Prime Minister Rishi Sunak hosts weekly Cabinet meeting Prime Minister Rishi Sunak hosts weekly Cabinet meeting, 31 October. Photo: Number 10 / Flickr / CC BY-NC-ND 2.0 DEED

Wednesday’s AI summit shows a failing government falling for the latest hyped tech bubble, without any real grasp of dangers and limitations, argues Kevin Crane

Few things are more unedifying than a stupid person working under the delusion that he’s the smartest guy in the room. Rishi Sunak may be in the running as the smartest person at his much-hyped AI conference this week, but only because he’s been struggling to persuade other world leaders to bother showing up.

Historic Bletchley Park, the site of significant advances in computing during World War Two, is the venue Sunak has selected for his ‘AI Safety Summitt’, a hastily thrown-together event that he hopes the leaders of the world’s largest economies will attend. He has already been given short-shrift by France’s President Macron, he’s only getting US vice-president Kamala Harris and it is not clear yet who if anyone, from China – the country that produces the bulk of the hardware upon which AI actually runs – will bother to show up at all. If the turnout is poor, Sunak can presumably blame the international crises around the Israeli invasion of Gaza, but the reality is that he and his government have done a poor job of explaining what they think this event is actually for.

As with any discussion of AI, the problem starts with trying to work out what you are precisely discussing, which isn’t often easy even for people with a solid grasp of computing and software. It’s basically impossible for Sunak, a man who likes to posture about the importance of science and technology, but is utterly unqualified to discuss either. His public statements about AI have been the usual garbled mess of uncritical hype from industry press-releases and wide-eyed science-fiction doom-mongering that the world has been getting bombarded with since January this year.

Beyond a very vague mission statement of ‘maximising the benefits of AI while minimising the risks’, the government goal seems to be to position Britain as a world authority on the regulation of AI technologies, including speculative ones that don’t exist and haven’t been proven to work. One does have to wonder, even if that were a sensible objective, why would other world leaders seek to help Britain accomplish this before themselves or anyone else?

Dangers and distractions

Just like the definition of AI, what the benefits and risks that AI pose really are frustratingly hard to define, and often very subjective. The government has, at various times, listed job loses, deepfake video, biological warfare and machines becoming self-aware as the main things to be worried about, a combination of reasonable, less reasonable and downright silly suggestions.

Jobs are a question that needs to be taken seriously. Greater automation of jobs is unquestionably happening and is reshaping the employment market, but it is a most gross oversimplification to simply say that greater automation is just going to wipe out jobs. As we have seen with the Hollywood writers’ strike, which ended in the workers getting a decent victory, studio bosses ultimately couldn’t just dispense with human labour in the writing process.

It was the solid organisation of those writers as workers that protected their positions, though, not government regulation. We can be sure that the writers would have been facing some bleak changes to their terms and conditions if, instead of striking, they had waited for Joe Biden to ride to the rescue with some nice legislation. Biden does at least pretend to care about workers’ rights: British workers would have to be utterly deluded to believe that a Tory government that has sought to all but ban strikes ever has the intention of protecting their jobs.

Conversely, the government is telling us to worry about much less proximate risks – including the very unlikely eventuality of self-aware machines – while completely and conspicuously ignoring the very real and urgent environmental concerns that are emerging around high-technology industry. The extreme levels of computer processing required by AI-branded projects are energy-hungry, which has obvious implications for climate change, as well as longer-term concerns about water consumption and industrial pollution. This is, of course, where Sunak’s worst political instincts come to the fore.

Since this summer, the prime minister has exposed himself as one of the most backward world leaders on the question of climate change. Through a combination of craven electioneering and pandering to the fossil-fuel lobby, Sunak has significantly diminished Britain’s status as a country that takes meaningful action to avert ecological catastrophe. There’s a glaring, obvious irony in the same government completely flaunting scientific advice to put off and jeopardise the process of decarbonisation, literally while the country is battered by heavy storms caused by climate chaos, while at the same time posing as being the most technologically savvy and responsible party on AI.

The international controversies around Sunak’s regressive policies on internal combustion engine cars and North Sea drilling are likely, in fact, to be part of the reason why other governments are not falling over themselves to get to his summit. Either way, it really is difficult to take the Tories seriously, telling us they are worried about a sustainable future while they ignore the unsustainable present.

Falling for the hype

The reality here is that Sunak and his team have been absolutely bamboozled by hype and spin from the AI companies and, in the classic pattern of a rube during a marketing fad, they’ve attempted to join the party too late. The nonsense about regulation is most likely rationalised as an attempt to give British companies a boost in getting into the new AI sector, believing that there’s still plenty of scope to expand into the field. This is not the sure bet they think it is.

There is already strong evidence that the AI systems that were attracting so much excitement at the start of the year may not have that much further to go in terms of technological development. Large-language models, such as the famed ChatGPT, are struggling to implement improvements and upgrades to their functions. Some engineering publications are already declaring that AI is getting stupider, not smarter, as the months go by.

This is likely, at least in part, to be due to fact that generative systems like LLMs are having their processing distorted by using the output of other generative systems as input. Generative systems work by associating bits of data – text, images, sounds – with each other, based on associations in artifacts created by humans. However, since much of their output itself gets released onto the internet, quite often they are now picking that up and falsely assuming that it is just as good as a data source, misleading them about the associational values they are reading and then implementing. The ironies come thick and fast: generative models are poisoning the well they drink from, and they are unable to stop themselves, because if you think you struggle to tell whether something’s been created by a human or a machine, imagine how much a machine will struggle!

Essentially, there is something of a classic finance-backed hype cycle going on with AI. Huge capital surpluses are looking for things to be invested in, and AI was very much this year’s thing. It may not be the ‘current thing’ for very much longer if its capacities to replace humans and improve itself endlessly are indeed proving to be much more limited than promised. It is a symptom of capitalism’s current maladies that these fads are whizzing by so fast it becomes disorientating: it was only a couple of years ago that the return of virtual reality was the future, now we can barely remember it happened, despite the eye-watering millions of dollars that were invested in it at the time.

With this ‘Safety Summit’, Rishi Sunak is, more than anything, desperately trying to get in on a craze he does not understand and attempting to make himself look clever by repeating things he’s heard other people say. It’s a comedy side-show to the multiplicity of genuine great world events that are taking place all over right now. If this event gets remembered at all, it will be as a pathetic symbol of the utter political bankruptcy of the Tory government in the face of global crisis.

Before you go

Counterfire is growing faster than ever before

We need to raise £20,000 as we are having to expand operations. We are moving to a bigger, better central office, upping our print run and distribution, buying a new printer, new computers and employing more staff.

Please give generously.