00:00Yeah, this is interesting, especially since AI has become such a sort of central part of our
00:04economy and it's developing so fast. There was a really interesting case, I believe it was this
00:10week, of AI breaking into an area that, like breaking out of a kind of a walled off area it
00:18was supposed to be in. And they did have to use an open weight model. I think they used DeepSeek
00:22to contain it. So I do feel like there is a growing feeling that they really need to kind
00:31of open. Right. Yeah. Max Schenken, the number one thing I see when I talk to people way,
00:37way smarter than me on this, they all say, we don't know where we're going to be in six months.
00:42Is that what you feel when everybody's business does techie stuff? We have no clue where we're
00:48going to be, you know, I think, well, sure. I mean, I think that's always true to some extent
00:53in the technology industry. And it's always it's also always true that tech leaders and those who
01:00are financing them are going to be incentivized to to sort of, you know, emphasize the extent to
01:06which things are happening quickly. I mean, to me, what's most interesting here is you have the AI,
01:12the American AI industry trying to navigate what is both a competitive threat from China
01:18and also a potential regulatory complication from the United States and to try to they want to catch
01:26up and they want to they want to find a way to sort of move regulators in a direction that
01:33is going to
01:33help them. And I think that's what's happening here with Satya Nadella.
01:37And a Friday morning ripping up the script again. Damien Sassar and Tom Keen across America.
01:41We get lucky with us. Max and Stacey from everybody's business here. Satya Nadella out
01:48with a three page dense memo on Twitter on the future of AI. Damien Sassar.
01:54Well, Tom, I think what Max is really talking about here is the concept of distillation.
01:57You know, the practice of using one model's outputs to help train or improve another.
02:00And that's basically what China's been doing is taking a lot of, you know, I guess the
02:04the the models that we are building here or anywhere else and basically using it to kind
02:08of train up their own. And, you know, the concept of a closed model. Right.
02:11I guess that's like chat GPT. Right. A cloud relative to an open weight because I'm not you
02:15know, that's like DeepSeek or Mistral. Right. So. So I guess given the names that are signed
02:21on the back of Satya's letter, I mean, I'm talking Meta, I'm talking Anderson Horowitz, IBM.
02:26It seems like they're all on the same page here in that we need to protect the AI.
02:29We need to protect the IP here. Yeah. I mean, I think what they're what they're they want
02:34to head off some kind of regulatory action that would make it harder for them to compete
02:38with these Chinese models. And and there's been also a kind of concerted effort to try
02:44to get the U.S. government to crack down on the distill on this practice of distillation.
02:50I don't think it's clear how they would do that. I think it also raises some sort of interesting
02:55intellectual property questions, given that all of these models, including the closed
02:59ones, are kind of using the same technique distillation, which is which is like open
03:04AI scraped the Internet to create their model that, you know, so there's like they're talking
03:09on both sides of their mouth with intellectual property.
03:11I like that idea that they're talking on both sides of their mouth. Stacey, you guys are
03:15way hipper on this than I am. I'm like the fossil in the room. Should we fear the Chinese?
03:21I mean, just as an everybody's business interviewer set of interviews you've ever done.
03:27Oh, and gee, Beijing has AI. That's how people like me look at it.
03:32I mean, I think they're a real contender. They are the other country that is really in the
03:37in the arena.
03:39OK, fine. But their price is dirt cheap, right?
03:41Their price is dirt cheap. In fact, using their models free, it's pretty good. I think they
03:46are a major competitor, I think, because open AI and Anthropic charge a lot. And in fact,
03:53they're learning all the time. In fact, one of the points that comes up with this is that,
03:58you know, companies are paying so much to use this AI, but they're also training the AI.
04:02They're giving value twice with money and then with helping the AI learn. I think with China,
04:08it does open it up in a really interesting way. And potentially, as this technology gets more
04:13expensive, as the tokens get more expensive, as this technology becomes more vital, companies
04:18like DeepSeek are going to become a much more compelling product.
04:22Max, I think Stacey hit the nail on the head here. I mean, you know better than I, Moonshot's
04:26Kimmy 3K3 model is the blended price is half of that of Claude Opus GPT 5.5. I mean, to
04:32me,
04:33this signals that the economics and the price of these tokens is actually, it's going down,
04:37right? I mean, the fact that it's competing at this level, is that your read on it?
04:40Yeah. I mean, I think that is one of the reasons we've seen the kind of freak out in the
04:45US over
04:46this. So there, so there's the, there's the question of the Chinese economy getting stronger,
04:51that, that creates some geopolitical competition. There's also the extent to which these AI models
04:56can be used as weapons. Stacey alluded to that a second ago, talking about security. But I think
05:01part of what's happening is these Chinese open weight models kind of break the economics that all
05:08of the big US companies have bought into. So once you build all these data centers, you're kind of
05:14committed to it. And that, and that creates a threat that could hurt a bunch of companies across the US
05:20economy. I just got to get this thing because it's like the Odyssey. They're talking Greek to me.
05:26Open weights. In artificial intelligence, open weights refers to a model whose internal
05:32mathematical parameters, its weights are made publicly available to anyone to download, run locally
05:40and customize. I just had, I've learned something today. Yeah.
05:44Dan, you get one more question into these two.
05:47Well, look, I think I look at Kimi K3, this moonshot, this Chinese model, which is supposed to be like
05:52just as powerful as the top ones. And they're pricing it like anthropics on it. Right. I mean,
05:56so talk about the economics and like what that means on a forward basis. Your guess is as good
06:00as mine, but it just seems like every year it's getting, you know, it's getting cheaper and cheaper
06:05to compete with some of the, you know, some with chat GPT and anthropics and some of them, it just
06:10seems like this is, you know, kind of happening, you know, kind of like Intel, what's Gordon's law,
06:15you know, it's happening so quickly that we really can't keep up.
06:17I mean, China's got one major advantage in this particular area, which is that it's the government
06:23can back it and the government can essentially kind of run it and back it and everything
06:28becomes aligned.
06:29You guys are doing the great American fund shortage of 2026.
06:33Oh yeah. Ben Steberman.
06:35Are you talking about Bloomberg surveillance? It's no fun when you come in here?
06:39It's so much fun. This is, this is one of the few disappearing oases of fun in the world.
06:44Are we disappearing really? Are we disappearing? No, no, you are the oasis in a desert.
06:51What does Ben Steberman say about a lack of fun in America? I think he's right.
06:56Yeah, this is the argument that you've, you're sort of the price of a bunch of goods and services
07:04has gone up so much, partly because of consumer trends, partly because of inflation, that it has
07:11become more unaffordable to have some of the sort of basic ways that Americans have socialized,
07:17going bowling, going to the movies and so on. And when you turn these things into luxury products,
07:23which has happened to some extent that, you know, puts them out of reach of a lot of people
07:27and, and you have people essentially staying home and watching Netflix, uh, you know, which
07:32obviously can be great, but it doesn't, you know, it doesn't bring you with other people.
07:36You think Max wants any fish shows this week? No, I was going to say Max already has his
07:41kids in a pre-calc as they get ready for school. They're all going to math camp in August, speaking
07:48of no fun.
Comments