AI Nationalism is a topic that has been around as long as we have been hearing talk about how the US needs to be beat China to AGI. But an op-ed in the NY Times brought some issues together in an interesting way for me (not that I agree with the author).

As I write this though, I want to remind myself of Lee Vinsel’s concept of criti-hype, which I have encountered through Doctorow. As Doctorow puts it, we don’t need to run around arguing about Mark Zuckerberg’s “mind control ray,” when it doesn’t exist as such. Put more broadly, in our critical awareness, we need to be careful of adopting the narratives and terms that we mean to analyze.

The basic criti-hype concern in AI nationalism lies in slippages that occur in where the nationalism lies. If we want to argue that the US needs its own AI, for example, then what does that imply about the “always already” ideological process/content of AI? In turn, what does that say about the knowledge produced with these AI?

It’s almost as if we are suggesting that knowledge is shaped by the tools we use and the cultures which designed those tools and in which we use them. Imagine that all ye veterans of the Science Wars? Unfortunately I think it is mostly criti-hype.

In “The Hidden Cost of China’s Free AI,” Tal Feldman is responding in part to the recent release of models from China, but his argument has a longer foundation in the concept of AI nationalism. As he observes, China’s government censors its AI models. As such, he feels confident in describing China’s AI as ideological. And presumably so are American models. Feldman asserts that “China is winning a vital soft-power competition by providing transformative A.I. to the world,” and the problem with that is that it isn’t the US winning that vital soft-power competition. Fair enough. Those politics are as old as a James Bond film. And I don’t deny they exist, both the politics and the films.

More interesting, I think, is how AI nationalism has become a central trope in frontier AI discourse. Politicians and candidates talk about sovereign funds and public investment. AI systems are integrated into national military and security operations. Between the frontier AI corporations and the state there is certainly some shared nationalist agenda. I don’t think anyone wants the corporations running off on their own.

But there is an elision, a slippage, from believing that having the most powerful AI is in our national interest to asserting that a nation’s AI will embody its ideology (i.e., ethics, behavior norms, perspectives, assumptions, etc.) and promulgate it to the world. Do American frontier AI corporations bring “American intelligence and reasoning” to the world?

I admit that it does sound a little sci-fi, a little criti-hype perhaps, to raise these concerns. Feldman writes, “Imagine an employee inside an American company who was smart and hard-working and whose goal in life was to promote the interests of a foreign government. We would call that person a foreign agent. We are now hiring millions of them through the use of Chinese A.I.” That is strictly criti-hype in my view. He is importing a significant amount of agency to the Chinese AI bots. We have certainly seen AI agents go off the rails and do unexpected things. I don’t think we’ve seen them be James Bond.

As I understand it these current models remain incapable of alignment or reliability at a level that would make them economically valuable at scale. Or at least I haven’t seen any evidence that they are. I have no doubt that Chinese AI models are different from American or other models. That doesn’t mean that the Chinese AI can align them more powerfully than the Americans can. It also seems likely that the US government will remain interested in using an American-built AI system and keeping those systems “competitive” (whatever that story is). The same will likely be true for China. (I mean unless this whole speculative adventure proves to be ill-advised.) If AI proves important to national security, then other nations (or blocs like the EU) will want their own AIs as well. What if AI becomes another state-level adversarial technology? Does the AI arms race bankrupt the US? Of course, I’m just writing criti-hype science fiction with that line of thinking, imagining that pursuing this technology is an existential matter.

For a different, maybe less inflated story, what if AI turns out to be a commodity, more like energy generation? It won’t become this all-encompassing “being.” Instead of being a unique discovery, it will become more localized and differentiated. It won’t matter that your AI came from America any more than it matters that the gas in your car came from Saudi Arabia, which is to say that it matters on a geopolitical scale but not to your car. It isn’t forcing your car to obey Sharia law.

This is where we can go back to the Science Wars story though. Maybe it is criti-hype to imagine that ChatGPT is an American AI machine. But it isn’t imaginary to recognize that frontier AI are tools that have been shaped by national, economic, and cultural interests. And it matters. It doesn’t matter if a microscope or a test tube is made in China, but it does matter if your AI is. That indicates very clearly that scientific output is always already ideological.