SCSP is thrilled to launch Memos to the President, a new series dedicated to providing bold, actionable ideas to the incoming administration. Our first episode features none other than our chair, Dr. Eric Schmidt, offering his invaluable insights on navigating this complex landscape.
In this first episode of Memos to the President, Eric Schmidt, in conversation with Ylli Bajraktari at SCSP, discusses his new book "Genesis" co-authored with Henry Kissinger and Craig Mundie. He emphasizes the rapid advancement of AI, particularly highlighting the development of agents capable of complex tasks and the potential for AI to revolutionize programming and mathematics. Schmidt stresses the urgency for the US to lead in the AI race against China, emphasizing the need for a national initiative to accelerate AI development and ensure the US maintains its competitive edge. He also discusses the implications of AI for warfare, drawing on lessons from current conflicts, and makes the case for increased autonomy and affordability in defense technology.
Transition Memos to the President | Episode 1: Eric Schmidt
00:00:08:23 - 00:00:28:00
Ylli Bajraktari
Hey, everybody. Thanks for coming to SCSP. A lot of friendly faces. A lot of, friends of SCSP. It's always good to have Eric here in town. Eric, thank you for coming. Good to be here, as always. This is the recording of the first episode of a new podcast. But there was no better time than actually talking about your new book, Genesis. And I told you in the car but I'm going to share with the audience here. But the way I prepare for this interview, I asked Gemini advanced to go into YouTube and pull all their recent interviews about the book and pull the questions he received and buckle them in different threads. And it's amazing. It's amazing. Like, I got seven categories of questions that I have to ask or Eric was asked. So I'm trying not to repeat some of the questions. Obviously next, next week marks six years since I started working for you. And I've been in a lot of meetings, both classified and unclassified with you. But there are two stories that always intrigued me when you tell them that I would like you to share with the audience. Number one is, Kissinger's reaction when you brought him to Google, for the first time. And the second one is the AlphaGo in, South Korea.
00:01:21:00 - 00:01:46:20
Eric Schmidt
So, so Ylli, first, congratulations for building this extraordinary group. And, you know, and you you're always kind to give other people credit, but you built it and you worked with Doctor Kissinger before his death. So you inspired by what he did in the Rockefeller Commission way back when? So the, so the first question was.
00:01:46:22 - 00:01:49:07
Ylli Bajraktari
How you convinced Kissinger to visit Google?
00:01:49:09 - 00:02:08:14
Eric Schmidt
Well, I had met him at, a Bilderberg Conference, which is a strange conference of old people. And, he looked at me and he said, I need how does Google work and why are the results so bad? Let's just, you know, in his German accent. And I explained it to him. I didn't like the explanation, so I invited him to Google. So he shows up and this is in 2008, and he goes, he stands up in front of the entire company. He announces that Google is a threat to civilization. And everyone at Google loved it, because all of a sudden they became more important. I think today, you know, in that story, you see that he foresaw what we're seeing now with the impact of social media. And AI, all of the usage around the tech companies in that one narrative.
00:02:40:14 - 00:03:02:09
Ylli Bajraktari
The other question the other, I think answer that always. Is interesting is obviously you won with Demis Hassabis in South Korea. AlphaGo remarkable moment, I think, in human history. And so, for those of you that have not watched, it's a great documentary, but also, I know you were a force behind executing that.
00:03:02:11 - 00:03:26:19
Eric Schmidt
So in Korea, so what had happened was we had bought DeepMind and, this is a little startup, of very smart people who talked a really good game. And this was before we really understood the power of AI, before language models and so I figured that I should go and watch the game. So, what happens is it's in the Four Seasons in Seoul, perfectly nice.
They have all the cables and so forth, and you have all the humans, right? And all the Koreans. And they have the Koreans are all organized in their rooms, and they're all very big go players, and they're extremely confident. And then you have the engineers in their little room with the control panel. And I'm going like, it's the five of them against the 500 Koreans. And the Koreans are, this is like it. You know, this is like combat. And so there's roughly three games. And the first game, it was clear that the human made a mistake. Right? And everyone said, well, he'll correct it. But in the second game, as I remember, there was a move where we should call the move 37 in the game where people, the Koreans went crazy because this move made no sense at all.
And they were convinced that Google would - DeepMind would lose, the human would win. And, you know, there's a perfectly nice human who's all of 23 years old was like, brilliant at playing go. And so what happens is they take the move. So I'm with the Koreans because, you know, I'm going I'm subtle doing subtle diplomacy. So I go talk to the humans.
On the Korean side and they say, oh, you know, disaster. So I go into the lab where they're actually doing and they said, here's how we run the game. We're really optimizing our we're basically making sure that our probability of winning is always greater than 50%. So as long as our probability is always greater than 50%, we always win. And that move is the probability, because the computer could see a move that had never been invented in 2500 years. Ultimately, we won and the Koreans were destroyed and all that, and that was a big deal. Subsequently, we did the same thing in China, near Shanghai and same thing, argue a better player, in China than in Korea.
We have the first game. It's going okay. Then in the second game, it's clear that our algorithm is working. And guess what? The Chinese turn off the television. There were 300 million people watching of some kind so that the Chinese could not see their own feat, which we ultimately did. After that, Google retired to play. There was no reason to to keep going. And that team ultimately became the team that invented AlphaFold, which is the 200 million proteins for the human bodies, which got the leaders the Nobel Prize. So it's a really nice story. You start in games like who cares about go, well, awful lot of people in Asia care about go. But more importantly, all of that, the tree structures that were invented and so forth were then used to do something really important for humanity.
00:06:15:05 - 00:06:41:22
Ylli Bajraktari
You know, for the last six years, every time you talk about AI, you've always said to us, at least, we're at the beginning of the beginning. And I think the genesis really captures that sentiment. And, you know, we have a video here from SCSP that ends with your code that says, you know, development of alternative intelligence will be, I think quoting precisely you, something new in human history. And so just to deep dive a little bit of Genesis, you know, we wrote this is Doctor Kissinger's last book. You know, I've been on some of the sessions you were working on. Where did the idea come from? Why Genesis.
00:06:55:17 - 00:07:21:20
Eric Schmidt
Well, we had written our first book together right before ChatGPT, and we got most of it right. And we featured GPT three, but we did not because it hadn't happened yet. Understand that that what a chat GPT moment was. Remember, That was only two years ago. Just think about it. And it's only two years. And so, the book was out and so we thought we should update it. And this produced this argument. And Genesis is really about what happens when you have polymaths in people's pockets. It's really what happens when the, the, this alien intelligence shows up next to each and every one of you. And this is going to happen much quicker than even even we talk about it in the book. And I can explain why.
And you're beginning to see the first vestiges. So, for example, you, you know, you did your research to ask some questions, so forth, that for me, the most interesting is just to do a plug for Google. If you haven't played with notebook LM, from Google, you should, it's analogous to what you describe. You can take any form of multimedia and put it together. And it does. You can essentially ask queries. And the key thing is it's for multimedia. So it it watches the videos. It, you know, listens to the podcasts, does the text. And you can do this. So if you have like a complicated thing which almost everybody in this room is working on, some incumbent put it all in notebook LM and see how it can help you. The amusing part about it is it has a thing where it can generate a podcast, and if you're confused as to as to why why this is so funny type notebook LM not human. There's a podcast where the hosts are told that they're not human and see their reaction. And the thing that is remarkable is not that the voices are fantastic, but rather the entire discussion is generated. Fake, right? The query is the hosts have told you you're not human. And in UX, things like the man has a wife and the wife is unhappy with him. You know, all the sort of narratives. Now, how does it come up with that? Well, remember, it's read every fiction book. It's also read every nonfiction book, right? So it understands human narrative, human jokes, so forth and so on.
These are just the beginning.
00:09:20:21 - 00:09:38:18
Ylli Bajraktari
You know, you're always good about predicting the next wave. So from your standpoint, what are the next wave of AI? You know, we talk about a genetic AI, you know, the path towards AGI. So what is your sense of where all these trends are going now?
00:09:38:18 - 00:09:53:19
Eric Schmidt
So, so one of the things that you're doing is you're creating you have an AI task force, and you are now going to do an AGI thing here, which obviously I'll be heavily involved with. And so here's roughly how it goes. The first thing I want to say very, very clearly is our society is not ready for this. The human systems, the governance, the rules. You're not ready for this arrival. And it's happening. It will happen much more quickly than you think. And I'll give you a timeline, which I'm now confident is correct. So last year's discussion is language to language. So you input something. It does. This amazing thing gives you an answer. Now why is language important.
Well language turns out to have an awful lot of knowledge in it. Words have a great deal of meaning, especially in context. So you can cover an awful lot of things in concepts in human thinking about using language. So language language is pretty impressive that only one technology is called next word prediction. Technically what it's doing is it's looking back across millions of events and it's looking back at the most recent set of words and saying, what's a reasonable next word? That's why it writes so well. It's sort of an amazing invention in and of itself. And the same is true for any multi-scale prediction problem physics, biology, math and, and programing. So right now we're in a situation where we have very large context windows, context windows where you take you essentially have a you think of it as short term memory. You basically say the following things are true. Now, how long can a context window be a million words so you can feed it books and say, given these books are correct, do something. Most of us don't do that. We type ten sentences, but that automation is important. The second one is a the agent revolution. Agents are systems where you have language and it goes to a language model.
The language model generates scenarios, the scenarios are evaluated, and then the agent is modified internally using a technique, involving memory. So it essentially can learn stuff. So here's an example where the agent wakes up. It's in this room. And we say turn on those lights. And turn off those lights. Now it has lots of dials. It doesn't know what they are. So it turns them on. Well first thing it does is it turns off the heating. Oh that was an error. The second thing is it turns on the air conditioning. That's an error. Then it locks the doors. That was an error. Then it turns on the wrong lights. That's an error. And then it gets the right light. It says good.
And it has learned that. By the way, that all happens in a microsecond. It's not human learning, it's computer learning. So then it can learn something else and learn something else and so forth. And eventually it's the digital manager. Everything in her life. So, so the point is it only has to learn something one and then assemble it. Agents are important because you can concatenate them. They currently are designed to be language to language. So, here's another example I happened to. I grew up in northern northern Virginia. I happen to like it. I want to buy a farm in the outskirts of McLean. And so I say to it, organize a farm by McLean, designing a three bedroom house, make sure everything's legal. Don't spend more than X dollars. That's the command. So the agent manager organizes the following. The first is it does a survey of what's available. The second is it consults the building code. It then decides the building code. And this plot is available. It, evaluates the quality of it. It then negotiates the price. It then finds an architect, looks at the architect's work, rejects their work, does its own design, proposes that bids it out, and monitors the construction.
Now you sit. There you go. Wow. Can you do that? It's easy. It's called workflow. Now, the example I used is characteristic of kind of everything that happens in a business. So businesses are essentially there's something that happens. You have to respond. You have to deploy something. You have to calculate it. You have to build somebody. You have to deliver a product. Right. All of that gets automated through, agents very quickly. And there's a whole bunch of startups that are doing that. And the third one is generally known as text to code, and it's essentially generating programing. So today, the current models, if you say write the code to do something, it can do a pretty good job, although they weren't trained on programing.
So there's now a whole bunch of startups and and likely very successful efforts to build automatic programmers. And I'll explain why it is so important in a minute. And there's also a bunch of people who are working on automatic math proving. Now, why would I focus on programing in math? Well, I've just said that these systems are next word prediction and human language is immense and plus. We're also kind of running out of things to train against. But software, the verbs are relatively simple to learn, right? They're not that many verbs in any software language. This is what I do. And also there's a clear objective function. It either works or it doesn't. So write the code test. It doesn't work, modify it doesn't work, modify
It doesn't work. Modified. It works. Good. Done. Move on. People believe that in the next year or two, you'll have super programmers. I'll come back to why this is important. The same turns out to be true of math. The math verbs are relatively simple and you have an objective function. Does the proof work, and is a technology which I and others have funded called lean, which is an open source, format for this kind of learning, which seems to be taking off. I was at a dinner last week on AI safety. He, one of the mathematicians, said, I think we'll have a super mathematician in one year. Okay, that's pretty aggressive and super mathematician, meaning a mathematician better than the best mathematicians and best mathematicians are in, you know, Incalculably hard to understand, right? So you can imagine what this thing is like.
Why do I emphasize math and programing? Everything in our digital world starts with math and programing. So the path to AGI, is you have these agents and then you have these math and programing things, and then you start building AI scientists. So here's how it actually works. Let's say this is the slope of current progress which is pretty good by the way. Right. Like OpenAI is now $190 billion. Look at the valuations. All of the stock market games in the US appear to me to be completely a function of the AI craze across every sector. Everyone knows about Nvidia. That's pretty good. So let's say you have 1000 AI programmers, which Google certainly has, OpenAI has, Microsoft has so forth. What happens when you release a million non-human AI people to work with the AI people? The slope changes. Okay, how steep does it get? In my view, the company that gets to that or the country that gets to that will have a monopoly on innovation for some amount of time. Now we can debate how long, and the same is true of countries and the alert for this group, I want to say is the Chinese have caught up.
I spent my last trip with Kissinger was to China, and my assessment at the time was that they were two years behind. They are now six months behind in one year because of incredible invention and investment on their part. They there's lots of evidence that they've evaded the sanctions. And, you know, all the things that you could imagine they would do, that's at least what the evidence says. But the important thing is they've caught up. I have one company I'm funding who's doing their training on Chinese models because they better than American models. I said, Holy crap, that's like a big deal. And so it sure looks to me like we're now in a race. And this is my my message to to you all. We're now in a race between China and the United States for dominance in AGI.
It's just beginning. It's crucial that we win. And here's why. In my world, if your slope is like this, you keep gaining. And then the second guy, second country company, whatever comes in, it's doing it too. But this one is staying ahead. Now what does this mean? Why would it matter? Well, what is a problem we have? Climate change. Okay. What's another problem? We have? We decide we want. Let's say we're an adversary. We want to attack the American financial system. This system kind of can design and detect zero day exploits that have never been done before, because it can invent them. What about material science? New batteries, new energy? Cool. It can do all of that.
What about progress in quantum something which we care a lot about? It can do all of that. If you go through the list of all of the stuff that we care about in a competitive market, we have to win this one. It's crucial for America that we win this one.
00:18:46:19 - 00:19:07:20
Ylli Bajraktari
Thanks. If I Gemini, what was the most question you have received in the last six years since you were with a commission? And now with SCSP, this whole competition between us and China? I'm glad you answered, because it feels like every six months, 12 months, the dynamic changes. You know, we're ahead. They're catching up. We're ahead. But what about our allies Eric? I know you spent a lot of time in London and in Paris and, in Asia. Where do they stand in this group?
00:19:14:03 - 00:19:38:16
Eric Schmidt
So here's what you have to. In order to win this battle, you have to have the following. An infinite amount of energy, these things that you cannot imagine, the amount of energy, these things take. You have to have a market which will give you $10 billion to build a data center without having any current customers. Right? So you need a financial market and an energy market. You need access to the top chips, and you need to have all of these research scientists in Europe, you know, close allies to America. They just don't have it. You can't just get that amount of power, that amount of money, that amount of people, you might be able to get in in India because the Indians are really, really smart.
But again, where they have their own challenges and so forth. China's out of the question. Russia's out of the question. Where else can you get that kind of scale? It's going to be us. And I'm sure that they so one of the core questions is, what do they do? The most likely scenario is that these large foundation models that are known and will be owned by proprietary companies in various ways, but they'll be licensed to partners in our industry There is a debate over open source for open weights. Technically, and an open and closed source. The technical question is, can you release the weights of one of these models and compete? Can that be misused? And so forth? The Chinese have largely taken the 400 billion model, which is open weight, out of llama and meta Facebook and their most of their gains have been in that area.
So I used to think which is like three months ago, that the big question would be proliferation of these open source models. Recently, Deep Seek, which is one of the Chinese companies, has produced its own models without using the llama stuff, which is innovative and good and very powerful in what is called planning. So it sure looks to me like the Chinese competition can occur with or without open source. The danger of open source is that because we don't know what it's being used for. If you put guardrails in the open source like these models know how to do - so here's an example, the model no models, no nuclear secrets there no military secrets, no weapon secrets. You know, health secrets.
All of those things are eliminated in the training done by the companies. And there's a voluntary agreement among the companies to to make to implement those training regimes, pretty much universally, which is why we're safe. So if you ask the, ask any of the models a sort of exciting question, it'll say, I can't answer that question. That includes things like copyright and things like that. It's relatively easy with open source to reverse those guardrails out. And that's the concern.
00:22:19:04 - 00:22:45:20
Ylli Bajraktari
It's been almost ten years since Ash Carter asked you to help Department of Defense first, the Defense Innovation Board. Then Congress has to with the AI Commission, you move from observing and criticizing to recommending action. Can you tell us how the spirit has changed in your mind for the last ten years? How is AI impacting, now warfare? And just your personal experience in this journey?
00:22:53:03 - 00:23:14:18
Eric Schmidt
Well, so, I had the privilege of working for the Secretary of Defense for many years. And spent a fair amount of time listening to the discussion about what is called the third offset, which the military people here in the room know, you know, the history here. It's essentially the arrival of precision weapons. We don't understand it. And the belief in the military is that there's a next wave. Now, it's quite obvious to me what the next wave is. It's autonomy, autonomy, autonomy, and inexpensive, inexpensive, inexpensive. And what does the US military do? Not autonomous and very expensive. So? So the disconnect between what my community sees and what the military doing is, is profound. So my favorite example is that we're building a sixth generation fighter. Now, these fighter jets last for 50 years. So do you really think in the year 2080, we're going to have humans in fighter jets in dogfights? And the reason is because you don't fight, you don't have dogfights.
You shoot precision missiles over from aircraft over long distances, which is easily done by drone aircraft. So there's a fundamental problem in the conception of our national security and defense, which is we're not moving to autonomy, autonomy, autonomy, and inexpensive, inexpensive, inexpensive, fast enough. Autonomy is, from your perspective, code for AI. So an example would be talk about Ukraine for a minute. Ukraine is interesting. Aside from the fact that we want we as America should defend freedom, right? You know, freedom is you have to fight for freedom. I'm sorry to say I don't I, I'm certainly not trying to encourage war. You have to fight for freedom. Otherwise you lose it. And Ukrainians are fighting for their freedom. So when the war started, Ukraine had no air force and no navy.
So what do they do? They build sea drones, which cleared the Black Sea Fleet. They are technically, an example is called Sea Baby. They are essentially speedboats with rockets on the front and a navigation system. And based on the azimuth and the sea state, they can shoot a missile or rocket into the target, which is a Russian ship. The first time this happened, they sunk the ship. Second time the helicopters came and destroyed their ships. So it turns out you need a whole bunch of those, but eventually enough of them will get through to destroy the ships. Ships are big targets. I know everyone likes ships. I happen to like big ships. Everyone likes aircraft carriers, they are very impressive.
A single one of these autonomous drones that I'm describing. Underwater drones can destroy an aircraft carrier. This is like we shouldn't be building more of them. I'm sorry to say. I know a lot of people like them, but I'd much rather have automated equivalent of aircraft carriers that are AI powered in the Air Force. They didn't have an Air Force to speak of. They had a couple of jets which are kind of not working, and we didn't give them F-16s until a year into it till they built drones. So the first phase of the war, basically the Russian tanks were coming in. And a simple rule is a $5,000 drone can kill a $5 million tank. Our tanks cost $30 million, by the way.
So same is true there. Two, tanks I think are largely over as a war weapon. And my idea was that in Ukraine you'd see the end of artillery, mortars and tanks. Today, artillery is still being used. Mortars are not really being used because they're not precise enough. Artillery is bad enough and tanks are not being used except for APCs armed personnel carriers. And the reason is that the drone collisions, if you will, the drones over the flot, as it's called, are so ubiquitous that you can't move. The drones have daytime cameras and nighttime cameras. They are not particularly automated. So part of my interest in this is to understand what happens when you add intelligence. So here's roughly how it works.
You want precision ISR as it's call, which means you need out drones in this case in the in the ground along whatever their contested line is. Let's assume they're always there. Current drones can last 12 hours. They're specific long range drones Ukrainians have built. They can fly around for 12 hours, and they have very good cameras. The cameras are, typically Japanese. They have 30km, 40km vision. They also have night vision. Various radar things and so forth. That allows you to see what's going on. And then you have other drones, which are attack drones which carry a weapon. There are designs in Ukraine which are both an ISR drone and an attack drone, although I think that that separation is probably makes sense because the ICR drone is sort of sitting up there, whereas the attack drone is either a kamikaze drone or a return drone to get refilled
But the way is in the future will be land wars will be fought. Is you have an automatically controlled ISR drones, and then you have drones, that are killer drones, if you will, that are dispatched by the action of the ISI drone. Making the latency between target identification and the destruction of that target in seconds, remember, So these are sounds kind of straightforward. I just gave you the simple diagram. Let me tell you why. It's hard. An old rule about war is the enemy gets a vote to. So, today, for example, there are many drones that are there are not, armed. They're dummy drones. How do you distinguish between the two of them?
Another example is that the GPS is denied. When I say deny it, I mean, it's just like often bad and so forth. Many of the American weapons at the beginning of the war would miss their targets because the Americans had been either stupidly built it this way, or they were required to build systems that use GPS navigation first rule of war is G.P.S. doesn't work. So you have to come up with a way of knowing where you are. There are techniques, but fundamentally you have to solve that problem. And the second one is that in this case, the Russian electronic warfare is the best in the world. And by that I mean the jammers, the suppression of signals and so forth. So I watched some of the missions, and what happens is, just to give you a stereotypical, it's in Kyiv, it's nighttime,
It's like 9:00. The soldier is drinking coffee with his screen, and he's running a war. In this case south of Zaporizhzhia. So the drone takes off and you can see the drone going along. It's perfectly fine. And then you begin to see the blocking. And in and out and in and out. In this case, the drone eventually got to its target. But they kept trying to block it. So you can see when you watch the war, the drones, you can see with the DJI drones, for example, which they used ubiquitously. You can see the GPSs go from 20 to 0 when you get that. So you need a solution to blocking. But you have another problem. The systems are not automatic enough to hit the target without a human telling it where to go.
So when they get close, you lose sight. Well, then you don't have confirmation that this is your target. And the rules of law say you have to have to know where the target is. You know, you have to actually do this. You lose the signal, you have to come back. So you have to figure out a way to get through that. If America wants to win any next war, we have to collectively solve the problems I just described. And that learning is occurring in Ukraine because the US is not allowed to be in Ukraine and US soldiers, which is a good idea, by the way, are not there
They're not seeing us. They're not watching the evolution and development of a new war. And that means that the transition from the old model to the new model in America will take much longer.
00:30:54:20 - 00:31:32:20
Ylli Bajraktari
Thanks. I'm gonna ask one quick question, and then I'm going to turn it over to the audience for a couple of questions, if you don't mind. We are in transition now. We will have, an incoming president, soon in January, at the end of January. This podcast is called memo to the president. What would be some of your advice for the new president on the air is we have to double down and accelerate, both on AI, but also, given how you passionately talk about Ukrainian warfare, what would be some of the things you would say to the incoming administration that should do?
00:31:32:22 - 00:31:54:09
Eric Schmidt
Thank God I don't have to be president. Imagine the situation where you have your president and you're trying to achieve some objective. And every single group is organized around its own special interests and bureaucracies. The first thing bureaucracies do is they self perpetuate. It's just a rule of bureaucracies. So I think you have to have some philosophies. I think the first has to do with, as a country, we need to help defend democracy. We need to fight for freedom. And I think that's pretty obvious. Now, that doesn't mean we have to fight, but we have to assist other people in the fight, which is why Ukraine is important. I think that there are many other examples.
I'm just using that as the one I'm familiar with. And that specifically means providing the necessary assistance for them to hold the Russians back until there's some kind of a peace thing. But what I have also learned in my wandering around the military is that as much as I would like to say, defensive weapons are the correct answer, in fact, the way you win and the way deterrence works is you have to you have to be able to win at offense. And the reason is offense can always overwhelm defense with enough offense. I didn't understand that. So that's a learning for me. I think the second point goes back to this. I think if you believe as I do, that the arrival of this non-human intelligence is the single most important thing that will happen in our lifetimes. It is the creation of a new epoch of history, right? Because these things aren't going to stop when we're all dead. They'll get smarter and smarter and people will still be here. Our children and grandchildren will be dealing with these things. So winning the war in terms of leadership, winning and winning here doesn't mean breaking them in the sense of China. It means being ahead of them, is crucial.
I believe that the country should have a national initiative around getting there. Now, what would that a national initiative look like? We don't know. There have been a number of proposals, but it goes something like this. The majority of the developments will be in the private companies. They need things that the government has, starting with energy. The government needs things from the private companies, including knowledge of what they're doing, access, national security things, classified things, so forth and so on.
There is a deal to be made. If I were Trump with the tech companies and I guess Elon will help here, maybe, where it's basically access for energy. Right? It's a good partnership, right? We all have the same goals. And there are a number of other tactics, including making sure that we have the world's top researchers and things like that.
00:34:15:22 - 00:34:21:07
Ylli Bajraktari
Thanks, Eric. Turn over to you, for any questions you might have, but, my friend.
00:34:21:09 - 00:34:22:05
Eric Schmidt
Yes. Go ahead.
00:34:22:07 - 00:34:43:08
Guest Question 1
What would be a meal that used to be Israel's national security? Yes. And, the things that, you said was something a lot, part of the. I want to, Hello. As about the last thing you said and tied up one of the first sentences that, governance isn't ready. I'm not sure access will be has. If there was no putting it standing of how to use this unless it so happens. That's the thing is, governance is more. It's more than access. They control progress. They can be number two. But if they're better in weaponizing those, technologies, then being number one on the private sector might not be enough. So just if you can elaborate a little bit more, you talked about energy and access.
Would that be enough for the US to lead the world?
00:35:11:16 - 00:35:30:12
Eric Schmidt
Well the American model is phenomenal, right? I mean, you know, before we get too depressed about this, like a the innovation model of America is the, darling of the whole world, right? Look at what we've achieved as a country. And and thank you for partnering with us and all these amazing ways. Your model is also very impressive.
00:35:30:12 - 00:35:55:10
Eric Schmidt
I'm a big investor in Israel. So to me, America will be good if we get our act together on this. Now, what does it mean? China is likely to productize this technology much better than America because their costs are lower and they work harder. So an example would be robotics. I worked a lot on robots for 20 years. The Chinese robots are now, in many cases better, right? You know, it's just they're just better. I'm sorry to be honest. So we have to have answers to each of these areas, and we have to win that. In the case of Israel, by the way, if you look at the horrific Hamas attack, had you had two kinds of drones, you would have been in a much better situation.
The first drone is an ISR drone that has a bomb on it. The Israeli drone sees the Hamas rocket taking off and destroys the target immediately. That would have helped. Another one would be tunnel clearing robots. It might talks with the military. And I'm sure you I'm sure you were in the military. So you must have an opinion about this. The notion of a human going into a tunnel is the scariest thing I can imagine. As a human right. I want the robot to do that. I want the drone to do that. And I think these techniques are relatively straightforward with an A, with a view of autonomy and ubiquity. And they would have really helped.
00:36:53:18 - 00:36:55:04
Ylli Bajraktari
Thank you. Rye.
00:36:55:06 - 00:36:56:21
Eric Schmidt
Go ahead. Good to see you again.
00:36:57:01 - 00:37:02:22
Guest Question
Good to see you. Thank you. Thank you. Right back out with honor. The last time that I saw.
You in person was about a month before Secretary Kissinger passed away. It was shortly after the October 7th massacre. So it's been about a year. It's an extraordinary partnership that you had in creating your first book and now this one. How might he, over the last year, have challenged you in your thinking or advanced your thinking?
What type of questions might he have asked over these last 12 months, when so much has continued to happen in the world?
00:37:30:02 - 00:37:55:05
Eric Schmidt
Well, Kissinger, you know, it was literally my best friend. He had a remarkable ability to put things in a historical context and see 2 or 3 moves ahead of me, and I think anybody else. So he was extremely concerned before his death, not that Ukraine would win, but that Russia would lose. Now, today, that's not a conversation. And he was concerned because his view is that the that the Russia is fragile. I'm just saying what he believed and that if they begin to lose in this, the various wacky parts of Russia that are not in Moscow and Saint Petersburg get sort of unruly. You can have internal civil wars, lack of control, things like that. So it's that kind of insight that I miss.
If you look at Syria today, Syria fell, right. He worked, he spent a fair amount of time with the Shah of Iran. You know, the 1970s when he was secretary of state, he would have a very interesting answer as to how to unify the Shia in a way that's friendly to us and keep them under some level of control with balance of power, with the Sunnis. So that's the kind of insight that we're missing. I would, all of us would benefit from that kind of insight.
00:38:48:10 - 00:38:51:11
Ylli Bajraktari
Tammy, I know you're going to. You have you always have questions.
00:38:51:15 - 00:39:22:09
Guest Question 3
Hi. Hi, Eric. I have a question, about human in the loop and the Defense Department and our friends from Israel had so many tough decisions to make, but some of the reporting of, the activities, military activities said that there were there was an automation bias that soldiers were making decisions about, where to strike, how to strike, not just in Gaza, Israel, but in Ukraine, etc. And I just wanted you to address the whole idea of human in the loop and how how do you train militaries? How do you educate people mean what's the answer? And just like in the larger picture, because you've dealt with automation bias for a long time.
00:39:37:09 - 00:40:05:04
Eric Schmidt
So I'll give you my favorite example. So you and I are the captains of a destroyer, and we've got a whole bunch of weapons and a radar, and the computer system says in 30 seconds because we can't see it, a hypersonic missile is going to come and it's going to kill us. What would you like to do? Okay. 29 28 27 press here to launch.
Okay. How many people will not press that button like in about two seconds? So I think we have a problem that in, you know, in real dire situations, people are going to shoot because they're going to trust they have no other source, they can't see it, they have no other indication in the AI system. And that could lead to tragic to tragic outcomes. You know, we saw this with in, the Cuban Missile Crisis. There were there were a couple of cases that are similar to that. The military is now using the term meaningful human control because the targeting of the weapons is so good now that basically the human authorizes it. But the target, the weapon can go find the target and destroy the target.
So I believe the way this will resolve itself is that the human has to take responsibility in a legal sense of what's going on and has to be competent enough to understand what can happen. And the system has to have enough safeguards to make sure the thing, for example, doesn't turn around and go the wrong way, but think that's how it's going to be resolved. I don't think you're going to, by treaty, reduce the accuracy or automatic targeting. Remember, these drones have fast enough processors and good enough imaging that autonomously they can do search and destroy. Right. So where is the boundary? You say there's a tank over there. Take that tank out. That's illegal. What happens if there's a tank somewhere in this world?
Go look for it. Right. Where is that boundary? We haven't found that yet.
00:41:41:03 - 00:41:45:02
Ylli Bajraktari
Thanks Eric, last question. Here. Thanks.
00:41:45:04 - 00:42:01:03
Guest Question 4
Thank you so much. So yeah, [name] from Georgia ISF fellow here work on, information warfare at the Fletcher School, blind diplomacy. You mentioned Ukraine and Georgians care about Ukraine. Can we have our own freedom fighting going on as the.
00:42:01:03 - 00:42:05:00
Eric Schmidt
You are also the victims of the most interesting cyber attacks in history? Yes.
00:42:05:05 - 00:42:07:01
Guest Question 4
I'm sorry. Fortunately.
00:42:07:03 - 00:42:10:02
Eric Schmidt
I'm sorry.
00:42:10:04 - 00:42:51:14
Guest Question 4
What I wonder is, What? Ukraine, as they have demonstrated, some of the latest, potential of, warfare technology on the battlefield can, teach to the Americans or to the rest of the world. And how do you, as a second leg to it, how do you imagine, as you we see a, big emergence of entrepreneurs, in Ukraine and developers and Ukraine, you know, known for its, human capacity in that field, how do you expect Ukraine to emerge as a new marketplace for development, that especially along the lines of those talents that you mentioned.
00:42:51:16 - 00:43:12:12
Eric Schmidt
So, in my most recent visit, to Ukraine? There are and it's fun to take the train and it's a great country. There are more than 600 drone startups in Ukraine, partly because if you're a man and you're in a drone startup, you're not being drafted into the war. So there may be oversupply of labor, which is a good thing. I'll give you two examples of two companies that are now growing quickly profitable, raising money, Skyfall and Air Logix, both of whom make domestic drones that are quite powerful in different segments of the market. Those are two of of many that are likely to become exporters to, for example, NATO and why would NATO buy it?
They have no way of testing. They have no way of understanding what it really takes to do these things. So I think you will see, assuming Ukraine survives a set of such things which will be available to the West, the, the Ukrainians are very clever with hardware. They're not very good, if I may say this, at huge volume. So, for example, they don't have assembly lines with PC boards and PC manufacturing that you would find, for example, in Asia. So they're dependent upon external things like that. And they're not particularly good in the higher functions of AI software. I think the military AI stuff is going to get built in America by the startups, many of which DIU, and others are helping fund.
So this is really important. When I talk to my friends who are in those startups, their problem is that their customer doesn't get why they're so important, but they're not allowed to sell to anybody except for the US, for obvious reasons. And the US system is so slow right now. There are there's a group called, Replicator, which is a project which is trying to solve this. These are infinitesimal compared to the $90 billion budget for the US military. So one of the things that has to happen in the US is we have to have some, some space to build these kinds of unmanned, uncrewed, the the term is uncrewed weapons, which is sort of I just call them drones.
00:45:00:09 - 00:45:06:20
Ylli Bajraktari
Eric. Okay. Thank you. I know you have a horse. Though, I want to thank you for dropping by.
00:45:06:21 - 00:45:09:21
Eric Schmidt
No, no. Thank you, thank you. All right.
00:45:09:23 - 00:45:15:06
Eric Schmidt
So, and I hope I'm wrong. But, one, I hope that you enjoy the books. And because we worked hard at it, and it is Henry's last book, he finished it on his deathbed. So he literally finished it about a week before he died. That's how committed he was to this. So in his honor, it's it's good. And the second thing is that we should talk about this AI task force we're putting together. We want to we collectively at SCSP, we want to drive this thinking throughout every aspect of the US national security complex. And you all can help us, right? If you if you agree.
00:45:48:07 - 00:46:05:11
Ylli Bajraktari
Thank you, Eric. AThank you for your support. You have been the friends and family of SCSP from day one. So, happy holidays to everybody. And thanks again Eric.