البيان رقم صفر

Prolegomenon to a strategy to become the third AI power

1,221 words

Some time ago, I wrote a rather irritated screed about Tony Blair.1 You can find it if you like, but, in this note, I shall attempt to be more serious.

Note
1
Yes, if you are reading this footnote, you have probably worked out how to find it; and, yes, I know that this post contains no numbers, but at least there is a sketch of where you could put them in to make the argument reasonably rigorous.

Here is a sketch of an ideological divide.

  • Leftists are denialists about AI who do not understand that their denialism will leave Britain impotent and impoverished. They therefore have luxury beliefs—most notably, opposition to data centres—and an energy policy that accordingly rejects e.g. construction of new gas plants to reduce costs.
  • Pragmatists (Blairites, Tory wets, venture capitalists and so on) understand, and therefore will build lots of gas power stations to power data centres. We should probably stay out of the European Union too. Something good will come from this.

I may later bother to substantiate the claim that anybody believes in this sketch, but I think enough people reading this will either agree with it or know others who agree with it for me to proceed on that assumption.2 It’s also very convenient. Would it be a surprise if leftists were to ignore an important technology to meet its climate goals? Probably not. However, since I spiritually belong to Labour’s soft left, but I also have a brain, I’d like to come up with a fairy-tale to explain why I don’t actually think building five billion gas peakers is a good idea.

Note
2
I had intended to write this post once I’d seen Phillips, The Manchester Consensus.

The basic claim I want to make is that the marginal penny (or equivalent in political currency) is most effective spent on people rather than hardware, and that energy policy has very little to do with our ultimate fate.

I first want to address the obvious worry: security. Chinese researchers abroad can be pressured or induced to steal things. Admittedly, so can British nationals, but Chinese researchers are more susceptible to such threats, on average. We know that China is trying to do this in certain sensitive industries (UK-China Transparency, Behind MI5’s warning to British universities). There are two very natural ways of circumventing this problem. The first is to securitise only in industries where there is something to steal. Fine, we are probably ahead on whichever bits of the F-35 we make, assuming the parts lost in Hong Kong weren’t that important (Bloomberg, 7 October 2026). Securitise that bit. Are we ahead on open models? Well, no; I don’t even know what the best British open model is or when the last time someone made one was. Second, liberalise industries where stealing is pointless. In competing on open weights, there is no question of theft. Helpfully, China is very far ahead of us.

But why think this strategy would work? The argument that we needn’t worry about security only applies because we are behind. Conveniently, this does not apply to the people with whom we are trying to catch up: the Americans. With fairly elementary techniques (such as actually recording the Unicode rather than merely the Hanyu Pinyin of Chinese applicants’ names) we can make Britain considerably less intimidating as a destination than the US without meaningfully compromising our security.3 Since the Americans are ahead, they cannot sensibly compete with us in being a welcoming destination (well, it would be silly for them to do this). When you are catching up, you can use the whole world as a talent pool; but they are not catching up. There is good evidence that a freer and less securitised environment here would encourage many of these researchers to leave the US (Xie et al., ‘Caught in the crossfire’). Some other suggestions:

Note
3
A very cynical suggestion is that we should cause a moral panic in the US about Chinese researchers, to make them come here.
  1. anecdotally, the Home Office could be faster, so I shall vaguely gesture at some sort of fast track that works;
  2. if we really want to spend lots of money, instead of spending it on data centres, we could match US academic salaries;4
    Note
    4
    For instance, if you know where to look, the new labs UKRI is funding are attracting attention partly because their salaries for postdocs are a complete joke.
  3. there are probably clever tax cuts or other incentives we could find to make it financially easier to found start-ups; and
  4. if we do all these things, we can certainly advertise this fact.

Suppose, then, that we have the people to host a British DeepSeek or Z.AI. Would this be preferable to having more hardware? I think so. This is simply a point about comparative advantage. It is probably easier to build and power large datacentres in, say, Spain or Australia than in Britain. It is helpful that these countries are extremely unlikely to even aspire to catch up in the same way that we—a middle power of the first rank—might. It is easier to convert a British DeepSeek into a mutually beneficial deal with such countries than it is to fix our own planning problems; at least, that is a more efficient use of financial and political capital.

Since writing the first draft of this, I have had time to consider a few objections.

  1. Would China let them leave? First, I’m not actually sure that it’s obvious China would mind. Internal competition in the Western camp might be viewed as a good thing. Moreover, even if we succeed in convincing ourselves that there’s no security risk, China might still view this as an espionage opportunity. Second, there are lots of Chinese nationals in the US. If the Democrats remain as hawkish as they were under Biden, a significant opportunity will remain. Finally, even in the case that China would eventually start to worry, it would only do so after a significant exodus, which could well be enough to make a good start.
  2. Would the Americans let them leave? I don’t think the Americans have the imagination to set up a hybrid of шарашки and internment camps.
  3. How would this actually help Britain? Would we control the technology? The most optimistic case is that those who leave would become British and tie their fates to ours. This could happen to a surprising extent—the medium-term prospects for political change in China are not good—but let us suppose it doesn’t. It is still useful that they are here and not in China. Labs in Britain hire far more British people than labs in China. They generally improve the quality of the local ecosystem: their alumni will found new startups, they will attract more capital to Britain generally, and they may fund university research. More generally, it is almost never a bad idea for British universities to contain more intelligent people. These benefit can accrue with fairly minimal requirements e.g. on hiring British citizens—even if many of those who leave China to stay in Britain stand in the same relation to Britain as they might e.g. as expatriates in Dubai. (Indeed, Dubai is perhaps a singularly amusing example of a place that succeeds in attracting human capital with very little effect on the local population’s skills.)

References

Capaccio, Anthony, ‘UPS Worker Missed Security Email, Letting China Get F-35 Parts’, Bloomberg (2026).

Phillips, James W., The Manchester Consensus: The Manchester Consensus: Why the UK must ‘look up’ on AI, James W. Phillips’ Newsletter, 2026.

UK-China Transparency, Behind MI5’s warning to British universities, UK-China Transparency, 2026.

Xie, Yu, Xihong Lin et al., ‘Caught in the crossfire’: ‘Caught in the crossfire: Fears of Chinese–American scientists’, Proceedings of the National Academy of Sciences 120.27 (2023), p e2216248120.

AI • Britain • China