The President Wants US to Dominate AI – That is Highly Unlikely to Happen
China has pretty much caught up with us. With the president hobbling frontier AI by claiming national security, that is just making things worse.
Last week Chinese company Moonshot released Kimi K3. While not on par with the best America has to offer, it is damn close. Close enough for most purposes. Unlike OpenAI and Anthropic, which have to answer to investors, Moonshot is allowing anyone to download their models. Even if the US attempts to shut that down, all that will do is put America even further behind and it will cause people in other countries to look at it even more closely. All of this causes concern around the business models of the big US LLMs.
Given that both OpenAI and Anthropic are on the verge of IPOs, the news from China is not good for them. The president hoped there would be a moat around US AI software. That is not the case and it is very unlikely that he can fix that. Ever.
Gary Marcus, an AI blogger, made these predictions about 18 months ago:
- The US would not achieve a decisive victory in AI over China if we continued to focus largely on LLMs
- That we would converge on a tie
- Companies like OpenAI would remain unprofitable (note that Google, Microsoft and Meta have alternate revenue streams)
- Nvidia would become vulnerable (we are already seeing that)
- The CHIPS act would backfire and not stop China (it is not)
- Models will get better but will still hallucinate
- Racing endlessly around LLMs would sap resources that could allow us to win the race
Look at the list and see how many of these are already true.
Part of the problem is the government’s love affair with Silicon Valley. Since the Valley has been right before, the government assumed they would be right this time.
What could we possibly do now?
- Let companies like OpenAI and Anthropic sink or swim on their own; don’t bail them out.
- Outlaw open source (good luck on that)
- Try to regulate a moat
- Bail out the big AI companies (the president seems to like the communist model of the government owning the means of production which is why he has bought ownership stakes in a number of US companies already, so this is not so far fetched as it seems. He has bought a stake in US Steel, MP Materials, Intel, Lithium Americas and others)
- Ban Chinese models altogether (see item 2 above)
- Buy companies like AI Labs, OpenAI and Anthropic completely. Maybe, Gary suggests, for two cents on the dollar. (see item 4 above)
- Give up on the winning the AI war and move towards a global good (China is making noise in this direction and SURELY we can trust China (NOT!))
Check out the link at the end for more details on these options.
Bottom line is that US industry is going to face ongoing steep competition from China and others. That doesn’t mean we are going to lose, but it does mean the big guys should probably rethink their business models.
While OpenAI and Anthropic are heavily dependent on their LLMs, other companies like Microsoft and Google are not, so the AI industry is not going to disappear. But, some companies may, ultimately, be in for hard times.
Find more details here:
by 