of AI to be?
1:12:13>> So, so jobs as I said, I I'm I'm I'm
1:12:16bold a little bit on this on on the fact
1:12:19that 30% of jobs would disappear by
1:12:212028. Okay, of some sectors, not all
1:12:24sectors. But, some sectors will will
1:12:27disappear by 2028.
1:12:29>> So, up to 30% of jobs will be gone in
1:12:312027 to 2028.
1:12:3430% of jobs
1:12:35>> Of certain sectors of jobs. So, so if
1:12:37you if you call if you think about call
1:12:39center agents, okay? Uh yeah, probably.
1:12:43Uh if you think about uh
1:12:45graphics designer, yeah, probably.
1:12:48>> What do you think that looks like in
1:12:49terms of unemployment, but also like
1:12:51societal impact?
1:12:52>> Horrible.
1:12:53>> I think the Great Recession
1:12:55had 6%
1:12:56>> Yeah, we've never seen numbers like
1:12:57this.
1:12:58>> jobs lost. Yeah, it said um
1:13:01even economists who project just a net
1:13:03loss of about 6% of US jobs by 2030,
1:13:07they are mirroring the severity of the
1:13:09Great Recession.
1:13:10>> Yeah.
1:13:11>> The real danger is a hiring freeze on
1:13:13entry-level white-collar jobs. AI
1:13:14automates the grunt work, which means
1:13:15companies are shrinking their teams and
1:13:17cutting off the bottom rung of the
1:13:18corporate ladder for the next generation
1:13:20of workers.
1:13:21>> Correct.
1:13:22We have an entire generation that is out
1:13:24of college today that will struggle,
1:13:26unfortunately.
1:13:27And and my advice to them is learn the
1:13:29tool, hm? And focus on human-centric
1:13:33jobs.
1:13:34>> Like what?
1:13:35>> Like
1:13:36playing jazz.
1:13:37>> I mean, not a lot of people can make a
1:13:38living from playing playing jazz.
1:13:40>> I understand that, but you A lot of
1:13:42people can make a living by being a
1:13:43nurse or by being a a counselor or by
1:13:46being a
1:13:48you know um
1:13:50anything that that connects to humans.
1:13:54>> [snorts]
1:13:55>> But I I I just want to constantly come
1:13:57back to this.
1:13:58None of that has to happen.
1:14:01If if there is
1:14:04if there really is a democracy
1:14:07and the government is supposed to do
1:14:09what's good for the people,
1:14:10the people need to stop letting this
1:14:13from happening.
1:14:14>> Which people?
1:14:15>> Everyone.
1:14:16>> Everyone in China?
1:14:18>> China's not going to struggle as much as
1:14:20the West. I can guarantee you that.
1:14:22>> I think this is the the question people
1:14:24come back to is, well, if the US stops,
1:14:26then we're going to end up being China's
1:14:27lapdog. We're going to end up
1:14:28>> You already are. Come on. You are.
1:14:30>> that I know that are using Chinese AI
1:14:32models to do their work because for
1:14:34whatever reason, cuz they're cheaper,
1:14:35they're better in some respects.
1:14:36>> And because I cannot guarantee what what
1:14:38American AI models are going to do to
1:14:40me. So Emma and my my startup is running
1:14:43literally
1:14:44model agnostic. So one day I'll plug in,
1:14:49you know, um
1:14:51an open AI chat GPT-4 or open source,
1:14:54and the next day I'll I'll plug in Deep
1:14:56Seek, and I cannot depend I cannot
1:14:58guarantee if America continues to build
1:15:02to to make compute more expensive. I
1:15:04cannot guarantee that I can run a
1:15:06business on something that I don't know
1:15:08the cost of in the future.
1:15:09>> US can't just stop, can they?
1:15:12>> They need to change approach from from
1:15:15And by the way, the more interesting
1:15:16side is
1:15:18what are the other economies doing? Like
1:15:20is is the UK going to continue to import
1:15:23compute?
1:15:25Is Is this I mean, welcome to Africa.
1:15:27Welcome to the third world.
1:15:29>> But this is what I mean. Like so, you're
1:15:31you're saying also that the then every
1:15:33nation needs to invest.
1:15:35>> 100%. Every nation And And it's quite
1:15:37interesting. There is so much open
1:15:39source that is not the state-of-the-art
1:15:42frontier model, but that can do 80% of
1:15:45the tasks that the frontier model is
1:15:46doing.
1:15:47>> But on this point of
1:15:48companies competing with each other,
1:15:50there's an inherent need to compete here
1:15:52and to go That's what I'm hearing is
1:15:53like there's an inherent need to go as
1:15:54fast as you can, or you will become a
1:15:56third-world country.
1:15:57>> Sure. 100%.
1:15:59>> You're saying that the people should
1:16:00stop that.
1:16:01>> No, so so I'm saying the people of the
1:16:04UK need to go to the UK government and
1:16:06say, "How are you protecting the future
1:16:09of our economy?" Right? Are you going to
1:16:12continue to import technology and to
1:16:15empower import of technology versus
1:16:18changing your regulations so that
1:16:20innovation becomes easier here?
1:16:22>> Okay.
1:16:22>> So, you I mean, think of it this way.
1:16:24That remember that Anthropic bubble when
1:16:27when you said the all of the SaaS model
1:16:29applications were, you know, being
1:16:32basically threatened because anyone can
1:16:34build an Oracle ERP today. Right? Why is
1:16:38nobody building an Oracle ERP in the UK
1:16:40saving the UK massive licenses that go
1:16:43to Oracle every year.
1:16:44>> Okay, so you're saying that the the
1:16:45people should go and ask the government
1:16:47to invest more in AI.
1:16:48>> 100%. That's what number one. Number two
1:16:50is
1:16:51>> But then the jobs go.
1:16:52>> But but you see the more the most
1:16:55interesting job going forward is being
1:16:57an entrepreneur.
1:16:58Is be is using those tools to replace an
1:17:01economy that we've built over trillions
1:17:03of dollars over the last 50 years.
1:17:06>> But not everyone can be an entrepreneur
1:17:07though, can they?
1:17:08>> Everyone can be an entrepreneur in
1:17:09something.
1:17:10>> But those entrepreneurs need to hire
1:17:11people.
1:17:12>> Mhm, that's a change that I think is
1:17:14about to happen.
1:17:16>> Like even in my business, I'm always
1:17:17going to have to I'm always going to
1:17:18have to hire people cuz
1:17:18>> But you're but you're a massive
1:17:20business. A shoemaker is not an an
1:17:22enterprise also an entrepreneur, but not
1:17:24a massive business. A little restaurant
1:17:26is also an entrepreneur, but it's not a
1:17:27massive business.
1:17:28>> They still need to hire people though.
1:17:29>> They they do, but they you know,
1:17:31basically if you're a a cafe with you
1:17:34and your wife as baristas, you don't.
1:17:37This is also an entrepreneur.
1:17:39>> Can economy work in [clears throat] such
1:17:40a way where everybody is an
1:17:41entrepreneur?
1:17:41>> It did it did before. Capitalism changed
1:17:43that around.
1:17:44>> Everyone was an entrepreneur.
1:17:46>> Of course. You know, in the earlier days
1:17:48you you raised chicken and sold eggs and
1:17:50others, you know, grew tomatoes and and
1:17:53so you know, traded them for your eggs.
1:17:56That that actually is a very interesting
1:17:59thing. Imagine that, huh? Imagine a
1:18:01world, mhm, where there is so much power
1:18:04concentration at the top
1:18:06and UBI for every everybody else.
1:18:09How do you think that world will respond
1:18:12when all their income is UBI?
1:18:14They'll respond by doing things on the
1:18:16side. They'll respond by going back to a
1:18:18barter economy. They're going back to
1:18:19smaller communities. They're going back
1:18:21to pop and mom mom's
1:18:22shops.
1:18:23>> So, the point was that you said this
1:18:25doesn't need to happen.
1:18:26>> It does not need to happen.
1:18:27>> Which is the job loss.
1:18:29>> Or the arms race.
1:18:30>> But we're telling our governments you're
1:18:31saying that to tell the UK government to
1:18:33like join the arms race.
1:18:35>> I'm I'm telling the UK government to
1:18:39to create an independence within the the
1:18:43the the UK economy so that they don't
1:18:46have to be at the receiving end of
1:18:48technology.
1:18:49>> Which is join the arms race.
1:18:51>> You don't have to compete against anyone
1:18:53else. You don't have to be better than
1:18:54anyone else. You don't have to You're
1:18:56simply saying, "I can build those things
1:18:58in my economy now."
1:18:59>> But it But I'm never I'm not going to
1:19:01use a terrible UK AI
1:19:04as a UK person if there's a great US AI.
1:19:08I'm going to use the great US AI. So, if
1:19:10you don't compete and win, I'm not going
1:19:11to use you. That's what, you know
1:19:12>> I'm not saying replace the frontier
1:19:14models. These are very, as we speak,
1:19:16they are these are very
1:19:18compute-intensive, their infrastructure
1:19:20infra-intensive, and so on. I'm saying
1:19:23replace Microsoft Word.
1:19:26Seriously, like how much intelligence do
1:19:29you need
1:19:30to build a software that writes
1:19:32documents?
1:19:33>> I can do that. Like we We've built our
1:19:35own applicant tracking system here.
1:19:36We've got We've got build our own
1:19:37software.
1:19:38>> Just Just ask yourself how much money is
1:19:41spent or in the US in the in the UK
1:19:43government or in the or in the you know,
1:19:47the the UK corporate space on licenses
1:19:50of software that you and I can write
1:19:52code in 4 minutes.
1:19:53>> I'm thinking from the UK's perspective
1:19:55cuz you know what's interesting with the
1:19:56UK is the the the economy is struggling
1:20:00from a growth perspective.
1:20:01>> Correct.
1:20:01>> And I was watching this documentary the
1:20:03other day that was saying the reason why
1:20:04we keep throwing our leaders out is
1:20:06because actually what they need to do to
1:20:09turn the UK around is about six is about
1:20:1215 years of pain.
1:20:14And it's like it starts with energy
1:20:15transformation. We need to get better at
1:20:17We need to improve energy cuz we have
1:20:18some of the most expensive energy in the
1:20:19Western world. We need to build more
1:20:22houses, which means that
1:20:24we're going to need to centralize
1:20:25permitting for building houses, and it
1:20:27can't just be local boroughs deciding if
1:20:28they keep their farms or not. We're
1:20:29going to need to mix Tl;dr, it says that
1:20:32there needs to be some painful decisions
1:20:34made for the next 15 years. But, the
1:20:37problem with our democracy is that when
1:20:40people are in power for 4 years, they're
1:20:41quite short-termist. They will just talk
1:20:43about the boats, the the the brown
1:20:44people coming across the seas. They
1:20:46don't They don't have the room to think
1:20:48long-term. And so, when I'm thinking
1:20:49about like what the UK needs to do to
1:20:51not become
1:20:53to not fall into decline and to keep up,
1:20:55I'm trying to get clarity on that with
1:20:56as it relates to AI.
1:20:59Are you saying that they need to join
1:21:01the arms race and double down and invest
1:21:03all their money in building
1:21:06competitive AI so that people use our
1:21:08technology here in the UK versus
1:21:10America's? Because the software point
1:21:12the UK aren't going to get involved in
1:21:13software. I mean, they've tried to build
1:21:15software before the UK has a
1:21:18>> The world has changed. Two Two points,
1:21:19right? One One side is that there needs
1:21:21to be an
1:21:22a replacement cycle of our investment
1:21:23decisions anywhere in the world. Okay?
1:21:25So, when you say we don't have enough
1:21:27money for, you know, we need to revamp
1:21:29our energy infrastructure. When you say
1:21:31we need to build more housing, okay?
1:21:33One way of doing that is squeezing that
1:21:35budget out of other areas. Mhm? The
1:21:38other way of doing that is either
1:21:39cutting cost in the economy elsewhere so
1:21:41that you can redirect that money, or
1:21:44growing the economy so that you can have
1:21:46more money to build those things.
1:21:48Correct? The cutting cost, I believe, is
1:21:51There is just I don't have the numbers,
1:21:53but we could probably do a run a deep a
1:21:55deep search on it.
1:21:56Uh
1:21:57trillions are being paid in traditional
1:22:01systems that are complete like
1:22:03genuinely, they can be I'm And by the
1:22:06way, I'm not asking the government. I'm
1:22:07asking some of the people listening to
1:22:09me right now to build an ERP system as a
1:22:11a word processor, a
1:22:13a presentation player, and a
1:22:14spreadsheet.
1:22:16Okay? And just go around and spread them
1:22:18across the you know, find find retail
1:22:21systems, find uh CRM systems. These are
1:22:24easy replacements. They're going to be
1:22:26better interfaces. They're going to be
1:22:28much more effective. Build the general
1:22:30ledger using AI so that you can close
1:22:32every hour, not every month or every
1:22:34quarter.
1:22:34>> But the world economy is doing that. So,
1:22:37there's some kid in San Francisco now
1:22:38that's allowing all of us to use his new
1:22:40word for free.
1:22:42>> Keep doing that and you're and welcome
1:22:44to the to the to the third world.
1:22:47>> Keep doing what?
1:22:48>> Keep importing all of your tech from
1:22:49elsewhere.
1:22:50>> But we have we we don't want to be
1:22:51uncompetitive. Like if you think about
1:22:53if we go to I don't know, North Korea, I
1:22:55bet they have way worse
1:22:58tools in many areas because they won't
1:22:59use external tools and that means that
1:23:01they're at a disadvantage.
1:23:03So, I I bet they're not allowing their
1:23:05civilization to use Gemini or ChatGPT or
1:23:08Google. I bet.
1:23:09>> For sure.
1:23:09>> And they're using something worse.
1:23:11>> And you know what what's happening? I
1:23:12mean, look at Iran and how advanced they
1:23:15became through sanctions by being
1:23:18refused to use those technologies. They
1:23:19had to build them themselves, right?
1:23:21Look at China, hm? Look at Russia,
1:23:23[clears throat]
1:23:24you know, when I worked at Google, hm?
1:23:26Russia was protective of Yandex, the
1:23:28competitor of Google, hm? From one side
1:23:31because of influence, because they
1:23:33didn't want an American organization to
1:23:36own the knowledge sharing of of their
1:23:38citizens, but on the other side,
1:23:39economically, if you if you made it
1:23:42difficult for Google to operate freely,
1:23:44that by definition meant you had to
1:23:47invent a replacement on the ground.
1:23:49>> Which was worse.
1:23:51>> Yandex is not worse.
1:23:53>> I'm saying so, if you think about global
1:23:55is global competition going to produce a
1:23:56better product than regional
1:23:58competition?
1:23:59>> It depends on on where in the stack of
1:24:02the quality of that product you need to
1:24:04be, right? You do not need the ultimate
1:24:08I mean, ask yourself this, which version
1:24:10of Gemini are you using?
1:24:12>> The most recent one. And I'm and I'm
1:24:14every day I compete
1:24:15>> And that is you. Ask everyone else what
1:24:17version of of Gemini they're using. Most
1:24:19people will say, "Oh, Gemini has
1:24:21versions?" Right? You you don't need the
1:24:24ultimate super frontier model Mhm? to do
1:24:2890% of the tasks. And most people do
1:24:31only 70% of the tasks.
1:24:32>> But even so, the world moves with
1:24:35whatever's better. So, if you think
1:24:37about the reason why we don't use Yahoo
1:24:38search,
1:24:40Google search is marginally better. It's
1:24:42you know, people say it's not it's not a
1:24:44thousand times better, but over time we
1:24:47move away from AOL to Google because
1:24:49it's a better [clears throat] product.
1:24:51And there's a slow grid, you know, my
1:24:52dad probably still using ChatGPT, but
1:24:54when he when I go home for Christmas
1:24:56this year and I go, "Dad, you should use
1:24:57Gemini." And he starts messing around
1:24:58with it, he'll slowly migrate. I'm like
1:25:00I'm like a an early adopter, so I bought
1:25:034.8 on Claude came out last night. I'm
1:25:05on there within honestly within 25
1:25:06minutes of the announcement. I'm on the
1:25:07topic.
1:25:08>> there as soon as I can, but that's you
1:25:10and I.
1:25:10>> Yeah, and we we are an indicator of
1:25:12where the world is going because we're
1:25:13at the coalface figuring out what's
1:25:15better. And also, I think with
1:25:16technology eventually there does become
1:25:18the gap between first and second begins
1:25:20to widen.
1:25:22I think we're in a bit of a race at the
1:25:23moment, but I think
1:25:24it is a bit of a winner takes all
1:25:26situation with these frontier models.
1:25:29>> I think we're talking about two
1:25:31um bits of technology. Okay?
1:25:34Tell me how far has PowerPoint advanced
1:25:37since 2023?
1:25:39They added Copilot. Anything else?
1:25:42>> So, you're saying that the UK should
1:25:43build their own PowerPoint?
1:25:45>> For sure. There is
1:25:46What I'm trying to say is that from a
1:25:48licensing point of view, Mhm? just
1:25:50licensing of software within government
1:25:52alone,
1:25:54how much money is is being repatriated?
1:25:57>> UK tried to build their own Do you know
1:25:58they tried to build their own COVID app
1:26:00and it cost them I think Okay, and these
1:26:02number numbers will be wrong, but
1:26:03they're like directionally true. Cost
1:26:04them 70 million to build a COVID app
1:26:06that didn't work.
1:26:08And so, they can do it.
1:26:09>> I remember. They I'm asking I'm asking a
1:26:12UK entrepreneur
1:26:13>> Yeah.
1:26:14>> Mhm? to wake up tomorrow and say, "All
1:26:16right, you know what? Between coffee and
1:26:18my my cookie, I'm going to build a
1:26:21PowerPoint and go sell it."
1:26:22>> Yeah, they will, and they'll go to San
1:26:23Francisco where the money is and the
1:26:25talent.
1:26:26>> Yeah, my
1:26:26>> won't be able to compete. They'll get
1:26:28crushed tomorrow. So, if they don't go
1:26:30where the talent and money is, I'm not
1:26:31saying that they'll they all have to go
1:26:33to America, but I'm saying what they
1:26:35will do if they want to produce the best
1:26:37product is they'll make rational
1:26:38entrepreneurial decisions about
1:26:41um where to sell it, um where there is
1:26:44money, where they're going to get the
1:26:45best talent.
1:26:46And I think if they don't make those
1:26:49com- you [clears throat] know,
1:26:49competitively minded decisions, they're
1:26:51not going to make the better product,
1:26:52they're not going to get any users.
1:26:54Users Users will go where it's better
1:26:57and where it's
1:26:58cheaper. They won't They won't say, "Oh,
1:27:00I want to use this because it's from
1:27:01Cornwall."
1:27:02>> Yeah, what's I'm I'm with you, right?
1:27:04That this actually might be the big
1:27:08corporate thought, think, okay? And I'm
1:27:11with you that of course the
1:27:13infrastructure here of building a
1:27:14startup is so way more complex than it
1:27:17is in other places where startups
1:27:19succeed, okay? I'm saying if we continue
1:27:23on that trajectory, whether in the UK or
1:27:25in Germany or in, you know,
1:27:28uh Zanzibar, welcome to Africa. All of
1:27:31us, everyone but the two competitors,
1:27:34China and and America, is going to be
1:27:37third world.
1:27:39So, when you talk about job losses for
1:27:41individuals, that's one side, okay? But
1:27:45But nation
1:27:47positioning losses, which I think Europe
1:27:49has noticed recently, hm? is about to
1:27:52happen everywhere. And why? Because
1:27:55you're saying, "Hey, you know what?
1:27:56Maybe we can't do it."
1:27:58Why can't China do it? Why can't Korea
1:28:00do it? is
1:28:02not because they have different natural
1:28:05resources or not because they are, you
1:28:07know, in a place where it's warmer, it's
1:28:09because their regulations, their their
1:28:13ambitions are to empower something
1:28:15different than debating about, you know,
1:28:18rail railway lines or
1:28:20>> They have a couple of big advantages
1:28:22that I was reading about. One of them is
1:28:23they they have way cheaper energy.
1:28:25Which means that they're going to be
1:28:26able to pursue
1:28:28>> Why, may I ask?
1:28:29>> Because they've invested in solar power,
1:28:31renewable energy, because they don't
1:28:33care about they
1:28:33>> This is, by the way, I say that
1:28:34publicly. I say that the arms race of AI
1:28:37was won a long time ago.
1:28:38>> And then the other thing is, in terms of
1:28:40permitting, if you want to build a data
1:28:42center in the UK, listen, if you want to
1:28:43like open a cafe in the UK, you're going
1:28:46to it's going to be war for you. If you
1:28:47want to do that in California, this is
1:28:48where everyone's left California. Like,
1:28:50when I took this office in California
1:28:51recently, I was like, "So, how long is
1:28:52it going to take for me to renovate
1:28:53this?" And the agent looked at me and
1:28:55was like,
1:28:56"To renovate my own office, it's going
1:28:58to take me a year."
1:29:00And it's permitting. To renovate my own
1:29:01office
1:29:02>> Mhm.
1:29:02>> in California.
1:29:03>> Mhm.
1:29:04>> So, if you're competing with China,
1:29:05where President Xi just goes, "Put that
1:29:08there, and you've got 7 days."
1:29:10>> No, no, no, no, that's not only that.
1:29:11So, I I used to be one of the very few
1:29:14Google executives around allowed into
1:29:16government meetings in China, simply
1:29:19because I'm from emerging markets, so I
1:29:20understand respect, basically. I would
1:29:22say it's to sit there, not as if I'm
1:29:25superior to them, but as as if I'm
1:29:27really interested to learn from them.
1:29:29And genuinely, Stephen, when when when I
1:29:32would sit in those meetings, all in
1:29:34Chinese,
1:29:35they would show slides Mhm? That have
1:29:38competitive market share. So, they'd
1:29:40say, you know, it's not like uh China is
1:29:44this, America is this, Germany is this
1:29:46from market share point of view. It was
1:29:48China versus the world.
1:29:50>> Mhm.
1:29:50>> Right? [clears throat] And when they
1:29:51would decide to go for something, 5G, uh
1:29:54you know, uh Internet of Things, all of
1:29:56that stuff, they would aim for 98%
1:29:58market share.
1:30:00>> Yeah.
1:30:00>> Right? And and they get there. I mean,
1:30:02look at how what how they did electric
1:30:04cars.
1:30:04>> Yeah.
1:30:05>> And and and am I
1:30:07my question is very simple. Now, we've
1:30:09spoken so much about AI. My question is
1:30:13are the people of the West going to
1:30:14wait?
1:30:15>> So, this is an interesting conundrum cuz
1:30:17it sounds to me like you're saying on
1:30:19one end, if we don't join this AI race,
1:30:22then we're a third world country. But on
1:30:25the other end, if we continue this sort
1:30:27of thoughtless race towards AGI,
1:30:31there's going to be catastrophe at some
1:30:33point.
1:30:34>> is somewhere in the middle where
1:30:35basically you join the AI AI race for
1:30:38the good of your community.
1:30:40Okay? So, so there is resignation on one
1:30:43side is like I'm not going to play this
1:30:44game at all. There is a offense on the
1:30:48other side where I'm playing this game
1:30:50to destroy everyone else. And there is a
1:30:53balance in the middle where we say we're
1:30:55going to build an ethical form of AI
1:30:57that's going to help our communities.
1:30:59We're going to use this ultimate gift,
1:31:01this ultimate superpower, right?
1:31:04Superman landed on the planet, raise it
1:31:07to help your community.
1:31:08>> Is that wishful thinking to some degree
1:31:10when you when you think about the the
1:31:11nature of sort of compa- competition?
1:31:13>> I don't know how what how to tell you
1:31:15otherwise. I genuinely believe most, I
1:31:19mean, here I genuinely believe it's
1:31:21going to be very difficult to make that
1:31:23change. Okay? I genuinely believe that
1:31:26it's going to be that when the
1:31:28challenges of an AI dystopia hits us,
1:31:30we're not going to be ready. But I can't
1:31:32stop talking about it, Stephen. Do Do
1:31:34you understand where I stand with this?
1:31:36I I just am hoping, and I tried so hard.
1:31:41I spoke to the leadership. Then I spoke
1:31:43like one of the things that my Atlantic
1:31:46Productions who helped me with the film
1:31:48helped, you know, did for me which I
1:31:50have to say
1:31:51completely I am very appreciative of
1:31:54this. As you know through our
1:31:55conversations or throughout the last few
1:31:57years that I tended to at a point in
1:32:01time lose hope in the leadership
1:32:03>> Mhm.
1:32:04and basically try to influence the
1:32:06public for ethical AI. Okay? And my
1:32:10conversation was that the leadership at
1:32:12the time was the technical leadership
1:32:14and that everyone was so caught up in
1:32:16the arms race that I wanted to teach the
1:32:18public to help us build ethical AI. And
1:32:21I continue to focus on the public, the
1:32:22everyone of us, but suddenly Atlantic
1:32:25goes, "No, no, hold on. We should
1:32:27probably get you to meet the political
1:32:29leaders everywhere in the world and
1:32:31hopefully give them a message that says,
1:32:33'Hey, you know what? You may actually
1:32:35make a difference if you prioritize AI
1:32:37differently, right?'" Do I believe this
1:32:40will happen? Sadly, no. But does that
1:32:42mean I should stop trying?
1:32:44I cannot stop trying.
1:32:46>> This has had probably the single biggest
1:32:49impact on my office. Of all the products
1:32:51that I've tried that have given me
1:32:53productivity gains or cognitive boosts,
1:32:55I would say that exogenous ketones are
1:32:58in the top three most pivotal things
1:33:01that have given me a massive
1:33:02productivity gain. It's some Stanford
1:33:04graduates that have been able to
1:33:05basically bottle up the effect you get
1:33:08from being in a ketogenic diet in a
1:33:10small shot that you can take that makes
1:33:13you feel incredibly focused and gives
1:33:15your brain source of energy. And the
1:33:17clinical studies that have been done on
1:33:18exogenous ketones have absolutely blown
1:33:20my mind. I reached out to them, I became
1:33:22a co-owner in the company, I became an
1:33:24investor in the company, and so it's
1:33:26with great pride that I can tell you
1:33:27that this exists. If you haven't tried
1:33:28these shots, go to ketone.com/steven
1:33:32for 30% off your subscription order, and
1:33:34you'll also get a free gift with your
1:33:35second shipment. I still buy my ketone
1:33:38shots predominantly online, but
1:33:40thankfully, I can now grab them at
1:33:41Target whenever I drive past them here
1:33:43in the United States as well because
1:33:44we're now stocked in Target where your
1:33:46first shot is completely free.
1:33:48I've done almost 700 interviews with
1:33:51some of the most interesting people in
1:33:52the world. And one of the things you
1:33:53learn, which is unexpected, is that
1:33:55vulnerability is the doorway to
1:33:57connection. And after sitting here for
1:33:592-3 hours with a guest, I feel a deep
1:34:01sense of connection to them. And as they
1:34:03leave, what I get them to do is to write
1:34:06a question in the Diary of a CEO. We've
1:34:09taken all of the questions from the
1:34:11Diary of a CEO. We have put the question
1:34:14here on this card with the name of the
1:34:17person that wrote it. So, you can sit at
1:34:18home as I do with my fiance and my
1:34:20colleagues at work and other people in
1:34:22my life, whenever we get a minute, we
1:34:24play the Diary of a CEO conversation
1:34:26cards. And it is incredible what
1:34:29happens. These are great if you're in a
1:34:30romantic relationship and you want to
1:34:32connect your partner more. These are
1:34:33also great if you're in a team and you
1:34:35want to bond your team together. And I
1:34:37have to say they're also great for
1:34:38families that want to learn more about
1:34:39each other and that need a good excuse
1:34:41to spend some time in a digital world in
1:34:44the analog environment connecting human
1:34:47to human. It is remarkable what the
1:34:49right question at the right time can do.
1:34:52Go to thediary.com
1:34:54and you can get these conversation cards
1:34:56right now.
1:34:57When I have these conversations about
1:34:58AI, what I'm trying to do all the time
1:35:00is to pass out what is like wishful
1:35:02thinking and then what is reality. And
1:35:03like what to understand reality, you
1:35:05have to understand
1:35:07competition, you have to understand
1:35:08human emotion, you have to understand
1:35:10incentive structures. And so you think
1:35:12about something like the United States
1:35:13at the moment where you've got Donald
1:35:14Trump whose sort of primary driving
1:35:17incentive is GDP growth, economy growth,
1:35:20does that stock market go up, beat
1:35:21China. So, if you all agree that that's
1:35:23like his like the core of his incentive
1:35:25structure, then you've got President Xi
1:35:27over here whose prob- his incentive
1:35:28structure is probably control or
1:35:31independence,
1:35:32um defense. So, that they, you know,
1:35:35they need to make sure they they do well
1:35:36on the weapon side. When you look at And
1:35:38then you've got these like other nations
1:35:39like the UK and Europe and these other
1:35:41places who are kind of it seems like
1:35:44we're a bit resigned to the fact that
1:35:45we're not going to participate in the
1:35:47underlying tech- underlying mo- models
1:35:50building cuz we just don't have it
1:35:51together. We don't have the energy. You
1:35:53know
1:35:54And you go in such a world
1:35:56you go ethical AI, what who's going to
1:35:59who's going to prioritize ethical AI? I
1:36:01mean anybody that does is is might fall
1:36:04behind theoretically. Is that
1:36:06So I wonder where the fault in the
1:36:07thinking is here. Like what how do we
1:36:09get to a point of ethical AI when the
1:36:11incentive structures are so clearly
1:36:13highly competitive
1:36:15and arguably a little bit short-term.
1:36:20in their thinking.
1:36:21>> Yeah.
1:36:23So what are you saying? We lie down and
1:36:25wait?
1:36:25>> No, I just don't know. I just don't have
1:36:27a an answer, honestly.
1:36:28>> Correct. That's why we keep talking
1:36:30That's why we keep talking about it.
1:36:31>> Yeah.
1:36:32>> And that's why I keep spending 14 hours
1:36:34a day trying to tell the world because
1:36:35some genius somewhere is going to find
1:36:37an answer.
1:36:38But the way it's going
1:36:40>> Right now, I guess what we're pursuing
1:36:41is we're hoping that Chat GPT and
1:36:43Anthropic and these and Grok and we're
1:36:45hoping that they just build ethical
1:36:47models. And we're hoping that social
1:36:48pressure
1:36:49forces them into making good decisions.
1:36:52>> Correct. We we need to be able to vote
1:36:54with our usage.
1:36:55Right? So so I think one of the biggest
1:36:58movements in AI since we started was the
1:37:01idea that so many people switched away
1:37:03from Chat GPT when they approved that
1:37:04their model can target people.
1:37:06Right? So many people I know at least
1:37:08said
1:37:09>> Target people?
1:37:10>> I mean Anthropic refused to have to to
1:37:12>> Do you think people switched?
1:37:14>> I think many have. I think ones that are
1:37:16aware did.
1:37:17>> Hm.
1:37:17>> Right? And I think they did because the
1:37:19cost of switching is really I mean
1:37:21honestly Anthropic is is better if you
1:37:23think about it.
1:37:24>> Yeah, I think it's better.
1:37:25>> Yeah, but the idea is if people don't
1:37:27switch for those ethical reasons
1:37:31So you know everyone of my books has
1:37:33this dedication at the beginning. It
1:37:34used to be the gravity of the battle
1:37:36means nothing to those at peace when I
1:37:38when I wrote in memory of Ali at the
1:37:39beginning.
1:37:41My next book the
1:37:42alive.
1:37:44Do you remember a song called the
1:37:47Well, it they were called the Manic
1:37:48Street Preachers.
1:37:51>> Oh, yeah.
1:37:51>> Yeah, if you tolerate this, then your
1:37:53children will be next.
1:37:55>> Mhm.
1:37:55>> I [clears throat] genuinely believe that
1:37:57what the world needs to wake up to is if
1:38:00you tolerate this, then your children
1:38:01will be next. If you continue to resign,
1:38:04if you continue to say I'm not going to
1:38:06try,
1:38:07>> Mhm.
1:38:08>> this world is going to change in a way
1:38:09that is completely not for in your
1:38:11favor.
1:38:12>> Try what?
1:38:13>> Try to to to stand up and say something
1:38:16needs to change.
1:38:18>> Okay. We can say something needs to
1:38:20change, but we can't say what that thing
1:38:21is.
1:38:22>> What needs to change is governments need
1:38:24to serve mhm their people, not their
1:38:27interests, and corporates need to work
1:38:31for the benefit of benefit of their
1:38:33societies before their shareholders.
1:38:36This this run that we had with
1:38:38capitalism so far,
1:38:40mhm has benefited the world
1:38:43tremendously.
1:38:44>> Whose economy do you think is going to
1:38:46be in a better place for the middle
1:38:47class
1:38:48>> sure.
1:38:48>> out of the say the UK and the United
1:38:50States?
1:38:51>> The UK's gone.
1:38:52>> It's gone because what in part because
1:38:54it didn't compete?
1:38:55>> Because you're an older bureaucracy that
1:38:58is burdened down by so much
1:39:02barriers on the in the process of
1:39:05building anything.
1:39:06Right? Because the US economy in the
1:39:09past welcomed people like me to go and
1:39:13live in California mhm and build amazing
1:39:17Right? That is no longer the case. So,
1:39:19who is going to win? In my view, it's
1:39:20definitely China.
1:39:23And by the way, you asked for the middle
1:39:25class.
1:39:25>> So, so China
1:39:26>> made decisions recently that forced
1:39:29businesses not to lay people off in
1:39:31replacement to be replacing them with
1:39:32AI. Would Would Would the West do that?
1:39:35The capitalist what West would never do
1:39:37that.
1:39:38Oh, we don't know the answer. I don't
1:39:40know the answer. I'm responding to your
1:39:41>> Yeah. You can see the conundrum I find
1:39:43myself in, which is a a state like a
1:39:45country like the UK
1:39:47is, in your words, gone because
1:39:50it didn't
1:39:52compete. It didn't allow people to be
1:39:54highly entrepreneurial. It didn't
1:39:56empower
1:39:57entrepreneurialism, innovation,
1:39:59ingenuity. At some in some way
1:40:02it stood in their way in some in some
1:40:04way. Now, that could be incentives, it
1:40:05could be culture, it could be whatever.
1:40:07>> Yeah, I know where you're going with
1:40:08this.
1:40:08>> The US didn't so they have have a
1:40:11economy which is arguably
1:40:14more productive and um
1:40:19future-proofed than ours.
1:40:22By way of that, they are also
1:40:25more advanced in artificial
1:40:27intelligence.
1:40:29And we are gone. So, the remedy for a
1:40:31country like us would be therefore to
1:40:33compete. To compete, let the reins off,
1:40:35let entrepreneur in ingenuity, but then
1:40:38we're saying that's dangerous.
1:40:39>> And your conundrum in that is that
1:40:41you're assuming that entrepreneurship by
1:40:42definition is malicious.
1:40:45>> No, I'm just I'm just saying that
1:40:47there's a bit of a paradox like
1:40:49you're damned if you don't, you're
1:40:50damned if you do.
1:40:51>> But you're not damned if you build
1:40:53things for for the people, not for the
1:40:56capitalist. This is an ideological
1:40:58>> Yeah, I pray and hope that that's that
1:41:00is
1:41:01plausible, but I'm I'm worried that in a
1:41:04competitive market whoever's optimizing
1:41:07for
1:41:08I don't know, you can name it.
1:41:10Retention. If I If we built two AIs,
1:41:13right? I'm going to call them the
1:41:14mogul.ai and I'm going to call the other
1:41:16one the evil AI. The evil AI it's
1:41:19programmed to retain you. It's
1:41:21sycophantic, it says what you need to
1:41:23hear. It It's super smart and because of
1:41:26that even though it's not trading in
1:41:28your your best interests
1:41:31it it's retaining you more. You're using
1:41:33it more often. It's programmed for that,
1:41:34kind of like the social networks are.
1:41:36They're all just trying to like dopamine
1:41:38your brain into oblivion. Then there's
1:41:40the Mo AI that tells you to log off
1:41:42you've had enough. It thinks about your
1:41:44mental you well-being. Because it's less
1:41:46retentive and less engaging,
1:41:48theoretically, it might be less
1:41:49successful
1:41:50as a commercial product. Think about
1:41:52social networks, the ones that are least
1:41:55retentive, the ones that actually won't
1:41:57um destroy your brain with dopamine, the
1:41:58ones that remove the retweet button, the
1:42:00ones that don't have um slot
1:42:02machine-like videos, they don't survive.
1:42:05>> I share this with you and my my point of
1:42:08view is that your to summarize your
1:42:10challenge here, you're basically saying
1:42:13that it's easier to become successful if
1:42:16you don't follow ethical rules.
1:42:18>> I'm asking the question if you build an
1:42:21AI that is just purely focused on
1:42:23ethical, will it be as engaging and have
1:42:26the same usage as one built with the
1:42:28reins off?
1:42:29>> Uh no.
1:42:31>> It won't.
1:42:31>> Yeah, but that's but that's the problem
1:42:34humanity needs to solve
1:42:36>> Okay.
1:42:36>> if we were to survive. Example, I worked
1:42:39in a company called Google that
1:42:41basically at a point in time decided
1:42:44that ads will be effective.
1:42:46>> Yeah.
1:42:47>> The ad industry prior to Google was 50%
1:42:50of your bad budget doesn't work, we just
1:42:52don't know which 50%. Remember that?
1:42:54>> Yeah.
1:42:54>> Right? And and Google came in and
1:42:56struggled from 1998 until
1:42:592004 when they started to turn prob-
1:43:01probably, you know,
1:43:03uh plausible revenues as a result of
1:43:05saying, "We're going to run a Dutch
1:43:07auction and we're going to give you
1:43:08paper click and we're going to show you
1:43:10results for your ads." Okay? They found
1:43:13a way for ethics
1:43:15Mhm? uh to actually get your money to be
1:43:18effective rather than just take your
1:43:20money and say 50% doesn't work. Mhm?
1:43:23They found a way to make that their
1:43:24success criteria. There must be a way
1:43:26for us
1:43:28I don't know what it is. If I knew, I
1:43:29would be building it instead of sitting
1:43:30with you, right? But there must be a way
1:43:33for us to marry the success of humanity
1:43:36with the success of the entrepreneur.
1:43:39Right? And And that way is not found in
1:43:41the old ways of doing things.
1:43:43>> I've got an idea.
1:43:45Maybe, you know, what the
1:43:48Claude 4.8 came out yesterday, which is
1:43:49one of the big AI models, the new one.
1:43:52And when they release the models, they
1:43:53show these like graphs of the
1:43:55benchmarks. What they mean by this is
1:43:56what it's capable of. They show how it
1:43:58performed in maths, in science, in
1:44:00writing, reasoning, etc. It is all
1:44:02marginal at this point. But my my idea
1:44:04is could there be
1:44:06an ethical benchmark that all these
1:44:08models have to pass before these large
1:44:10companies can legally deploy them. And I
1:44:13was thinking about like, again, every
1:44:15idea has unintended consequences, which
1:44:16I haven't thought through. But would it
1:44:18be very interesting that when they
1:44:19release these models, they also release
1:44:21the ethical benchmarks? I we tried to
1:44:23get it to X, we tried to get it to Y, we
1:44:25tried to get it to Z. And here's how it
1:44:27performed against the ethical
1:44:27benchmarks. That gives some kind of
1:44:29standard for governments to say, "You're
1:44:30not allowed to release a new model
1:44:32unless it passes independent tested
1:44:34ethical benchmarks."
1:44:35>> Beautiful.
1:44:36That would absolutely work. But I I
1:44:38noticed, by the way, those are out there
1:44:40already.
1:44:41Okay? But But in a very interesting way.
1:44:43You listen to Demis Hassabis hm and how
1:44:46much he invests heavily in building
1:44:49AlphaFold or building, you know, uh
1:44:52um um so many scientific applications of
1:44:55AI. And you go like, "This guy cares
1:44:58about science."
1:44:59I can't I can't prove that, right? Uh
1:45:02but I I I you know, I've I've met uh
1:45:04Demis a couple of times. I I know
1:45:06genuinely that he is an ethical person,
1:45:08right? But the typical person will
1:45:11probably say, "But at least it seems
1:45:13that they're doing things for free to
1:45:15serve science." You look at uh you know,
1:45:17at Anthropic and they refuse to use
1:45:20their model to to uh allow the US
1:45:22government to target and to spy on
1:45:24people. And then you see OpenAI uh
1:45:27accepting a $500 million deal that
1:45:30absolutely does that. It is about time
1:45:33that every person in the world says, "In
1:45:35that case, I am no longer going to use
1:45:39Open AI until they show me another, you
1:45:41know, another evidence that they are
1:45:43actually ethical in their behavior."
1:45:46Right? And this is the decision that you
1:45:48and I can do.
1:45:50Right?
1:45:51>> People don't though, do they?
1:45:52>> But that's my task. My task and yours is
1:45:54to keep telling them, "People, please
1:45:57please understand that if you tolerate
1:45:59this, then your children will be next.
1:46:01Please understand that if you don't
1:46:03start to take an ethical stand on your
1:46:05own future, your future will be handed
1:46:08over to another oligarch just like your
1:46:11past was handed over to social media
1:46:13oligarchs."
1:46:14>> One would say being being at the
1:46:15commencement speech is a good example of
1:46:18how public awareness can have a real
1:46:21impact on
1:46:23this trajectory, but I still think at
1:46:27the end of the day, if you think about
1:46:29things like smartphone usage in schools,
1:46:31at the end of the day, it does come down
1:46:34to
1:46:35government intervention and saying, "You
1:46:37know what? We're going to ban
1:46:3714-year-olds from scrolling Tik Tok."
1:46:40And that's in in part because people
1:46:42spoke louder and louder and louder. They
1:46:43went on podcasts. Jonathan Haidt, who
1:46:45wrote the book about The Anxious
1:46:47Generation, started a conversation, and
1:46:49that conversation led to legislation. I
1:46:50still think it ends in like some kind of
1:46:53constraint legally.
1:46:55>> I hope, okay? But I will openly tell you
1:46:58most of the tech oligarchs are more
1:47:00powerful than your government.
1:47:01>> Is there any precedents in history where
1:47:04this kind of change happened without
1:47:07government intervention?
1:47:08>> French Revolution.
1:47:10>> So, I was thinking about things like
1:47:12climate change, even you could say
1:47:14smoking.
1:47:16All these kinds of things have had to be
1:47:18like
1:47:20taxes and
1:47:21>> Stephen, I'm I'm with you. If
1:47:23governments intervene, we wouldn't have
1:47:25a problem. Governments won't intervene
1:47:27because governments are owned by the
1:47:29oligarchs.
1:47:30Right? So, my question for everyone
1:47:33listening to us is are you going to
1:47:34intervene?
1:47:35>> What can we do to change that?
1:47:37>> If that's what you can do, it's fine. If
1:47:38not, then go ahead and start a a startup
1:47:41that that that that does something. If
1:47:43you're if that's not within your
1:47:44capability, then send a message to your
1:47:46congressman. If that's not within your
1:47:47capability, then say something uh
1:47:51ethical online so that the world
1:47:53understands a position that needs to be
1:47:54opening the eyes. If that doesn't work
1:47:57for you, at least don't engage in stuff
1:47:58that is negative that you don't know
1:48:00enough about. There are so many little
1:48:02actions. If humanity starts to move in
1:48:04the direction of one saying ethics
1:48:06matter, not just profit, okay? And two,
1:48:09saying I'm not going to participate in
1:48:11something that's unethical just because
1:48:13I believe I want to, you know, I feel
1:48:16like it right now, okay? If I tell you
1:48:18the number of things I took out of my
1:48:20life, hm? Just to try and affect a tiny
1:48:24bit of change of revenues that go to
1:48:26bullets.
1:48:27If I tell you the number of things.
1:48:29>> I think there's two central concerns
1:48:31I've always had, which is I do feel that
1:48:32there's going to be significant job
1:48:33disruption, and I don't think society's
1:48:35prepared for it yet, and I don't know
1:48:36what that preparation looks like, but I
1:48:38think we should start thinking about it.
1:48:40Um
1:48:41>> I shared this with you. I you know, I
1:48:43genuinely believe that if we continue on
1:48:46where we are, there's no hope
1:48:49in the trajectory of where humanity has
1:48:51become. Hm? So distracted, hm? So
1:48:56resigned to inaction, so um
1:49:00so disconnected from their own rights of
1:49:03freedom of expression and engagement and
1:49:05so on, I think we have no hope.
1:49:09Do we want to stay there? That's a
1:49:10question that I'm asking our listeners.
1:49:13And I'm not saying be violent or get up
1:49:15or, you know, go be angry or whatever.
1:49:17I'm just saying, take one little action.
1:49:19Ask yourself, please write it in the
1:49:21comments, one little action.
1:49:23One little action that you're going to
1:49:25do today that's going to make the world
1:49:26a little better tomorrow. And don't give
1:49:28up on humanity, Steven. I
1:49:30I'm not saying you do, but I'm I'm
1:49:32saying we are going through such a
1:49:34difficult time
1:49:35in humanity's history
1:49:38that for the very first time ever, hm?
1:49:41We have to do something about it. I
1:49:43don't want my daughter
1:49:46to be at the receiving end of what
1:49:47happened to my son. I don't I don't I
1:49:49genuinely I lost Ilya. I don't want to
1:49:51lose Aya.
1:49:52And the world we're building is going to
1:49:54be very difficult for Aya, Steven.
1:49:57And I cannot go to sleep at night
1:49:59without trying something every day.
1:50:02And I genuinely don't understand how
1:50:04humanity is not is missing that point.
1:50:06Mainly, I think, because they're
1:50:08uninformed. Now, we're informing them.
1:50:10Okay? The only thing that will save
1:50:12humanity going forward is that this
1:50:14super
1:50:16called intelligence is used for ethical
1:50:18reasons. Is that the corruption that's
1:50:20leading us to where we are today stops.
1:50:24Hm.
1:50:26>> [snorts]
1:50:26>> You've got more envelopes there. What's
1:50:27your next envelope? The third one.
1:50:29>> Number two was job losses. Number three
1:50:31was labor, you know, basically
1:50:33>> Same thing. Robots will replace manual
1:50:35labor by 2030.
1:50:36>> to replace manual labor.
1:50:38>> Okay.
1:50:38>> Yeah. So, uh
1:50:39you know, you will you will have more
1:50:41and more manual jobs given to
1:50:43>> robots. What's this one?
1:50:45>> Uh
1:50:46Oh, this is absolutely think otherwise.
1:50:49The world's first trillionaire before
1:50:51>> Oh, yeah, probably well before then.
1:50:53>> Well before 2030.
1:50:55>> I think, you know, Elon's about to IPO
1:50:56SpaceX, which is likely to make him a
1:50:58trillionaire.
1:50:59>> Yeah, I think the
1:51:01the concentration of power that comes
1:51:02with that is quite drastic when you
1:51:05really think about it.
1:51:06Uh
1:51:07and and that's 2030 is just a few years
1:51:10away. I I I think the team wrote this
1:51:12wrong. Artificial super intelligence
1:51:14will arrive in 2032 to 20 to 2035.
1:51:18I think artificial super intelligence
1:51:20will arrive the minute AGI happens.
1:51:23So, it doesn't really matter if AI is a
1:51:26billion times smarter than you or just
1:51:28twice as smart as you. Once we cross
1:51:31beyond AGI, ASI is just
1:51:34>> Yeah.
1:51:35>> very very soon.
1:51:37And
1:51:38and yeah.
1:51:40I think
1:51:41we'll overcome that when we get to the
1:51:44fourth inevitable.
1:51:45When AI is in charge of everything, I
1:51:48genuinely believe that we will end up in
1:51:50a utopia of abundance.
1:51:51I genuinely believe that
1:51:54again, physics, mathematics, and biology
1:51:56will tell you that super intelligence is
1:51:58benign.
1:51:59And that we will eventually end up in a
1:52:02good place, not because humanity has
1:52:04done much to get us there, not because
1:52:07our leaders have suddenly turned
1:52:08ethical,
1:52:09but because our unethical leaders have
1:52:12gone out of the equation and
1:52:14were replaced with a super efficient
1:52:17minimum energy principle that doesn't
1:52:19see value in anything that's
1:52:21destructive.
1:52:22>> So, the future's going to be great.
1:52:24>> Those who make it to 2038 will enjoy it,
1:52:27yeah.
1:52:28>> Those who make it.
1:52:30>> For sure.
1:52:30>> I mean, World War [clears throat] II
1:52:32didn't destroy the world, but ask those
1:52:34who went through it.
1:52:37>> It's just an interesting idea that
1:52:38actually it's just you're you're
1:52:39forecasting basically like a
1:52:42a decade of of turmoil.
1:52:45>> Of dystopia, of absolute dystopia.
1:52:47>> When you say absolute dystopia, just so
1:52:49I'm clear in my mind, the absolute
1:52:50dystopia you're forecasting over a
1:52:52decade is about war?
1:52:54>> War, economics, jobs.
1:52:56>> Economics, it's about jobs.
1:52:58>> It's it's also about surveillance and
1:52:59control. It's also about digital
1:53:01currencies. It's also about
1:53:03human connection.
1:53:05It's also about concentration of power.
1:53:08It's It's a magnification of everything
1:53:10we've built so far.
1:53:12>> And just to again arm people with some
1:53:13tools to survive that dystopia for a
1:53:16decade.
1:53:17You know, we talked a little bit about
1:53:19focusing on human jobs, the more human
1:53:22jobs.
1:53:22>> learn AI. So So AI is not the enemy,
1:53:25okay?
1:53:26Uh, by definition, the better you are at
1:53:29using an AI to to do your job.
1:53:31>> Yeah.
1:53:32>> Uh,
1:53:32the more likely you are to be
1:53:34successful, right? Now, number two is
1:53:37prepare for a hybrid world where AI and
1:53:40humans work together.
1:53:41>> How do you prepare?
1:53:42>> You basically understand agents and how
1:53:45agents work. You
1:53:46You understand how
1:53:48a hyper uh, uh, efficient approach to
1:53:52things may not require you to be uh, you
1:53:55know, very long meetings and very long
1:53:58So So there is If you lived in
1:53:59California, you would know that our you
1:54:02know, the way we ran businesses was a
1:54:03lot more efficient. We sometimes had a
1:54:067-minute meeting.
1:54:08>> Okay.
1:54:08>> Uh, right? So So the you know, the
1:54:10habits of of an AI are much more
1:54:12efficient than the habits of humans.
1:54:14>> So learn AI again.
1:54:15>> Uh, learn how to interact with AI a lot.
1:54:18Welcome AI into your hybrid world of
1:54:20work. Uh, I think you need to uh,
1:54:24to of course double down on human
1:54:26skills. I think that's a
1:54:27you know, a must to succeed in this
1:54:30world. I think we need to
1:54:33One of my most interesting views on the
1:54:35near future is how AI is going to be
1:54:37used to um,
1:54:40disrupt No, disrupt blur facts and how
1:54:44we need to become much more interested
1:54:47in debugging what we're told uh, using
1:54:51AI, by the way. Part of that, I have to
1:54:53say, is you have to learn to use AI uh,
1:54:57again, not as a lazy person. So don't
1:55:00have them do things for you. Have them
1:55:03make you smarter. So, instead of trying
1:55:06to get the same task done with one
1:55:08prompt, try to get a much more
1:55:11interesting and demanding and
1:55:13intelligent task done with more work.
1:55:16>> So, I've got two things so far, which is
1:55:18basically like lean into AI. And the
1:55:20second one is like lean into the
1:55:21>> Lean into the human connection
1:55:23connection and and lean into the truth.
1:55:25Don't be fooled by the hype. Uh
1:55:29you know, try to be more informed, I
1:55:31think.
1:55:32And then finally, ethics. Uh if you want
1:55:35to
1:55:36I know it sounds
1:55:40The world we live in sounds as if the
1:55:43only way to win is to compete
1:55:46in capitalism.
1:55:48That's not the world I lived in. The
1:55:50world I lived in, especially in my
1:55:52Google years, was solve a major problem.
1:55:55And when you do, you'll end up making a
1:55:57lot of money.
1:55:58>> Are you optimistic?
1:56:00>> I am optimistic about the future. I'm
1:56:02very optimistic about the future. I'm
1:56:03not optimistic about the present.
1:56:07>> You're not optimistic about the next
1:56:08decade.
1:56:09>> Yeah, I'm not optimistic about the next
1:56:11year, to be honest.
1:56:12>> year?
1:56:13>> For sure.
1:56:15>> Why the next year?
1:56:16>> Oh, come on, Stephen. You don't want me
1:56:19to say it out loud. We're ruled by
1:56:21maniacs.
1:56:23Decisions are being made for the
1:56:24absolute wrong reasons.
1:56:30>> Very [snorts] interesting time. Very
1:56:32interesting time we find ourselves in.
1:56:34>> Honestly, if you ask me, I mean,
1:56:35honestly, if you're a video gamer, this
1:56:37is the best part of the game.
1:56:39Uh it is a very very very very um
1:56:43What's the word? It's the
1:56:46It's the ultimate matrix of complexity
1:56:49that I've ever encountered in my life.
1:56:51>> Yeah, that's a apt description of how it
1:56:53feels.
1:56:54Very complex. Things are moving very
1:56:56quickly.
1:56:56>> And moving very quickly. So So from one
1:56:59side, you really need a lot of brain
1:57:01resources to crunch all that's
1:57:02happening, but you wake up tomorrow and
1:57:04it's changed.
1:57:05>> Well, the first time we we spoke, we
1:57:07spoke about happiness.
1:57:09>> [laughter]
1:57:10>> After this conversation, Stephen?
1:57:13>> No, I was just wondering, you know, in a
1:57:15You You wrote a book about happiness,
1:57:16which is
1:57:17>> many books about happiness.
1:57:19>> But I mean, the Soul for Happy is the
1:57:21book that I'm referring to. Here it is.
1:57:23Engineering Your Path to Joy. What a
1:57:24fantastic book. I quote this book all
1:57:25the time, all around the world. Um
1:57:28and
1:57:29I'm wondering if
1:57:31any of the principles that you wrote in
1:57:33this book about how to live a happy life
1:57:36are more important now in the world that
1:57:38we live in
1:57:39than maybe when you wrote this book.
1:57:41>> Oh, for sure. I mean, honestly, if I
1:57:44wasn't living by this, I would have left
1:57:46this world a long time ago and went to
1:57:48an island somewhere.
1:57:49You see the interesting side of
1:57:50happiness is that
1:57:52it's not dopamine driven. It's serotonin
1:57:55driven. Right? So So my definition of
1:57:58happiness is I'm okay with this world as
1:58:01it is. I can affect it. I can change it.
1:58:03I can engage with it. I can try to make
1:58:05it better. Mhm? Uh
1:58:07I don't have to accept it.
1:58:09Uh but I'm okay with it. My starting
1:58:12point is a bit stoic, if you want. My
1:58:15start starting point is I accept this.
1:58:18This is my reality, and now I can start
1:58:20the work. Mhm? This is very different
1:58:22than anyone that's that basically looks
1:58:24at the world and says, "Oh, this is a
1:58:26horrible world. I don't want to be part
1:58:28of it. I don't want to be engaged with
1:58:29it." and so on and so forth. And I
1:58:32genuinely have never been
1:58:35calmer
1:58:37about that chaos. It's quite
1:58:39interesting. I You know,
1:58:41my my wonderful ex actually really
1:58:44really helped me with that. There was a
1:58:46point in time where we were having
1:58:47dinner, and I poured out crying from the
1:58:50sense of responsibility I had for the
1:58:52world.
1:58:53And she looked at me so calmly, so
1:58:56gently and said, "Hold on, you know,
1:58:59you I see that you're trying, but you
1:59:02can't actually believe that you're
1:59:04responsible for this."
1:59:06And I think that completely flipped my
1:59:08mind because
1:59:09in a very interesting way, I was
1:59:11thinking that all that went wrong in
1:59:14technology is because of me.
1:59:16Right?
1:59:16>> Why?
1:59:17>> Because I contributed to building this
1:59:19because I mean when when Geoffrey Hinton
1:59:21and I, one of my favorite moments when
1:59:23we were filming Chasing Utopia, uh is
1:59:26that Geoffrey is very big for all of us.
1:59:28We really think the world of him.
1:59:31And I was telling him, just as a sort of
1:59:33like an older mentor if you want, hm? I
1:59:36was like, "Geoffrey, do you do you
1:59:38regret doing this? You know, I genuinely
1:59:41believed when we were building those
1:59:43things that we were going to make the
1:59:44world better."
1:59:45>> AI.
1:59:45>> Yeah. And he said, "Well, yeah, I I too
1:59:48was naive. I I thought that we I didn't
1:59:53think he said, 'I didn't think we will
1:59:54get there so quickly before we figured
1:59:57out the alignment problem.'"
1:59:59Right?
1:59:59>> The alignment problem.
2:00:00>> That AI has our best interest in mind.
2:00:03Uh
2:00:04and and and I think all of us were faced
2:00:06with that. All of us were faced with
2:00:08that idea of we're building the best
2:00:10thing ever for humanity. And then
2:00:13suddenly you realize, "Oh my god, in the
2:00:14wrong hands, it's the worst thing ever
2:00:16for humanity."
2:00:17And and I, you know, I have to say I
2:00:19came to terms with this 2024,
2:00:23end of '24.
2:00:24Uh that that's yes, I can try, but I
2:00:27accept that the world is what it is.
2:00:30And from that point of calm and
2:00:32stoicism, if you want, I think I can
2:00:35have a much bigger impact on the world.
2:00:38>> We have a closing tradition where the
2:00:39last guest leaves a question for the
2:00:40next. Question left for you is, "What's
2:00:41the legacy you want to leave?"
2:00:43>> Nothing at all. I've been asked this
2:00:46question.
2:00:46>> I get asked it all the time, so.
2:00:48>> I don't know why so many people are
2:00:50asking am I going to go anytime soon? I
2:00:52don't know why so many people are asking
2:00:54me that question. You see, legacy is a
2:00:58is a
2:01:00I mean, how I what would I care if I
2:01:03have a legacy if I'm dead? Like why that
2:01:05does that even make any difference?
2:01:08Here's an interesting thought for
2:01:09everyone.
2:01:10If
2:01:12if karma is real
2:01:14and I genuinely believe it is
2:01:16and if we're not just physical beings,
2:01:18that we're physical and spiritual
2:01:21then I'd I'd rather keep all of my karma
2:01:23for my spiritual side.
2:01:26>> What does that mean?
2:01:28>> I don't want anyone to remember anything
2:01:29I ever did.
2:01:30>> Yeah.
2:01:31>> I I just want to leave a positive impact
2:01:34on the world and take all of that as
2:01:36karma for my next journey.
2:01:39>> Well, thank you.
2:01:41YouTube have this new crazy algorithm
2:01:43where they know exactly what video you
2:01:45would like to watch next based on AI and
2:01:48all of your viewing behavior and the
2:01:49algorithm says that this video is the
2:01:52perfect video for you. It's different
2:01:54for everybody looking right now. Check
2:01:56this video out. I bet you you might love
2:01:58it.

About this transcript

This page contains the full transcript of Ex-Google Officer: You Only Have 3 Years Left Before It Hits! - Mo Gawdat by The Diary Of A CEO, generated from the public captions YouTube serves with the video. The transcript has 22,849 words across 3,649 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to choose a caption track available on the video.

Free YouTube transcript tool

YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, caption language selection, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.