00:00Let's start with China.
00:01You focus a lot on China and the growth of AI there vis-a-vis what's happening here.
00:05We learned this week that some H200 chips have been sent to China, a small amount according
00:10to the Commerce Department.
00:11How big a deal is that when you look at the way that our country is interacting with China
00:16when it comes to AI?
00:17Yeah.
00:17Thanks for being here with us.
00:18I'm so happy to be sitting here in this beautiful place.
00:21Beautiful.
00:21We're about to kick off the conference.
00:22And yeah, China's going to be front and center.
00:25And the technology race with China, we are really at an inflection point right now.
00:30And it's no longer just about chips because the Chinese open source models are so close
00:37to our most advanced frontier models, like what just came out, GLM 5.2, you all heard
00:43about Deep Seek, all of those things, especially in the cyber area.
00:48They're nearly as good as Mythos and some of our best ones.
00:51So we're getting into an era where you're about to see extreme cyber warfare enabled
00:57by AI.
00:58Deeply worrying.
00:59Conventional wisdom here has been that China has been behind us because they haven't had
01:02access to these technology, the kind of infrastructure that we have.
01:05Is that misplaced?
01:06Is that faith in that misplaced?
01:08No, it's not misplaced.
01:10And I do think the Biden administration was exactly right to have export controls on the
01:15most advanced chips and the semiconductor manufacturing equipment.
01:19But you can only hobble your opponent for so long, you've got to run faster.
01:24And so that's the race we're now on.
01:26They're basically caught up.
01:28And I would say I have to give the Trump administration some real credit.
01:32Just in the last couple of weeks, you've heard a lot of quiet conversations behind the scenes
01:36in Washington within the administration and with advisers outside thinking about how
01:42do you test these models?
01:43How do you make sure they're safe from a national security perspective before you put them out there?
01:47And I think a corollary to that is they also are going to probably want to test the Chinese
01:52models and tell American and Western companies is this safe to use or not.
01:57I think back on when we became aware of DeepSeq and the amount of surprise that was associated
02:01with that, now many months since then, do we have a more comprehensive, a better sense
02:06of sort of what China has been doing and is doing?
02:08Yes.
02:09And what's led to that?
02:10What gives us the confidence that we know sort of where they are in their development?
02:13Well, I mean, our technologists, I live in Silicon Valley, so you can't say a sentence
02:18without saying AI.
02:20It's required.
02:20Yeah, exactly.
02:22You can't look down the street without seeing a Waymo.
02:24But so our scientists are often, and the people in the firms, know what their competitors
02:29in China are doing.
02:30They just make it a point, just like the Chinese know what we're doing.
02:33There are some scientists that go back and forth, far fewer than you would think anymore.
02:38But of course, you're in a business of knowing what your competitors are up to.
02:41How would you characterize the U.S.'s level of engagement with China and with all that's
02:44happening with AI in China?
02:46I think back on the first Trump administration, there was such a focus on competition with
02:51China.
02:51There was a more hawkish stance, I think, than we have now.
02:54Has that changed?
02:55Are we talking about China enough, do you think?
02:57Yeah.
02:57I do think that some of the serious China hawks that worked in the first Trump administration
03:02aren't in this time.
03:04And so you have more disparate voices giving advice on China here.
03:09And the president obviously wanted to have a good meeting with Xi Jinping.
03:13Things were quiet.
03:14Behind the scenes, under the surface, the competition is on.
03:18The competition is on in the Indo-Pacific, as we'll also hear this week.
03:22It's on in technology.
03:24It's on in all sorts of areas.
03:26But I think on the top, people are hoping to pretend it's smooth sailing.
03:30Moving from Beijing to Washington, how would you assess Washington's interest in regulating
03:36this space?
03:37It sounds like you feel like there's more seriousness being devoted to that now.
03:41There is.
03:41What does that look like?
03:42What does that look like kind of in concert with industry?
03:44Yeah.
03:45So the contours are not yet known, but the Biden administration and actually the UK government
03:52did a really good start on this.
03:53They created these small AI safety institutes, deeply nerdy institutions that had scientists
04:00that were testing for national security threats.
04:03Can these AI models build bio weapons, chemical weapons?
04:07Can they be used in cyber?
04:09However, those have been stood down, but there's now some conversation about having the national
04:15labs test these models.
04:16I think that's a really good model because it's what our national labs do.
04:19They test technology for the American government.
04:21So it doesn't matter so much who does it and how.
04:26It matters that we're doing it at all.
04:28There will be conversations this week here centering on the use of AI in the context of national security.
04:33How are you feeling just about the way in which that's been deployed thus far and how folks
04:37are thinking through sort of the policy implications of having AI driving advanced weather?
04:42Yes.
04:42Exactly.
04:43I know that the Pentagon is thinking about this very carefully.
04:47And I think there are some really good people from the tech world who are now in the Department
04:52of War.
04:53So that part is good.
04:56Look, we're all learning and we're all learning super quickly.
05:00You have the Belgian defense minister here.
05:01You have actually a number of defense ministers from Europe and beyond.
05:04And all of the informal conversations I've had so far has been around this.
05:09What do we do about drones?
05:10What do we do about counter drones?
05:12How do you take them down without having a whole lot of debris?
05:15How do you jam?
05:17How do you can you build new surface ships or are they just sitting ducks to see drones?
05:23So none of it's final yet because we're at such a strange hinge point in history.
05:28But I've been really encouraged by the number and the sophistication of the conversations
05:36that are happening.
05:37This brings us to the Middle East and we've seen the U.S. kind of wage war, make strikes
05:42with kind of traditional weapons.
05:43But I'm curious of how you see all of this playing out, all of the back and forth surrounding
05:48the Strait of Hormuz.
05:49But yes, sort of the lessons learned thus far from the conflict that we've seen.
05:53Yeah.
05:53Turns out that mines and drones are pretty effective because if you're, I used to be
05:58a shipping company director, you're not going to send anything through the Straits of Hormuz
06:02if there's any chance that it's going to be hit.
06:04Your insurance rates are sky high and you're just not going to take the chance.
06:08So, the United States and our friends and allies have to get it right 100% of the time.
06:14The Strait has to be 100% open and Iran only has to get a couple ships.
06:19And so that is what's making this incredibly difficult.
06:23And I think if you and I had chatted two months ago, I don't think anyone would have predicted
06:27that A, this blockade and the closing of the Straits would be going on so long and B,
06:34how much the world has managed to muddle through, for lack of a better word.
06:40Like, we haven't shut down the world economy even though this has been very, very difficult.
06:44I've seen this analogy of kind of Schrodinger's box to Schrodinger's straight that you wouldn't
06:47take a ship through if you thought there's any indication or any chance that there could
06:50be mines there.
06:51What will it take in practical terms to give somebody who is running a shipping company,
06:55say, or a captain the confidence to navigate that straight in the way that he might have
06:59before?
07:01You need to be 99% sure that there are no mines in your path and 99% sure that
07:08no drones
07:08or missiles are going to hit your deck.
07:11And we are not in a position yet to do that.
07:14Let me ask you, lastly, just sort of where we go from here.
07:17So initially, President Trump floated this 20% tariff the U.S. might place on ships to kind
07:23of help them navigate the strait going forward to protect it.
07:25Now that's become kind of an investment deal perhaps with Middle Eastern nations.
07:30I'm just curious sort of what you think the region is looking for and how transformative
07:33this moment has been for Middle Eastern allies of the U.S.
07:37Oh, my gosh.
07:38Well, you should ask them.
07:39Some of them will be here.
07:40But it's hugely transformative.
07:43I think the quiet behind the scenes working together of the Israeli military with the Emiratis,
07:50for example, the fact that the split between Saudi and the Emirates became so apparent, how
07:59difficult it has been to defeat Iran.
08:02It's all we're seeing the architecture of a totally new Middle East.
08:06And I think we're nowhere near done seeing it.
08:08I think there's one encouraging that I'm going to leave us with the silver linings, which
08:14is really hard to do in the Middle East.
08:15But there is this moment where the Lebanese government actually could stand up and try
08:22to kick out Hezbollah.
08:23And I think part of the difficulty, really the poison pill that unraveled the MOU, as you know,
08:29that Israel wasn't willing to give up the fight in southern Lebanon because they were
08:33not willing to be hit by Hezbollah.
08:34So you could imagine a positive future where the Lebanese government gets stronger.
08:41They put down Hezbollah.
08:42You have a better northern neighbor for Israel.
08:45And you start building stability from that side.
08:47But let's hope we get there.
08:49We'll leave it on that.
08:50Thank you very much.
08:51Appreciate it.
08:52Great to be with you.
Comments