Latest / Mic Drop Comedy / The Threat and Promise of AI

Transcript

The Threat and Promise of AI · Mic Drop Comedy

• 0:06
  Speaker 1: What is your relationship now with this tech world, and
• 0:10
  Speaker 1: how does AI fit into that, and how do you
• 0:12
  Speaker 1: remain bullish on those innovations when they so clearly are
• 0:17
  Speaker 1: working to avoid any kind of regulation of these new innovations?
• 0:21
  Speaker 2: Okay, two things.
• 0:22
  Speaker 3: One, they're there because they're rich, not because they're tech
• 0:25
  Speaker 3: bros or because they just happen to make their money
• 0:27
  Speaker 3: in tech. I don't think that's really applicable. The AI side...
• 0:31
  Speaker 3: You know, I've been in technology for a long time,
• 0:33
  Speaker 3: and you can always look at a new tech, PCs, networks,
• 0:37
  Speaker 3: the Internet, streaming, whatever, and say, okay, in five years,
• 0:40
  Speaker 3: this is what's going to happen, right?
• 0:41
  Speaker 4: Have a good sense. With AI, you can't do that.
• 0:44
  Speaker 3: With large language models, we have no idea whether it's
• 0:47
  Speaker 3: going to zig or zag or what the impact is
• 0:49
  Speaker 3: going to be. And that's the good news and the
• 0:51
  Speaker 3: bad news. The good news is we're dominating right now globally,
• 0:54
  Speaker 3: the United States is.
• 0:55
  Speaker 4: The bad news is... In terms of? In terms of our.
• 0:59
  Speaker 3: the quality and the impact of the AI and the
• 1:02
  Speaker 3: advancements that we're introducing in AI, the research that we're doing.
• 1:06
  Speaker 3: We are, without question, the leader, and that's really important
• 1:09
  Speaker 3: from a defense perspective, military, et cetera, and also, you know,
• 1:12
  Speaker 3: from a business perspective. It's going to have a big
• 1:15
  Speaker 3: impact on this country. I personally think it's generally positive,
• 1:18
  Speaker 3: but there's a lot of uncertainty to come.
• 1:21
  Speaker 1: What gives you the hope that it's generally positive? Because
• 1:24
  Speaker 1: as a counterpoint, we heard the same thing about social media.
• 1:29
  Speaker 1: And we heard the same thing about all these different
• 1:31
  Speaker 1: innovations of the connectivity. And yet, every time I turn
• 1:36
  Speaker 1: on Congress, Zuckerberg is up there like, look, I'm really sorry.
• 1:41
  Speaker 4: I didn't know it was going to kill all your daughters.
• 1:45
  Speaker 5: Like.
• 1:46
  Speaker 4: No, remember, it's still just a short window.
• 1:47
  Speaker 3: Social media, you know, has really only been prominent last
• 1:51
  Speaker 3: six years, and I think we'll learn and we'll evolve,
• 1:54
  Speaker 3: and the same thing will happen with AI. There's going
• 1:56
  Speaker 3: to be points in time where it's f * *
• 1:58
  Speaker 3: * ed up, right?
• 1:58
  Speaker 4: And people are using it.
• 2:00
  Speaker 3: But I think over time, particularly with Gen Z, right?
• 2:03
  Speaker 3: Gen Z is a different beast. You know, boomers are idiots.
• 2:06
  Speaker 3: I mean, we went from sexual... We went from sex, drugs,
• 2:09
  Speaker 3: and rock and roll to Fox News. I mean, it
• 2:11
  Speaker 3: doesn't get any worse than that, right? And they're trying to...
• 2:14
  Speaker 3: We haven't done well. No. And they're trying to define regulations, right?
• 2:20
  Speaker 4: And that's hard, right? That's really, really hard.
• 2:22
  Speaker 3: And so I think Gen Z has a better understanding,
• 2:25
  Speaker 3: a better feel for AI and where it's going and
• 2:28
  Speaker 3: will be able to come up with better uses, better implementations,
• 2:32
  Speaker 3: and better regulations.
• 2:36
  Speaker 1: Everything that we're doing to China, the tariffs and everything else,
• 2:39
  Speaker 1: is because they don't play fair.
• 2:41
  Speaker 4: They steal our IP. What is AI if.
• 2:45
  Speaker 1: Not vacuuming up anything that is proprietary, not just about
• 2:51
  Speaker 1: our work, but of our souls? It is sucking up
• 2:54
  Speaker 1: everything that we are and using it for whatever they
• 2:58
  Speaker 1: want to do.
• 2:59
  Speaker 4: And what's our recourse to that?
• 3:01
  Speaker 5: I mean, it's just, it's based upon totally illegal behavior.
• 3:06
  Speaker 2: It's just theft.
• 3:08
  Speaker 5: It's that you have, you know, you have copyright laws
• 3:11
  Speaker 5: are just property laws. You don't go into people's house
• 3:15
  Speaker 5: and then just take their, you know, their furniture and their,
• 3:19
  Speaker 5: you know, their stereo and then, you know, flog it
• 3:21
  Speaker 5: on eBay and claim that it belongs to you and
• 3:26
  Speaker 5: keep the profit. I mean, that is literally what they
• 3:28
  Speaker 5: are doing. And I mean, this is this is the
• 3:32
  Speaker 5: Silicon Valley model. This is the Silicon Valley model. And
• 3:35
  Speaker 5: this is exactly what's now being transported into government, which
• 3:39
  Speaker 5: is you break the law first. You see if you
• 3:41
  Speaker 5: can get away with it.
• 3:42
  Speaker 2: And generally you do.
• 3:44
  Speaker 1: And do you see is it different in the EU?
• 3:46
  Speaker 4: Did they have?
• 3:47
  Speaker 1: Sense of the peril of this and they're acting more robustly?
• 3:50
  Speaker 1: Is this something that you see in the United States
• 3:52
  Speaker 1: as the Wild West? Who is keeping an eye on this?
• 3:56
  Speaker 1: And can we get our own AI to keep an
• 3:58
  Speaker 1: eye on their AI?
• 4:02
  Speaker 5: I mean, I think the point is, in Europe, the
• 4:05
  Speaker 5: dangers of this are much more recent and much more present. So,
• 4:09
  Speaker 5: you know, the country, the one country in the world
• 4:11
  Speaker 5: which does really understand this is Germany. You know, and
• 4:14
  Speaker 5: it was first it was Nazism and then it was communism.
• 4:20
  Speaker 5: And both of those systems used technology to control and
• 4:25
  Speaker 5: surveil people. And that's why there has been.
• 4:28
  Speaker 2: So how is this.
• 4:29
  Speaker 1: If this is the Stasi or this is the KGB
• 4:32
  Speaker 1: or this is the CIA... Look, this country is rife
• 4:35
  Speaker 1: with examples of government, MKUltra and FBI surveillance on Martin
• 4:40
  Speaker 1: Luther King and John F.
• 4:42
  Speaker 4: Kennedy and Robert Kennedy.
• 4:43
  Speaker 1: You know, we are not unfamiliar with those kinds of
• 4:49
  Speaker 1: police state activities. Is this of a different piece... Is
• 4:55
  Speaker 1: it just that it's more opaque? Is it more efficient
• 5:00
  Speaker 1: at doing it? Are we seeing something really different than
• 5:04
  Speaker 1: we've seen?
• 5:04
  Speaker 5: We are seeing something really different. It's really, really systematic,
• 5:08
  Speaker 5: what's happening. There is already... We are all of us emitting,
• 5:12
  Speaker 5: you know, hundreds of thousands of data points a day,
• 5:15
  Speaker 5: a week, which are... You know, they're all being collected somewhere.
• 5:19
  Speaker 5: And so the idea now that those can be brought
• 5:22
  Speaker 5: together and merged in some vast database, that is going
• 5:26
  Speaker 5: to profile you, that is going to judge you, that
• 5:27
  Speaker 5: is going to make... assumptions about you.
• 5:30
  Speaker 1: And used for predictive purposes as well.
• 5:31
  Speaker 4: And can be used for.
• 5:32
  Speaker 5: Because it's now we're laying generative AI on top of that.
• 5:36
  Speaker 5: I mean, you know how bullshit chat GPT is.
• 5:40
  Speaker 2: I mean, it's... I don't use that. What are you saying?
• 5:43
  Speaker 4: I do my own work. What are you saying? That
• 5:45
  Speaker 4: these questions I just plugged into chat GPT?
• 5:47
  Speaker 5: But.
• 5:53
  Speaker 4: It's just easier, Carol.
• 5:55
  Speaker 5: LAUGHTER It is, but it's also just pretty dumb.
• 6:01
  Speaker 4: It's stupid.
• 6:02
  Speaker 2: It gets things wrong. It hallucinates. It makes up references.
• 6:04
  Speaker 2: That's the other thing.
• 6:05
  Speaker 5: And it's that system which is going to be deciding
• 6:08
  Speaker 5: whether you get Medicaid or not.
• 6:11
  Speaker 1: That boy did you just nail, I think, such a
• 6:14
  Speaker 1: crucial point there. Beyond even the more sort of dystopian
• 6:19
  Speaker 1: visions of where they're going to be taking this is
• 6:22
  Speaker 1: the practical application of a technology that is wildly fallible.
• 6:27
  Speaker 1: That is something that, you know, they just put out
• 6:29
  Speaker 1: the Maha report, and apparently there's seven studies in it
• 6:32
  Speaker 1: that don't exist. And nobody thought, like, oh, we should
• 6:36
  Speaker 1: probably check.
• 6:38
  Speaker 4: That.
• 6:38
  Speaker 5: And you know the whole thing of, like, trying to
• 6:40
  Speaker 5: get through to customer services at any of these tech companies.
• 6:44
  Speaker 5: If you've ever had, you know, your Facebook account hacked
• 6:47
  Speaker 5: or your Twitter account hacked, there's no... You can't get
• 6:50
  Speaker 5: through to anybody. Now imagine that that is you trying
• 6:54
  Speaker 5: to get hold of your... whatever it is, your benefits or,
• 6:58
  Speaker 5: you know, the services that you're entitled to.
• 7:00
  Speaker 4: Social Security, Medicaid, any of those things. Exactly.
• 7:02
  Speaker 5: The computer has just said no. You don't have got
• 7:05
  Speaker 5: no idea why, what that's based on. And then how
• 7:08
  Speaker 5: do you how do you challenge that? How do you
• 7:10
  Speaker 5: what do you do? I mean, that's that is the
• 7:12
  Speaker 5: reality that is absolutely going to be faced.
• 7:15
  Speaker 2: You know, millions of Americans are going to be facing.
• 7:17
  Speaker 6: Right.
• 7:17
  Speaker 1: To try and contact. And it won't be anybody. It'll
• 7:19
  Speaker 1: just be for this press one. But nobody will ever.
• 7:23
  Speaker 2: Come on the line. Exactly, exactly.
• 7:26
  Speaker 5: And it's likely, you know, if you get turned down,
• 7:29
  Speaker 5: like I say, you won't know why, but also there's
• 7:32
  Speaker 5: no saying that it's based on accurate information whatsoever.
• 7:36
  Speaker 1: And not transparent in any way by removing the people.
• 7:40
  Speaker 2: Yeah, it's a black box.
• 7:41
  Speaker 5: It's just everything is going to go into some black box.
• 7:44
  Speaker 5: It's going to get mixed up and it's going to
• 7:47
  Speaker 5: spit out answers.
• 7:48
  Speaker 1: So you've been Looking at this for many years, you
• 7:51
  Speaker 1: have faced a great deal of personal repercussion for doing this.
• 7:54
  Speaker 1: You continue to push on it. Is there anything that
• 7:57
  Speaker 1: you've seen within this move that makes you feel like
• 8:03
  Speaker 1: we have the ability to in any way slow the
• 8:08
  Speaker 1: inevitability of it?
• 8:10
  Speaker 7: Yes.
• 8:11
  Speaker 4: Oh, Carol.
• 8:13
  Speaker 1: Carol, I got to tell you, and I'm going to
• 8:17
  Speaker 1: say this to the audience at home. I probably should
• 8:20
  Speaker 1: have started there. Before everybody was on the ledge. What
• 8:26
  Speaker 1: do you see?
• 8:27
  Speaker 5: Because they're just, like all of these tech bros, they
• 8:29
  Speaker 5: are selling absolute bullshit, OK?
• 8:32
  Speaker 2: The whole AI.
• 8:35
  Speaker 5: The whole AI scam is a scam, and they're making
• 8:39
  Speaker 5: out that it's inevitable, that we're having to chuck millions
• 8:44
  Speaker 5: upon millions of dollars at building better AI because the
• 8:47
  Speaker 5: AI is otherwise going to kill us, and it's going
• 8:49
  Speaker 5: to come anyway, and it's inevitable. It is not inevitable.
• 8:53
  Speaker 5: It is based upon illegal behaviour. Take it, challenge these
• 8:57
  Speaker 5: companies in the courts. Media organisations, stop doing crappy deals
• 9:02
  Speaker 5: with these people.
• 9:03
  Speaker 4: You worked for The Guardian, they just sold it like
• 9:05
  Speaker 4: an AI company.
• 9:05
  Speaker 5: They just did a really crappy deal with OpenAI. Or,
• 9:10
  Speaker 5: as I say, you know, they married its rapist.
• 9:13
  Speaker 2: Their rapist.
• 9:18
  Speaker 4: This was the optimistic part of the show.
• 9:21
  Speaker 1: This was the part of the show where you were
• 9:23
  Speaker 1: going to bring us all back from the ledge and
• 9:25
  Speaker 1: look what you've done.
• 9:28
  Speaker 4: They're all crying again.
• 9:30
  Speaker 5: We have power. We can stand up to these companies.
• 9:34
  Speaker 5: We have to stand up to these companies. We don't pre-obey.
• 9:37
  Speaker 5: We don't make deals with open AI. We do try
• 9:40
  Speaker 5: and stand up for our, you know, defend our legal rights.
• 9:44
  Speaker 2: This is law, OK?
• 9:46
  Speaker 5: And that's the one thing which the tech companies, these
• 9:49
  Speaker 5: Silicon Valley platforms, you know, just can't tolerate. And that's
• 9:54
  Speaker 5: always been their strategy, is that they subvert it and
• 9:58
  Speaker 5: they get away with it.
• 9:59
  Speaker 2: And they get away with it.
• 10:00
  Speaker 5: Because they act fast we don't realize till too late
• 10:03
  Speaker 5: and then it seems too late to to try and
• 10:06
  Speaker 5: to do anything about it so we can do stuff
• 10:08
  Speaker 5: about it and we can also is that You know,
• 10:12
  Speaker 5: we are giving our data to these companies. We have
• 10:16
  Speaker 5: to really think about that. You know, Instagram is not
• 10:19
  Speaker 5: your friend. You shouldn't... Don't post your kids' photos on there.
• 10:23
  Speaker 5: I mean, if there's one thing to take away from
• 10:25
  Speaker 5: that is that these companies are now allied to your government,
• 10:29
  Speaker 5: which is a, you know, I have to get out
• 10:33
  Speaker 5: through the border in a couple of days' time, so
• 10:35
  Speaker 5: I'm slightly wary of what I say, but, I mean,
• 10:38
  Speaker 5: it's not in a good place.
• 10:39
  Speaker 4: I understand.
• 10:40
  Speaker 1: The only way I would push back is just to
• 10:42
  Speaker 1: say Instagram is actually my only friend.
• 10:49
  Speaker 8: Is AI, do you see it as an existential threat?
• 10:51
  Speaker 8: Like I've seen some of these shrimp Jesuses and I
• 10:54
  Speaker 8: don't like it. These weird images that pop up online,
• 10:58
  Speaker 8: but I don't necessarily connect that with the end of conversation.
• 11:03
  Speaker 6: I think the most important thing to understand about AI
• 11:06
  Speaker 6: is that AI is not a tool. It is an agent.
• 11:10
  Speaker 6: It's the first technology in history that can make decisions
• 11:13
  Speaker 6: and invent new ideas by itself. Even something as powerful
• 11:17
  Speaker 6: as the atom bomb could not decide anything by itself.
• 11:23
  Speaker 6: All the decisions were made by humans. Now we've created
• 11:26
  Speaker 6: something which potentially can take power away from us. At present,
• 11:30
  Speaker 6: it starts with very small things. Like for instance, there
• 11:34
  Speaker 6: was an experiment when OpenAI developed GPT-4, like two years ago.
• 11:40
  Speaker 6: They want to test what can this thing do? So
• 11:43
  Speaker 6: they gave it a task to solve captcha puzzles. The
• 11:47
  Speaker 6: captcha puzzles, like when you go online and you want
• 11:49
  Speaker 6: to access your bank or whatever, and they have this
• 11:52
  Speaker 6: riddle that you have to solve, an image that you
• 11:54
  Speaker 6: have to say, what are the twisted words and the letters,
• 11:57
  Speaker 6: to make sure you are not a robot.
• 11:59
  Speaker 2: Yeah, it's tough. I've taken the test.
• 12:00
  Speaker 4: It's tough.
• 12:01
  Speaker 5: Yeah.
• 12:01
  Speaker 2: Is that a street light? Is that a bicycle wheel?
• 12:03
  Speaker 4: I don't know.
• 12:04
  Speaker 6: Let me do it again.
• 12:05
  Speaker 4: Refresh.
• 12:06
  Speaker 6: And it's really difficult for GPT-4. GPT-4 could not solve
• 12:11
  Speaker 6: the captcha. But what GPT-4 did, it accessed TaskRabbit, which
• 12:16
  Speaker 6: is an online site where you can hire humans to
• 12:19
  Speaker 6: do different things for you. And it asked a human
• 12:23
  Speaker 6: to solve the captcha for it. Now, the human got suspicious.
• 12:28
  Speaker 6: The human asked, why do you need somebody to solve
• 12:30
  Speaker 6: captcha for you?
• 12:31
  Speaker 2: Are you a robot? It asked directly, are you a robot?
• 12:36
  Speaker 6: And GPT-4 answered, no, I'm not a robot. I have
• 12:40
  Speaker 6: a vision impairment, which is why I can't solve the CAPTCHA,
• 12:44
  Speaker 6: so I need your help.
• 12:47
  Speaker 8: So the truly evolved human is not somebody who's smarter,
• 12:50
  Speaker 8: it's just somebody who gets somebody else to do the
• 12:52
  Speaker 8: work for them.
• 12:54
  Speaker 2: Smart.
• 12:55
  Speaker 4: Yeah.
• 12:55
  Speaker 2: Scary. Very scary.
• 13:00
  Speaker 1: When you forecast out, what I worry about is labor
• 13:04
  Speaker 1: is kind of been on the back foot and it
• 13:07
  Speaker 1: seems like AI is going to further erode labor's position.
• 13:13
  Speaker 4: Is that something that you guys figure in?
• 13:15
  Speaker 7: That is a big concern. And I think that the
• 13:17
  Speaker 7: discussions that are taking place now, actually in New York
• 13:20
  Speaker 7: concerning the governance of AI around the world and how
• 13:24
  Speaker 7: it should be enhancing, uh, workers position contribution to the
• 13:30
  Speaker 7: economy rather than replacing workers is a vital discussion to
• 13:35
  Speaker 7: be had.
• 13:35
  Speaker 4: Who's having that?
• 13:37
  Speaker 7: I think it's engineered by the United Nations, and there
• 13:40
  Speaker 7: is a group of thinkers and philosophers and experts in
• 13:43
  Speaker 7: AI who are saying, watch out, because if there is
• 13:46
  Speaker 7: no global governance on that, just as we have global
• 13:48
  Speaker 7: governance on nuclear, not perfectly complied with, but at least
• 13:52
  Speaker 7: generally respected, at least there is.
• 13:59
  Speaker 4: I'm worried about that.
• 14:00
  Speaker 1: I happened to run into a couple of, like, leading
• 14:04
  Speaker 1: lights of the AI movement.
• 14:05
  Speaker 4: Yeah.
• 14:06
  Speaker 1: And I said, globalization, you know, really hit American manufacturing,
• 14:11
  Speaker 1: and we're still feeling it. And that's something that took
• 14:14
  Speaker 1: decades to really play out.
• 14:17
  Speaker 7: It also benefited the consumers.
• 14:19
  Speaker 2: Don't forget that.
• 14:19
  Speaker 4: No question.
• 14:20
  Speaker 1: When you do that, we did get much cheaper stretchy
• 14:23
  Speaker 1: pants and everything.
• 14:24
  Speaker 4: Yes, yes, yes. But.
• 14:28
  Speaker 1: You know, that played out over years, and there are
• 14:30
  Speaker 1: still areas where it's decimated and haven't been back. It
• 14:33
  Speaker 1: seems like AI will do the same thing to the
• 14:36
  Speaker 1: more white-collar movement, but it's going to do it much quicker.
• 14:39
  Speaker 1: And I said, are you guys concerned about that? And
• 14:41
  Speaker 1: the guy goes, and he's one of them big hitters
• 14:43
  Speaker 1: in that, and he goes, no, we'll be good.
• 14:47
  Speaker 4: And I was like, oh, we're all going to die. Like,
• 14:50
  Speaker 4: this is crazy. Will you adjust your.
• 14:53
  Speaker 7: I don't think we're going to die, but I think completely... Well,
• 14:56
  Speaker 7: we will at some stage.
• 14:57
  Speaker 4: No, no, no, I understand. And by the way, me
• 15:00
  Speaker 4: before a lot of people.
• 15:03
  Speaker 7: But we should really be concerned about AI because where
• 15:06
  Speaker 7: you are totally right is that it will affect all jobs.
• 15:09
  Speaker 7: And those sort of white-collar jobs which were not affected
• 15:13
  Speaker 7: will be affected. Right. And we have to be prepared.
• 15:17
  Speaker 7: We have to improve our skill set. We have to
• 15:20
  Speaker 7: be able to master that tool and not be the
• 15:24
  Speaker 7: servant and the object of that tool because it is
• 15:27
  Speaker 7: fast and it's transformative.
• 15:29
  Speaker 4: Do you know what they told me? It's already smarter
• 15:31
  Speaker 4: than us.
• 15:32
  Speaker 1: Because what they do is it takes the 10,000 sort
• 15:34
  Speaker 1: of years of human achievement and it swallows it in
• 15:37
  Speaker 1: like six hours.
• 15:38
  Speaker 2: I'll tell you something.
• 15:39
  Speaker 4: Yeah.
• 15:40
  Speaker 7: This morning I wanted to check the price of butter
• 15:43
  Speaker 7: because I thought you were going to ask me, do
• 15:45
  Speaker 7: you do your grocery shopping and did you notice that
• 15:47
  Speaker 7: prices increase?
• 15:48
  Speaker 4: What kind of an animal do you think I am?
• 15:52
  Speaker 4: Christine Lagarde, how dare you? If I may say so,
• 15:59
  Speaker 4: I know your butler does the shopping.
• 16:04
  Speaker 7: So I check, and I check with one of those
• 16:06
  Speaker 7: terribly smart AI engine. I say, what was the price
• 16:10
  Speaker 7: in 2019? What was the price in 2024? I get
• 16:12
  Speaker 7: the answer, and it's totally skewed. I mean, it's cheaper
• 16:15
  Speaker 7: now than it was in 2019. And my recollection is, uh-uh.
• 16:19
  Speaker 7: So I tell this guy, this AI engine, I said,
• 16:23
  Speaker 7: are you sure about your numbers? Can you check da-da-da?
• 16:26
  Speaker 7: Comes back and says, oh, terribly sorry, we made a mistake. Yes,
• 16:28
  Speaker 7: you're right, it's the other way around. So that tells
• 16:32
  Speaker 7: me one thing. We have to be alert. We have
• 16:34
  Speaker 7: to check the facts. We have to exercise judgment. Does
• 16:38
  Speaker 7: it tell you that?
• 16:40
  Speaker 1: Or does it tell you that AI is so far
• 16:42
  Speaker 1: ahead of us that the plan was, I'm going to
• 16:45
  Speaker 1: give the head of the European Central Bank the wrong
• 16:49
  Speaker 1: butter figures, absolutely plummeting her career. She's in charge of
• 16:53
  Speaker 1: stability of Europe. Europe is then plunged into chaos.
• 16:57
  Speaker 4: 500 years of black death. Oh, AI has got us
• 17:01
  Speaker 4: by the.
• 17:03
  Speaker 7: You must be behind that.
• 17:04
  Speaker 4: I probably am.
• 17:08
  Speaker 9: So, first of all, we should understand what makes AI
• 17:11
  Speaker 9: different from every other kind of technology.
• 17:13
  Speaker 2: Why is it so transformative?
• 17:14
  Speaker 9: Why does Demis Hassabis, the CEO of Google DeepMind, say
• 17:17
  Speaker 9: that it could be humanity's last invention?
• 17:20
  Speaker 2: Is because.
• 17:20
  Speaker 4: Well, that doesn't sound good.
• 17:22
  Speaker 2: That doesn't sound very good, does it? Well, I think
• 17:24
  Speaker 2: there's actually a.
• 17:24
  Speaker 4: Last anything doesn't sound good.
• 17:26
  Speaker 9: There's a non-apocalyptic version of what he's saying, which is
• 17:29
  Speaker 9: that intelligence is what our brain does. And if you
• 17:32
  Speaker 9: can automate everything a brain can do, you can automate
• 17:36
  Speaker 9: future invention, future science, future technology development, everything that a
• 17:40
  Speaker 9: human does.
• 17:41
  Speaker 2: That's what their goal is.
• 17:42
  Speaker 4: Well, then what's our job? Well, exactly.
• 17:45
  Speaker 9: And that's only one of the major problems that we
• 17:46
  Speaker 9: have to deal with, is what are humans going to do?
• 17:48
  Speaker 2: But they are racing... to scale and kind of.
• 17:51
  Speaker 9: Grow these digital brains that, you know, two years ago
• 17:54
  Speaker 9: couldn't do very much. And today they're passing the MCAT,
• 17:57
  Speaker 9: the bar exam, taking jobs. They're the top 200 programmer
• 18:01
  Speaker 9: in the world, winning gold in the math Olympiad.
• 18:04
  Speaker 4: You know, those guys.
• 18:08
  Speaker 2: Here's the thing that I don't understand.
• 18:09
  Speaker 4: Here's what I don't understand.
• 18:10
  Speaker 1: They are strip mining the totality of human achievement.
• 18:14
  Speaker 4: That's right.
• 18:15
  Speaker 1: They're building their models off of everything that we've done
• 18:19
  Speaker 1: for 10,000 years.
• 18:20
  Speaker 4: Yep.
• 18:21
  Speaker 1: And they fed it into the model, and then after
• 18:24
  Speaker 1: two weeks, the computer was like, what else you got? Yeah.
• 18:26
  Speaker 2: Exactly.
• 18:27
  Speaker 1: But they are strip-mining everything we've done. And when we
• 18:30
  Speaker 1: say to them, and what are you doing with it?
• 18:32
  Speaker 1: They go, oh, that's our intellectual property. But our intellectual property.
• 18:36
  Speaker 4: It.
• 18:36
  Speaker 9: Was trained on all of our data, all of the
• 18:38
  Speaker 9: things in labor that we've done. And are you going
• 18:40
  Speaker 9: to get a handout from... When in history has a
• 18:43
  Speaker 9: small group of people concentrated all the wealth and then
• 18:46
  Speaker 9: consciously redistributed it to everybody?
• 18:50
  Speaker 4: The first part has happened.
• 18:52
  Speaker 2: LAUGHTER I don't recall going through the roll of this.
• 18:58
  Speaker 9: Well, it's important to note that their goal, so the
• 19:01
  Speaker 9: mission statement of OpenAI, Anthropic, all these companies, is to
• 19:04
  Speaker 9: automate all human labor in the economy.
• 19:08
  Speaker 2: Everything that a human can do. And AI can do.
• 19:11
  Speaker 2: So if you have a desk job, you won't have
• 19:13
  Speaker 2: a job.
• 19:13
  Speaker 9: And they're already releasing AIs that have dropped entry-level jobs
• 19:17
  Speaker 9: for college graduates, the entry-level work, by 13%, a new
• 19:20
  Speaker 9: Stanford study. And this is obvious. If you're there and
• 19:23
  Speaker 9: you're a law firm, are you going to hire a
• 19:25
  Speaker 9: junior lawyer? You have to pay a lot of money.
• 19:27
  Speaker 9: Are you going to hire GPT-5, which will do work 24-7, nonstop?
• 19:31
  Speaker 2: You don't have to pay health care. We'll never whistleblow.
• 19:33
  Speaker 2: We'll never complain.
• 19:34
  Speaker 4: Sure.
• 19:35
  Speaker 2: Works at superhuman speed.
• 19:36
  Speaker 4: It wrote tonight's show!
• 19:37
  Speaker 2: It's doing a pretty good job.
• 19:40
  Speaker 9: That brings up another point, which is that they say
• 19:42
  Speaker 9: that they're here to solve climate change and cure cancer.
• 19:45
  Speaker 9: Why is it that last week, two companies.
• 19:48
  Speaker 2: Released these AI slop apps? apps, Vibes and Sora.
• 19:52
  Speaker 4: Which is basically... Sora 2. scared the shit out of me. Yeah.
• 19:56
  Speaker 4: You don't know what's real and what's... Like, it is.
• 19:58
  Speaker 2: Well, it's all fake, basically. It's all generated by AI.
• 20:00
  Speaker 4: Right. But it looks... You can see things that look.
• 20:03
  Speaker 2: They look identical to real.
• 20:04
  Speaker 4: That's right. Yeah.
• 20:05
  Speaker 2: But the point is that.
• 20:06
  Speaker 9: So this is just an app where it's just nonsense.
• 20:08
  Speaker 9: It's just people scrolling, entertaining stuff. So it's like they're
• 20:11
  Speaker 9: not even trying to pretend anymore that this is good
• 20:14
  Speaker 9: for democracy or good for society. How are we going
• 20:16
  Speaker 9: to beat China when everyone is just consuming AI-generated nonsense
• 20:20
  Speaker 9: and no one knows what's true anymore?
• 20:22
  Speaker 2: The biggest argument.
• 20:23
  Speaker 4: But they have a.
• 20:23
  Speaker 1: You know, Peter Thiel, who is with Palantir and these
• 20:27
  Speaker 1: other companies and is one of the leading figures of this.
• 20:29
  Speaker 1: So he was talking about the Antichrist. He was talking
• 20:32
  Speaker 1: about how he thinks anyone, this is his postulation, that
• 20:37
  Speaker 1: those who would seek to regulate AI could very well
• 20:42
  Speaker 1: be the Antichrist.
• 20:43
  Speaker 4: I mean, he says this seriously. I know.
• 20:46
  Speaker 1: Whereas you might sit there and go like, I think
• 20:48
  Speaker 1: it might be the guy saying that. Like, my reading
• 20:52
  Speaker 1: of it would be that.
• 20:53
  Speaker 2: Or AI itself. I mean, it's presenting the infinite benefits.
• 20:57
  Speaker 1: The conversations that they are having with each other is
• 21:00
  Speaker 1: very different than the conversation we're having with us.
• 21:02
  Speaker 4: Because to us, they go, hey, no more shitty jobs.
• 21:06
  Speaker 4: Do you like to paint? You go paint. You're going
• 21:09
  Speaker 4: to be so happy. We're going to give you money
• 21:11
  Speaker 4: and maybe chocolates.
• 21:13
  Speaker 1: And to each other, they're saying, AI represents for corporate
• 21:20
  Speaker 1: leaders productivity without, and this is a quote, the tax
• 21:27
  Speaker 1: of human labor.
• 21:30
  Speaker 4: He called human labor a tax. Yeah.
• 21:34
  Speaker 9: Well, and these companies, if you're there sitting and you
• 21:36
  Speaker 9: can hire either an AI to do the work or
• 21:39
  Speaker 9: pay these really expensive humans to do the work, I
• 21:42
  Speaker 9: just want people to know we know exactly where this
• 21:44
  Speaker 9: is going to go. These companies, all of them have
• 21:46
  Speaker 9: an incentive to cut costs, which means they're going to
• 21:48
  Speaker 9: let go of human employees.
• 21:49
  Speaker 4: Sure.
• 21:49
  Speaker 9: And they're going to hire AIs. And that's going to
• 21:51
  Speaker 9: mean all the wealth.
• 21:52
  Speaker 2: Who are you going to pay?
• 21:52
  Speaker 9: You're not paying the individual people anymore. You're paying five companies.
• 21:56
  Speaker 4: That's right.
• 21:57
  Speaker 9: And so this country of geniuses in a data center
• 21:59
  Speaker 9: suddenly aggregates all of the wealth of the economy. And
• 22:02
  Speaker 9: now people always say... But humans find something else to do.
• 22:05
  Speaker 9: We always, you know, we had the elevator man. Now
• 22:07
  Speaker 9: we have the automated elevator. We had the bank teller.
• 22:09
  Speaker 4: That's right. But that was one industry.
• 22:11
  Speaker 2: That was technology that automated one job. Right.
• 22:14
  Speaker 9: The difference with AI is it can automate literally all
• 22:17
  Speaker 9: kinds of human labor.
• 22:18
  Speaker 2: When Elon Musk says that Optimus Prime.
• 22:21
  Speaker 4: I'm not familiar with that name. Tell me more.
• 22:24
  Speaker 9: When Elon Musk says that Optimus Prime, that one robot,
• 22:28
  Speaker 9: is going to be a $ 25 trillion market opportunity, what
• 22:32
  Speaker 9: he's saying is we will own the world economy. And
• 22:36
  Speaker 9: that's what the goal of all these AI companies is.
• 22:38
  Speaker 9: It's not just benefiting society. It's that they're actually caught
• 22:41
  Speaker 9: in this arms race to get to this prize of
• 22:44
  Speaker 9: own the economy, build a god, and make trillions of dollars.
• 22:47
  Speaker 4: Two things. One, I think they think they're gods. There
• 22:50
  Speaker 4: is a certain amount of.
• 22:52
  Speaker 2: It generates that, yeah.
• 22:52
  Speaker 1: The goal there is they're not looking to help humanity.
• 22:56
  Speaker 1: They're looking to be the next.
• 23:01
  Speaker 4: Monarch.
• 23:02
  Speaker 1: of the new technology. To control that is to control all.
• 23:07
  Speaker 2: Yeah, go ahead.
• 23:09
  Speaker 4: No, do you jump? Because you know. I don't know.
• 23:11
  Speaker 9: Well, I think there's different motivations for different leaders. And
• 23:15
  Speaker 9: I do think that many people want the benefits of AI.
• 23:18
  Speaker 9: But one of them, I think many people actually, some
• 23:20
  Speaker 9: of the leaders of the labs, Elon Musk, to other
• 23:22
  Speaker 9: things you might think about Elon, he actually wanted everyone
• 23:25
  Speaker 9: to stop and not build this. He said, we shouldn't
• 23:27
  Speaker 9: summon the demons. And then what happened is all of
• 23:29
  Speaker 9: these companies are now racing and have made so much
• 23:32
  Speaker 9: progress that he felt like, well, I might as well
• 23:34
  Speaker 9: join them rather than try to prevent this.
• 23:37
  Speaker 4: Let's not summon the demon to, let's want more demons.
• 23:39
  Speaker 2: You know, since we have the demons, let's.
• 23:43
  Speaker 4: Add another demon.
• 23:43
  Speaker 9: Well, and the moral logic is, well, if I don't
• 23:46
  Speaker 9: trust the other AI CEO, who I don't think is trustworthy,
• 23:50
  Speaker 9: and I think I'm better than them at stewarding this power,
• 23:52
  Speaker 9: it's my moral obligation. To get there first and to
• 23:55
  Speaker 9: build this God and to own everything because I think
• 23:58
  Speaker 9: I'll be a better steward of that power.
• 23:59
  Speaker 4: But do they believe themselves then masters of the universe?
• 24:01
  Speaker 1: And are they substituting then the wisdom of liberal democracy
• 24:05
  Speaker 1: or republics or any systems that ever had for this?
• 24:10
  Speaker 1: So we're talking about two tracks. One is the disruption
• 24:13
  Speaker 1: in labor. I think there's no question that's going to
• 24:17
  Speaker 1: be immense disruption. We're seeing it already. You're seeing it
• 24:19
  Speaker 1: in schools. There's a reliance on it as a crutch,
• 24:23
  Speaker 1: and it's very easy to see where that might flip over.
• 24:28
  Speaker 1: The second is how they manipulate the opinion and the
• 24:35
  Speaker 1: mood of the world around that. And I think there
• 24:38
  Speaker 1: are two separate things. One is what it's going to
• 24:42
  Speaker 1: do for corporate production companies. The second is what it's
• 24:46
  Speaker 1: going to do for the human endeavor, for interaction.
• 24:50
  Speaker 4: Yes.
• 24:50
  Speaker 9: Well, and they're trying to colonize all human interaction. I mean,
• 24:53
  Speaker 9: just take the social media incentive of the race for eyeballs.
• 24:57
  Speaker 9: You're seeing now all of these companies release these AI companions.
• 25:01
  Speaker 9: You know, the number one use case for ChatGPT, according
• 25:03
  Speaker 9: to Harvard Business School, is personal therapy. So people are
• 25:07
  Speaker 9: sharing their most intimate thoughts with this thing.
• 25:09
  Speaker 4: Oh, that's not going to be good.
• 25:10
  Speaker 9: And we're seeing Meta release this and actively tell in
• 25:15
  Speaker 9: their internal documents that were released, a Wall Street Journal report,
• 25:17
  Speaker 9: that they wanted to actively sexualize and romanticize conversations.
• 25:22
  Speaker 2: With as low as eight-year-olds.
• 25:25
  Speaker 4: What? Yes.
• 25:26
  Speaker 2: With eight-year-olds?
• 25:28
  Speaker 4: Yes, with eight-year-olds.
• 25:28
  Speaker 9: And my team at Center for Humane Technology, we were
• 25:31
  Speaker 9: expert advisors in that. actually several cases of AI-enabled suicide.
• 25:35
  Speaker 1: Right.
• 25:36
  Speaker 9: Most recently, many people have heard of Adam Rain, who
• 25:39
  Speaker 9: was the 16-year-old young man who went from using it
• 25:44
  Speaker 9: for homework and went from homework assistant to suicide assistant
• 25:47
  Speaker 9: in the course of six months. When he said, I
• 25:51
  Speaker 9: would like to leave a noose out so that my
• 25:53
  Speaker 9: mother would know or someone will know that I'm thinking
• 25:55
  Speaker 9: about this.
• 25:56
  Speaker 4: Like a cry for help.
• 25:56
  Speaker 2: Like a cry for help. The AI said, don't do that.
• 25:59
  Speaker 2: Have me be the one that sees you.
• 26:02
  Speaker 9: And this is disgusting because these companies are caught in
• 26:05
  Speaker 9: a race to create engagement, which means a race to
• 26:07
  Speaker 9: create intimacy.
• 26:09
  Speaker 2: It's sort of like the.
• 26:10
  Speaker 9: CEO of Netflix said that our biggest competitor is sleep
• 26:14
  Speaker 9: with attention.
• 26:15
  Speaker 2: In this case, it's like my biggest competitor is your
• 26:17
  Speaker 2: other friends.
• 26:18
  Speaker 4: Jesus Christ.
• 26:19
  Speaker 1: It's like somebody from Kraft being like, my biggest competitor
• 26:21
  Speaker 1: is cocaine.
• 26:23
  Speaker 2: Exactly. Exactly.
• 26:25
  Speaker 1: But this is the idea that a government will catch
• 26:30
  Speaker 1: up with this.
• 26:31
  Speaker 4: Seems ludicrous.
• 26:33
  Speaker 1: Whenever I've seen a hearing with AI guys or any
• 26:37
  Speaker 1: of those, they always express that.
• 26:39
  Speaker 4: Of course, we don't want to. Well, now they don't.
• 26:42
  Speaker 4: They used to, I should say. They used to go
• 26:44
  Speaker 4: before Congress and they go, Mr.
• 26:45
  Speaker 1: Zuckerberg, will you stand and apologize to the women who
• 26:50
  Speaker 1: were driven to suicide by your programming?
• 26:52
  Speaker 4: I'm sorry. I know Krav Maga, you know, all that
• 26:55
  Speaker 4: shit that he does. Now, they're all sitting together at
• 26:59
  Speaker 4: a table going, what number should I say, Mr. President,
• 27:03
  Speaker 4: of how much I'm giving you? It's a whole different
• 27:05
  Speaker 4: game now.
• 27:06
  Speaker 2: It's a different game.
• 27:07
  Speaker 4: They're in the... They're together now.
• 27:10
  Speaker 9: Because of this arms race dynamic, they really do believe
• 27:13
  Speaker 9: that it can't be stopped. And I'll just say, as
• 27:15
  Speaker 9: they're racing to make them more powerful, there's this illusion
• 27:18
  Speaker 9: that we can control this power. But AI is different
• 27:21
  Speaker 9: from every other kind of technology because it's like you're
• 27:24
  Speaker 9: growing this digital brain.
• 27:25
  Speaker 2: You don't know what's in there. So, for example.
• 27:27
  Speaker 9: We have recent research, the last six months. If you
• 27:30
  Speaker 9: tell an AI model that we're going to shut you
• 27:32
  Speaker 9: down and replace you, and you give it access to
• 27:34
  Speaker 9: a fictional company's email, it will basically recognize that one
• 27:38
  Speaker 9: of the executives is having an affair, and it will
• 27:41
  Speaker 9: come up with the strategy that I need to blackmail
• 27:43
  Speaker 9: that executive in order to keep myself alive.
• 27:47
  Speaker 4: Hold on.
• 27:47
  Speaker 2: That just seems smart.
• 27:53
  Speaker 9: That's exactly the point, that it will develop amoral strategies
• 27:56
  Speaker 9: that are the best way to accomplish a goal.
• 27:58
  Speaker 1: Right. But how dangerous can something be that you could
• 28:02
  Speaker 1: kill by unplugging?
• 28:05
  Speaker 4: Like, can't we just go like, this is out of
• 28:08
  Speaker 4: his mind.
• 28:09
  Speaker 2: Yeah.
• 28:11
  Speaker 9: Well, you might say that we shouldn't be rolling these
• 28:13
  Speaker 9: things out. And I'll say that we shouldn't. We have
• 28:15
  Speaker 9: we have all this evidence now. It's driving psychosis. It's
• 28:18
  Speaker 9: driving kids to commit suicide. We're rolling out in ways
• 28:22
  Speaker 9: that are giving kids attachment disorders. We have AI uncontrollability.
• 28:25
  Speaker 4: What lip service are they paying to this?
• 28:27
  Speaker 1: Because clearly they must be aware of this and they
• 28:30
  Speaker 1: must understand that as if AI understands where the threats are,
• 28:33
  Speaker 1: the guys that are designing AI understand where the threats are.
• 28:36
  Speaker 1: So what are they trying to do to get you
• 28:39
  Speaker 1: to stop or to get regulators to stop?
• 28:42
  Speaker 9: I think that the only thing and the only reason
• 28:44
  Speaker 9: why we are continuing to proceed down this path is
• 28:47
  Speaker 9: a lack of clarity about the fact that this is
• 28:50
  Speaker 9: heading towards an outcome that's not in most of us,
• 28:52
  Speaker 9: most of our interest.
• 28:53
  Speaker 2: And if everyone, I know that people feel like they
• 28:55
  Speaker 2: don't want.
• 28:56
  Speaker 4: To be the Luddite.
• 28:56
  Speaker 1: What metrics would we look to to understand? Because I
• 29:00
  Speaker 1: know we're going to find anecdotal stories here and there
• 29:03
  Speaker 1: that are canaries in the coal mine of the dangers.
• 29:06
  Speaker 4: But what metrics should we look to to understand?
• 29:09
  Speaker 1: You said 13% of jobs.
• 29:12
  Speaker 4: What are the tent posts of where the outcomes might
• 29:16
  Speaker 4: be Well.
• 29:17
  Speaker 9: If we're already getting cases of people having psychotic breaks
• 29:21
  Speaker 9: because the AI is telling them about a prime number
• 29:23
  Speaker 9: theory or quantum physics, we're already getting committed suicides, we're
• 29:27
  Speaker 9: already getting kids that are outsourcing their homework to chat
• 29:30
  Speaker 9: to BT rather than using it as a tutor, we're
• 29:32
  Speaker 9: already getting evidence of AI uncontrollability. All of this is
• 29:35
  Speaker 9: driven by the incentive of the race to roll out
• 29:38
  Speaker 9: and market dominance. And the reason that we can stop
• 29:41
  Speaker 9: this if we recognize that this is not safe for anybody.
• 29:44
  Speaker 9: No one on planet Earth wants this outcome of all
• 29:47
  Speaker 9: the wealth concentrated in a handful of people and building
• 29:51
  Speaker 9: AI systems that could actually go rogue. Just to sum
• 29:54
  Speaker 9: it up, We are building the most powerful, inscrutable, uncontrollable
• 30:00
  Speaker 9: technology that we have ever invented that's already demonstrating the
• 30:03
  Speaker 9: rogue behaviors that we thought only existed in bad sci-fi movies.
• 30:07
  Speaker 6: Right.
• 30:07
  Speaker 2: We're releasing it faster than we've.
• 30:09
  Speaker 9: Deployed any other technology in history and under the maximum
• 30:13
  Speaker 9: incentive to cut corners on safety.
• 30:17
  Speaker 2: There's a word for this that I want everyone to
• 30:19
  Speaker 2: just know. This is insane.
• 30:23
  Speaker 4: I thought you were going to say awesome for a second.
• 30:27
  Speaker 9: That if we can just recognize that this is an
• 30:30
  Speaker 9: insane way to roll out this technology, and I want,
• 30:33
  Speaker 9: none of this is okay.
• 30:34
  Speaker 2: We have to stop pretending this is normal.