Sir Tony’s latest essay is as empty as one would expect an article not titled in sentence case to be.

The problem with Sir Tony is that he knows very little about India, artificial intelligence or technology (to take three topics about which I do know something), but pretends otherwise. This is how he manages to avoid saying anything of interest in 5,600 words (despite his best efforts). Unfortunately, our political class, its nominal ideological diversity notwithstanding, is uniformly woke and politically correct in the way that matters, and therefore afraid to call anybody (other than immigrants with the wrong skin colour) thick. (Burnham comes closest: ‘[s]aying “AI” is not the same as having a plan’.) Thus we find Tim Shipman praising the ‘quality of analysis’ of the piece; Polanski says, in his usually inarticulate way, that he thinks ‘there are positive things. And it’s so important to put that nuance in there’. Even an intelligent Unherd columnist calls Blair the ‘adult in the room’; as leader, he ‘would have told the party…to turn the bulk of their attention to AI’.

Here’s Sir Tony on India.

The first epochal change is in the geopolitical order where America’s superpower status is now shared by China, in time to be joined by India. A sort of G2/3. These countries will be far ahead of whichever nation is in fourth place. By this calculation, everyone else including Britain is a middle power.

This is a vapid throwaway line mostly intended to sound profound rather than an indication of serious thought about the extent to which India will influence the world order.

What makes Sir Tony so confident that China will ‘in time…be joined by India’? Does he know who Arun Jaitley is? Does he know anything about demonetisation? When was the last time he looked at the progress of the Ahmedabad–Mumbai high speed rail line? When was the last time he looked up the average speed on India’s dedicated freight corridors? How many chief ministers can he name? (You, the reader, probably know nothing of this either, but even if you write Blairite screeds, people probably ignore them.)

Why mention India in the first place? Does it matter whether the G3 turns out to be a G2.5 or G2.2? He doesn’t say.

Sir Tony, like many other credulous middle-aged men, also feels the need to talk about AI despite his not knowing anything about it.

The problem with Sir Tony is that, ironically, he displays the sense of victimhood of a lazy postcolonial despot or unthinking university leftist. (I would prefer to say that lazy postcolonial despots take after Blair, but unfortunately the chronology doesn’t work.) This is the worldview of a drunk Iranian who blames a Franco–British conspiracy for the revolution or a Malian general who carefully explains how it’s actually the fault of the French that they succeeded in inviting in the only middle power that somehow would succeed in committing even worse human rights violations in a few short years. Sir Tony here has the mind of a vulgar cantonment drunk, if not the potbelly or fake medals. He is the sort of man who would have asked Mohammed Hanif who his sources for his satirical novel, A case of exploding mangoes, were.

The funny thing is, after the book came out, a lot of people — and some of them were heads of intelligence agencies — I've run into them at a party or at a social gathering, and they take me into a corner and say, 'Son, you've written a brilliant novel. Now tell me, who's your source?' I used to find it a bit scary at the beginning that, my God, these people are running my country and they actually believe all the lies that I've written.

In Sir Tony’s case, this mindset leads him to declare that AI will uproot everything, but we have no control over how.

There is no point in debating whether this technological revolution is a good or bad thing. Just know it is a ‘thing’.

‘Self-stultifying’ is a kind description. The unfolding of this ‘technological revolution’ is not yet set in stone. It may be pointless to debate whether the average of the set of paths before us is good or bad, but it is hardly pointless to debate the merits of each one.

This probably explains Sir Tony’s rather manichean view of technology regulation as bad and deregulation as good.

Technology policy is the critical factor. If European policy continues in the direction of addressing the dangers rather than seizing the opportunities, i.e. weights technology regulation against the technology sector, it will be impossible for Britain to go back fully into the European Union. We can’t argue that technological innovation and adoption is the key challenge of modern governance and tie ourselves to a technology environment essentially hostile to it.

It’s not hard to intelligently criticise European regulators for being extremely stupid. Sir Tony somehow manages to unintelligently criticise them. He has nothing to say about the space of actual AI regulations, because he knows nothing of that space in the first place. Does Sir Tony really think that, for instance, mandatory reporting requirements would seriously undermine our competitivity? Why? Has he read any legislation in this connexion?

Another nostrum:

[G]overning in the age of AI will be the principal challenge. And opportunity.

Some more vice-signalling on energy:

We must prioritise cheaper energy and electrification over net zero and use what is left of our North Sea oil and gas resources. This is essential for our competitiveness and for taking advantage of AI.

Where are the numbers? What does Sir Tony think of CfDs? What about nuclear power? Has he ever thought about storage? Or the grid? How exactly does he want to turn gas in the North Sea into lower strike prices? There is an obvious but very expensive and un-Blairite answer, viz. to nationalise the whole thing and segragate it from the global gas markets. Is that what Blair wants? Does he think we have the state capacity to avoid buggering it up? We don’t know. Much easier just to gesture at these points, as if Sir Tony has more meaningful thoughts about them than Burnham or Streeting. If he’s right, that’s by accident.

Another point is that it’s not even obvious that we should naïvely focus on hardware in the first place. The ‘extreme centre’ would have limited financial and political capital. Would it be a better use of that capital to double the number of GPUs in Britain or double the number of researchers from frontier labs? Note these decisions are orthogonal; researchers don’t need to sit next to GPUs to do things. Of course, we should expect this from someone who focuses on banalities like immigration policy for low-skilled workers. If AI is as important as Sir Tony thinks it is, the only important question about immigration policy is the number of researchers from frontier labs we attract. Alas, Sir Tony is easily distracted by brown people, and this time he can only deport rather than bomb them.

We should create a major new partnership with the private and voluntary sectors for apprenticeships and training – not just for the young and unemployed, but for the existing workforce whose jobs will be affected by AI and who need to learn AI adoption.

More nostrums. Replace ‘AI’ with any other important technology (‘the silicon chip’, ‘electrification’, ‘mass literacy’, and so on): the sentence loses none of its (negligible) incisiveness or force.

All governments for the foreseeable future will govern in the age of AI. Those which understand it will see their countries prosper; those which don’t, won’t. This is literally the challenge across all sectors including welfare and health (digital ID is just one, though vital, part of it).

This, at least, is right. It would also help to have some preliminary thoughts about how to avoid a scandal orders of magnitude worse than Horizon. But has Sir Tony ever read a single interpretability paper? Does he even know what interpretability is? What does he think of the government’s consultation on the common law presumption that computer systems operate correctly? (If he thinks that it’s woke sentimentalism to worry about making subpostmasters commit suicide and the only path to economic growth is to be scammed by software contractors, he should say so.) Alas, Sir Tony is not the man. It seems that invading Iraq wasn’t a sufficient education to dispassionately examine the precipitous and horrifyingly grave choices before us. Perhaps we should be thankful that he doesn’t ultimately advance any serious proposals on either India or AI.

What do I think? Well, it would help not to be credulously naïve China hawks.

Our earlier conclusions (A model of computation) might be summarised as making the following conjecture: physically plausible models of computation do not differ in the problems they can solve in polynomial time.

What has this to do with language? To a certain extent, the argument must be reconstructed, but its rough outlines are clear.

A first premiss is the so-called Cobham–Edmonds thesis.

The Cobham–Edmonds thesis is only meaningful relative to a model of computation. But, if the van Emde Boas invariance thesis is correct, any reasonable model of computation will yield the same class of feasibly computable functions.

I shall, at this point, make a

The Cobham–Edmonds thesis might seem to admit obvious counterexamples. For instance, an algorithm that takes $\mathrm{\Theta }(n^{2^{100}})$ time seems obviously infeasible. (For that runtime to be meaningful, it would have to be associated with some privileged subset of the ‘reasonable’ machine-class above; a good example is e.g. physically instantiated computers.) Less obviously, there is some reason to suspect that some problems that likely have no polynomial-time algorithm are feasible. In this case, we might nevertheless hold that, most of the time, the Cobham–Edmonds thesis is correct. For instance, there are surprisingly few ‘galactic’ algorithms, i.e. algorithms with polynomial asymptotic runtime that in practice are unusable; see e.g. Helfgott (Isomorphismes de graphes en temps quasi-polynomial (d'après Babai et Luks, Weisfeiler-Leman...)), Helfgott (Isomorphismes), and Lipton and Regan (David Johnson). We can therefore provisionally take the Cobham–Edmonds thesis to be a (defeasible) heuristic in the sense above.

What has this to do with language or cognition? There are, broadly, I think, two sorts of argument that relate asymptotic analysis to language.

For the sake of argument, I will assume that these arguments are right, and that, therefore, ‘language computations’ are amenable to asymptotic analysis.

Is the application of asymptotic analysis to cognition and language an idealisation? First, I shall give an example of computational analysis that plausibly is not an idealisation.

Let us suppose, with Searle (Rediscovery: 200), that

[T]here [is] some description of the brain such that under that description you could do a computational simulation of the operations of the brain…given Church's thesis that anything that can be given a precise enough characterization as a set of steps can be simulated on a digital computer…in the same sense in which weather systems, the behavior of the New York stock market, or the pattern of airline flights over Latin American can.

Turing (Computability: § 11) showed that no Turing machine decides the Entscheidungsproblem. If we follow Searle’s rendering of Church’s thesis, it follows that the brain does not solve the Entscheidungsproblem either. I suggest there is no obvious idealisation here. (Of course, the argument might be false; but on its intended reading, it should be read as approximately false in the way idealisations usually are.) In other words, a fairly natural computationalist view is that these results from computability theory apply straightforwardly to computation generally, and, therefore, to cognition—and so to language computations.

However, in a considerable number of the arguments to follow, some antecedent assumptions seem to be false, and, therefore, the arguments will have to be parsed as idealisations. To say that the antecedent assumptions are false is not to deny computationalism. Suppose we are given a learning task over some stream of data $x_1,x_2,x_3,...$ (e.g. linguistic input). It may turn out that the probability distributions from which $x_1$ and $x_2$, for instance, are drawn are not independent. For simplicity, however we may assume that $x_1,...$ are i.i.d. variables; making this assumption, we may find some asymptotic bound on the learning task. There is nothing inherently objectionabfle about the i.i.d. assumption, but it must be assessed as an idealised antecedent assumption, in addition to the computational model of language computations supposed.

I have translated an interesting piece from FoST France–Europe.

A friend recently asked what Chinese leaders are thinking about the Iran war, and noted their ‘message discipline’. My initial response was as follows: first, they are probably just confused; second, you don’t need much message discipline when the system is clamming up, because people don’t generally go out of their way to start posting; and, third, we should expect them to be confused due to a deep dysfunction in the transmission of information in the MFA and allied institutions (cf Taiwan, where all the information required is transmitted, but it’s simply not understood, because that would require understanding that sometimes Asian liberals aren’t merely CIA plants but might organically like not sliding towards North Korea).

My evidence for this general claim involves a close examination of Chinese policy in Burma and Pakistan. Eventually this will be written up here. For now, here’s an article from The Diplomat. You can fill in the blanks whilst I procrastinate.

David Jones infers from Pahlavi père’s despotism that Pahlavi fils ‘can’t save Iran’. Jones praises his rival, Maryam Rajavi, leader of the Mojahedin-e-Khalq (MeK). Like Pahlavi fils, Rajavi makes all the right noises, but has never governed. Unlike Pahlavi fils, but like Pahlavi père, she has held enough power to have a track record—of abusing members of the MeK through ‘prolonged incommunicado and solitary confinement, beatings, verbal and psychological abuse, coerced confessions, threats of execution, and torture that in two cases lead to death’ (Human Rights Watch). Yes, prospective dynasts are politically answerable for their fathers’ crimes. And Rajavi is personally answerable for her own.

This is very good (modulo gratuitous capitalisation).

‘There are now more Chinese nationals than British nationals studying physics, engineering, chemistry, mathematics or computer science at a postgraduate level at Britain’s top five universities for those subjects. In engineering, there are roughly twice as many Chinese nationals as British at that elite level.’

This is not.

Michael Jinghan Zeng’s fairly good FT Chinese piece, translated Yuxuan Jia at Pekingnology, expounds most of Samuel’s article’s problems fairly accurately. I propose to make these points more explicitly.

I make the following painfully dull assumptions (each of which should be read ceteris paribus for a reasonable range of starting values).

Samuel probably holds these (in my view correct) views for reasons I’d consider to be some combination of unenlightened and unsophisticated. Since I am lazy and cowardly, I shall only gesture at my obviously sounder reasoning, and decline for the moment to elaborate; but by way of explanation rather than justification, I permit myself to remark that I am a mostly deracinated évolué whose parents and grandparents did much of the évolution (see aside below); I preëmptively disclaim any responsibility for attendant misinterpretations.

Any sensible question to ask is of the form: what is the marginal effect of cutting the number of student visas available to PRC nationals—

1. by some given proportion $P$, and
2. holding fixed other parameters (concerning policy, the macroeconomic situation, etc.)?

For instance, given current policy in all other areas and some plausible medium-term economic forecast, what is the marginal effect of cutting the number of visas by one?

I have posed the question thus because Samuel does not particularly clearly distinguish the following two claims.

1. There are too few British STEM students.
2. There are too many students who are PRC nationals.

Obviously one can consistently hold both views. It seems Samuel takes each to partially strengthen the case for the other. This is obviously wrong. Zeng hints at the problem: ‘If the number of Chinese students were to decrease, it would not be the “Chinese places” that shrink first, but rather the overall size of the disciplines and the research capacity.’ To put it more explicitly: there is no overall fixed total number of students; there is no guarantee that when we throw out a Chinese trinmo, a Briton will replace them. It is equally possible that 0.1 British trinmos replace them and 0.9 places disappear, or some other combination thereof.

Zeng’s point is pretty simple: the presence of intelligent PRC nationals (both students and researchers) generally increases the attractiveness of Britain’s universities, and so usually works to the advantage of British students. That is to say, making the same assumptions above—

—decreasing the number of student visas available to PRC nationals would, with respect to 3 (growing the British economy) probably be a bad idea. Nor would it free up funds to encourage the study by young Britons of technically important subjects, because they financially contribute to the system. So this is not very useful.

What about the other points? I don’t think that British universities should conduct joint research with Chinese universities that seems more likely to be of military use to them than to us, for instance. And § 3 of the UKCT article mentions one such example. However, it also notes that it is possible to block research collaboration independently of visa policy, and that obviously it would be helpful not to utterly stupidly hobble investigations by e.g. only recording Pinyin names.

Samuel appears only to have read the first part of § 3, rather than its conclusion or § 4. In UKCT’s words: ‘those trying to devise policy approaches to support the public or national interest ought to think more creatively and constructively about how to address the concerns behind the ‘research security’ paradigm without resorting to “securitisation” per se.’ Although it is obvious that foreign students historically have played an important rôle in China’s technological development, it is far from clear that the marginal student at Imperial rather than Tsinghua will really make it any easier to invade Taiwan, given China’s recent development (§ 4).

Samuel blithely overlooks this point: ‘The problem is that hosting thousands of Chinese engineers is very much not the same as getting in bulk batches of Canadians or Germans. The Chinese state has an official policy of “military-civil fusion” whereby all civil technology is put at the disposal of defence and security needs as well as technological espionage and economic coercion.’

Samuel would obviously right to say that, in gross terms, admitting Chinese students will involve training some future military officials who might invade Taiwan in potentially useful skills. But this is the wrong question to ask; the policy question with respect to objectives (1) and (2) is: what difference does it make that they are here as undergraduates rather than in China?

The difference it doesn’t make is that they will be any better at e.g. coming up with stealth materials for fighter jets to invade Taiwan. We (that is, the West), have missed that bus. Samuel’s argument might have made some sense in 1996, but it hardly does now.

It is more instructive to ask why there are Chinese students in Britain in the first place. UKCT: ‘General push and pull factors are likely to remain supreme and include (but are not limited to) job opportunities, public safety, politics, family proximity, and all the other elements of quality of life, doubtless down to weather and food for some’. In other words, civil-military fusion ought to be on our minds in examining e.g. Sino–British research collaboration on militarily sensitive technologies, but is much less relevant in considering the marginal effect of kicking out the generic PRC national Imperial maths student.

Here is a conjecture: they are here because we are still winning in one respect, and one respect only. We are winning because we are still, broadly, democratically governed liberal polities, in which the penalties for (e.g.) speech (of course too high in Britain) are far more impartially and gently administered than in China. We are winning because we are far further away from turning into West Korea. We are winning because the voices of April can whisper in reading groups in the West.

I don’t have any particularly sound argument for this conjecture. Some reasons for hope in this connexion are that:

1. informal conversation with mainland Chinese students suggests that the Shanghai lockdown severely radicalised many overseas students against the government;
2. overseas dissident activity saw at least a brief uptick afterwards (e.g. at Tiananmen vigils, protests outside the embassy, and so on); and
3. at least for a while, there was a meaningful effort to start reading groups and other forms of overseas dissident activism, which I think are now sustainably at a higher level than before.

The marginal effect of reducing the number of visas available to PRC nationals, I therefore conjecture, is something like this.

1. It vastly diminishes the opportunity for political exploration that generally leads away from the ruling orthodoxy in China.
2. It cuts off a useful source of human capital for own own economies, using which need not (given a sensible understanding of civil-military fusion) meaningfully be to the advantage of Chinese military power.
3. At the same time, it cuts off a useful source of human capital for the Chinese state, which does meaningfully erode Chinese military power.
4. In the long term, it will diminish the community of people who might indigenise foreign ideologies other than Stalinism in the service of the world’s second-most populous country, and therefore make any prospect of bettering Chinese society when the question of the succession arises dimmer.

One obvious rejoinder is that transnational repression makes overseas dissident activity very difficult. A first point is that this is at most will mitigate rather than reverse the effects of such activity; the Chinese state has no means of making people more loyal abroad than they would be at home. A second point is that, in the worst case scenario, and without a response, the primary victims of transnational repression are mainland students themselves, who are usually better off here than at home. But this is right: it is hard for mainland students to know whom they can trust, and to have open discussions. The measures I have seen to stop such effects will certainly have some effect, but it is unclear how much.

So a more interesting question than the visa question is how host countries can best facilitate political and academic freedom amongst overseas Chinese students; that, at any rate, is my response to UKCT’s invitation to avoid naïve securitisation, with a view to taking China seriously rather than credulously.

The short answer: it’s complicated.

Jurisprudence quiz: who said this? ‘The Constitution is the mother law for all laws. So, I’d like to note we all need to abide by the Constitution. If one does not follow the laws, such laws must be revoked. I mean if it is the Constitution, it is necessary to revoke the Constitution. If one does not follow the law, the Constitution must be revoked.’

Quite obviously, not the brightest chappy about; more specifically, Senior General Min Aung Hlaing, who took three attempts to be admitted to the Defence Services Academy (Mclaughlin and Webb, Senior general emerges), has generals detained (Irrawaddy, Myanmar Junta Detains Generals Who Surrendered to Resistance in Laukkai) and executed for following his orders to surrender (Irrawaddy, Defeated Myanmar Junta Generals Given Death Sentences), and is notoriously superstitious (Irrawaddy, Rituals and Yadaya; Kavi, Behind the Boasts, Myanmar’s Junta Boss Is a Superstitious Mediocrity).

This, in addition to the junta’s inability to decisively crush even ex-urban bourgeois armed with, er, hunting rifles (Fishbein and Vahpual, ‘Our only option’: Myanmar civilians take up arms for democracy), has created the impression that the Sit-Tat does not really know what it’s doing.

It was with that assumption that I began my perusal an exposé of a supposedly régime-linked group of media outlets controlled by the Swe family (Win, The shadowy past—and present—of the Myanmar Times and Frontier). Before reading it, I sent it to two friends who know something of Burma, both of whom concurred with my initial assessment: surely they couldn’t be that clever?

To my perhaps unwarranted surprise, Win wasn’t making much up. One has to scroll to the section titled ‘Happy stories’ for the salient evidence, but the gist of it is that military intelligence was, by organisational design, able to censor articles to its desires. Unless these documents were made up, and Frontier’s response (Kean, The other side) makes no such suggestion, Win is clearly right.

In 2004, some years after the establishment of the first Swe-aligned paper, the Myanmar Times, the junta arrested and jailed Swe fils. One vindication of the Swes would be if they were to have broken with the junta. But such ‘assurances were called into question two years later when Thein Swe, now 80, and former junta spokesperson Hla Min toured China on behalf of the Paragon Institute—a secretive think-tank close to the junta. During the trip he reportedly discussed Myanmar’s “national security” and the deployment of private Chinese security firms, preparing the ground for Min Aung Hlaing’s first visit as junta leader a week later.’

This, however, leaves a question: what’s the point of running what is, by all accounts, a fairly impressive media outlet (qua régime critic), in the form of Frontier Myanmar, if the Swes really are aligned with the junta?

One explanation is that Swe fils has broken with his father to the extent he can. I should like to believe this, because I should quite like to imagine I’d be such a person. Unfortunately, I don’t.

Another is that the Swes were a relatively moderate or liberal faction in the régime. Maggie, then, would not have been wrong, just before her time: ‘I think there are probably two parts to the Khmer Rouge: there are those who supported Pol Pot and then there is a much more reasonable grouping within that title “Khmer Rouge”’ (3:40). This doesn’t strike me as an especially plausible proposition, given the Swes’ continued alignment with the junta after the coup.

My working explanation now is that the Swes are an unusually sophisticated faction of the régime. There is no point in regarding them as more ‘liberal’ ideologically, but they are certainly more sophisticated (e.g. Swe père appears to be competent enough at English in the leaked documents). The reading that follows is that Frontier was indeed a bona fide opposition-aligned outlet, but it was also an insurance policy. And if you were relatively intelligent and on a ‘dominant’ side incapable of decisively defeating urban bourgeois with homemade rifles, you might also want such an insurance policy.

Gaby Hinsliff (Dubai) asks who might fail to have observed that ‘living a few hundred miles as the drone flies from Tehran might have risks’. Topping my list are British columnists innocent of any salient expertise and even mere cartography, especially those who seemingly are unaware that Copenhagen is closer to St Petersburg (713 miles) than Dubai is to Tehran (767 miles), fail to wonder whether Iran might fire drones from the side of the country next to Dubai, and probably could not find Bandar Abbas (162 miles away) on a map.