Miles Saltiel is a former investment banker and a Senior Fellow at the Adam Smith Institute.
As everyone knows, AI programs are “smart autocomplete”. They train on vast volumes of already-written material, limited by the amount of training data, computing power (”compute” in insiders’ unlovely jargon) and electrical energy to run it. Maybe this is how our own brains work. I could be bonkers, but I sometimes sense something akin to “smart autocomplete” as I formulate my sentences on the fly – or “speak”, as we usually call it. And don’t we train our own brains on precedents: “great books”, recipes, mathematical formulas and so on? In any event, Chinese and US firms compete vigorously for new tricks to do more with less.
Are the blessed things intelligent? To my mind, it’s a meaningless question. If their outputs came from humans, we would see them that way. So what? Best to use them as tools, without falling into needless anthropomorphism. Are they any good? It’s a mixed bag. In closed domains like computer programming, mathematics and, increasingly, other narrowly rules-based systems such as law, they do a job that is somewhere between workmanlike and outstanding. A few months ago, AI helped me prevail in a legal dispute, and the experience was just as everyone says. It was like working with a new postdoctoral hire: energetic, resourceful, a bit creepy, innocent of much beyond its speciality and sometimes hopelessly wrong.
Now for concerns currently in the air, together with the implications for public policy and the UK. First, two areas at the heart of the technology, where Britain is on the sidelines.
In the last couple of weeks, we’ve been getting spooked by a torrent of warnings from AI industry leaders and the odd junior employee leaving the big outfits with “end-of-days” laments, including warnings that the technology will shortly take on a life of its own. Such hyperbole is familiar to anyone with a few miles on the tyres (please excuse the self-referential pun). I began doing business in Silicon Valley before it got the name. Since Fairchild set up shop there in 1957, it has been a weird place, full of bright, driven nerds, apt – as in any hermetic culture – to “get high on their own supply”: remember the worst aspects of the grotesques in The Big Bang Theory.
In addition, their bosses are hell-bent on cementing their lead by bamboozling government into regulation to restrict competition. These guys are old hands at manipulating the media. Right now, they’re putting on the frighteners, reinforced by the scare-stories attaching to the ambitious valuations sought in, for example, Anthropic’s IPO prospectus. Hats off to all involved in the studiously orchestrated leaks, with their relentless drumbeat: “never mind the cash-burn or the revenue concentration; feel how our capex will fuel a technology so overreaching as to pose an existential threat to humanity!” What investor could resist?
The policy issue would be an industry pause, possibly coupled with international regulation. US industry is for both, while Trump and China are against. The rest of the world (including the UK) doesn’t have too many votes, and so is mouthing the platitudes to be expected from the powerless. As Clement Attlee wrote to Harold Laski, “a period of silence … would be welcome”.
We hear worries that AI could seize control of events with “agentic operations” in the real world, above and beyond the interference in data systems discussed in (7) below. This is not yet feasible and won’t be for a while. Many will have seen the amusing clip of a Chinese robot that broke Usain Bolt’s record for the hundred-metre sprint and then fell over. To put it another way, and in the words of Max Bygraves, a veteran of the pre-war music halls whom I remember from my childhood, “You gotta have hands” – but see (8) below. In principle, controlling the deployment of AI-linkable devices operating upon the real world should serve as a chokepoint, but British tech is too far from the leading edge for us to be taken seriously on this.
Now to three issues where the UK has, or ought to have, something to offer, but is held back by self-harming policies.
It is commonly acknowledged that economic disruption is in prospect, though details are hard to anticipate. At present we are most aware of the middle-level intellectual jobs at immediate risk. Policy calls for adult retraining – what I have elsewhere termed “quaternary education”. The UK has an internationally respected university sector that ought to be able to take a lead, if only its education departments could be weaned from their obsession with identity politics.
Current stock valuations give rise to concerns about capital-market corrections, if not an outright blow-up. This cannot be dismissed. AI valuations are based upon premises that contradict each other: not everyone can win. Public policy ought to be the usual prudential regulation of companies and stock issues, coupled with readiness for counter-cyclical fiscal and monetary action. None of this is helped by the permissive mood of the US regulators and the near-bankruptcy of all the G7 nations bar Canada. Another reason for the UK to get its finances in order, though fat chance with the current crew.
Industry insiders muse about granting AI programs legal personality. To my mind, this is the worst sort of “high on supply” anthropomorphism. I would be extremely cautious of departures along these lines, as we have seen how the current climate of ever-extending rights enables bad actors to hide behind the courts. I would argue that the UK should mount muscular opposition to any such thing but, if the question arises soon, I fear our leaders won’t see it my way, given local reverence for juridical folly.
Photo by Emiliano Vittoriosi on Unsplash
Finally, three fields where the UK is in an immediate position to act.
The comment about China in (2) brings to mind the risk of some sort of undesirable breakout akin to that hypothesised from Wuhan with Covid. This can’t be ruled out: we have no control over, and imperfect knowledge of, goings-on in the Middle Kingdom, though we can’t avoid the results of a fierce industrial competition that puts us to shame. The policy prescription is forensic surveillance. The US is best equipped for such action, but it is not clear how much Trump is on board. This leaves it for Britain to leverage its listening posts at comms chokepoints in Cyprus and (oops!) Diego Garcia.
There are concerns that AI programs could hack power grids, air traffic control or military systems. It makes more sense to see them as adding to the bad guys’ toolkit. The answer is better security, notably “air gaps” (that is, breaking contact with public communications) for key systems. Everyone should be doing this anyway. The UK has skills in cyber security, so the field qualifies for national focus.
To look on the bright side, several spectacular industrial breakthroughs are within reach. Anything to do with complicated shapes is grist to AI’s mill: thus, warmish superconductivity (molecular combos), coolish fusion (plasma formations), quantum engineering (subatomic probabilities), gene splicing (protein surfaces) and, yes, robotics, which takes us back to the chokepoint proposals of (2) above. All these technologies are disruptive, but in a good way, promising cheaper computation, energy and manufacture, as well as profound advances in healthcare, if not in other aspects of the human condition. The UK’s policy should be to select an area for concentration in applied AI and promote it energetically, for choice with policies pushing competition hard. This probably means pharma, though possibly something else. It need not include data centres: our hand-wringing energy regime and high costs rule them out, but that is not much of a loss, as they are destined to become commodities.
To conclude, Pandora’s box is open. Regulation is inevitable but, as things stand, it will come neither domestically nor internationally on Trump’s watch. We can only hope that his military is keeping a beady eye on China. The UK and the other “middle powers” are all too often bystanders, their appetite for lippiness and unilateral regulation rendering them further marginalised. The policy prescription for Britain is silence where we have nothing to offer: international regulation, an industry pause and chokepoints for agentic operations in the real world. Instead, we should concentrate on the areas where we have (or ought to have) a part to play: opposing the grant of legal personality to inanimate technology, surveillance of China via our listening posts, quaternary education, cyber security, restored financial resilience and focused industrial applications. This calls for the country to raise its game in education, law and macroeconomics, not to mention reversing Starmer’s student-politics scheme for Diego Garcia.