The Great America Show - DOES A.I. SCARE YOU？ WILL A.I. MAKE US SMARTER, BETTER, MORE PRODUCTIVE OR WILL A.I. BE THE END OF CIVILIZATION, THE END OF THE HUMAN ERA？

00:00:00.000 Hello, everybody. I'm Lou Dobbs, and this is The Great America Show. We welcome you. Great to have

00:00:05.420 you with us. The big story of the week, that is right after the breaking installments from the

00:00:10.820 running drama at Fox News, is the 6 a.m. video announcement from the White House that the

00:00:17.160 world's greatest living puppet president, who is 80 years old, impaired and compromised,

00:00:23.060 has, after careful deliberation, deep prayer, earnest consultation with his family,

00:00:30.000 and humble counsel with President Xi Jinping and Vladimir Zelensky, he has made up his mind,

00:00:37.360 and Mr. Biden has decided to run for reselection in 2024. And part of the excitement around the

00:00:45.560 president's announcement, and let's face it, this wasn't entirely unexpected. Part of the excitement

00:00:51.740 is Mr. Biden has already thrown a changeup. And this is my insight, my analysis of what may not

00:00:59.180 be obvious to casual observers. Biden's video message is the message, I believe. Biden announced

00:01:07.920 he's running on video. And that's a big change from 2020, when he was doing most of his talking

00:01:14.980 and campaigning from his basement in Wilmington, Delaware. Now it looks to me like he's made a

00:01:21.340 solid commitment to being on video throughout his re-election campaign. A lot to look for there

00:01:28.420 in the coming weeks and months as President Biden seeks a second term and a way to energize and to

00:01:35.240 excite a base that, for the most part, isn't excited or perhaps even excitable. The 30 percent of

00:01:43.380 Democrats who do approve of President Biden's madcap policies of the last two and a half years

00:01:49.320 are cheering. We haven't, however, checked in with the other 70 percent who are opposed to the idea.

00:01:55.820 Biden's $7 trillion fiscal 2024 budget is a killer. There's no other way to say it. That is a heavy

00:02:04.840 load for any, any incumbent to carry to the voters. And so is the almost $2 trillion deficit

00:02:11.880 that results. For all practical purposes, Biden is leading a regime that has little resemblance to

00:02:19.480 the constitutional republic that preceded it for 240 years. Biden has gone rogue and he's getting

00:02:26.980 away with it. He leads by fiat. He issues executive orders and does exactly what he wants. Wide open

00:02:34.960 borders. He's sending more than $100 billion to Ukraine. He refuses to arbitrate a peace deal with

00:02:41.760 Russia and Ukraine. He's doing whatever China wants. Ends U.S. energy independence. Draws down

00:02:48.560 our strategic petroleum reserves to the lowest level in 40 years and sells some of that oil to the

00:02:55.120 Chinese, for crying out loud. Then drives Saudi Arabia into the arms of Vladimir Putin and the 0.94

00:03:01.080 Ayatollahs. Inflation is at 6 percent. Markets are volatile and the economy nearing recession.

00:03:06.960 This is an incumbent looking for re-election? Really? And President Biden and his family are

00:03:14.360 crooked, corrupt and dangerous, I'm told reliably. The House investigating committees already have 0.99

00:03:20.440 evidence that a dozen family members are profiting from Biden's influence peddling. And we know now for

00:03:27.360 a fact that the Marxist Dems, the Deep State and Biden campaign in 2020 conspired to create a cover-up

00:03:35.540 of Hunter's laptop with the help of the Obama intelligence chiefs and a deputy State Department

00:03:42.000 official who is now Biden's Secretary of State. Biden may not be held to account for what he did in

00:03:49.160 2020. But I do believe he doesn't have a chance of re-election. Mark my words, not a chance. Biden may,

00:03:57.380 of course, do more damage, and almost certainly he will. But this rogue Biden-Marxist-Dim regime

00:04:03.280 is in its last throes. But as they say, there's more. Not about Biden or even presidential politics,

00:04:11.160 but about what some have styled humanity's last invention, or the end of civilization,

00:04:17.180 the end of the human era. Now, I would call that somewhat sensationalist exposition. But what do I

00:04:23.760 know? What do any of us know, really, when compared to artificial intelligence, which promises to be

00:04:29.680 humanity's next big thing? Revolutionary, explosive, amazing, transformational. And only a few thousand

00:04:38.600 people on this planet seem to know what this thing called artificial intelligence really is,

00:04:44.480 or what it's likely to become. But we do know we're all about to be profoundly affected by it.

00:04:51.140 And I thought it might be useful if you and I started to learn more about AI, or AGI,

00:04:56.740 artificial generative intelligence, its benefits, its dangers. And our guest today is one of those

00:05:04.020 few on this planet who's been thinking about AI for some time. He co-founded, in fact, a machine

00:05:10.120 learning startup. The company was Geometric Intelligence, which he sold to Uber seven years

00:05:15.920 ago. He's a cognitive scientist, NYU professor emeritus of psychology and neural science,

00:05:22.040 and author of Rebooting AI, Building Artificial Intelligence We Can Trust. Available on Amazon,

00:05:29.640 and we recommend the book to you highly. And we welcome to The Great America Show, Gary Marcus.

00:05:35.780 Gary, good of you to be with us. We're almost all of us hearing a lot about AI. We know the big tech

00:05:41.840 companies are warning all of us about the dangers of AI, the dangers of AI just in the hands of the

00:05:48.020 corporations themselves. And we're told by the Biden Department of Homeland Security that they

00:05:53.980 want all the AI they can get to control people and to control information. And we don't like the sound

00:06:00.580 of that either, Gary. The truth is, most of us don't know whether to fear or cheer AI. So could we

00:06:07.620 please start, Gary, with just what is artificial intelligence?

00:06:11.520 It's actually hard to define. It's a little, I know it when you see it, but it's basically the idea of

00:06:17.100 machines doing smart things. You could argue and say a calculator counts as a little bit of artificial

00:06:22.200 intelligence. And you could say that the Star Trek computer that could talk about anything is maybe a lot

00:06:27.900 more artificial intelligence. Intelligence actually has many different aspects to it. But basically, we're talking

00:06:33.240 about machines that can do smart things, maybe replacing people, maybe augmenting people to do things they

00:06:38.300 couldn't do before. Augmentation. That's a that's an interesting word, because this concern right now being

00:06:45.680 expressed by that in that open letter, and now in follow up to that, there's, I will put it this way, AI

00:06:54.280 leaders, including yourself, talking about really the dangers that AI represents. And right now, that seems to be

00:07:03.900 more the focus than the benefits of AI. Do you agree? Well, I mean, they're both. I think a lot of

00:07:10.040 people are thinking every day about the benefits. So for example, these new systems save computer

00:07:13.940 programmers a lot of time. And increasing the productivity of programmers is a great thing. So

00:07:18.760 there's definitely some benefits here. They're also fun to play with and so forth. But you ultimately,

00:07:22.960 you want to ask a kind of cost benefit trade off. And there are certainly a lot of risks around these

00:07:28.460 systems. And things have moved so quickly that I would say, the first thing that a lot of us are

00:07:33.900 saying is maybe we need to slow down and understand where we are. So I don't think anybody can give an

00:07:39.160 honest question to do the benefits outweigh the risks, because nobody fully understands with these

00:07:44.720 new technologies, what the limits are and how they can be used. So probably someone else might make

00:07:49.780 the case for the positive. I've been focusing on the negative and trying to understand that.

00:07:53.600 My biggest short term concern is about misinformation, the ability of say, foreign 0.99

00:07:59.680 countries to disrupt elections by making up as much misinformation as they want in incredibly

00:08:04.640 plausible ways. Nobody can tell the difference. Just much greater volume. So the cost of misinformation

00:08:10.000 has gone to zero. And I think that's threatening democracy. So that's one concern I have. Another

00:08:15.360 concern is that cyber criminals can use this stuff to trick people. And there are now new tools like

00:08:20.580 auto GPT where one AI controls another. And we might see situations where people do like

00:08:25.940 fishing expeditions to get people's credentials, or there's something called a pig butchering scheme

00:08:32.260 where you pretend to be somebody's friend and eventually get them to send you money and you milk

00:08:36.760 them for money. And we may see those things automated in a way that we've never seen before.

00:08:41.080 So there are a lot of risks like that. I would call those nearer term risks. Another near term risk is that

00:08:45.880 people might trust these new search engines, don't really know what they're talking about,

00:08:50.020 aren't really that reliable with, for example, medicine or consult them almost like they would

00:08:55.880 consult a psychiatrist. And there may be problems there as well. Misinformation, bad advice. We've

00:09:02.380 already seen a situation where people are essentially in love with these chatbots. And then the chatbot

00:09:07.760 stopped basically having, what is the polite way to say this, having relations with them, verbal relations

00:09:14.740 with them. And people were really upset. And so like, we have attachment issues to it. Most people don't

00:09:19.540 understand that these systems aren't really in form attachments. So there are a lot of different

00:09:26.880 questions like that. And then there are longer term questions about what happens if you have a lot of

00:09:31.400 systems that aren't fully reliable, and you start hooking them up to more and more aspects of the

00:09:35.860 world. And they're really long term questions. I don't take so, so seriously, but I don't think we

00:09:40.940 have a full answer to like a kind of terminator scenario. Like what if they turn on us? And I don't think

00:09:46.000 that's very plausible. But I don't think we have a formal proof that it can't happen. And so I think

00:09:49.880 we do need to take it into consideration. And the reality is, although these new AI systems are very

00:09:55.520 interesting, we don't fully control them. We don't fully understand what they're doing. We call them

00:09:59.800 black boxes. We put in a lot of input data. We don't know exactly what comes out. We don't know

00:10:04.740 exactly what they do. And that's enough for at least some of us to say, maybe we should slow down.

00:10:09.920 You know, my TED talk yesterday, what I called for was for global governance for AI,

00:10:14.180 some kind of coalition where we bring governments together with the companies and

00:10:19.240 brought a representation around the world to try to figure out what should we do

00:10:23.300 about many different individual questions in AI in some kind of coordinated way.

00:10:27.900 Nobody really wants to have, you know, 195 companies, 195 countries with 195 sets of rules.

00:10:33.740 That's not even the interest of the companies. And the companies have to spend a lot of money to

00:10:38.040 train their models. If they do that, you know, uniquely for every country, that's not really good

00:10:41.980 for them either. So I think we're in a rare moment in political history where kind of everybody

00:10:46.820 actually wants the same thing, which is to figure out a regulatory framework where the tech companies

00:10:51.700 can do what they want, but where the citizens are protected and where governments still, you know,

00:10:57.280 have some power. And so it's an interesting and complicated moment in history, but I think it's a

00:11:02.160 good moment to try to work this out. And we don't want to do this like five years from now. We really

00:11:07.080 want to do it now. It's fascinating to start pondering all of the possibilities here to even

00:11:14.540 in restricting, if you will, containing artificial intelligence, restraining it still under human

00:11:21.440 control, if you will. And when I said that word augmentation is interesting because augmentation can

00:11:29.200 also suggest giving great power to the folks you just talked about, criminals, those who mean to

00:11:37.940 do harm in any fashion. And we ultimately are talking about policing, not AI, but ourselves

00:11:45.800 in that instance. And yet the long-term dangers, as you put out, and frankly, Gary, you're the only one

00:11:53.840 I know who's talked about this in both the near term and the long-term, talked about it in terms

00:11:59.220 of the immediate societal impact as well as the long-term. And I think that all of that has to be

00:12:07.560 discussed. But I really wonder when I see Time magazine saying, you know, we really can't,

00:12:15.740 we can't pause. We've, you know, we've got to pause, excuse me, and we've got to hold fire.

00:12:24.400 The Chamber of Commerce comes out and says, we can't pause. We've got to beat the doggone Chinese 1.00

00:12:29.700 or the Russians or whomever. That divide is proximate and it's critical because that's what 0.94

00:12:38.840 we're talking about now is competitive AI. And the nation states are the ones who are going to be

00:12:46.100 most competitive in terms of retaining power. The corporations most aggressive, given the fact

00:12:54.740 that they are competitive institutions and mean to win as well, but in a different realm, right?

00:13:00.780 I mean, there's a lot going on there. So there's definitely competition between companies,

00:13:04.860 there's competition between the countries. There's an argument that we shouldn't pause

00:13:08.960 because China will get way ahead of us. I'm not so worried about that particular thing

00:13:13.200 in the sense that the AI that we have now is still fairly limited. I think there are a lot

00:13:18.200 of fantasies that like if China gets to use GPT-5 three months before we do, that they'll invent, 0.86

00:13:24.180 I don't know, spaceships, or they'll invent some renewable energy that we don't have or something

00:13:28.720 like that. And these tools are not really good for that. There will be AI like that that's

00:13:33.100 sort of genuinely super intelligent, can do scientific reasoning, invent new technologies.

00:13:38.740 What we're talking about now is more like a productivity enhancer. I mean, people probably

00:13:42.920 play with chat GPT. You can use it to write boilerplate text for you and things like that.

00:13:47.980 It's not going to completely change the world if one nation can write boilerplate text faster than

00:13:52.280 the other, or even if one can code faster than the other for a few months. It might make some

00:13:56.720 differences in productivity. I don't think we're at the level of the technology that some people are

00:14:02.000 fantasing about. You know, China's not going to wake up and build interstellar travel that we don't 0.93

00:14:06.340 because they have this tool a few months earlier. But there are all these nearer term problems that

00:14:11.900 everybody faces, whether they have GPT-4 or GPT-5. I don't know that that's really the critical

00:14:16.140 variable. And nobody really is calling for the end of all AI research. The letter itself actually

00:14:22.580 called for a pause only on particular research on this one model, GPT-5, and actually encouraged

00:14:29.820 more research around safety, around making these systems trustworthy and reliable.

00:14:34.600 I think of the AI we have right now as sort of like a teenager, like it's powerful, but not very

00:14:39.400 well controlled yet. It doesn't have a prefrontal cortex to kind of tell it what's right and wrong.

00:14:44.140 And I think we should be mostly focused not on like who builds the biggest model fastest,

00:14:48.980 but who can figure out how to make this stuff tractable and reliable, have the stuff work in an

00:14:55.400 ethical way, in an honest way. They have a huge problem with hallucinations, making stuff up.

00:15:00.600 Nobody's calling for a ban on that kind of research. And I think we should focus more on

00:15:04.940 that, on how to make it so that these systems are things we can count on. I mean, like, you know,

00:15:09.320 it's a nightmare science fiction story when the computer goes out of control. What we really want

00:15:14.160 are computers that will do what we want them to do and that are aligned with our interests. And we don't

00:15:20.120 really have that so much yet. We're going to find out what we do have here next. We're talking with

00:15:25.660 Gary Marcus, a leading voice in artificial intelligence. Stay with us for this quick

00:15:30.740 message from our sponsors. We're coming right back. We're back now. We're talking with Gary Marcus

00:15:36.660 and Gary, you mentioned science fiction and I go back to 1969 and, you know, Space Odyssey and how

00:15:45.920 we're talking about something that's 50 years old is a very good, I think, metaphor for what all is

00:15:55.880 going on. An avatar of two is for that matter. How close are we to Hal?

00:16:03.400 Well, Hal had very good language, very good comprehension. I would say it's ahead of what

00:16:07.500 we have right now. What we have now gives an illusion of understanding language, but it doesn't

00:16:12.000 really. Sometimes I think of her, if you want to talk about science fiction, where the Scarlett

00:16:17.340 Johansson character was sort of all-purpose general assistant who understood a lot of things,

00:16:22.840 had a good theory of how human beings worked. I would say we're somewhat far from that. The latest

00:16:28.300 research shows these systems don't really have that. The thing that we have now that's probably

00:16:32.800 most impressive is these systems are general. They can work on many different things, but their level

00:16:37.260 of comprehension is still pretty poor. In that sense, I think we're still fairly far from Hal.

00:16:41.920 And of course, you know, I don't want to give away the plot for anybody who still hasn't seen 2001,

00:16:45.600 but let's just say that issues about control are important there. They weren't fully resolved in

00:16:50.680 the movie, and they're certainly not fully resolved in the real world right now.

00:16:54.600 And as we are, Sundar Pichai said, the CEO of Google, said he doesn't believe this is a decision for any

00:17:04.180 corporation. And he is calling, just as you intimated, for a broader discussion with ethicists,

00:17:12.780 philosophers, and we're talking about then, of course, the possibility of the intrusion. And I

00:17:20.440 know you're calling for government involvement, global governance. But right now, we have government

00:17:26.740 that I'm a product of the 60s, Gary. I don't trust government. And I am not the first person to

00:17:33.700 get to the line and say, you know, what we need here is more government and government control.

00:17:39.160 Your thoughts on the problems that would be created with one world governance?

00:17:45.660 I mean, I think that, you know, a critical part of what makes the US government work as well as it

00:17:50.300 does, which is not perfectly, is checks and balances. And I think we need some checks and

00:17:55.400 balances in the global governance of AI. We need to have both the companies and the governments

00:18:01.200 at the table. And they both have interests that are not necessarily truly aligned with the citizens'

00:18:06.580 interest. In an ideal world, maybe they would be, but we don't live in that ideal world.

00:18:10.840 And so I think we need a lot of stakeholders to balance a lot of things here.

00:18:14.280 I think that, you know, we don't want the companies to have all the power either, right? So nobody

00:18:21.380 really thinks about it, for example. But ChatGPT is sucking down lots of private information. We saw,

00:18:26.200 I think it was Samsung, we go, you know, people were typing in private company data, and then

00:18:30.820 suddenly OpenAI had its hands on. And so, you know, there's multiple concerns here about who has

00:18:37.860 control of data, about who makes decisions about the politics, essentially, of these systems.

00:18:44.280 There are concerns about whether we want any regulation about what you can release. So,

00:18:49.580 for example, you know, in the pharmaceutical system, we have phase one, phase two, phase three

00:18:53.720 trials. You don't just try something out on 100 million people without testing it first.

00:18:58.360 Probably don't want the tech firms to be able to do that. That's what they did 0.68

00:19:01.300 with Sydney. And they didn't really quite know what they were doing, as far as I can tell.

00:19:07.020 You know, they kind of just threw it out there and wanted to see what happens. And so,

00:19:10.840 we may want some regulation around that, for example, that the tech companies might not do

00:19:15.460 on their own. So, there are many different trade-offs that have to be made. But I think

00:19:21.680 to leave them entirely to the government or entirely to industry, neither of those models

00:19:25.100 really works. And so, we need some way of balancing those interests. And you mentioned,

00:19:29.860 like, having philosophers at the table. I think we need a lot of people at the table. We need economists,

00:19:33.860 we need philosophers, anthropologists, pretty much all fields. I think we really do want

00:19:39.100 global representation here to try to come to something that works for everybody across the

00:19:44.160 table.

00:19:44.380 Working for everybody across the table, the folks I worry about are the folks. That is,

00:19:51.580 people who are underrepresented right now, for example, in the United States. It's a government

00:19:57.260 that is divided between two parties, neither of which is arguably, well, I'll put it this way. In my

00:20:03.600 view, government is too much about growing government. We are very suspicious of what these policies are

00:20:12.040 leading to. We're very suspicious of China and its intentions toward the United States, and indeed,

00:20:19.720 world civilization. The threat of AI, and I understand that this is not proximate, that it is some time off,

00:20:29.920 but the way in which it's progressing seems geometric to me. What GPT-4 is, next up is five,

00:20:40.280 and how quickly do we get to 10 or 15? Is there a velocity multiplier here that we should also expect?

00:20:51.800 I'll talk about that in a second, but I'll go back first and say, I don't think that AI governance

00:20:56.540 actually should be a right-left issue. And I think it's interesting, for example, that Peggy Noonan,

00:21:01.380 who is, as you well know, was Reagan's speechwriter, one of Reagan's speechwriters,

00:21:05.720 came out in the Wall Street Journal saying, we need a longer pause. I think that everybody,

00:21:11.020 whatever their party is, should be concerned about tech companies having that much power to shape

00:21:15.200 our lives without any kind of government say over it at all, or any say for the people over it at all.

00:21:23.200 It's sort of like what we've seen with social media, but I think an even greater extent in terms of

00:21:28.580 invasion of privacy and control about what information we see and so forth. So I think

00:21:33.080 that we may actually see a surprising amount of unity between the right and left,

00:21:37.880 which, as we all know, has largely been a dysfunctional divide for a long time. But I

00:21:41.700 think on this issue, there's reason for everybody to care. On the acceleration issue,

00:21:47.860 it's not clear because the enormous energy costs, enormous expense of training bigger and bigger

00:21:56.120 models, it's not clear how long we can push it. I like to think of Moore's law. We all thought

00:22:00.700 you could just double the amount of transistors you had indefinitely forever and keep cutting the

00:22:06.780 costs. And it actually, by most people's accounts, started to slow down around the year 2000.

00:22:12.100 So Moore's law is not a physical law of the universe like gravity. It's just something that's

00:22:16.840 a generalization that we saw over time for a while, and it lasted for a while, and then it stopped.

00:22:21.180 It's not clear there's enough, let's say, electricity in the United States to actually train

00:22:25.740 GPT-10 or GPT-11 or something like that. So at some point, these things are going to

00:22:31.900 stop accelerating at the speed that they are. But they will continue for a while, and we're not that

00:22:39.240 good at projecting out what they'll look like, say, even two years from now.

00:22:42.920 And as we think about artificial intelligence, and we think about the cloud, can AI be contained?

00:22:54.520 Or will we see an array of computers that just gets double the volume every, you know, following

00:23:01.900 Moore's law, as you suggest, it perhaps won't last much longer than Moore's law. But the fact is,

00:23:08.740 it wouldn't have to do too much in the way of geometric progression to be just an unthinkable

00:23:16.080 and extraordinary regenerative artificial intelligence that would be working at light speed

00:23:25.540 and creating just dazzling results.

00:23:32.160 Well, yes and no. So like, I wrote a piece, an essay in my substack, Gary Marcus, that substack,

00:23:38.600 called What to Expect When You're Expecting GPT-4. And I predicted that it would have a lot of the

00:23:43.400 same problems as GPT-3, like hallucinating, making stuff up, having trouble understanding

00:23:49.460 the physical world, the psychological world. And all of those predictions were actually true.

00:23:53.820 Like, there's some ways in which these systems are better, and there's some ways in which they

00:23:57.360 really haven't improved at all. I'll give you another example. GPT-4 was trained on a lot of chess

00:24:02.580 games and on the rules of chess, but it can't even always follow the rules. And it doesn't play any

00:24:07.500 better than a chess computer from 1978. You wouldn't want to put it in a car to drive your

00:24:12.480 car. There are lots of ways in which these AIs are actually still pretty limited. And it's not

00:24:18.080 clear that doubling and doubling the current technology is actually going to solve those

00:24:21.820 problems. So you will see kind of more of what we have now of this kind of being able to write

00:24:27.620 boilerplate text and be able to do some interesting things. But I wouldn't assume that it's going to

00:24:31.760 be what we sometimes call, excuse me, artificial general intelligence that can solve any problem.

00:24:37.340 We're talking with Gary Marcus, a leading voice in AI. He's author of the book, Rebooting AI,

00:24:45.180 among the very first to be cautionary in terms of AI. And we're going to continue our discussion.

00:24:54.060 And if I may say, Gary, a fascinating discussion. Stay with us, please, for this brief message from

00:24:58.900 our sponsors. We're coming right back. We're back now talking with Gary Marcus. And Gary,

00:25:05.680 I have to say, thank you so much for being here today, because you're instructive and you're

00:25:12.880 illuminating. And we appreciate very much your time and your thoughtfulness. Elon Musk's warning of a

00:25:20.200 threat to civilization. It's a threat that you don't see, apparently, as...

00:25:25.120 Not as imminent. I mean, I think it's possible, but I don't, I'm not as concerned about that

00:25:31.080 particular one. I think we should have some awareness of it.

00:25:34.900 And as we look at what's involved here, there is an effort to talk about transhumanism and AI as if

00:25:42.300 they are, well, they're going to meld into one form. Your thoughts about that?

00:25:50.800 I think we will see more and more kind of augmentation. I don't know about transhumanism, 1.00

00:25:56.360 but like already, like my cell phone augments my mental life, right? Like it remembers all my phone

00:26:01.320 numbers and appointments for me. And we will see more and more of that where we rely on machines to

00:26:06.340 do more and more for us. And I think there's been a lot of talk about people losing jobs. So far,

00:26:11.240 AI has not taken that many jobs, but it's made a lot of jobs more powerful and more effective. So like

00:26:16.260 we heard five years ago that all the taxi drivers were going to lose their jobs and they didn't.

00:26:20.820 We heard that radiologists were all going to lose their jobs and they didn't. What radiologists do

00:26:26.100 now is they can do more work faster by using the AI, but there's still some human judgment there.

00:26:31.260 You know, in a hundred years, maybe machines will just do most of our work for us, but at least

00:26:35.460 in the near term, they're just going to make our jobs easier. And they'll change some people's jobs,

00:26:41.260 change the dynamics of it. But we will mostly be working together with the machines for a while.

00:26:46.120 Well, working together, that sounds good. I have to say some people, as you well know,

00:26:54.160 consider it to be the next step in human evolution, which is a fascinating concept,

00:27:00.820 but you can't get perfection from artificial intelligence, at least now. What is the future

00:27:07.980 as you see it? I guess it depends on the timescale. I don't think anybody can predict

00:27:12.840 what it's all going to be like a hundred years from now. I mean, think about all the things that

00:27:16.200 weren't here a hundred years ago. Like there weren't commercial airliners, there weren't cell

00:27:22.380 phones, there was no social media. I guess there wasn't television yet. And then maybe on the drawing

00:27:28.080 board, a hundred years is a long time. And I think in the AI world, it's particularly long. I'm just

00:27:34.340 looking at what happened in the last few months. I don't think we can really predict that. I think

00:27:38.740 we can predict that in the next decade, employment will still be pretty good, but maybe not as good

00:27:43.220 as it is now. I think we can predict that driverless cars are actually going to take a while yet,

00:27:48.520 that we're not really to the level of reliability to do that. And we can predict that AI is going to

00:27:53.780 be more and more of our daily life. And that's going to rapidly escalate over the next several years.

00:27:59.220 Uh, beneficial and helpful and, uh, extraordinary. I'm sorry, good. They are the good and the bad.

00:28:07.640 We're going to see more good and we're going to see more bad. We're going to see more of it. It's going

00:28:11.080 to be more of a focus of our lives. Going back to that word you used initially, augmentation, uh, the

00:28:17.420 choices, and you were talking about, it shouldn't be a red or blue thing, uh, a partisan matter. Uh, but we

00:28:24.120 always seem to devolve, uh, to ideological and partisan differences, uh, around the world. Uh,

00:28:32.140 and we know those differences between China and the United States now are widening, uh, and more

00:28:37.820 intense than ever. Give us your judgment about what is a safe way to proceed to have a, a geostrategic,

00:28:49.400 uh, advantage, uh, against this country's enemies. I mean, I think, you know, every country has to

00:28:56.620 continue to do the kinds of things that it's done in its defense. And, and, you know, the U S needs

00:29:03.080 to think about how, for example, its defense department can use these technologies, how it

00:29:08.060 can deal with the limitations of these technologies. Like, I don't think anything really changes there.

00:29:13.220 We're always trying to, to figure out how to maximize our use of new technologies. And we certainly

00:29:18.660 should be doing that. Um, I think we have to do it with eyes open. I think a lot of people

00:29:23.000 treat these technologies as if they're magic and really they're just a set of tools that have

00:29:27.540 strengths and weaknesses. So we need to be informed and nuanced about how we do it. Um, but I don't

00:29:32.680 think any of that changes fundamentally, but I think it is going to keep a lot of people really busy

00:29:38.020 because suddenly there are all these opportunities and all these risks, and it's going to take a lot of

00:29:43.360 work to really understand, you know, how does this technology work in the real world? What are the use

00:29:47.640 cases where it's actually helping me? What are the use cases where I can't really trust it?

00:29:51.640 So there's plenty of work. And I mean, the Chinese have to do that just like the people 0.95

00:29:55.000 in the United States. I'm an American citizen. Um, you know, everybody has to look at these new

00:30:01.360 tools and say, what are the risks? What are the benefits? How are we going to use them?

00:30:05.420 And when we look at the tools, we're talking about government, uh, and those in government

00:30:10.800 trying to make assessments, what strikes me and one of the reasons we're having this discussion

00:30:16.300 and this program will be discussing this issue a lot because of what you have been talking about

00:30:23.260 here today, the potentialities are tremendous. I don't think I can think of a, uh, of a development

00:30:31.180 that has any more powerful, uh, uh, than this, uh, seems to be, uh, and government is not

00:30:41.120 possessed of the minds that are necessary to comprehend and to, to shape that future because

00:30:47.980 of their technological, uh, yeah, there's a very serious problem that governments aren't

00:30:54.040 technologists and that they need technologists at the table and they need not only the technologists

00:30:58.620 at the big corporations who obviously have vested interests, they need, you know, smart

00:31:02.540 academics and researchers and so forth who have thought about these things too. Part of

00:31:06.800 the reason to have a global alliance, which is what I'm pushing for is to have a lot of

00:31:10.400 expertise on board so that, you know, individual governments that don't have that expertise to

00:31:15.420 have a place to consult and, and, you know, to ask, you know, what should I think of this

00:31:21.040 new technology? We have nothing like that. Most governments have no training in this, or they

00:31:25.540 have a few people that have a little bit of training. Um, and that means they're not really

00:31:29.620 up to speed. So, you know, governments need a place to turn. I think an international organization

00:31:35.360 that's neutral could be, could be a place for the governments to get informed. I think there's

00:31:39.440 also, we haven't talked about it, but a huge need for AI literacy around the entire globe,

00:31:43.240 um, for all citizens, for all governments. And that needs to be part of this too, is, is figuring

00:31:49.360 out how to get people up to speed on like, when do you trust these things? Why you shouldn't

00:31:53.760 treat them as humans, even though they seem like humans, um, you know, how fast are they

00:31:58.300 moving? What can they do? Where do they go wrong? Why do they hallucinate? Um, you know,

00:32:02.940 we need a lot of literacy around that. I'm going to start doing some like, uh, uh, animated

00:32:07.520 videos with, with one of the television networks to try to raise some AI literacy, but we need

00:32:12.620 to do this at every level from, you know, young kids all the way up to governments.

00:32:17.760 Well, and that's why we're, again, we're having this discussion here today. And one of the

00:32:22.760 reasons we're deeply appreciative is because this, this podcast is going to be dedicated

00:32:27.380 to, to raising that literacy and, and bringing, uh, information to our audience, uh, because

00:32:33.320 it's just, it's, it's, it's, uh, I'm a populist and I want, uh, you talked about people around

00:32:39.600 the table. I want the people around that table, uh, and not intermediate areas. I want them there,

00:32:45.320 uh, because the center of this country depends on, uh, on them. Uh, that's, that's where we live.

00:32:52.620 And our values shine brightest, uh, when this nation is at its best, uh, give us, if you will,

00:32:59.940 you're, as we're wrapping up here, give us, can I actually jump in for one second? Um, I don't

00:33:06.500 usually plug things so directly, but I think it's so relevant. I have a new podcast called coming out

00:33:10.900 called humans versus machines. And it's really designed to get people kind of a deep dive into

00:33:15.500 how all of these things work. So we'll, for example, talk about the rise of IBM Watson and

00:33:20.220 how they went in jeopardy and then how they overpromised and said they'd solve cancer and

00:33:23.580 how that failed. So that's humans machines comes out coming out next week. Um, and it's very much

00:33:28.360 designed to go to the people and teach people, um, how this all works.

00:33:32.300 And the title is humans and machines, humans versus machines versus machines. I think you've

00:33:38.980 got a great time. It looks, it looks like that's exactly what, uh, we're going to be looking at

00:33:46.040 talking about it in terms of, uh, whether appropriately or not. Uh, I think that's wonderful. And I, I wish

00:33:52.040 all of the best of luck. Uh, and I want to, uh, once again, it's humans versus machines, a podcast

00:33:59.480 starting, you said next week, Gary, Gary markets will be leading that hosting it. And we look

00:34:04.900 forward to that. And Gary, we also hope you will come back and join us on this podcast, uh, for more

00:34:11.960 discussions. It's fun. It's fun. It's really important. So anytime. And we always give our

00:34:17.260 guests the last word. Uh, so you're concluding thoughts, uh, here today. I think everybody needs

00:34:23.600 to come together around this left, right governments, corporations, citizens. We all need to make sure that

00:34:29.440 we get the value out of these things, but also that we have enough control over them that we can

00:34:33.700 trust them. Well said. And I want to say to you, thank you so much for being with us here today for

00:34:39.460 educating, uh, us. I will add, um, specifically myself. I appreciate it so much, Gary. It's just

00:34:47.600 been wonderful talking with you. I hope you'll come back soon. I'd love to real pleasure. Thank you very

00:34:52.560 much. Gary Marcus. Thanks for the tutorial. Thanks for being with us. I hope you found Gary is

00:34:57.320 interesting and instructive as I did. It's a tough subject, but one I think we all need to be

00:35:02.940 thinking about, and we're going to have a number of guests here to lead us through all of this.

00:35:08.820 Thanks everybody for being with us here tomorrow. Our guest will be former Trump presidential

00:35:12.840 assistant, Peter Navarro. Peter Navarro has been charged with contempt of Congress for honoring the

00:35:19.080 presidential executive privilege that the January 6th committee chose to utterly ignore. Peter's been in a

00:35:26.100 battle and he's fighting through. Please join us tomorrow. Till then, thank you and God bless you.

The Great America Show - April 26, 2023

DOES A.I. SCARE YOU？ WILL A.I. MAKE US SMARTER, BETTER, MORE PRODUCTIVE OR WILL A.I. BE THE END OF CIVILIZATION, THE END OF THE HUMAN ERA？

Episode Stats

Summary

Transcript