WEBVTT - Former Google CEO Eric Schmidt talks about new book & AI build out

0:00:02.400 --> 0:00:06.800
<v Speaker 1>Bloomberg Audio Studios, podcasts, radio news.

0:00:07.000 --> 0:00:09.880
<v Speaker 2>The former Google CEO Eric Smith joins us on his

0:00:10.000 --> 0:00:14.120
<v Speaker 2>latest book, Genesis, Artificial Intelligence, Hope and the Human Spirit.

0:00:14.400 --> 0:00:16.279
<v Speaker 2>Eric joins us now for more, Eric, Welcome to the

0:00:16.320 --> 0:00:17.919
<v Speaker 2>program Sir. I want to pick up on that one

0:00:17.960 --> 0:00:20.520
<v Speaker 2>word hope, and ask you how much hope I should

0:00:20.560 --> 0:00:22.880
<v Speaker 2>have that this goes right? And start with this quote

0:00:22.960 --> 0:00:26.880
<v Speaker 2>from Foreign Affairs magazine. It reads as follows. In cases

0:00:26.880 --> 0:00:30.160
<v Speaker 2>where some humans might face off militarily or diplomatically against

0:00:30.200 --> 0:00:34.360
<v Speaker 2>the highly AI enabled state or against AI itself, humans

0:00:34.400 --> 0:00:38.519
<v Speaker 2>could struggle to survive, much less compete. Such an intermediate

0:00:38.680 --> 0:00:42.960
<v Speaker 2>order could witness an internal implosion of societies and an

0:00:43.040 --> 0:00:46.960
<v Speaker 2>uncontrollable explosion of external conflicts. So you've got to calm

0:00:47.000 --> 0:00:49.519
<v Speaker 2>meet down, Eric. Should I have hope that this plays out?

0:00:49.560 --> 0:00:53.800
<v Speaker 3>Well? Well, let's start with and hope that either of

0:00:53.840 --> 0:00:58.880
<v Speaker 3>your show is incredible. Previous guests talked about two percent productivity.

0:00:58.920 --> 0:01:02.120
<v Speaker 3>Imagine that productivity goes to five percent a year because

0:01:02.120 --> 0:01:03.000
<v Speaker 3>of this technology.

0:01:03.080 --> 0:01:04.800
<v Speaker 1>So there's every reason to think that there's.

0:01:04.920 --> 0:01:09.399
<v Speaker 3>Enormous land businesses, inventions, science, health, and so forth to

0:01:09.400 --> 0:01:11.920
<v Speaker 3>get invented. But the same is also true for the

0:01:11.920 --> 0:01:14.800
<v Speaker 3>invention of conflict and war. And if you think about it,

0:01:14.800 --> 0:01:17.240
<v Speaker 3>it makes absolutely no sense to put somebody in a

0:01:17.280 --> 0:01:18.920
<v Speaker 3>fighter jet up in the air and have it be

0:01:18.959 --> 0:01:21.839
<v Speaker 3>shot at by another missile. It makes much more sense

0:01:21.840 --> 0:01:23.720
<v Speaker 3>for that fighter pilot to be sitting on the ground

0:01:24.240 --> 0:01:25.560
<v Speaker 3>and having a swarm.

0:01:25.280 --> 0:01:26.720
<v Speaker 1>Of the equivalent of those jets.

0:01:26.800 --> 0:01:30.280
<v Speaker 3>Let's just call them powerful robots, powerful drones that act

0:01:30.360 --> 0:01:34.679
<v Speaker 3>in synchrony to achieve the objective of this attacker or

0:01:34.760 --> 0:01:38.480
<v Speaker 3>to defend yourself. And the future of war is autonomous

0:01:38.560 --> 0:01:40.040
<v Speaker 3>and networked and AI driven.

0:01:41.000 --> 0:01:43.200
<v Speaker 4>This is something that a lot of people have focused on,

0:01:43.440 --> 0:01:46.440
<v Speaker 4>in particular how the US is going to evolve in

0:01:46.480 --> 0:01:48.280
<v Speaker 4>this manner, as well as how China is going to

0:01:48.280 --> 0:01:50.400
<v Speaker 4>evolve in this manner, how they're going to use some

0:01:50.440 --> 0:01:54.600
<v Speaker 4>of the space satellites to help coordinate that effort. Eric,

0:01:54.640 --> 0:01:58.360
<v Speaker 4>do you feel like right now one superpower is winning

0:01:58.440 --> 0:02:01.240
<v Speaker 4>versus the other? What is the best way to gain

0:02:01.280 --> 0:02:03.840
<v Speaker 4>preeminence in a military AI sphere?

0:02:04.840 --> 0:02:07.440
<v Speaker 3>But when doctor Kissinger and I went on his last

0:02:07.520 --> 0:02:10.840
<v Speaker 3>trip to China, I was convinced from that meeting that

0:02:10.960 --> 0:02:13.160
<v Speaker 3>China was two to two and a half years behind us.

0:02:13.720 --> 0:02:17.079
<v Speaker 3>But I am now wrong turns out that in the

0:02:17.160 --> 0:02:20.359
<v Speaker 3>last couple of weeks, China has brought out software models

0:02:21.320 --> 0:02:24.359
<v Speaker 3>what are called language large language models that are rivals

0:02:24.400 --> 0:02:26.639
<v Speaker 3>of the best American ones, which I never thought would

0:02:26.680 --> 0:02:29.359
<v Speaker 3>be possible. One is called Quinn, another word's called hong Yong.

0:02:30.400 --> 0:02:33.840
<v Speaker 3>And it looks like they've caught up or they're very

0:02:33.840 --> 0:02:37.520
<v Speaker 3>close behind. And it looks like China has decided that

0:02:37.560 --> 0:02:40.040
<v Speaker 3>it's another part of its industrial policy along with this

0:02:40.160 --> 0:02:43.840
<v Speaker 3>focus on cellular dominating the battery industry, dominated in the

0:02:43.840 --> 0:02:44.800
<v Speaker 3>car industry, et cetera.

0:02:44.880 --> 0:02:46.200
<v Speaker 1>So they're willing to spend the money.

0:02:46.680 --> 0:02:49.520
<v Speaker 3>What's interesting is that it does frame the narrative of

0:02:49.639 --> 0:02:52.079
<v Speaker 3>China versus the US as the defining narrative.

0:02:52.440 --> 0:02:54.320
<v Speaker 1>And one of the more interesting questions, which we.

0:02:54.280 --> 0:02:56.639
<v Speaker 3>Discussed at some length in the book, is what happens

0:02:56.680 --> 0:02:58.360
<v Speaker 3>to the deterrence and what happens to all the other

0:02:58.400 --> 0:03:00.919
<v Speaker 3>one hundred and ninety five countries that are not China

0:03:00.960 --> 0:03:04.400
<v Speaker 3>and not the US. These questions are really important because

0:03:04.680 --> 0:03:09.720
<v Speaker 3>the power of this technology upends society in so many ways, economically,

0:03:10.000 --> 0:03:12.280
<v Speaker 3>the way we govern, the way we reach language, and

0:03:12.320 --> 0:03:13.120
<v Speaker 3>so forth, and so on.

0:03:13.720 --> 0:03:15.360
<v Speaker 4>So there are a lot of questions within this, and

0:03:15.480 --> 0:03:17.920
<v Speaker 4>this could be a half an hour discussion, but one

0:03:18.000 --> 0:03:20.000
<v Speaker 4>of them is who should be financing and who should

0:03:20.000 --> 0:03:24.320
<v Speaker 4>be leading the national efforts to gain pre eminence in

0:03:24.400 --> 0:03:26.760
<v Speaker 4>this sphere. And you have to wonder, you talk about

0:03:26.840 --> 0:03:30.720
<v Speaker 4>China as a state organized, state funded type of effort

0:03:30.840 --> 0:03:33.680
<v Speaker 4>that is coordinated in that manner, how should it be

0:03:33.720 --> 0:03:36.920
<v Speaker 4>directed in a place that does consider itself capitalist like

0:03:36.920 --> 0:03:37.760
<v Speaker 4>the United States.

0:03:38.960 --> 0:03:42.040
<v Speaker 3>Well, the Chinese model is what is called civil military fusion,

0:03:42.240 --> 0:03:45.840
<v Speaker 3>where they subsidize their biggest companies. No one in America,

0:03:46.000 --> 0:03:48.080
<v Speaker 3>not even in the companies, thinks that the thinks that

0:03:48.160 --> 0:03:51.720
<v Speaker 3>the largest American companies need to be subsidized. But let

0:03:51.760 --> 0:03:55.480
<v Speaker 3>me make an argument that if general intelligence, that is,

0:03:55.600 --> 0:03:59.520
<v Speaker 3>human intelligence, is to be invented, it should be invented

0:03:59.720 --> 0:04:02.440
<v Speaker 3>in the United States and under control of Americans and

0:04:02.440 --> 0:04:05.280
<v Speaker 3>the American government. And I don't mean managed, but I

0:04:05.320 --> 0:04:08.520
<v Speaker 3>mean under the legal control of our country and our

0:04:08.560 --> 0:04:12.160
<v Speaker 3>citizens at our democracy and getting there first is a

0:04:12.160 --> 0:04:13.880
<v Speaker 3>big deal. A number of us have talked for some

0:04:14.040 --> 0:04:18.320
<v Speaker 3>time that there should be ANAI for America initiative where

0:04:18.320 --> 0:04:21.960
<v Speaker 3>the government works closely with the private companies and figures

0:04:21.960 --> 0:04:24.160
<v Speaker 3>out a way to make sure that we are ahead

0:04:24.160 --> 0:04:25.119
<v Speaker 3>of ahead of time.

0:04:26.000 --> 0:04:26.520
<v Speaker 1>In truth.

0:04:26.560 --> 0:04:29.479
<v Speaker 3>However, the financial industry that you all invented, for which

0:04:29.520 --> 0:04:33.240
<v Speaker 3>we are incredibly grateful, has been incredibly generous in terms

0:04:33.240 --> 0:04:35.479
<v Speaker 3>of the ability to fundraise the billions of dollars that

0:04:35.480 --> 0:04:38.760
<v Speaker 3>are required, and id state by the way that you

0:04:38.800 --> 0:04:40.599
<v Speaker 3>guys have done such a good job in raising money,

0:04:40.640 --> 0:04:43.239
<v Speaker 3>and we've done such a good job in actually getting

0:04:43.240 --> 0:04:45.279
<v Speaker 3>the computer scientists around the world and the data and

0:04:45.320 --> 0:04:47.600
<v Speaker 3>so on and so on. What we now need is energy.

0:04:48.440 --> 0:04:51.240
<v Speaker 3>We are prolific consumers or energy, and we're going to

0:04:51.279 --> 0:04:53.520
<v Speaker 3>run out. One estimate is that US will run out

0:04:53.560 --> 0:04:56.200
<v Speaker 3>of all sources of energy by twenty twenty eight at

0:04:56.200 --> 0:04:57.120
<v Speaker 3>our current growth rate.

0:04:57.680 --> 0:04:59.800
<v Speaker 5>There's another topic you discussed in this book, which is

0:05:00.400 --> 0:05:01.240
<v Speaker 5>to check AI.

0:05:01.839 --> 0:05:06.880
<v Speaker 1>How can that work? What do you think about it.

0:05:07.160 --> 0:05:09.719
<v Speaker 3>One of the things that happens is in AI systems

0:05:09.800 --> 0:05:12.640
<v Speaker 3>is you have compositional generations. What happens is you take

0:05:13.000 --> 0:05:15.680
<v Speaker 3>one piece and another piece and you interlick them and

0:05:15.720 --> 0:05:19.000
<v Speaker 3>they've never been linked together before, and it looks like

0:05:19.040 --> 0:05:20.600
<v Speaker 3>when you combine those two.

0:05:20.480 --> 0:05:24.080
<v Speaker 1>You get emergent properties. Right, So that's in and of

0:05:24.120 --> 0:05:24.960
<v Speaker 1>itself interesting.

0:05:25.360 --> 0:05:27.240
<v Speaker 3>So the question is what is the limit of that

0:05:27.320 --> 0:05:29.600
<v Speaker 3>how far can can that system go up? I think

0:05:29.680 --> 0:05:33.000
<v Speaker 3>it will get there, and I think that it's probable

0:05:33.040 --> 0:05:36.680
<v Speaker 3>that we can build systems that are the technical term

0:05:36.720 --> 0:05:39.920
<v Speaker 3>is super intelligent, where you have a simple system, sorry,

0:05:39.960 --> 0:05:45.320
<v Speaker 3>a single system that is at the nineteh percentile of physics, math.

0:05:45.520 --> 0:05:48.880
<v Speaker 1>Chemistry, and arts and so forth. No human can do that.

0:05:49.320 --> 0:05:52.120
<v Speaker 3>It looks like these systems will be not only available

0:05:52.160 --> 0:05:54.880
<v Speaker 3>in the next five years because we already have examples

0:05:54.920 --> 0:05:57.680
<v Speaker 3>of passing these tests already, but also that they'll be

0:05:57.720 --> 0:06:01.760
<v Speaker 3>broadly available for all of society. That has huge implications.

0:06:01.800 --> 0:06:04.680
<v Speaker 3>We've never had an experiment where each and every one

0:06:04.720 --> 0:06:07.640
<v Speaker 3>of us has a polymath, you know, the Einstein type

0:06:07.680 --> 0:06:12.880
<v Speaker 3>scientists at our back and call. And furthermore, that scientist

0:06:13.279 --> 0:06:15.280
<v Speaker 3>whatever you want to call him or her or it,

0:06:16.240 --> 0:06:19.480
<v Speaker 3>is capable of writing code, doing agents, and making things happen.

0:06:19.800 --> 0:06:22.120
<v Speaker 5>How do you think the United States competes with China

0:06:22.320 --> 0:06:24.960
<v Speaker 5>when we have an incoming administration that wants to put

0:06:25.000 --> 0:06:27.680
<v Speaker 5>up tariff walls, and even this administration who's been putting

0:06:27.680 --> 0:06:30.799
<v Speaker 5>export controls on some of our high tech semiconductors.

0:06:32.120 --> 0:06:33.919
<v Speaker 1>I am not in favor of broad tariffs.

0:06:33.960 --> 0:06:36.040
<v Speaker 3>I never have been I'm the son of an economist

0:06:36.279 --> 0:06:40.040
<v Speaker 3>and a grandson of an economist, and tariffs are essentially taxes.

0:06:40.440 --> 0:06:43.480
<v Speaker 3>What I am in favor of is restrictions or limitations

0:06:43.520 --> 0:06:47.320
<v Speaker 3>for strategic reasons. It's really important that we win in

0:06:47.320 --> 0:06:51.240
<v Speaker 3>this race against China. And so the things that the

0:06:51.279 --> 0:06:54.800
<v Speaker 3>Trump administration Biden administration to limit, for example, hardware access,

0:06:54.839 --> 0:06:57.080
<v Speaker 3>were really smart. And I know because I was part

0:06:57.080 --> 0:07:00.840
<v Speaker 3>of a commission that recommended it way back when in

0:07:00.839 --> 0:07:01.719
<v Speaker 3>fact was adopted.

0:07:01.800 --> 0:07:02.360
<v Speaker 1>Great job.

0:07:02.800 --> 0:07:05.440
<v Speaker 3>So the question here is ignoring the tariff question, which

0:07:05.440 --> 0:07:06.719
<v Speaker 3>I just don't like, and I think this is just

0:07:06.760 --> 0:07:10.120
<v Speaker 3>another tax. How do you control how do you control

0:07:10.160 --> 0:07:13.600
<v Speaker 3>and limit what China does? And how do you amplify

0:07:13.640 --> 0:07:16.600
<v Speaker 3>what America does? And you sit there and you go, well,

0:07:16.680 --> 0:07:19.000
<v Speaker 3>you're doing pretty well, right. Everybody's making a lot of money.

0:07:19.000 --> 0:07:20.960
<v Speaker 3>There's all this growth, and video is doing well. All

0:07:20.960 --> 0:07:23.280
<v Speaker 3>the tech companies are doing well. But let me explain

0:07:23.320 --> 0:07:26.080
<v Speaker 3>what happens. You have a slope that's like this, and

0:07:26.120 --> 0:07:28.720
<v Speaker 3>it keeps going up and up and up with humans.

0:07:29.120 --> 0:07:32.240
<v Speaker 3>At some point, the industry believes that there will be

0:07:32.320 --> 0:07:36.080
<v Speaker 3>AI scientists, that is, non human scientists, and he or

0:07:36.120 --> 0:07:38.440
<v Speaker 3>she who gets there first gets a slope like this

0:07:38.560 --> 0:07:39.760
<v Speaker 3>and all of a sudden boom.

0:07:39.960 --> 0:07:42.560
<v Speaker 1>Yeah right, we're really really growing fast.

0:07:42.960 --> 0:07:45.600
<v Speaker 3>These network effect businesses are what we do in our

0:07:45.640 --> 0:07:49.160
<v Speaker 3>in our competitive environment against each other under a US regulation.

0:07:49.360 --> 0:07:51.040
<v Speaker 1>I want that to be true globally for.

0:07:51.000 --> 0:07:52.800
<v Speaker 2>The US, Eric, I can tell you we're all in

0:07:52.840 --> 0:07:55.200
<v Speaker 2>favor of having your back soon to continue the conversation.

0:07:55.320 --> 0:07:57.920
<v Speaker 2>Congratulations on the new book, so we appreciate your time.

0:07:57.920 --> 0:08:00.000
<v Speaker 2>The former Google CEO Eric Smith