Podcast: Back to the Future: Historical Lessons of U.S. AI Policy, With Arthur Herman

(Ed. Note: The “Innovation Fact of the Week” appears as a regular feature in each edition of ITIF’s weekly email newsletter. Sign up today.)

The United States has been a leader in artificial intelligence (AI) since the 1950s. But AI and other advanced industry leadership in the United States has been threatened by increased competition with China. Rob and Jackie sat down with Arthur Herman, a senior fellow and director of the Quantum Alliance Initiative at The Hudson Institute, to discuss how AI leadership in the United States has eroded and what policymakers can do to save it for the future.

 

Mentioned

Related

Auto-Transcript

Rob Atkinson: Welcome to Innovation Files. I’m Rob Atkinson, founder and president of The Information Technology and Innovation Foundation. We’re a DC based think tank that works on technology policy.

Jackie Whisman: And I’m Jackie Whisman. I head development at ITIF, which I’m proud to say is the world’s top ranked think tank for science and technology policy.

Rob Atkinson: This podcast is about the kinds of issues we cover at ITIF, from the broad economics of innovation to specific policy and regulatory questions about new technologies. Today we’re going to talk about artificial intelligence and the importance of the US defending its lead in this critical technology.

Jackie Whisman: Our guest is Arthur Herman, a senior fellow and director of the Quantum Alliance Initiative at The Hudson Institute. His research programs analyze defense, energy, and technology issues. Dr. Herman is the author of nine books, including a New York Times bestseller, a Pulitzer Prize finalists, and one of The Economist’s best books of 2012.

Rob Atkinson: And I would add one of those books, Freedom’s Forge, really enjoyed that book. I would encourage all listeners to get it. It’s really a lesson about how American industry essentially won the war. What was FDR’s fame ... The Arsenal of Democracy.

Arthur Herman: That’s right. And as I explain in Freedom’s Forge, he swiped that phrase from the person who’s the real hero of that book, and that was General Motors CEO Bill Knudsen, who had come over to Washington from Detroit at Roosevelt’s request to say, to answer the question. How in the world are we going to get the world’s most powerful industrial nation geared up to face the threat of modern mechanized war, first against Nazi Germany, but then of course against imperial Japan?

Knudsen, by the way, is not only a main figure in that book, Freedom’s Forge, but I should add he appears as a character also in the new book, the one ... Jackie mentioned nine books. This is the 10th one that came out in August called The Viking Heart: How Scandinavians Conquered the World, because Knudsen was an immigrant from Denmark, who had come and had worked at the Ford Motor Company, helped devise the system of flexible mass production that the automobile industry depended upon. But he’s also I think belongs in the list of great Scandinavian Americans, as well, alongside such stalwarts as Charles Lindbergh and Andrew Volstead, the author of the Volstead Act and prohibition. So it may be an ambiguous legacy in many ways with the Scandinavians, but in Knudsen’s case, I think we can say it’s an entirely positive one, both in the private sector with General Motors and Ford, but then also, as Rob was just saying in terms of his mobilizing interesting for the great ramp up for World War II.

Rob Atkinson: Another ... And I want to wonk out here Arthur, and then we’ll get to the topic. Another book that reminded me of your book, and actually, I’ve asked the author, Stefan Link, who’s a historian at Dartmouth, to be on the podcast. And he’s going to do that later in the semester. But a new book called Forging Global Fordism: Nazi Germany, Soviet Russia, and the Contest Over the Industrial Order, just a fantastic book. And it talks a lot in that book about Knudsen, but also about how frankly, both Soviets and Germans came over and were almost trained at these American auto plants in the 30s before the conflict because we just didn’t have that sense that they were going to be adversaries.

Arthur Herman: And you know something too, also one of the great untold stories is the large numbers of American engineers and American engineering companies who went to the Soviet Union in the ‘20s and ‘30s and basically built out the infrastructure, the whole Soviet economic miracle, which Stalin and his cronies were happy to attribute to socialism as the alternative to capitalism, not at all. It was capitalism all the way, 100%, that built those factories, built those hydroelectric dams, transformed Russia’s metallurgy and mining industry. And the story I always like to tell, Rob, is the famous tractor factory at Stalingrad with the famous fighting between Soviet and Nazi troops.

Rob Atkinson: Oh, yeah. Back and forth, back and forth, back and forth.

Arthur Herman: That entire factory was all built by Americans. It was 10,000 guest workers came from America.

Rob Atkinson: Wow.

Arthur Herman: Working for American engineering companies, and built that place to start with. And so it really is the great patriotic war. It’s another monument to American capitalism and what it can accomplish when it puts its mind to it.

Rob Atkinson: So my favorite story of all time related to that is, I don’t know if any of you listeners have read Khrushchev’s autobiography. Fantastic book.

Arthur Herman: Very interesting book.

Rob Atkinson: But there’s a great, very interesting line in there. I think this was in the late ‘40s when he was commissar of industry. And he noticed that his driver was always putting on new tires to their limo. And he goes up, “Comrade, I’ve noticed you’re putting on new tires. Well, what’s the deal?” And the comrade driver says, “Well, Comrade Khrushchev, I’m sorry to say this, but the Soviet Union doesn’t make very good tires. They wear out quite quickly.” So Khrushchev says, “Okay, let’s drive down 400, 300 kilometers to the tire factory.” They get down there to the tire factory built by Goodyear. And it was a great tire factory, really modern. The only problem was that the steel industry, they couldn’t get enough steel because of the central planning, so they had limits on the steel, so the scrimped on the steel sidewalls, the steel fibers in the thing.

Arthur Herman: That made the tires vulnerable to blowouts. Is that it?

Rob Atkinson: Yeah. So the tires would always ... And so Khrushchev’s lesson-

Arthur Herman: That’s a great story.

Rob Atkinson: His lesson, it was great, he was like, “Well, this was a clear lesson. We need better central planning.”

Arthur Herman: Oh, that’s lovely. Drew exactly the wrong conclusion.

Rob Atkinson: Exactly.

Arthur Herman: Right data. Love it.

Rob Atkinson: Exactly. Yeah. All right. So why don’t we move on to the actual topic?

Arthur Herman: Let’s talk artificial intelligence.

Jackie Whisman: Welcome to Rob Atkinson’s book club. We’re happy to have you here. Thank you for being here. But starting back on our topic, you’ve written a lot about AI from both a technological perspective and a historical one. And I think we’d like to focus on the latter first. The US has been the center of AI research going back decades, but you say we’re in real danger of losing this leadership to China. Why is that?

Arthur Herman: Well, it’s not just me who says that. It’s also no less a person interested in Eric Schmidt, who had drawn everyone’s attention to this in the last couple of years, to the fact that America’s leadership in artificial intelligence and machine learning was eroding as China stepped up its effort to catch up and surpass in this area because China, and this is one of the things we have to understand about why artificial intelligence, machine learning, which is really simply the driving process that makes artificial intelligence work as an intelligent process, that they understand that the future of a strategy in the world, the future not just of the economy, but also national security, is who has access and who can process data faster. Data will become the new strategic commodity. It already is become the new strategic community in the world.

And data is the grist of the artificial intelligence mills that will allow the Chinese military and the Chinese intelligence services to really get a complete and 3D picture about what America is doing, what America and its allies, what their strategic posture is, what their economic future is, what their vulnerabilities are in the national security sphere. So China totally understands how important maintaining a lead and having artificial intelligence [inaudible 00:08:43] is to its goals. And one of the problems that we’ve had is that we haven’t had that kind of focus. We have treated artificial intelligence and machine learning as just interesting and cool technologies, which have some applications that are important for national defense, for our intelligence services, for our national security, but there certainly hasn’t been, despite a great deal of talk, and I know Rob, like me, has been invited to many conferences and working groups and discussions about: What are we going to do on the artificial intelligence front?

But for all of the talk that’s gone on, there hasn’t been the kind of, in my opinion, the kind of focused attention and strategic attention that’s really needed for artificial intelligence. And instead, if I may, Jackie, add this, that what we’ve seen in this country instead of this kind of strategic focus, what we’ve seen instead is a muddled debate over the ethics of artificial intelligence that arises from all of these kinds of fears and panic that somehow artificial intelligence and machine learning is going to lead to a generation of Arnold Schwarzeneggers, a rise of machines, nightmare scenario in which artificial general intelligence takes over the world and machines are going to be making decisions that human beings should be making, and even deciding at some point in the future that human beings are obsolete and ought to be replaced by even more intelligent machines.

That’s the kind of scenario which anyone who knows anything about artificial intelligence, what it can do, what it can’t, will tell us is completely fiction, science fiction fantasy. And yet, we’ve allowed that debate to dominate discussion about artificial intelligence, in the media, in our discussion of the ethics of advanced technologies, not just artificial intelligence, but others like 5G, in ways that the only real winners in this debate are going to be Beijing and possibly Moscow because they understand that we’re being distracted by all these rise of machines scenarios and debates when the real importance of artificial intelligence keeps eluding both public opinion and also I think many in our government and Congress.

Rob Atkinson: See, I would argue there would be another winner, which is fear mongering TED Talk presenters. They’re going to do really, really well.

Arthur Herman: Yeah. They score big on this. Don’t they?

Rob Atkinson: They score big on this, and they got the ponytail going. They’re like, “AI’s going to do this thing.” When you said these big machines and all, it reminds me of one of my favorite groups, Steely Dan, they have this album, maybe a song or an album, IGY, which is a take off on I think it was 1958, International Geophysical Year, and Arthur, as you know, the very first AI came out of that Dartmouth conference from all these MIT people.

Arthur Herman: That’s true. That did, didn’t it? Yes.

Rob Atkinson: And this, ‘56 maybe, something like that.

Arthur Herman: ‘56 or ‘58, I don’t remember the exact date.

Rob Atkinson: Yeah. So there was this conversation at the time, and they actually, Marvin Minsky and others, they actually thought that real AI was going to happen fairly soon, really.

Arthur Herman: Yes, they did.

Rob Atkinson: And so this Steely Dan song has a great line, basically it’s about the year 2000. And one of the lines is, we just have great big machines that make great big decisions, or something like that. In other words, we’ll have these machines now that are going to make decisions for us, and it’ll be safe and wonderful. And that was nonsense then and it’s nonsense now.

Arthur Herman: Still is.

Rob Atkinson: It’s really striking how people say this. I always like to say there’s sort of two major groups of AI misleading people. One I would call the Euro followers, which is AI is just biased. It’s going to steal your privacy. It’s all these other terrible things, so we need preemptive precautionary regulation, which we can talk about. But the others, I might call them the singulartarians, after singular, singularity university, who think AI is super powerful. And it’s just a matter of time when you’re going to get to what AGI is, artificial general intelligence.

Arthur Herman: Right. And so both of those groups is sort of the Ray Kurzweil on the singularity side, and also the Luddites, and Luddites who have a deep-seated resentment against any kind of anything that is sort of machine based, and who perpetuate this fear that machines will take over and dehumanize, dehumanize the social and cultural landscape. Both of them in the end see the same outcome. Don’t they? And that is that machines take over. The difference being that the Kurzweilians, the singularity people think that’s a great thing, they can’t wait, and the Luddites, those who spread the panic about AI and AGI, artificial general intelligence, think it’s a very, very bad thing. The truth is that such an outcome is highly unlikely, given the very nature of machine learning itself, what it does, which it doesn’t replace human thinking. It is in fact subservient to human thinking in order to build the pattern recognition models on which machine learning and therefore artificial intelligence depend.

The one indispensable commodity that machine learning depends upon, I come back to this again, is access to data. The more data it has, the better, the more precise its modeling is, and the more accurate its predictions about human behavior as well as behavior of other types of phenomena becomes. And so in the end, the real worry we should have, I believe, is not so much about who is it who’s building artificial intelligence and machine learning as a sector of our economy, or as a part of the digital world. But rather, who is it who’s controlling access to all of that data? Control from the point of view both of protecting privacy, which I think is an important issue, and is a matter of legitimate concern. But the other one, again, we go back to the Chinese model of the way in which this use of data can be used to build artificial intelligence apparatus and applications, which are really, really evil. And this is what we’re seeing in China, that the very things that those, as you were talking about those who fear the loss of privacy and the role of AI in the total surveillance state. Well, guess what, it’s already taking place.

It’s taking place in China. China has created an exportable model, which the Russians have taken on, the Iranians have taken on, and they export to other countries, whose dictators are very much afraid of their populace and want to exercise as much total control over everything that they do. The whole system of what we call social credits that the Chinese have pioneered, and that is of being able to survey what your choices are on the internet, in your daily life, what you read, who you meet with, what views you may express on the telephone, or in social media, as being ways to basically deny you certain privileges as a citizen, and to render you as a second class person, or even cancel you out altogether. This is happening. This is happening. It’s underway in China. And it’s become an export industry, the social credit system. There are private companies that say, “I would love to know what my employees are really thinking, not just about the company but about life in general.” These are all things that artificial intelligence applications are able to generate. It’s not the technology. It’s who runs the technology.

Rob Atkinson: Exactly. This is what drives me crazy about this debate. We hear this all the time. The Chinese are doing their dystopian social credit system, therefore, we have to worry about AI in this country. My argument back to that is the Soviets are using weapons to kill innocent people in Ukraine. Should we ban all steel?

Arthur Herman: For example, yes.

Rob Atkinson: Because those tanks are made out of steel. The Chinese put people in jail, I’m sure they have steel bars. Aren’t they abusing steel? Look, the reality is we have such a system, not just laws, but deep, deep traditions and values in the US about freedom, civil liberties, rights, that the idea is somehow that this technology will be used for dystopian purposes is to me, is just ludicrous. If they’re going to use it for dystopian purposes, we got a lot bigger problems than the government using AI. That means the government is out of control in the US.

Arthur Herman: I think that’s true. It’s an appropriate analogy in many ways. But I think even more so, Rob, and I know you agree with me here too, it’s not something that is inevitable or inherent in the technology itself.

Rob Atkinson: Right.

Arthur Herman: That’s the point. It’s who’s using it, who’s designing the programs, what they’re looking for, and how its applications are being used for good or for evil, for defense of freedom and democracy, or as a means of undermining that. That’s where the real debate should take place. And when we see it in that clear light, then the issue of where should we focus our attention on artificial intelligence and machine learning as a technology takes on an entirely new complexion. Doesn’t it? Because what we really see now is that far from retreating from the development of artificial intelligence and machine learning as a technology for fear that it’ll run out of ... It’ll be a runaway, a runaway tech.

Instead, what we need to do is to double down on developing these technologies so that what the Chinese and what the Russians are using it for won’t be able to deter us in using that technology in ways that protect our national security, help to defend our allies, secure our economy and our privacy at the same time. AI can be very powerful too for protecting privacy as well as for intruding on it, and then also as a way in which I believe, and this is one of the things that I’ve been working on at Hudson with regard to all of these advanced technologies, is looking for ways in which we can both overtake and surpass China in this sector that our job in stage one is to stay ahead or catch up in areas like, for example, AI and areas like 5G wireless technology. But also, how do we overtake and move ahead? What’s our offset strategy with regard to advanced technology and dealing with an adversary such as China? That’s where we need to be thinking for stage two in these developments in my view.

Jackie Whisman: And what are some things that the US government should be doing to help us maintain and expand our leadership here?

Arthur Herman: That I think is the big question.

Jackie Whisman: The million-dollar question.

Arthur Herman: And one certainly that again, as I say, we have good deal of debate takes place in the US right now. But what’s really required I think is ways in which we link up artificial intelligence as a technology to its two most important applications, both for national defense and national security, but also I think too, from an economic standpoint. The first is artificial intelligence and its use in developing autonomous systems. One of the most important applications for artificial intelligence is to allow machines to make the decisions that human beings don’t have to waste their time on, whether it’s on the battlefield, or in economic affairs, or in cybersecurity arenas. Artificial intelligence allows for quick decision making about a range of options that the machines do see more clearly, and given sufficient amounts of data, see more accurately than human beings can make similar judgment.

What we’re talking about is not the machines taking over human judgment. What we’re talking about is an interface, what the experts call the human machine interface. And in developing that and making that more sophisticated and making that interface more seamless, we actually get the most out of our artificial intelligence technology. The second area, and I come back to this question about data, the second area is increasing and securing access to data in order to speed up and to make more efficient our artificial intelligence technologies, that this is the way in which you turn questions of big data and big data analytics and move it to the next level.

And the next level is sifting through those millions and billions of bits of data in order to arrive at a very clear picture of whether you’re talking economic forecasting, weather, natural disaster information, it’s usually important. But then also too, in terms of our national security and securing our intelligence networks as well. So what I would like to see from Washington’s point of view is not just more money to be spent to develop AI. AI is already developed. The commercial companies, it’s there. Unlike my other subject that I work on, quantum technology, which is still an emerging technology, AI technology is an emergent. It’s there. It’s really just a question of unleashing its potential in those two areas in terms of supporting autonomous systems, both in the military and civilian spheres, including smart cities. It’s not just the military which autonomy can be hugely important and valuable, but also in terms of how infrastructure functions at a civilian level as well, whether you’re talking about ... Or air traffic control systems, including ... And driverless cars. But at the same time in which artificial intelligence can be hugely important for that, so that’s number one.

But then also for looking for ways in which we can speed up and secure access to the data to make those artificial intelligence mills faster and more accurately than ever. And from that point of view, Jackie, I think it’s very important that we see our debate about artificial intelligence as part of a whole, that it includes on the one side, our debates about autonomous systems and the role of autonomy as a technology, both for national defense and in the civilian sphere, and with regard to data and how we get access to data. And that leads us to 5G and wireless technology.

Without 5G, we’re simply not going to be able to maintain the artesian well of data that will be necessary to drive civilian infrastructure as it would function in the 5G networks. And that’s all the more reason why we have to make sure that our 5G networks and those of our allies and friends are not integrated into China’s 5G networks as being built and engineered and maintained by their stalking horse company, which is Huawei, the world’s largest telecommunications equipment giant, which has really taken the lead and jumped out well in front of the United States in the race for 5G. But even 5G, we have to look ahead to what happens at the next level. Now we’re calling 6G. And I have heard experts regard to driverless cars for example, who say the amount of data that’s really going to be necessary to have truly driverless vehicles or autonomous vehicles here, even 5G ain’t going to cut it to give us the kind of instantaneous access to data that those artificial intelligence applications will need in order to maneuver your car driverlessly through normal traffic. We’re going to have to raise the level of wireless speed and use of bandwidth that 5G opens the door to the next generation.

These are the kinds of debates which you really should be directing towards and thinking about, not just the funding issue. And which government agency should get the funding? And which companies do we look to, to lead the way? But also: How does artificial intelligence fit into the whole landscape of advanced technologies for America, both for national security and for our economic future?

Rob Atkinson: Arthur, that was great. We have to wrap up. Just a couple of closing thoughts. Yeah, we used to have the world’s leading telecommunications equipment company in the world, Lucent. And I wrote a long piece during COVID shutdown.

Arthur Herman: It’s a great piece, by the way, Rob.

Rob Atkinson: Thank you.

Arthur Herman: It’s a wonderful dissection of how American high-tech leadership can be squandered and eroded through neglect and through failure to keep everybody’s eyes on the ball.

Rob Atkinson: Right, exactly. We can’t make that mistake again. And then secondly, we’ve done work on our 5G rollout. And actually, it’s interesting. We’re doing better than a lot of people think, partly because the Chinese count cell sites in a very different way. They might have five radios on a cell, say we count that as one, they count it as five. In fact, there was a new study I saw yesterday that said that China’s a little bit ahead of us now, but the estimate in four or five years, we’ll be ahead of them. So I think our carriers are doing a pretty good job with 5G. Hopefully, this nonsensical, crazy thing with the airplanes and 5G interference, it seems like that’s been resolved. But we should be able to move forward. So Arthur, that was so great. We’d love to have you back again. Really enjoyed the conversation.

Arthur Herman: I’d look forward to it very much. Thanks, guys.

Jackie Whisman: And that’s it for this week. If you liked it, please be sure to rate us and subscribe. Feel free to email show ideas or questions to [email protected]. You can find the show notes and sign up for our weekly email newsletter on our website, itif.org. And follow us on Twitter, Facebook and LinkedIn at @ITIFdc.

Rob Atkinson: We have more episodes and great guests lined up. New episodes drop every other Monday, so we hope you’ll continue to tune in. 

Twitter Image: 
Podcast: Back to the Future: Historical Lessons of U.S. AI Policy, With Arthur Herman