Trump's Super Intelligence Summit, AI Safety Accord, GDP Beats, Midterm Predictions

All-In with Chamath, Jason, Sacks & Friedberg

The hosts analyze President Trump's Super Intelligence Summit and the resulting White House Accord, emphasizing its practical approach to AI safety through mandatory internal controls and

Key takeaways

• President Trump convened the White House Accord on Superintelligence, requiring frontier AI companies to implement internal controls verified by external auditors, with board-level fiduciary duties enforced via D&O insurance.
• Dario Amodei of Anthropic attended the summit and dined with Trump, marking a notable shift for an organization previously resistant to such government engagements.

Main topics

• White House Accord on Superintelligence
• AI governance and board-level audits

Notable quotes

"whoever wins super intelligence wins."

Conclusion

The episode concludes that while political narratives and media framing may skew public

Transcript preview

Speaker 3 (0:00) Another huge shout out to Starlink. How good do I sound right Speaker 4 (0:02) now? It's actually great. It's flawless. It's flawless. It's shocking. It's amazing. Speaker 3 (0:06) Nick, show this behind the scenes shot of what it took to get us going. Starlink is bananas. Speaker 4 (0:12) That's amazing. Do you guys want to know something about the Starlink network that I think is just kind of coming out now? Basically, the Starlink network, because it's putting out radio waves and it can pick up radio waves and you've got so many of them, you've basically got a phased array telescope looking down on the Earth. And you can pick up what would otherwise be stealth aircraft from above. So this Speaker 3 (0:34) is a UFO detector, you're saying? Speaker 4 (0:36) It's a stealth airplane detector. It's basically a Starlink. It could be a UFO detector. But the Starlink network basically becomes this incredible radar system that sits above the Earth to look down on things that would normally be stealth from the surface of the Earth. Speaker 1 (0:51) So does this mean we don't have to do a science corner today because you just did it? Speaker 1 (0:56) Jack. Speaker 4 (0:57) Great to see you, Sox. As one Speaker 3 (1:00) of the architects of Bretton Woods 2.0, Sox is on a high. He's coming in hot. Speaker 2 (1:22) All right, everybody, welcome back to This Week in AI. Sorry, all-in podcast. And we've got a ton of news to get to on the docket. Speaker 3 (1:30) You're such an Speaker 2 (1:30) idiot. Speaker 3 (1:30) You're an idiot. Speaker 2 (1:31) I mean, it's just the, every week is AI on top of AI on top of AI. You have a show called This Week in AI, and you cover nothing about AI. No, no, we cover AI, but thank you for the plug. Thisweekinai.ai, sign up for the newsletter. This Week in Speaker 1 (1:43) Superintelligence. Speaker 2 (1:44) I'm going to change the name. This Week in... chat dj what you should say Speaker 1 (1:49) is we have four super intelligences here four Speaker 2 (1:52) super intelligences here to super intelligence the week's news Speaker 1 (1:56) intelligently discuss the Speaker 2 (1:58) news yes and what a week it was we were sitting here just a couple of weeks ago or sitting in los angeles i should say and there was a bit of a chaotic moment hey should we be should we be auditing or de-selling with all this ai and Three weeks later, President Trump held a super intelligent summit in DC. Many of the besties were there, or three or four, I should say. On Tuesday, President Trump convened the biggest names in AI, Elon, Jensen, Zuck, Dario, Sundar, Satya, yada, yada, yada, all the way down the line, and everybody from the government as well. I could ramble on about this, but Sacks, I have a... feeling you were instrumental in setting this up. So maybe I'll just hand it off to you. Tell us how this went down so quickly, and then what the outcome is of this superintelligence summit. Speaker 1 (2:57) Well, the reason it happened so quickly is because President Trump convened all of these people to the White House. And I think he's the only person who could get all the CEOs of all these companies in one room. I mean, these are the leading companies in the world across data centers, across chips. hyper scalers, and then obviously frontier models. And many of them, frankly, don't like each other. They have disputes with one another. We all know this. They may have lawsuits with one another. All of that didn't matter because President Trump called them to the White House for this conversation. And I've called it the Bretton Woods of super intelligence. This was the first gathering of really all the critical players across the entire stack of the industry. And I think we had a really good, open and honest conversation about both the opportunities. and the risks of superintelligence. And as a result of that meeting, I'd say during the meeting, we worked out this agreement that has come to be known as the White House Accord on Superintelligence that was then signed by the six major frontier model companies. And I think we'll now apply a new rigor and new set of controls on the development of superintelligence so that the public can be more confident that this will be done safely. And if I could just highlight a few of the key features. So first of all, the companies who are developing these products acknowledge that they are responsible. They have to be the responsible actors. Let's not deflect it onto someone else or anthropomorphize the models or talk about the United Nations. These companies accept responsibility, number one. Number two, they agree to have internal controls and to have those internal controls verified by an internal team that does that. But we don't just leave it to quote unquote self-policing. It starts there. But then you have external auditors who will then validate and verify that the internal controls are working. And these companies will set up an independent committee of the board to then receive those reports. And as Gavin Baker pointed out this morning, when the board receives a report from an external auditor, you can't disregard that. It's not optional. The board has a massive fiduciary duty to then act on that. Otherwise, their DNO policy can get canceled, right? So now the board is going to receive the report of the external order, and they have to make sure as a board that they're overseeing, they're providing oversight to make sure that the frontier development is happening safely. And then finally, sitting behind all of that, you still have the FTC and the SEC, which have the ability and the responsibility to enforce the public commitments of these companies. Although this agreement was entered into voluntarily, the governance that follows from it is not voluntary. It is now day factor required. It's fiduciary. This is a critical point. Speaker 2 (5:44) It Speaker 1 (5:44) becomes fiduciary. Speaker 2 (5:45) Because Speaker 1 (5:46) it's an audit, Speaker 2 (5:46) because, and D and O for people who may be not familiar with the directors and officers, that's the liability insurance that that board of directors has in case something goes wrong. So this has got teeth to it. Even though it was voluntary, it has a deep, deep... hooks into the board of directors. Speaker 1 (6:03) That's right. And just one final point on this is that, again, this White House Accord on Superintelligence, it was convened by the president. It was as a result of that discussion and that conversation. We got the major companies to sign it. And I think everyone in the room, there was basically a consensus around it. And these controls can now be implemented basically immediately or very, very quickly. And it doesn't require new legislation. It certainly doesn't require some new international agreement that's never going to happen. So this is something that can be done right away. I think it's a credit to the president. And I think this should, in large part, I think, address many of the concerns that the public has about the safe development of superintelligence. And I think there's a lot of Democrats right now who don't want to give the president credit around this because they want to keep using it as a campaign issue. But the reality is that I think the president dealt with it swiftly and sensibly and practically. This is way more practical. than trying to declare a pause on frontier development that China would never go along with or creating some DMV for models that, again, would slow us down to the point where China just wins. So this is very practical and it's immediate and the president deserves credit, even though the Democrats won't give it to him. But I'll notice that even Jacob Coxon. where, you know, in a way you could say that the Speaker 2 (7:18) whistleblower, quote unquote, the so-called rage quit after six weeks, who is working with a PR firm for the years. Speaker 1 (7:25) Yes. I would say fake whistleblower because he was working with a PR firm the whole time, even though he said he wasn't official whistleblower or the artificial whistleblower. But look, even he said that he was happy to read this. He was giving credit to the president. So maybe everybody else should. I mean, Speaker 2 (7:39) even the most cynical person, Shamath, looking at this would say great first step, right? Like if you just. wanted the global UN, United Nations of AI, this would be a great precursor to that, a first step where everybody's opting in to be regulated, to regulate themselves, as we discussed on this very podcast with the MPAA analogy. And here we are three weeks later. Take us behind the scenes. There was one notable addition to the group of AI executives, Dario from Anthropic. is obviously very anti-Trump. He has never shown up for these things. The organization is 96 % anti-Trump, left-leaning, et cetera, but he participated. So how did Dario wind up participating in this? How did he wind up doing dinner one-on-one with the president? Take us to the palace intrigue here. Dario was a conscientious objector. He did not want to participate in any of this. And now he was front. and center. I can tell you what I heard. And I just think it Speaker 3 (8:39) speaks to how conscientious and gracious the president is. I think there was like a couple of previous events. He noticed that Dario wasn't invited and actually just called and said, you know, this is an oversight and invited him to dinner. And I think that's how it started. And they had dinner, I think, two nights before that was widely reported. And then he stayed and he was at the dinner. So I think it's just. very smart, strategic thing to do. It was also a very conscientious and very polite thing to do as well, just to make sure that this is a broad coalition of people that were included. And I think that's sort of been the general pattern that the president has exhibited. Let me maybe give you a couple of other... things that happened at the luncheon that had an impact on me. The first was, it sounds like a very simple statement, but the gravity of it kind of has built in my mind as I thought about it. He says, whoever wins super intelligence wins. And then he just stopped talking. That was the first line of his prepared remarks. And then he said it again. And when you think about it, he's completely right. There is no world in which you're going to have this multinational corpus of people trying to allocate wins and losses on this thing. This is on the order of the Industrial Revolution. In fact, Sachs had some really nice comments at the luncheon, which framed it as even bigger than the Industrial Revolution. He went through the numbers and it's bigger than. the Industrial Revolution, and like three or four other things all added up together, this thing is still Speaker 1 (10:07) bigger. The railroads, the canals, and the electrical grid combined. And Speaker 2 (10:11) the pyramids and the Great Wall of China. That was in the Speaker 1 (10:14) Wall Street Journal last week. Speaker 3 (10:15) When you think about it that way, it's very important that we win. And so he started by setting the stage that way. He had a couple of other specific things, which was, it's really important that the extremely positive use cases. that exist out of AI and now SI are demonstrated. And he said, you have a responsibility to lead with the benefits. And then he said, you also have a responsibility to impact the local communities in which you operate. And he pointed at Meta as an example of a company doing really good things with teacher bonuses and hospitals. And we heard about that from Dina at the All In Summit. And he said, you need to continue to do these things. But meanwhile, we have to have the checks and balances, the trust and verify. After the luncheon, I had to fly to Orlando because we at 8090 have a long-term partnership with Ernst & Young. And I was speaking to their 3,400 partners. And one of the things that Ernst & Young and 8090 were doing was building an entire audit infrastructure for superintelligence. And what's so cool is that We had been building it for EY as our first customer. But on the heels of this, I think we're going to be able to take that infrastructure and roll it out to everybody, as Zach said, because everybody's going to need it. And there's really three things that you're going to need in an age of superintelligence so that boards are protected, CEOs have visibility, and all of the other compliance infrastructure and governance infrastructure of a company like DNO insurance and all these other things are kept in place, which is end-to-end traceability. Right. Which is to say, when you deploy superintelligence, you need to know what's going on. You need to map it to the policies so that you understand what the risks are. All companies have standard operating procedures, but you need to know exactly when you're deploying superintelligence, how that builds on top of your standard operating procedures. So you need to map policies to risk. And then the third is you need to have auditable evidence, which is your way of showing all of your downstream ecosystem, here's what I've done. So if anybody taps you on the shoulder, whether it's a lawyer, whether it's an auditor, whether it's a regulator, whether it's in a lawsuit, whether it's a customer, hey, how did this happen? You can present something. So these are like very practical, sane things. And you can expect folks like EY now to deploy this broadly and quickly to all of the companies that they work with. And this is just one extra step. But if you think about how fast this happened, this is like three weeks. And you have this solution where the alternative was hit the pause on trillions of dollars of GDP so that we could sort this out in a multinational body. The alternative was crazy. So I think what happened was historic and important. Speaker 2 (12:56) Was there any pushback on the doomerism narrative? Take us behind the scenes. Speaker 3 (13:01) Jensen was really powerful. He said something to the effect of alarmism without solutions is unproductive, which I thought was helpful grounding. But he said, alarmism with solutions is helpful. Everybody had the same perspective in the room, to be honest. Was that directed Speaker 2 (13:20) at Dario, that comment? Speaker 3 (13:21) It was just generally said, but I think, you know, if we had to guess who was Speaker 2 (13:25) straight. Did Dario respond after that? I mean, Speaker 3 (13:29) Susie and I were sitting right across from him, so it was very fun just to watch his reactions in general. Speaker 2 (13:35) Please unpack it. What was his reaction to that comment? I'd be like, he listened. Got it. Who moderated the discussion, Sachs? How is this moderated? Were you the moderator? I noticed I didn't get an invite. The world's greatest moderator wasn't there, but Trump's a pretty good moderator himself. Who moderated the Speaker 4 (13:50) discussion? Oh, Speaker 2 (13:50) no, the president Speaker 1 (13:51) moderates it. The president moderates Speaker 2 (13:53) it. The president actually actively moderates it. Oh, yeah. Speaker 1 (13:55) No, no, no. Give an example Speaker 2 (13:57) of his moderation. I would like to take some notes here. Well, first when Speaker 1 (14:00) he got there, you know, this is in the East Room of the White House, I believe. And through the back of the East Room, you get to where this new corridor to the ballroom is going to be. And so we did a quick tour out there and he showed us what they're building. And it's really amazing. And I predict that when it's Speaker 2 (14:13) - Speaker 1 (14:13) This Speaker 2 (14:14) is Speaker 1 (14:14) the new Speaker 2 (14:14) ballroom. Speaker 1 (14:15) The new ballroom and the new east wing of the White House. Speaker 2 (14:17) Did you have to wear hats? No, no. Did you go into the construction zone? You were on the periphery. Yeah, we walked Speaker 1 (14:20) out there. What's Speaker 2 (14:23) the Speaker 1 (14:23) status of the foundation? Speaker 2 (14:24) They are making Speaker 1 (14:25) record progress. You know, the thing is like seven stories. So the ballroom itself is two stories, but it goes five stories down. Oh, Speaker 3 (14:33) wow. Speaker 1 (14:33) So there's all sorts of stuff that they're building there. The bowling alley, the sauna. No, no, no. For the military. I know, I know. There's like all sorts of vaults that they want. But in any event. And he's going to have it done in like a year, under budget and on schedule. By the way, when it's done, I predict that it'll be so nice that all future presidents will really want to use it. And it's things that the White House has needed for a long time. They're going to look back and say, why did they give him such a hard time doing this? It's just a Speaker 2 (15:02) distraction too, by the way. The fact Speaker 1 (15:03) that we had a builder in the White House who just enjoys doing it and is so good at it. I mean, why wouldn't you take advantage of those skills? Speaker 2 (15:12) All in annual dinner party there. We'll have an all-in gala. Speaker 1 (15:17) Yeah, anyway, it's going to be a big upgrade. So anyway, he took us on a little tour of that, and then he presided over the conversation. First, he gave some remarks, and he spoke to the whole group. And then he started calling on people to give their opinion. And so he always moderates these things. What was the most Speaker 2 (15:34) challenging? Did you have a challenging moment where you kind of went at somebody, a follow-up question, anything like that? Well, Speaker 1 (15:40) you just kind of ask people what they think and there's a follow-up. And then we had a breakout session of the six frontier model companies in the Roosevelt room, I believe. And that's where we kind of hammered out the accord. Although it already- When Speaker 2 (15:56) you're hammering out that accord, are you literally there typing it in Google Docs? And who's responsible for this typo? There was a typo in the tweeted version that became viral. Speaker 1 (16:06) I don't want to say you're responsible for that. But if that's the only thing people can find to object to, that's not so bad. Speaker 2 (16:12) Was Satya in Microsoft Word and he just was typing too fast? Speaker 1 (16:16) Yeah, I don't know. But basically... What happened is Zuck and Jensen talked through some of the high-level principles in the main luncheon. And then Mike Johnson, the Speaker of the House, actually had a copy and he read the whole thing. And then we discussed it and there was general agreement and accord with it. And then we did the breakout session and there was some wordsmithing, not much though, to be honest. Everyone was good with it. Now we had, I mean, the White House had a statement that had been circulated to the groups in advance that it was a couple of paragraphs. But frankly, this other statement had more meat on the bone. You know, the original statement had the statement of responsibility and it had the creation of a working group to discuss, you know, safety issues on an ongoing basis and share best practices. The new part were the internal controls, the external audits, and the committee of the board. And I think it's been publicly reported now, so I'm not really giving anything away, that Zuckerberg and then Jensen, I think, had kind of come forward with that idea. And everyone was good with it. Any Speaker 2 (17:21) notable moments where somebody was like, hey, I need a comma here, or I want to put an end here, I want to add something to this? Any notable edits? I Speaker 1 (17:30) mean, like I said, I mean, people were generally good with it. and they read it and everyone's like, yeah. Which part Speaker 2 (17:35) do you think has the most teeth, will be the most effective at making things more secure? The external Speaker 1 (17:41) audits. The external audits. Who's going to do an Speaker 2 (17:43) external audit then, I guess? Is it going to be an Ernst & Young type thing, PricewaterhouseCoopers, or KPMG? It might be they should Speaker 1 (17:50) be professional auditors, not these NGOs. That's my view. Speaker 3 (17:53) You're going to see EY in short order here or down something. Speaker 2 (17:57) And Friedberg, when you look at this in terms of... the speed and velocity it happened, and you're, I guess, on the President's Science and Technology Council now. Maybe you could speak to, you know, how quickly this is all occurring and what the impact might be in terms of America's competitiveness, because you've had concerns, hey, maybe slowing down here means China catches up. So maybe you could talk a little bit to your prediction, if any, of the alignment here of all of the companies. Also of note, just for the audience, people are like, Sam Altman's not there. His partner, Greg, was there. So Sam was doing, I guess, the launch of some new products they had. But OpenAI was obviously represented. So maybe you could talk a little bit about what you think the impact this will have on maybe the American public and their view of AI. I mean, look, I think we are... Speaker 4 (18:47) We always assumed in the sci-fi novels that we would be in an era of having robot wars. And I think we failed to recognize that the software came first and we are actually having the AI wars. And there's a war that I think we're in the middle of right now where there's kind of one side that wants to centralize, control. Speaker 4 (19:13) manage, regulate, whatever the term of the day might be, the technology. And there's another side that I think is cognizant of the reality of the technology, which is that it is effectively ubiquitous. We talked about this in the past due to open source, open weight models proliferating. The fact that you can run models on small local machines, that you don't need large clusters to do things, and that we're seeing model after model come out now that are not under the jurisdiction of some centralized system or agency, therefore, it becomes very hard and practically impossible to actually regulate the development of the software. That's the truth. It is what it is, you may not like it. But it is what it is. And those two sides are going to continue to fight. I think that the risks will rise. And as that happens, I'm going to make two