00:00For that, we're joined by Bloomberg Executive Editor for Global Technology, Peter Elstrom.
00:05So, Peter, how do you read this moment?
00:07Do you think that it's a DeepSeq version 2, or do you think it's something different?
00:11Because now a lot of the tech companies kind of had an idea of what DeepSeq could do,
00:16but this is moonshot, and maybe it's something new altogether.
00:18Yeah, I think the tech companies in the U.S. in particular thought they knew
00:23where the competition was coming from in China, and this was a bit of a surprise.
00:27We had seen the DeepSeq moment last year, as you had talked about.
00:31There were some other players that were kind of prominent, including Alibaba,
00:34the e-commerce company that has a strong model, and Gwen in particular.
00:38But then over the weekend, we got this news from moonshot that they had introduced Kimi K3,
00:44which is a very powerful model.
00:46It outstrips most of the frontier models on many parameters,
00:50except for the two leading ones from Anthropic and OpenAI.
00:54So it's a very strong performer. It has 2.8 trillion parameters,
00:59which is sort of like equivalent to brain synapses.
01:01It's a very powerful model, and I think that surprised people,
01:04that there was yet another Chinese company that was able to come
01:07and compete with the best from Silicon Valley on a number of parameters
01:11at costs that people really don't understand at this point.
01:14Many of these Chinese models are also open source or open weight,
01:17so they're much less expensive.
01:19They're offered very broadly.
01:21If you don't want to run them on somebody else's computer systems,
01:24you can run them on your own.
01:25Also, you can take the model and then run it,
01:28so that can avoid some of the security concerns around Chinese models.
01:32But it is a very serious threat for some of these frontier models in the U.S.
01:36that had been planning on charging premium prices.
01:39Now there may be some pricing pressure on what they're able to charge their customers,
01:43and that may make it a little bit more difficult as we see OpenAI and Anthropic
01:47try to go public in the near future.
01:49Both of them want to get valuations that are above a trillion dollars each.
01:53That may be a little bit more challenging today.
01:55Right, and it challenges the whole CapEx thesis and the whole AI stack.
02:00Peter, Isabel made a really good point earlier when we were talking on our pre-show call,
02:03basically comparing this to BYD.
02:05I mean, BYD, you have this Chinese technology that's just as good,
02:09it is not better than other EVs in the U.S., and it's cheaper.
02:13But notably, in America, you can't buy a BYD.
02:17Do we hear any noise coming from government officials
02:19about trying to curtail the usage of some of these Chinese models?
02:23There are questions about this.
02:25I mean, you make a very good point.
02:26China has been able to compete in a number of different areas,
02:29not just electric vehicles, but also solar panels, steel, for example.
02:33So you do have some of these historic cases where, in different industries,
02:38Chinese companies have come out, they've been able to do things at much, much lower prices.
02:41Yes, that is happening in AI models right now, and you've got a whole series of them,
02:46DeepSeek, Moonshot now, there's Minimax,
02:49there are others that are coming out with these less expensive models.
02:52And there has been some discussion about whether we should try to
02:55bar those companies from competing in the U.S.
02:58I think the challenge there is that you may be hurting American companies
03:03by forcing them to use more expensive models or less capable models.
03:07There are open source models in the U.S. too,
03:09but I think when you look at the leading edge of open source right now,
03:12these less expensive models, they're really Chinese companies at this point.
03:15So if you take away a tool that companies want to use to be able to incorporate AI into their
03:21businesses,
03:22you may end up hurting American businesses that are not the AI leader.
03:25So you may perhaps help the AI frontier companies,
03:29but you may punish other companies in other parts of the economy.
03:34Peter, so interesting.
03:34Thank you for your coverage.
03:35That's Bloomberg's Peter Elstrom.
Comments