June 9, 2023 - 7:30am

Rishi Sunak is in Washington DC to promote the UK as a global leader in regulating artificial intelligence. The Prime Minister’s office has already announced a conference, but also seeks to found “a nuclear-style global AI watchdog based in London”. For this, he needs President Joe Biden’s blessing, and the two are clearly ready to forge close Anglo-American ties, striking the “Atlantic Declaration” to reinforce economic security on Monday afternoon.

For a nation that pioneered the computer, invented packet switching networks — on which the design of today’s internet was based — and devised the most popular microprocessor architecture currently used in the world, this is quite a change of emphasis. Once a pioneer, the UK now seeks to be a bureaucratic overlord, the chief junketeer. But then talking about technology, rather than creating it, is where the political and administrative class sees the UK making a difference worldwide. A future conference will no doubt continue this theme.

Civil service recruitment places far greater value on communication skills than deep technical knowledge or STEM qualifications. For instance, the Government’s AI Council is led by a BA in graphic design and advertising, whose qualification for the role was hosting an AI conference the previous year. Typical of what we may expect to see is The Global Partnership on Artificial Intelligence, which the UK helped create three years ago, and which studies how people discuss AI.

Is Sunak’s regulatory zeal justified? In February 2019 OpenAI, then a non-profit, declined to release a large language model project to the general public. This was GPT2, and the lab feared negative social and economic consequences from the low-cost production of fakes. When its successor GPT3 was released, OpenAI’s own researchers again warned of “misinformation, spam, phishing, abuse of legal and governmental processes, fraudulent academic essay writing and social engineering pretexting“. Then, in November, OpenAI decided to release ChatGPT anyway.

There are other motives for regulation. Ominously for creators, a review led by Sir Patrick Vallance calls for copyright to be weakened to help the AI companies, so publishers and artists could no longer decline to have their material ingested and reproduced. Intellectual property has bounced from the Department of Digital, Culture, Media and Sport (DCMS) to the Department for Business, Energy and Industrial Strategy (BEIS), where it is considered an inconvenience. But beyond the vagaries, it is unclear what the government is actually trying to regulate.

Sunak’s impulse to control generative AI might be noble, then, but he and his advisers have no idea how to do it, while the problem has been compounded by a reliance on self-interested or poorly-equipped “experts”. The burgeoning AI ethics community, fretting instead over bias, diversity and representation issues, was also caught napping. So we shouldn’t be surprised when we hear the wolves agreeing to regulate the sheep. They might even agree.