Tech giant Eric Schmidt warns China is catching up to U.S. in AI
In this episode of "Intelligence Matters," National Security Commission on Artificial Intelligence Chair and Former CEO of Google Eric Schmidt joins Michael Morell to discuss the importance of investing in artificial intelligence as a national security priority. Schmidt believes China is likely to catch up to the U.S. in a few years in its artificial intelligence capabilities. He outlines how intelligence and national defense can benefit from superiority in these technologies and the benefits of holding A.I. to American values.
HIGHLIGHTS
China is catching up to the U.S. in A.I. capabilities: "Where we are today with A.I. is that we judge America still ahead, but China investing very heavily and likely to catch up very soon. We don't say what soon is, but my personal opinion, it is a few years, not five years."
Future of intelligence and defense is A.I.: "Computers are extraordinarily good at finding that needle in a haystack using artificial intelligence principles. I cannot imagine the CIA or the National Security Agency or any of the DNI functions not having very powerful A.I. systems to assist our human intelligence officers from watching things. That's a first use. The next use is probably autonomy. The future of our national defense is not more aircraft carriers and more tanks, it's more autonomous ships and more autonomous tank equivalents, which we can send into battle to defend ourselves without loss of human life on our side."
A.I. should be done with American Values: "One of the things that we concluded is that any use of A.I. should be done with American values. An obvious example is we don't believe in discrimination. So systems that would automatically discriminate are not good. Things which have inherent biases that we dislike in our country are not ok with us. The military, the DOD in particular, has also adopted an A.I. ethics plan, so we say that all of these A.I. activities need to be done consistent with a set of ethics principles that are published and understood."
Download, rate and subscribe here: iTunes, Spotify and Stitcher.
INTELLIGENCE MATTERS - Eric Schmidt
PRODUCER: Paulina Smolinski
MICHAEL MORELL: You're joining us today to talk about the recent report of the National Security Commission on Artificial Intelligence, on which you served as chairman. I think this may be one of the most important episodes that we've done in our three years of doing Intelligence Matters. Let me ask you to start by telling us what artificial intelligence is? I'd love for your answer to be A.I. 101 because I think there's a lot of people who talk about A.I., but I'm not sure they really understand what it is. What is A.I.?
ERIC SCHMIDT: The term was invented in the 1950's, and it took more than 50 years to develop something that was akin to the original idea, which was human level, artificial intelligence or what seems to be a human level ability in computers. There's no question today, we have artificial intelligence in our devices. Every time you use something that translates from one language to another. That's a good example of A.I. today.
MICHAEL MORELL: Can you explain how A.I. relates to, or perhaps not, all the other key technologies that we hear about from semiconductors to 5G to advanced manufacturing to quantum to bio?
ERIC SCHMIDT: Each of the ones that you named are the basis of what we believe are multitrillion dollar industries of the future. As a matter of national pride, national security, national economics, all of us believe that the United States needs to be the leader in each of the categories that you just named. In some such as 5G we are not already. In others, such as A.I. we believe we're ahead, but we have a very strong challenger in China.
MICHAEL MORELL: Why is A.I so important to national security? Why is it so important to national security that we created a national commission to look at it?
ERIC SCHMIDT: There are things that computers can do that are better than humans because, frankly, computers don't get bored. They don't have to sleep. They have infinitely better memories. The initial use of artificial intelligence in national security is probably Invision Systems. If you think about national security in your previous roles, you spent an awful lot of time just watching things, watching for this, watching for that, watching for someone, watching for something to happen, watching for something to be said. Computers are extraordinarily good at finding that needle in a haystack using artificial intelligence principles. I cannot imagine the CIA or the National Security Agency or any of the DNI functions not having very powerful A.I. systems to assist our human intelligence officers from watching things. That's a first use. The next use is probably autonomy. The future of our national defense is not more aircraft carriers and more tanks, it's more autonomous ships and more autonomous tank equivalents, which we can send into battle to defend ourselves without loss of human life on our side.
MICHAEL MORELL: How do you think about the ethics involved here? You hear an awful lot about, particularly in a military context, the ethics. How did you think about that as a commission?
ERIC SCHMIDT: We spent a lot of time talking about where the boundaries for A.I. ethics were. It's useful to know that every major corporation in America now has an ethics panel, an ethics set of rules about how this stuff should be used. One of the things that we concluded is that any use of A.I. should be done with American values. An obvious example is we don't believe in discrimination. So systems that would automatically discriminate are not good. Things which have inherent biases that we dislike in our country are not ok with us. The military, the DOD in particular, has also adopted an A.I. ethics plan, so we say that all of these A.I. activities need to be done consistent with a set of ethics principles that are published and understood. The Law of War is published, and one of the goals in the United States is for us to maintain adherence to that. We're required to do so, and the military is very committed to doing that.
MICHAEL MORELL: You talked a little bit about where we are relative to China. Where is the U.S. relative to China in terms of A.I. capabilities? Which way are the trends headed?
ERIC SCHMIDT: A couple of years ago, the Chinese government announced a set of initiatives which are generally known as made in China 2025 and the A.I. Strategy for 2030. Those agreements or announcements that they made said that it is their objective to be the world's leaders and indeed dominate in some translations, the following fields: A.I., semiconductor, high energy, quantum, synthetic biology, and a few others. All of those technologies are the basis of American exceptionalism globally. The development of the platforms that those imply will drive an enormous amount of jobs, wealth, huge new very large companies in the United States. Where we are today with A.I. is that we judge America still ahead, but China investing very heavily and likely to catch up very soon. We don't say what soon is, but my personal opinion, it is a few years, not five years. And in my industry, a few years is nothing. Our report specifically says that we must be ready as a nation by 2025 in all of the implications of A.I., training, systems, protection, weapons, everything that you care about, in order to meet this challenge.
MICHAEL MORELL: Do we know enough about where the Chinese are or is there some level of estimation here in terms of where they are? What's our degree of visibility into what they are up to?
ERIC SCHMIDT: Because we were a government commission, we had access to the classified data that the intelligence community uses in this area. Without going into the details, I can tell you that the claims we make are supported by a great deal of evidence that is coming out of the secret world, as well as open source analysis that we did. Plus, frankly, we talk to the Chinese, the people that we know, and asked them what's going on. And they told us.
MICHAEL MORELL: What's behind the trend of them catching up and potentially passing us if we don't do the right things? Why is this happening?
ERIC SCHMIDT: The Chinese are a top-down autocracy, best way to describe them, and they have these multi-year quests and they put millions of people behind them. China has many more engineers than the United States does. They've invested heavily in basic research, the so-called STEM fields, and in that sense, they've decided to become real competitors to the United States. Many countries have made these claims, and we've been able to stay ahead of them because of the wonderfulness of America's governance, the way our universities work, the intelligence of our people, the fact that we allow high skilled immigration into our country. But China is a different kind of competitor, and we collectively believe that China will be at parity quite soon unless we act fairly quickly. I'll also notice that in the last few days, President Xi announced that they were an equal to the United States. He didn't mince words. He said you have to deal with us now as an equal. So it's clearly coming from the very, very top of the country.
MICHAEL MORELL: So it's what they're doing, but it must also be what we're not doing. It must be a combination of the two. In general, what aren't we doing that we should be doing?
ERIC SCHMIDT: We have lots of recommendations. In fact, we have seven hundred and fifty pages of recommendations. We took the liberty of writing the laws for the government, so they could just say yes. That's how serious we are about this threat. There's a set of very important ones. One is to increase the funding in research and development in the United States, roughly doubling it every year or so for the next three or four or five years. We need to get our funding up so that it's competitive with the money that China is putting in. We also have said that talent is everything. That while our government servants are hard workers, they don't necessarily have the right background. This A.I. stuff is hard even for me and I have a Ph.D. in it. It's important that we take special steps to get special skills into our government so people can understand what's going on. We've proposed, for example, an academy that would take young people who want to serve the government and give them a four year technical education in return for five years of service in the federal government, in the DOD or in the IC or similar groups. We have a number of initiatives like that. America is strongest when all the parts of it work together against a common cause and the special talent is in the universities and the private sector. We need to find a way to get that special talent into the government and have the government not lose them because they mismanage them.
MICHAEL MORELL: So big categories recommendations, R&D and talent. What else?
ERIC SCHMIDT: We talked earlier about the importance of ethics, and we talk about that we have to do things consistent with America's values. We would never want the A.I. stuff that we fund look foreign to us since it's us. We talk a lot about partners. It's traditional in the United States to think we're everything because, of course, after World War II, we really were everything. But today, the best safety and the best security is to work with our democratic partners. There's a fairly well understood list of countries which, by the way, include India, Japan, South Korea, Britain, Germany, France, Israel, who have the technical capability to partner with us to really move this technology forward. We talk about the need for building a research network for universities. One of the problems with this technology is it tends to be controlled by a small number of very large firms, including Google, which I used to be part of. We think it's important that we create a large number of new research efforts and new startups to compete with them, but also to explore the many, many things that A.I. can offer. I haven't even talked about the benefits of A.I. There are many, starting with medical devices and medical discovery.
MICHAEL MORELL: You hear a lot of discussion, including from President Biden, about a new relationship between the government and the private sector in a whole host of areas. How do you think about that broadly as it relates to A.I.? How do you see that relationship?
ERIC SCHMIDT: I start from the premise that I want to win for America, and I want America to do whatever it takes to build the next generation of platforms in technology that cause trillions of dollars of new companies to get created. The American system is not just companies, it's also the government and the universities. When I was a graduate student, I was funded by the U.S. government to get my degree. I wouldn't be here if I had not received that funding. That's how important the synergy between them are. Corporations have both a moral responsibility to help the United States, but also business imperative to do so. They're going to be an awful lot of great business opportunities for businesses to work with the U.S. government as this A.I. revolution unfolds. For people who criticize that, one of the benefits of a free country is that you can choose not to do it. No one's forcing you to work on it. But I can report to you that there are thousands and thousands of patriotic scientists, patriotic people who want to help our government, and they want to do so in a way that advances their career and their technology as well.
MICHAEL MORELL: What kind of reception have you gotten to the report and to its recommendations, both from the administration and the Hill and then even from the private sector?
ERIC SCHMIDT: We've had a lot of bipartisan support from Congress and also support from the White House and the Defense Department in our meetings so far after releasing the report. After all, we were commissioned by the Congress, so they asked us to do this. One of the questions for me is what is the mechanism by which these things happen? Normally there is a process which is called the NDAA, which you're well familiar with, which is a roughly yearly process of legislative changes around national security. We have proposed changes to the NDAA, which we think have a high likelihood of going through.
There's also a large proposal around infrastructure spending, including a large proposal around semiconductors, something we highlight in the report as well. We hope that those two initiatives will make a material difference in the reception of this. Put another way, and I say this to you as a government veteran, if there is no money supporting these proposals, they're unlikely to have the kind of impact. But with material new money, we think the government can get its act together and do this on the timeline that we propose. It's going to take multibillion dollar funding, which we think is possible in the current climate.
MICHAEL MORELL: So the authorization to do the kind of things that need to be done would be through the National Defense Authorization Act, the NDAA. Do you hope to get some of the funding you need in the infrastructure bill?
ERIC SCHMIDT: We do. The infrastructure bill, as you know, is not fully baked yet, so we're working hard to find the places to hook the necessary language to get the money. I think we'll see a good chunk of what we've asked, not because our arguments are so good, but because in the political structure, everyone agrees on one thing, which is Americans leadership in these technologies must be maintained.
MICHAEL MORELL: It's interesting to me that that wasn't the view five years ago. We weren't even talking about this. People weren't focused on it in government, people weren't focused on it on the Hill. It's really changed dramatically in the last few years. What driven that? Has it been China's gains?
ERIC SCHMIDT: If you go back to 10 years ago, there was a discussion about the pivot to Asia. So the people who've paid attention to this have known this was coming. In the last five years, the biggest change has been a much, much more aggressive posture from China on important things that we care about. The most obvious one being Taiwan. I think that plus the various trade deals and trade negotiations, which were somewhat inconclusive, highlighted the general understanding of the Chinese problem. I think for the technical community, one of the things that happened was that a group at Google called Deep Mind went to China and won using A.I., a game called Go. It's not played in the United States, but it's the central game of China. It's about two thousand five hundred years old. Many of us believe that that win where the computer beat the smartest, Go player in the world, was for them a Sputnik moment in understanding the power that A.I. could provide. It's very clear that right after that they organized these responses, the funding and so forth, in A.I.
There's another interesting thing that has happened, which we don't talk that much about, which is the rise of Huawei, the rise of TikTok, the concern over chips, which a lot of people have begun to worry about. When I would go talk to a senator or congressman or woman. They would ostensibly be there to talk about A.I., but then they would start asking me about 5G and TikTok. 5G is a good example where China, using a set of techniques, ended up with a company which had a stronger product offering that America did not have a good answer to where America cared about its national security involving telecommunications. That created a conundrum for the government because they didn't want the Western allies to use Huawei, but they didn't have a constructive alternative. I think the sum of all of those caused everyone to understand that we have to change our game. There are categories of technology of which semiconductors, A.I., synthetic biology are examples where we must win. We don't want to be in the position that we are in now with Huawei and 5G where we don't have a competitive offering. We're working on it, but there may be a factor of 10 ahead of us. That's not OK.
MICHAEL MORELL: My sense from the people that I talk to around the world is that if we have the capability here, they would want to do business with us. But if we don't have the capability here, they have no choice but to go to China. Does that resonate with you?
ERIC SCHMIDT: That's true for the countries that are Western. It's not true for what are called the BRI countries. The Belt and Road Initiative countries get financing and essentially just adopt whatever China tells them. So if you follow that line of reasoning, one future structure of the world will be China and her allies and then the West and its allies together fighting over these technology platforms. Now you say, what's a technology platform? Why does it matter? Well, imagine if Google were Chinese? So the Chinese rules about information and censorship were applied in the U.S. Well, that wouldn't be acceptable. Imagine if Apple was a Chinese company, and you were always wondering if the Chinese were listening to you on your phone. That's a problem. When you start going through this, because of the nature of communications and semiconductors in our lives. It's important that the supply chain of software and hardware be consistent with the national security of the United States.
MICHAEL MORELL: Are semiconductors the most important parallel technology here to A.I.?
ERIC SCHMIDT: We believe so. In our report we say that it's very important that the United States take a position and puts money behind it. It's lots of money, by the way, that the United States stay two semiconductor generations ahead of the Chinese. Now, today, we use chips that come out of Korea, the United States, and especially Taiwan and a company called TSMC, which is just a megalith company. This is an enormous corporation that builds really state of the art chips. TSMC, as an example, just announced this past week that they are going to spend a 100 billion, with a B dollars, to build a new set of plants to do their new and even faster technologies. The current proposals in the United States are only 35 to 40 billion. It gives you a sense of the amount of money involved in this semiconductor issue. I've been looking at the semiconductor thing for 20 years. This was a big issue under the Obama administration, and no one really knows how to catch up with Taiwan. We need to solve that problem. China has had an activity for the last 30 years where they've been trying to catch up with Taiwan. And again, they have not been able to either, even though they're 100 miles across the strait and there are many TSMC factories that are in mainland China. That shows you how difficult this technology is.
MICHAEL MORELL: They've been poaching TSMC engineers and stealing intellectual property and they're still struggling. I want to ask you one more time, but in a different way, why does this matters? If you were talking in front of a civic group in Kansas City or St. Louis and somebody says to you, why does this matter to me?
ERIC SCHMIDT:
If I would be speaking to an audience in the middle of the country, I would start by saying, does it matter that America leads the technology world? The answer is absolutely. The five most valuable companies in the United States are technology companies, with more coming. Even if you don't care about anything else, you care about your retirement plan, that wealth, that stock market wealth is directly tied to the innovation that we do. If you care about health, A.I. and the areas that I'm discussing here, will revolutionize the drugs and the treatments that you get as you age and deal with the infirmities of age. If you care about national security, do you really want opponents, in particular China, looking at what you're doing, understanding how your processes are, and actually stealing or otherwise interfering with the things that you're doing? Of course, you don't.
This is the next challenge. It was the nuclear challenge. Then it was the naval challenge. Then it was various conflicts. Now it's all about leadership and making sure that the A.I. tools are used in the appropriate ways for our national security, but also our business security. We tend to focus on national security as intelligence or weapons. But our national security is just as dependent upon our economic security. We desperately need strong growing global businesses that are exporting, that are getting our values. One of the greatest things that we've done is allow American firms to build global platforms that have exported American culture, our movies, the way we think, our values, the way we treat people, our diversity. It's good for the world. Imagine if we were prevented to doing so. It would be a real problem.
MICHAEL MORELL: I'm going to jump back to semiconductors again. Do you think it's possible to catch up with TSMC?
ERIC SCHMIDT: We don't know. Historically in technology markets, the leader in a category remains the leader as long as they are executing because they get what are called positive network effects. They get the top people, they get more money, they get more bases and so forth. It's almost certain that we can build a system which is as good as what TSMC did last year. The question is, can we do what they're doing for next year? That's a much harder challenge in technology. It's true in every platform, not just in semiconductors as well.
MICHAEL MORELL: If you were advising the Biden administration on how to implement the Chips act, would you advise them to make the subsidies and incentives available only to U.S. controlled firms or to foreign firms as well?
ERIC SCHMIDT: The language that we use is that we should make these systems available to American firms and authorized foreign firms. The specific language is the DOD language about partners that are in other democratic countries. And there are a couple. But I'm assuming that by the time we're done, the vast majority of the Chips Act will go to American firms. I'm actually much more concerned about having that money go to the waste. It's very easy for a political leader without understanding the subtleties to say I'm going to spend this amount of money to get what they're doing now. I would much rather have a group of scientists and thinkers be appointed by the White House to say, 'OK, you guys are so smart. You tell us what it's going to take to win in semiconductors. Tell us precisely what are the technologies? How do you control for the semiconductor risks of some kind?' There are huge new programs. The United States has funded a micro-electronics research program and platforms for the last decade which may give us some new ideas. If we just copy what other people are doing, we won't get what we want. This has to be driven by the top scientists in our country. When I was working with the DOD, I had the privilege of visiting some of the DOD labs that do the nuclear work. You have physicists there who are at the top of their, game that are serving our nation. Because the stuff is so hard, we need a similar set of devotion and candidates and people who are at the top of their micro-electronics game working with and for the White House and the government to make sure we don't waste this money.
MICHAEL MORELL: This competition with China across all of these key technologies, are you optimistic? Are you pessimistic with regard to the outcome?
ERIC SCHMIDT: I am a realist at this point. I think there are opportunities before us. I'm really pleased at the reaction that we're getting. It's not just money, remember, it's specific people and specific decisions. We have shown a roadmap, but it's possible that roadmap could be poorly implemented. One of the things that we need the government to do is to monitor this and make sure you're actually getting what you want. Our government sometimes as a general statement seems to be happy just to spend money without judging the outcome. The outcome here is we have to win. We have competitors and we have to benchmark against the competitors and do what it takes to make sure we stay ahead of them.
MICHAEL MORELL: The 9/11 Commission kept on reconvening to look at how the government was implementing its recommendations. Has there been any discussion about you guys reconvening and being public about how the government is responding to your recommendations?
ERIC SCHMIDT: Not yet, we would certainly be willing to do that or do it in an appropriate form at the time. It's just easy to write a report and throw it over the transom. It's much harder to get the right people aligned. The commission is spending a lot of time now with our counterparts, getting them educated. We're sending some of the staff that wrote the report into the government so they fully understand what the report is. So they're embedded literally inside of each of the secretary's offices. We're doing everything that we can think of. But I think at the end of the day, I would have some form of external review on A.I, and I would do the same thing on semiconductors. I think it's too important not to do it.
MICHAEL MORELL: Is there a window in which we need to act or it'll be too late? Or will it never be too late?
ERIC SCHMIDT: We're in that window now. We have to act now. In our report, we say we have to be ready by 2025. We think that if we're not in that configuration, the leadership that China and other countries- we haven't spoken much about Russia and her activities. But each of them is on a path in these areas. If they get well ahead of us, it will be very hard for us to catch up.
MICHAEL MORELL: So Russia is also threatening our leadership.
ERIC SCHMIDT: It is. It's much smaller and much less broadly focused, but they have very significant cyber activities underway which have been discussed in the media. We spend a lot of time in our report talking about making sure that our cyber defenses are strong and that we understand what's going on in the networks around us. And that's the Russia connection.
MICHAEL MORELL: How does A.I apply to cyber?
ERIC SCHMIDT: One of the things A.I. is very good at is watching for unusual things. If you sit in your house, you notice a change in your house that you didn't see because you remember what it looks like. A.I. I can do that a million times a millionth of a second. A.I. is going to be needed to watch the networks, to see an unusual activity. Then typically what happens is the A.I. says 'hey, guys, there's an unusual activity. Why don't you take a look at it?' There are more powerful A.I. systems which can also begin to diagnose what the problems are, and that's the state of the art right now.
MICHAEL MORELL: A not insignificant share of my listeners are undergraduate and graduate students and young professionals, how should they think about their role here?
ERIC SCHMIDT: This is a great time to be entering these fields because all of that stuff that everyone learned about in national security and global geopolitics is about to get upended again by these technology dominance platforms. Who will be the leaders in the next five to ten years? It'll be the countries and the partnerships, that is the alliances that can assemble their technical resources to fight this game, to fight for global leadership. Indeed, the primary competition with China in the next decade will be in this area. We love to talk about steel and food and all of those things, but those are commodities. The ability to control a platform is the sexy thing in geopolitics. If you control a platform, you can control behavior, you can control economics, you can control thought in the most extreme examples in a nation state characteristic. I feel very strongly that we need to take our young people and get them excited about this. In my world A.I has taken over computer science, and computer science is now the number one major in all the leading universities in the United States. Within computer science, every undergraduate does a machine learning course. Not only is this important, but our universities and our students have responded correctly with enormous supply of brilliant people to work on these problems.
The issue here is not that we don't have the talent, we have the talent coming. The problem is we're not organized to win. We say very, very clearly in our report that the government is not organized nor resourced to win the technology competition against a committed competitor that's focused on A.I. We have -if you're interested in geopolitics, which both of us are, we actually spend quite a bit of time suggesting that we build partnerships such as an international digital democracy initiative to work precisely on these things across the democracies of the world. This is our future.