[0:00] So, ChachiT kind of set off the starting
[0:02] pistol for this this whole race and
[0:05] subsequently it appears that every other
[0:07] major technology company now is
[0:09] investing godly amounts ungodly amounts
[0:12] of money in competing in this AI race
[0:14] and they're pursuing this thing called
[0:16] AGI which we hear this word used a lot.
[0:18] >> Yes.
[0:18] >> What is what is AGI and how is that
[0:20] different from what I use at the moment
[0:21] on chatb or Gemini?
[0:23] >> Yeah. So that's the thing that people
[0:25] really need to get is that these
[0:27] companies are not racing to provide a
[0:29] chatbot to users. That's not what their
[0:31] goal is. If you look at the mission
[0:32] statement on OpenAI's website or all the
[0:34] websites, their mission is to be able to
[0:36] replace all forms of human economic
[0:39] labor in the economy. Meaning an AI that
[0:42] can do all the cognitive labor meaning
[0:44] labor of the mind. So that that can be
[0:46] marketing, that can be text, that can be
[0:48] illustration, that can be video
[0:49] production, that can be code production.
[0:52] Everything that a person can do with
[0:54] their brain, these companies are racing
[0:56] to build that. That is artificial
[0:59] general intelligence. General meaning
[1:01] all kinds of cognitive tasks. Demis
[1:04] Hassabis the co-founder of um Google
[1:07] DeepMind used to say first solve
[1:09] intelligence and then use that to solve
[1:12] everything else. Like it's important to
[1:13] say why why is AI distinct from all
[1:15] other kinds of technologies? It's
[1:17] because if I make an advance in one
[1:19] field like rocketry, if I just let's say
[1:22] I uncover some secret in rocketry, that
[1:24] doesn't advance like biio medicine
[1:27] knowledge or it doesn't advance energy
[1:28] production or doesn't advance coding.
[1:31] But if I can advance generalized
[1:33] intelligence, think of all science and
[1:35] technology development over the course
[1:36] of all human history. So science and
[1:38] technology is all done by humans
[1:41] thinking and working out problems.
[1:42] Working out problems in any domain. So
[1:45] if I automate intelligence, I'm suddenly
[1:47] going to get an explosion of all
[1:49] scientific and technological development
[1:51] everywhere. Does that make sense?
[1:53] >> Of course. Yeah. It's foundational to
[1:54] everything.
[1:55] >> Exactly. Which is why there's a belief
[1:57] that if I get there first and can
[1:58] automate generalized intelligence, I can
[2:01] own the world economy because suddenly
[2:04] everything that a human can do that they
[2:05] would be paid to do in a job, the AI can
[2:07] do that better. And so if I'm a company,
[2:10] do I want to pay the human who has
[2:12] health care, might whistleblow,
[2:13] complains, you know, has to sleep, has
[2:16] sick days, has family issues, or do I
[2:18] want to pay the AI that will work 24/7
[2:20] at superhuman speed, doesn't complain,
[2:23] doesn't whistleblow, doesn't have to be
[2:24] paid for healthcare. There's the
[2:26] incentive for everyone to move to paying
[2:29] for AIs rather than paying humans. And
[2:32] so AGI, artificial general intelligence,
[2:35] is more transformative than any other
[2:37] kind of of technology that we've ever
[2:39] had and it's distinct.
[2:41] >> With the sheer amount of money being
[2:44] invested into it and the money being
[2:46] invested into the infrastructure, the
[2:47] physical data centers, the chips, the
[2:49] compute,
[2:51] do you think we're going to get there?
[2:54] Do you think we're going to get to AGI?
[2:55] >> I do think that we're going to get
[2:56] there. It's not clear uh how long it
[2:58] will take. And I'm not saying that
[3:00] because I believe necessarily the
[3:01] current paradigm that we're building on
[3:03] will take us there, but you know, I'm
[3:05] based in San Francisco. I talked to
[3:06] people at the AI labs. Half these people
[3:08] are friends of mine. You know, people at
[3:09] the very top level. And you know, most
[3:13] people in the industry believe that
[3:15] they'll get there between the next two
[3:16] and 10 years at the latest. And I think
[3:19] some people might say, "Oh, well, it may
[3:20] not happen for a while. Phew. I can sit
[3:22] back and we don't have to worry about
[3:23] And it's like we're heading for so much
[3:26] transformative change faster than our
[3:28] society is currently prepared to deal
[3:29] with it. The reason I was excited to
[3:32] talk to you today is because I think
[3:33] that people are currently confused about
[3:35] AI. You know, people say it's going to
[3:36] solve everything, cure cancer, uh solve
[3:39] climate change, and there's people say
[3:40] it's going to kill everything. It's
[3:41] going to be doom. Everyone's going to go
[3:43] extinct. If anyone builds it, everyone
[3:44] dies. And those those conversations
[3:46] don't converge. And so everyone's just
[3:48] kind of confused where how can it be,
[3:50] you know, infinite promise and how can
[3:52] it be infinite peril? And what I wanted
[3:54] to do today is to really clarify for
[3:56] people what the incentives point us
[3:58] towards which is a future that I think
[4:00] people when they see it clearly would
[4:02] not want.
[4:03] >> So what are the incentives pointing us
[4:05] towards in terms of the future?
[4:08] >> So first is if you believe that this is
[4:10] like it's metaphorically it's like the
[4:12] ring from Lord of the Rings. It's the
[4:13] ring that that creates infinite power
[4:16] because if I have AGI, I can apply that
[4:19] to military advantage. I can have the
[4:20] best military planner that can beat all
[4:22] battle plans for anyone. And we already
[4:24] have AIs that can obviously beat Gary
[4:27] Kasparov at chess, beat Go, the Go Asian
[4:30] um board game, or now beat Starcraft. So
[4:33] you have AI that are beating humans at
[4:34] strategy games. Well, think about
[4:36] Starcraft compared to an actual military
[4:39] campaign, you know, in Taiwan or
[4:40] something like that. If I have an AI
[4:42] that can out compete in strategy games,
[4:44] that lets me out compete everything. Or
[4:46] take business strategy. If I have an AI
[4:48] that can do business strategy and figure
[4:49] out supply chains and figure out how to
[4:51] optimize them and figure out how to
[4:52] undermine my competitors
[4:54] and I have a, you know, a step function
[4:56] level increase in that compared to
[4:57] everybody else, then that gives me
[4:59] infinite power to undermine and out
[5:01] compete all businesses. If I have a
[5:03] super programmer, then I can out compete
[5:06] programming. 70 to 90% of the code
[5:08] written at today's AI labs is written by
[5:10] AI.
[5:12] >> Think about the stock market as well.
[5:14] >> Think about the stock market. If I have
[5:15] an AI that can trade in the stock market
[5:16] better than all the other AIs, because
[5:19] they're currently there's mostly AIs
[5:20] that are actually trading on the stock
[5:21] market, but if I have a jump in that,
[5:23] then I can consolidate all the wealth.
[5:25] If I have an AI that can do cyber
[5:26] hacking, that's way better at cyber
[5:28] hacking in a step function above what
[5:30] everyone else can do, then I have an
[5:32] asymmetric advantage over everybody
[5:33] else. So AI is like a power pump. It
[5:37] pumps economic advantage. It pumps
[5:40] scientific advantage and it pumps
[5:42] military advantage. Which is why the
[5:44] countries and the companies are caught
[5:46] in what they believe is a race to get
[5:48] there first. And anything that is a
[5:50] negative consequence of that, job loss,
[5:53] rising energy prices, more emissions,
[5:56] stealing intellectual property, you
[5:58] know, security risks, all of that stuff
[6:00] feels small relative to if I don't get
[6:03] there first, then some other person who
[6:06] has less good values as me, they'll get
[6:08] AGI and then I will be forever a slave
[6:10] to their future. And I know this might
[6:12] sound crazy to a lot of people, but this
[6:14] is how people in at the very top of the
[6:17] AGI AI world believe is currently
[6:20] happening. And that's what
[6:21] >> conversations.
[6:21] >> Yeah.
[6:24] >> You you've had I mean know Jeff Hinton
[6:26] and and Roman Ylonsky on and other
[6:28] people Mogadat and they're saying the
[6:30] same thing. And I think people need to
[6:31] take seriously that whether you believe
[6:34] it or not, the people who are currently
[6:36] deploying the trillions of dollars, this
[6:38] is what they believe. And they believe
[6:40] that it's win or take all. And it's not
[6:42] just first solve intelligence and use
[6:44] that to solve everything else. It's
[6:45] first dominate intelligence and use that
[6:48] to dominate everything else.
[6:49] >> Have you had concerning private
[6:51] conversations about this subject matter
[6:52] with people that are in the industry?
[6:55] >> Absolutely. I think that's what most
[6:58] people don't understand is that um
[7:00] there's a different conversation
[7:02] happening publicly than the one that's
[7:03] happening privately. I think you're
[7:05] aware of this as well.
[7:05] >> I am aware of this.
[7:06] >> What do they say to you?
[7:10] >> So, it's not always the people telling
[7:13] me directly. It's usually one step
[7:15] removed. So, it's usually someone that I
[7:17] trust and I've known for many, many
[7:19] years who at a kitchen table says, "I
[7:21] met this particular CEO. We were in this
[7:23] room talking about the future of AI.
[7:25] this particular CEO they're referencing
[7:26] is leading one of the biggest AI
[7:28] companies in the world and then they'll
[7:29] explain to me what they think of the
[7:31] future's going to look like and then
[7:32] when I go and watch them on YouTube or
[7:34] podcasts what they're saying is they
[7:36] they have this real public bias towards
[7:38] the abundance part that you know we're
[7:40] going to cure cancer
[7:41] >> cure cancer universal high income for
[7:43] everyone
[7:44] >> yeah all this all this stuff
[7:46] >> doesn't work anymore
[7:47] >> but then privately what I hear is is
[7:49] exactly what you said which is really
[7:51] terrifying to me there was actually
[7:52] since since the last time we had a
[7:54] conversation about AR and podcast, I was
[7:56] speaking to a friend of mine, very
[7:58] successful billionaire, knows a lot of
[7:59] these people, and he is concerned
[8:02] because his argument is that if there's
[8:04] even like a 5% chance of the adverse
[8:09] outcomes that we hear about, we should
[8:12] not be doing this. And he was saying to
[8:13] me that some of his friends who are
[8:15] running some of these companies believe
[8:16] the chance is much higher than that, but
[8:18] they feel like they're caught in a race
[8:20] where if they don't control this
[8:21] technology and they don't get there
[8:22] first and get to what they refer to as
[8:25] um takeoff, like fast takeoff.
[8:27] >> Yeah. Uh recursive self-improvement or
[8:29] fast takeoff, which basically means what
[8:31] the companies are really in a race for
[8:33] you're pointing to is they're in a race
[8:35] to automate AI research. Um because so
[8:38] right now you have open AI, it's got a
[8:40] few thousand employees. Human beings are
[8:43] coding and doing the AI research.
[8:45] They're reading the latest research
[8:46] papers. They're writing the next, you
[8:48] know, they're hypothesizing what's the
[8:50] improvement we're going to make to AI.
[8:51] What's a new way to do this code? What's
[8:52] a new technique? And then they use their
[8:54] human mind and they go invent something.
[8:57] They they run the experiment and they
[8:58] see if that improves the performance.
[9:00] And that's how you go from, you know,
[9:01] GPT4 to GPT5 or something. Imagine a
[9:05] world where Sam Alman can instead of
[9:07] having human AI researchers can have AI
[9:11] AI researchers. So now I just snap my
[9:14] fingers and I go from one AI that reads
[9:17] all the papers, writes all the code,
[9:18] creates the new experiments to I can
[9:21] copy paste a 100 million AI researchers
[9:24] that are now doing that in an automated
[9:26] way. And it the belief is not just that,
[9:29] you know, the companies look like
[9:30] they're competing to release better chat
[9:32] bots for people, but the what they're
[9:33] really competing for is to get to this
[9:36] milestone of being to automate an
[9:38] intelligence explosion or automate
[9:40] recursive self-improvement, which is
[9:42] basically automating AI research. And
[9:44] that, by the way, is why all the
[9:46] companies are racing specifically to get
[9:49] good at programming because the faster
[9:51] you can automate a human programmer, the
[9:54] more you can automate AI research. And
[9:56] just a couple weeks ago, Cloud 4.5 was
[9:59] released and it can do 30 hours of
[10:02] uninterrupted complex programming tasks
[10:05] at the at the high end.
[10:07] That's crazy.
[10:09] So right now one of the limits on the
[10:10] progress of AI is that human humans are
[10:12] doing the work but actually all of these
[10:14] companies are pushing to the moment when
[10:16] AI will be doing the work which means
[10:17] they can have an infinite arguably
[10:19] smarter zerocost workforce that's right
[10:22] scaling the AI. So when they talk about
[10:24] fast takeoff they mean the moment where
[10:26] they where the AI takes control of the
[10:27] research and it and progress rapidly
[10:29] increases
[10:30] >> and it self-learns and recursively
[10:31] improves and invents. Um, so one thing
[10:34] to get is that AI accelerates AI, right?
[10:37] Like if I invent nuclear weapons,
[10:40] nuclear weapons don't invent better
[10:41] nuclear weapons.
[10:42] >> Yeah.
[10:43] >> But if I invent AI, AI is intelligence.
[10:46] Intelligence automates better
[10:47] programming, better chip design. So I
[10:49] can use AI to say, here's a design for
[10:51] the NVIDIA chips. Go make it 50% more
[10:53] efficient. And it can find out how to do
[10:54] that. I can say AI, here's a supply
[10:56] chain that I need for all the things for
[10:58] my AI company. And it can optimize that
[11:00] supply chain and make that supply chain
[11:01] more efficient.
[11:02] >> Mhm. AI, here's the code for making AI.
[11:04] Make that more efficient. Um, AI, here's
[11:07] training data. I need to make more
[11:08] training data. Go run a million
[11:10] simulations of how to do this and it'll
[11:12] train itself to get better.
[11:14] >> AI accelerates AI.
[11:15] >> What do you think these people are
[11:16] motivated by the CEOs of these
[11:18] companies?
[11:19] >> That's a good question.
[11:20] >> Genuinely, what do you think their
[11:21] genuine motivations are when you think
[11:23] about all these names?
[11:26] >> I think it's a subtle thing.
[11:29] I think
[11:31] there's um it's almost mythological
[11:35] because
[11:37] there's almost a way in which they're
[11:38] building a new intelligent entity that
[11:41] has never before existed on planet
[11:43] Earth. It's like building a god. I mean,
[11:45] the incentive is build a god, own the
[11:48] world economy, and make trillions of
[11:49] dollars, right? If you could actually
[11:52] build something that can automate all
[11:54] intelligent tasks, all goal achieving
[11:58] that will let you out compete
[11:59] everything. So that is a kind of godlike
[12:02] power that I think relative imagine
[12:05] energy prices go up or hundreds of
[12:06] millions of people lose their jobs. That
[12:09] those things suck. But relative to if I
[12:11] don't build it first and build this god,
[12:14] I'm going to lose to some maybe worse
[12:15] person who I think in my opinion, not my
[12:17] opinion, Tristan, but their opinion
[12:19] thinks is a worse person. It's it's a
[12:21] kind of competitive logic that
[12:26] self-reinforces itself, but it forces
[12:28] everyone to be incentivized to take the
[12:31] most shortcuts, to care the least about
[12:33] safety or security, to not care about
[12:36] how many jobs get disrupted, to not care
[12:37] about the well-being of regular people,
[12:39] but to basically just race to this
[12:42] infinite prize. So, there's a quote that
[12:44] um a friend of mine interviewed a lot of
[12:46] the top people at the AI companies, like
[12:48] the very top, and he just came back from
[12:50] that and and basically reported back to
[12:52] me and some friends, and he said the
[12:54] following.
[12:56] In the end, a lot of the tech people I
[12:58] talk to when I'm when I really grill
[12:59] them on it about like why you're doing
[13:01] this, they retreat into number one,
[13:04] determinism,
[13:06] number two, the inevitable replacement
[13:08] of biological life with digital life,
[13:10] and number three, that being a good
[13:12] thing. Anyways, at its core, it's an
[13:15] emotional desire to meet and speak to
[13:17] the most intelligent entity that they've
[13:20] ever met. And they have some ego
[13:22] religious intuition that they'll somehow
[13:23] be a part of it. It's thrilling to start
[13:26] an exciting fire. They feel they'll die
[13:28] either way, so they prefer to light it
[13:30] and see what happens.
[13:33] >> That is the perfect description of the
[13:35] private conversations.
[13:36] >> Doesn't that match what what you have
[13:38] description,
[13:38] >> doesn't it? And that's the thing. So,
[13:40] people may hear that and they're like,
[13:40] "Well, that sounds ridiculous." But if
[13:42] you actually
[13:42] >> I just got goosebumps cuz it's the
[13:44] perfect description. Especially the part
[13:46] they'll think they'll die either way.
[13:47] >> Exactly. Well, and um worse than that,
[13:52] some of them think that in the case
[13:54] where they if they were to get it right
[13:55] and if they succeeded, they could
[13:57] actually live forever because if AI
[13:59] perfectly speaks the language of
[14:00] biology, it will be able to reverse
[14:03] aging aging, cure every disease. And and
[14:07] so there's this kind of I could become a
[14:09] god. And I'll I'll tell you um you know,
[14:11] you and I both have know people who've
[14:12] had private conversations. Well, one of
[14:15] them that I have heard from one of the
[14:17] co-founders of one of the most, you
[14:19] know, powerful of these companies when
[14:22] when faced with the idea that what if
[14:24] there's an 80% or 20% chance that
[14:27] everybody dies and gets wiped out by
[14:29] this, but an 80% chance that we get
[14:32] utopia. He said, well, I would clearly
[14:34] accelerate and go for the utopia.
[14:37] Given a 20% chance,
[14:41] >> it's crazy. People should feel you do
[14:44] not get to make that choice on behalf of
[14:45] me and my family. We didn't consent to
[14:49] have six people make that decision on
[14:50] behalf of eight billion people. We have
[14:53] to stop pretending that this is okay or
[14:54] normal. It's not normal. And the only
[14:57] way that this is happening and they're
[14:58] getting away with it is because most
[15:00] people just don't really know what's
[15:02] going on.
[15:03] >> Yeah. But I'm curious what what do you
[15:04] think when they
[15:05] >> I mean everything you just said. It's
[15:07] that last part about the 8020% thing is
[15:09] almost verbatim what I heard from a very
[15:11] good very successful friend of mine who
[15:13] is responsible for building some of the
[15:15] biggest companies in the world when he
[15:16] was referencing a conversation he had
[15:17] with the founder of maybe the biggest
[15:21] air company in the world. And it was
[15:22] truly shocking to me because because it
[15:25] was said in such a blasé way.
[15:27] >> Yes. It wasn't Yeah. That that's what I
[15:28] had heard in this particular situation.
[15:30] wasn't like
[15:33] a matter of fact.
[15:33] >> It was a matter of fact, it's just easy.
[15:34] Yeah, of course I would do the I would
[15:36] take the I roll the dice.
[15:38] >> And even Elon Musk said he actually said
[15:41] the same number in an interview with Joe
[15:42] Rogan. Um and if you listen closely when
[15:45] he said, "I decided I'd rather be there
[15:48] when it all happens. If it all goes off
[15:50] the rails, I decided in that worst case
[15:52] scenario, I decided that I' I'd prefer
[15:54] to be there when it happens." which is
[15:56] just is justifying racing to our
[15:58] collective suicide. Now, I also want
[16:00] people to know like you don't have to
[16:02] buy into the sci-fi level risks to be
[16:04] very concerned about AI. So, hopefully
[16:05] later we'll talk about um the many other
[16:08] risks that are already hitting us right
[16:10] now that you don't have to believe any
[16:11] of this stuff.
[16:12] >> Yeah. The the Elon thing I think is
[16:14] particularly interesting because for the
[16:16] last 10 years he was this slightly hard
[16:19] to believe voice on the subject of AI.
[16:21] He was talking about it being a huge
[16:23] risk and an extinction level.
[16:25] >> He was the first AI people. Yeah. He was
[16:26] saying this is more dangerous than
[16:28] nukes. He was saying I try to get people
[16:29] to stop doing it. This is summoning the
[16:31] demon. Those are his words, not mine.
[16:33] >> Yeah.
[16:33] >> Um we shouldn't do this. Supposedly, he
[16:35] used his first and only meeting with
[16:37] President Obama, I think in 2016, to
[16:40] advocate for global regulation and
[16:42] global controls on on AI um because he
[16:44] was very worried about it. And then
[16:46] really what happened is um Chachi BT
[16:50] came out and as you said that was the
[16:52] starting gun and now everybody was in an
[16:54] allout race to get there first. He
[16:56] tweeted words to the effect I'll put it
[16:58] on the screen. He tweeted that he had
[17:01] remained in I think he used a word
[17:04] similar to disbelief for some time like
[17:06] suspended disbelief but then he said in
[17:08] the same tweet that the race is now on.
[17:11] >> The race is on and I have to race
[17:12] >> and I have to go. I have no choice but
[17:14] to go. And he tried, he's basically
[17:15] saying, "I tried to fight it for a long
[17:17] time. I tried to deny it. I tried to
[17:18] hope that we wouldn't get here, but
[17:19] we're here now, so I have to go."
[17:21] >> Yeah.
[17:21] >> And
[17:23] at least he's being honest. He does seem
[17:26] to have a pretty honest track record on
[17:28] this because because he was the guy 10
[17:29] years ago warning everybody. And I
[17:31] remember him talking about it and
[17:32] thinking, "Oh god, this is like 100
[17:33] years away. Why are we talking about
[17:34] that?"
[17:34] >> I felt the same, by the way. Some people
[17:35] might think that I'm some kind of AI
[17:36] enthusiast and I'm trying to ratch I I
[17:38] didn't believe that AI was a thing to be
[17:40] worried about at all until suddenly the
[17:42] last 2 three years where you can
[17:44] actually see where we're headed. But um
[17:48] oh man there's just there's so much to
[17:49] say about all this and so if you think
[17:52] about it from their perspective it's
[17:55] like best case scenario I build it first
[17:58] and it's aligned and controllable
[18:01] meaning that it will take the actions
[18:02] that I want. It won't destroy humanity
[18:04] and it's controllable which means I get
[18:06] to be god and emperor of the world.
[18:09] Second scenario, it's not controllable
[18:12] but it's aligned. So I built a god and I
[18:14] lost control of it but it's now
[18:16] basically it's running humanity. It's
[18:17] running the show. It's choosing what
[18:19] happens. It's out competing everyone on
[18:21] everything. That's not that bad an
[18:23] outcome. Third scenario, it's not
[18:26] aligned. It's not controllable. And it
[18:28] does wipe everybody out. And that should
[18:30] be demotivating to that person to an
[18:32] Elon or someone. But in that scenario,
[18:36] they were the one that birthed the
[18:37] digital god that replaced all of
[18:39] humanity. Like this is really important
[18:41] to get because in nuclear weapons,
[18:44] the risk of nuclear war is an omni
[18:47] lose-lose outcome. Everyone wants to
[18:49] avoid that. And I know that you know
[18:51] that I know that we both want to avoid
[18:52] that.
[18:54] So that that motivates us to coordinate
[18:56] and to have a nuclear
[18:57] non-prololiferation treaty. But with AI,
[19:01] the worst case scenario of everybody
[19:03] gets wiped out is a little bit different
[19:06] for the people making that decision.
[19:08] Because if I'm the CEO of DeepSeek and I
[19:12] make that AI that does wipe out
[19:13] humanity, that's the worst case scenario
[19:15] and it wasn't avoidable because it was
[19:17] all inevitable. Then even though we all
[19:20] got wiped out, I was the one who built
[19:22] the digital god that replaced humanity.
[19:23] And there's kind of ego in that. And uh
[19:27] the god that I built speaks Chinese
[19:29] instead of English.
[19:30] >> That's the religious ego point.
[19:32] >> That's the ego.
[19:33] >> Such a great point because that's
[19:34] exactly what it is. It's like this
[19:35] religious ego where I will be
[19:36] transcendent in some way.
[19:38] >> And you notice that it it all starts by
[19:39] the belief that this is inevitable.
[19:41] >> Yeah.
[19:42] >> Which is like is this inevitable? It's
[19:44] important to note because
[19:47] if you believe it's if everybody who's
[19:49] building it believes it's inevitable and
[19:50] the investors funding it believe it's
[19:51] inevitable, it cocreates the
[19:54] inevitability.
[19:55] >> Yeah.
[19:55] >> Right.
[19:56] >> Yeah.
[19:57] >> And the only way out is to step outside
[20:00] the logic of inevitability because if if
[20:04] we are all heading to our collective
[20:05] suicide, which I don't know about you, I
[20:08] don't think that I don't want that. You
[20:10] don't want that. Everybody who loves
[20:12] life looks at their children in the
[20:13] morning and says,"I want I want the
[20:15] things that I love and that are sacred
[20:16] in the world to continue." That's what
[20:19] night that's what everybody in the world
[20:20] wants. And the only thing that is having
[20:24] us not anchor on that is the belief that
[20:26] this is inevitable and the worst case
[20:27] scenario is somehow in this ego
[20:29] religious way not so bad if I was the
[20:32] one who accidentally wiped out humanity
[20:34] because I'm not a bad person because it
[20:36] was inevitable anyway. M
[20:37] >> and I think the goal of of for me this
[20:40] conversation is to get people to see
[20:42] that that's a bad outcome that no one
[20:43] wants and we have to put our hand on the
[20:46] steering wheel and turn towards a
[20:48] different future because we do not have
[20:50] to have a race to uncontrollable
[20:52] inscrutable powerful AIs that are by the
[20:55] way already doing all the rogue sci-fi
[20:57] stuff that we thought only existed in
[20:59] movies like blackmailing people uh being
[21:02] self-aware when they're being tested
[21:04] scheming and lying and deceiving to copy
[21:06] their own code to keep themselves
[21:07] preserved. Like the stuff that we
[21:09] thought only existed in sci-fi movies is
[21:11] now actually happening. And that should
[21:14] be enough evidence to say
[21:17] we don't want to do this path that we're
[21:19] currently on. It's not that
[21:22] some version of AI progressing into the
[21:24] world is directionally inevitable, but
[21:26] we get to choose which of those futures
[21:28] that we want to have. If you love the D
[21:30] CEO brand and you watch this channel,
[21:32] please do me a huge favor. become part
[21:34] of the 15% of the viewers on this
[21:37] channel that have hit the subscribe
[21:38] button.