Latest / DEEP INSIGHT / 'Worst possible president in charge right now' to handle A.I. Author

Transcript

'Worst possible president in charge right now' to handle A.I. Author · DEEP INSIGHT

• 0:00
  Speaker 1: All right, let's enter the group chat. Ahmed Baba is here,
• 0:02
  Speaker 1: author of AhmedBaba.News. Josh Terengel, staff writer for The Atlantic,
• 0:05
  Speaker 1: executive producer of the Instadoc series you must see on Netflix.
• 0:09
  Speaker 1: The latest episode, The Prediction Games, is a must see.
• 0:13
  Speaker 1: But he is also the author of AI for Good.
• 0:17
  Speaker 1: Alex Witt joins us now. Longtime MSNOW anchor and Sean Thomas,
• 0:20
  Speaker 1: MSNOW political director.
• 0:21
  Speaker 2: Josh, you just wrote the book on this.
• 0:23
  Speaker 1: OK, you see the many benefits of AI. But what
• 0:27
  Speaker 1: is your take on the president's task force?
• 0:31
  Speaker 3: I actually want to pause and tell you we're seeing
• 0:33
  Speaker 3: something pretty extraordinary. On the one hand, if you look
• 0:37
  Speaker 3: at all the polling, right, Americans somewhere between 70 and
• 0:41
  Speaker 3: 85 percent want to slow down to ensure the safety
• 0:44
  Speaker 3: of advanced artificial intelligence.
• 0:47
  Speaker 2: That's a huge number. And we are four weeks out
• 0:49
  Speaker 2: from a no action. Pretty incredible.
• 0:51
  Speaker 1: We rarely agree on anything.
• 0:52
  Speaker 2: Rarely agree on anything. This is pretty, it's pretty solid. Okay.
• 0:56
  Speaker 3: And you have a president who for really the past
• 0:58
  Speaker 3: six months, although it's heated up a lot in the
• 1:00
  Speaker 3: last six weeks, has said, yeah, we don't know.
• 1:02
  Speaker 2: What are you talking about? We need to slow this down.
• 1:05
  Speaker 3: And so what's fascinating is that the president has picked
• 1:09
  Speaker 3: quite the hill to die on. Now, I would be
• 1:13
  Speaker 3: very entertained, except for the fact that we are also
• 1:16
  Speaker 3: all on that hill. And there is also, based on
• 1:18
  Speaker 3: the model makers, a 10 percent chance we might all die.
• 1:21
  Speaker 3: But what you're seeing is a president who is so invested,
• 1:25
  Speaker 3: maybe it's belief, maybe it's the GDP contribution of all
• 1:28
  Speaker 3: the AI models. But the thing that he's suggesting is
• 1:31
  Speaker 3: patently not going to work.
• 1:33
  Speaker 1: Explain.
• 1:34
  Speaker 3: So voluntary, morally binding commitments are. in a race to
• 1:40
  Speaker 3: get to superintelligence with companies that have trillion-dollar valuations and
• 1:44
  Speaker 3: are actually looking to add another zero to those valuations,
• 1:48
  Speaker 3: do you think they're going to look at each other
• 1:49
  Speaker 3: and be like, oh, okay, well, if you're slowing.
• 1:51
  Speaker 2: I'll slow.
• 1:51
  Speaker 3: And so the model that I've looked at and that
• 1:54
  Speaker 3: I know you know well is finance, right? So let's
• 1:57
  Speaker 3: go back to the financial crisis where there was existential
• 2:01
  Speaker 3: risk in the system. We didn't know quite how much.
• 2:04
  Speaker 3: We didn't know who held it. And so the people
• 2:07
  Speaker 3: in charge at that point, Geithner and Paulson and Bernanke said,
• 2:11
  Speaker 3: all of you, all of you are about to comply
• 2:13
  Speaker 3: with this because if you don't and one of you
• 2:16
  Speaker 3: goes down, you're going to die.
• 2:18
  Speaker 2: We're not saving.
• 2:20
  Speaker 3: And so that's called the prisoner's dilemma problem. And that's
• 2:22
  Speaker 3: why government exists, is to take these decisions away from
• 2:25
  Speaker 3: the actors themselves, to say this is the rudiment by
• 2:29
  Speaker 3: which you're going to comply, and to stand between people,
• 2:32
  Speaker 3: 70% of whom want this to happen, and companies who
• 2:36
  Speaker 3: are incentivized to keep going at a reckless pace.
• 2:39
  Speaker 1: But Josh, what is the name of that person in government?
• 2:43
  Speaker 1: What is the role of that person who's going to
• 2:45
  Speaker 1: do that? Because here's the thing. If we look at
• 2:48
  Speaker 1: what the shining star is in the economy or when
• 2:51
  Speaker 1: the president talks just a week ago when he said
• 2:53
  Speaker 1: to reporters, I've made you twice as rich over the
• 2:56
  Speaker 1: last year.
• 2:56
  Speaker 2: The only thing he's.
• 2:57
  Speaker 1: Talked about is the stock market. And if you actually
• 2:59
  Speaker 1: look at the stock market, there are scores of companies.
• 3:02
  Speaker 2: You want to look at Home Depot?
• 3:03
  Speaker 1: Sure.
• 3:03
  Speaker 2: You want to look at Nike? Sure. What are those
• 3:05
  Speaker 2: companies do?
• 3:05
  Speaker 1: Their stock has suffered over the last year. The stock
• 3:08
  Speaker 1: market is being completely propped up by these very tech companies.
• 3:12
  Speaker 1: So unless they're soaring and their valuations are doing the same,
• 3:15
  Speaker 1: then the president's whole argument for the extraordinary economy, it
• 3:19
  Speaker 1: falls apart.
• 3:20
  Speaker 2: Correct.
• 3:20
  Speaker 3: And that's what that's what we're seeing here is that
• 3:22
  Speaker 3: the self-interest of the president politically is I have got
• 3:25
  Speaker 3: to keep the economy propped up and the market propped up.
• 3:28
  Speaker 2: And it's being propped up by one thing. Yeah, it's entirely.
• 3:32
  Speaker 3: I mean, we don't really have growth in the economy
• 3:34
  Speaker 3: without AI investment. At the same time, you have this
• 3:37
  Speaker 3: massive existential risk as relayed by the handful of people
• 3:41
  Speaker 3: running the most forward-thinking AI labs.
• 3:46
  Speaker 2: It's not hard to figure out what's going on.
• 3:47
  Speaker 3: When you ask who can do anything about it ordinarily
• 3:50
  Speaker 3: in a functional political system, the president would issue an
• 3:53
  Speaker 3: executive order saying a very simple thing. There is a
• 3:56
  Speaker 3: mandatory duty to report that a rogue AI system. You
• 4:00
  Speaker 3: have 72 hours. That's how we do it in the FDA.
• 4:03
  Speaker 3: Lots of other things where we regulate. You start there
• 4:06
  Speaker 3: and then you begin to build out. I think what's
• 4:08
  Speaker 3: likeliest to happen is an absolute slaughter in the midterms
• 4:13
  Speaker 3: that may draw the attention of moderate Republicans to get
• 4:16
  Speaker 3: behind a bill and may compel some of these labs
• 4:19
  Speaker 3: who will speak.
• 4:20
  Speaker 2: Privately about the need for federal regulation, if not publicly.
• 4:23
  Speaker 2: And we may actually get there. But it's going to
• 4:25
  Speaker 2: take a big number for people to be compelled to
• 4:28
  Speaker 2: do that.
• 4:29
  Speaker 1: Take us to that campaign trail.
• 4:30
  Speaker 4: Well, I think it's interesting, and I think we have
• 4:31
  Speaker 4: Kevin Fry on a little later, or he was on earlier,
• 4:34
  Speaker 4: but he was in Michigan. He was talking to Alyssa
• 4:37
  Speaker 4: Sklotkin and Abdul El-Sayed.
• 4:38
  Speaker 2: We're going to share that interview.
• 4:39
  Speaker 4: And one question he asked them was, what's something you
• 4:42
  Speaker 4: actually think, if Democrats take control of the Senate and
• 4:45
  Speaker 4: the House, what's one thing you actually think you can
• 4:48
  Speaker 4: maybe possibly get done, maybe possibly get to the president's desk?
• 4:52
  Speaker 4: And the answer they had was actually AI regulation of
• 4:56
  Speaker 4: some form or fashion. That's sort of what Slotkin said
• 4:58
  Speaker 4: to him was, you know, I can basically see how
• 5:01
  Speaker 4: some Republicans would come along. And if this election cycle
• 5:05
  Speaker 4: proves what we think of the polling that we have seen,
• 5:07
  Speaker 4: especially this morning, showing is that possibly she's right. I
• 5:12
  Speaker 4: think part of what the president has done here is, OK,
• 5:15
  Speaker 4: DNI Clayton, AI czar. Does does that satisfy all of you?
• 5:19
  Speaker 4: Is that enough? Is that enough for us not to
• 5:21
  Speaker 4: have a slaughtering in the election? Even though there's no
• 5:23
  Speaker 4: follow through, there's nothing to back that up.
• 5:25
  Speaker 1: People have no idea what that czar is, who Jay
• 5:28
  Speaker 1: Clayton is or what's on the menu. No, they're panicked
• 5:31
  Speaker 1: about what AI is doing to their communities in the future.
• 5:34
  Speaker 4: And if his idea is that AI will continue to
• 5:37
  Speaker 4: prop up the economy. And I understand what you're saying
• 5:39
  Speaker 4: about the stock market. We come back and I've said
• 5:41
  Speaker 4: this every week on the show. People we're talking to
• 5:44
  Speaker 4: in the field are not feeling it. The polling shows
• 5:47
  Speaker 4: they're not feeling it. The number one issue, when you
• 5:50
  Speaker 4: look at the battleground polling from the New York Times
• 5:52
  Speaker 4: and Siena this weekend, the number one issue for people
• 5:55
  Speaker 4: was their cost of living as well as the economy.
• 5:59
  Speaker 4: Those two things were at the very top of the list.
• 6:02
  Speaker 4: There is no changing that in this polling. Nothing is changing.
• 6:05
  Speaker 4: People do not feel it, and especially independents do not
• 6:09
  Speaker 4: feel that enough is being done. So if they think
• 6:11
  Speaker 4: that putting a czar in there that, as you point out,
• 6:15
  Speaker 4: no one knows who Clayton is, that is not going
• 6:18
  Speaker 4: to change any minds or hearts. And people are, they're
• 6:20
  Speaker 4: not voting on AI, but there's enough going on between
• 6:23
  Speaker 4: data centers and everything else that they're scared about it.
• 6:25
  Speaker 4: And that comes back to the jobs issue, that they're
• 6:28
  Speaker 4: scared about their jobs.
• 6:29
  Speaker 5: So I'm wondering about the whole play for it being
• 6:32
  Speaker 5: a national security issue.
• 6:33
  Speaker 2: This is why we have to develop AI and own it.
• 6:36
  Speaker 6: In the race terms.
• 6:38
  Speaker 2: You guys are the experts in this area. How true
• 6:41
  Speaker 2: is that? I mean, us versus China.
• 6:45
  Speaker 7: Yeah, I mean, I think that's a big question, right?
• 6:47
  Speaker 7: Because when you look at Jay Clayton, who had the
• 6:48
  Speaker 7: task force here, he compared this to the nuclear arms race, right?
• 6:52
  Speaker 7: And while doing that, this other side of his mouth,
• 6:55
  Speaker 7: he said he doesn't want to pause. Last time I checked,
• 6:58
  Speaker 7: endless nuclear proliferation does not end well, right? And I think,
• 7:04
  Speaker 7: you know, we shouldn't sacrifice... our humanity's well-being for the
• 7:08
  Speaker 7: sake of of this race and i think china they're
• 7:10
  Speaker 7: like if you had a president in power right now
• 7:12
  Speaker 7: that actually cared about this issue you could have a
• 7:16
  Speaker 7: joint pact here like i think we need a nuclear
• 7:18
  Speaker 7: situation like my background was tech before journalism so i'm
• 7:21
  Speaker 7: pretty familiar with these tools and i think what you
• 7:23
  Speaker 7: need is like what bill gates has suggested like nuclear
• 7:26
  Speaker 7: inspection level uh embeds in in these companies you can
• 7:29
  Speaker 7: have these kind of PACs With these foreign entities, you
• 7:32
  Speaker 7: can bring China to the table.
• 7:34
  Speaker 6: China has expressed more worry.
• 7:35
  Speaker 7: Xi himself in a speech recently has expressed more worry
• 7:38
  Speaker 7: about AI getting out of control than Trump himself. So
• 7:41
  Speaker 7: my point is that we have the worst possible president
• 7:44
  Speaker 7: in charge right now. And I think you need a
• 7:47
  Speaker 7: situation where you have these risks taken seriously because, for example, superintelligence, right?
• 7:52
  Speaker 2: Why are we calling it that?
• 7:53
  Speaker 6: That's what I'm trying to say. What he's doing. What
• 7:55
  Speaker 6: Trump has done here has done a linguistic propaganda game.
• 7:58
  Speaker 6: Superintelligence itself. is the risk that everyone's warning about.
• 8:07
  Speaker 7: So by trying to rebrand artificial intelligence as the risk itself,
• 8:11
  Speaker 7: he's minimizing and downplaying the risk.
• 8:14
  Speaker 6: And that is not functional. It's not what the American
• 8:16
  Speaker 6: people want.
• 8:17
  Speaker 7: And I really think where you have to take these
• 8:19
  Speaker 7: agentic AI, these rogue AI seriously, and we have to
• 8:22
  Speaker 7: follow a framework in which, like, for example, to your point,
• 8:26
  Speaker 7: and to all our points here about the bipartisan nature
• 8:28
  Speaker 7: of this regulation, Hawley and Murphy, They have a bill
• 8:32
  Speaker 7: on the table to potentially expand the Computer and Fraud
• 8:34
  Speaker 7: Abuse Act so that you can actually hold these companies accountable.
• 8:37
  Speaker 6: So you have to have liability and you have to
• 8:39
  Speaker 6: have regulation.
• 8:40
  Speaker 2: OK, now we're talking.
• 8:41
  Speaker 1: Isn't the whole enchilada here about liability? When they talk
• 8:46
  Speaker 1: about regulation, unless these companies are actually held liable for
• 8:50
  Speaker 1: the work that they do.
• 8:52
  Speaker 2: Going to do anything. And so ultimately, here's the issue, right?
• 8:54
  Speaker 3: There's been a lot of people who, one, I applaud
• 8:58
  Speaker 3: you guys for your sincere belief that the president is like.
• 9:02
  Speaker 2: Well, he's trying this. But I applaud you guys for
• 9:10
  Speaker 2: taking even.
• 9:10
  Speaker 3: This momentarily seriously. In the end, this is about the
• 9:15
  Speaker 3: fact that the companies know that if something goes horribly wrong,
• 9:19
  Speaker 3: they will have someone to depend on them, right? Everybody
• 9:21
  Speaker 3: actually wants kind of the same thing but doesn't want.
• 9:24
  Speaker 2: To say it. What does that mean? I don't know
• 9:25
  Speaker 2: what that means.
• 9:26
  Speaker 3: The companies do not want to be held liable or responsible.
• 9:29
  Speaker 3: In my conversations with the leaders of these companies, they
• 9:31
  Speaker 3: do not want to cause the extinction of humankind.
• 9:35
  Speaker 2: They definitely don't want to be held responsible for it, right?
• 9:39
  Speaker 3: We have one very short-sighted president who refuses to get
• 9:44
  Speaker 3: off the sauce, right? There is an incredible sugar high
• 9:47
  Speaker 3: of the money being poured into the economy.
• 9:49
  Speaker 2: These are solvable problems, right?
• 9:51
  Speaker 3: It is true that the tech is very powerful and
• 9:54
  Speaker 3: does have national security implications.
• 9:57
  Speaker 2: It is also true we can walk and chew gum.
• 9:59
  Speaker 2: And that.
• 9:59
  Speaker 3: by the way, mutually assured destruction has been one of
• 10:03
  Speaker 3: the great checks on power in the last 80 years.
• 10:06
  Speaker 3: And so we can continue to develop, but we also
• 10:09
  Speaker 3: can continue to say, and if it goes wrong, here's
• 10:12
  Speaker 3: your liability, and you have a mandatory duty to report,
• 10:16
  Speaker 3: and we're going to inspect. And that gets you some
• 10:18
  Speaker 3: time while you're dealing with a much more complicated thing,
• 10:21
  Speaker 3: which is dealing with China and figuring out what.
• 10:23
  Speaker 1: Do they want?
• 10:24
  Speaker 2: Hey, it's Rachel.
• 10:25
  Speaker 5: I am so glad to see you at the end
• 10:26
  Speaker 5: of that clip because this is where I get to
• 10:28
  Speaker 5: tell you about a new way to watch all our
• 10:30
  Speaker 5: clips and all of our shows and everything on MSNOW
• 10:35
  Speaker 5: on your phone or your tablet or your laptop.
• 10:37
  Speaker 2: You can watch things live or recorded.
• 10:39
  Speaker 5: Plus, you can ask me and all the other hosts
• 10:41
  Speaker 5: and reporters questions in our weekly live chats. You can
• 10:45
  Speaker 5: go to ms.now slash membership or just click below.
• 10:50
  Speaker 1: Thanks.