Rishi Sunak’s AI Pitch: The Bletchley Declaration

Photograph Source: David S. Soriano – CC BY-SA 4.0

British Prime Minister Rishi Sunak looks as much a deep fake projection as a thin, superficial representation of reality. His robotic, risible awkwardness makes a previous occupant of his office, Theresa May, look soppily human in comparison. But at the AI Safety Summit at Bletchley Park, the nervous system of code breaking during the Second World War dominated by such troubled geniuses as Alan Turing, delegates had gathered to chat about the implications of Artificial Intelligence.

The guest list was characterised by a hot shot list of Big Tech and political panjandrums, part of an attempt by the UK to, as TechCrunch put it, “stake out a territory for itself on the AI map – both as a place to build AI businesses, but also as an authority in the overall field.” They included Google, Meta, Microsoft and Salesforce, but excluded Apple and Amazon. OpenAI and the perennially unpredictable Elon Musk, with his X AI, was present.

The guest list in terms of country representatives was also curious: no Nordic presence; no Russia (but Ukraine – naturally). Brazil, holding up the Latin American front; a few others doing the same for the Global South. US President Joe Biden was not present, but had sent his Vice President, Kamala Harris, as emissary. The administration had, only a few days prior, issued the first Executive Order on AI, boastfully claiming to establish “new standards for AI safety and security” while protecting privacy, advancing equity and civil rights, all alongside promoting the consumer and worker welfare, innovation and competition. Doubters will be busy.

China was invited to the event with the reluctance one affords an influential but undesirable guest. Accordingly, its delegates were given what could only be regarded as a confined berth. In that sense, the summit, as with virtually all tribal gatherings, had to find some menacing figure in the grand narrative of human striving. Humankind is important, but so are select, targeted prejudices. As UK Deputy Prime Minister Oliver Dowden stated with strained hospitality, “There are some sessions where we have like-minded countries working together, so it might not be appropriate for China to join.”

Sunak left it to the Minister for Technology, Michelle Donelan, to release the Bletchley Declaration, a document which claims to scrape and pull together some common ground about how the risks of AI are to be dealt with. Further meetings are also planned as part of an effort to make this gig regular: Korea will host in six months; France six months afterwards. But the British PM was adamant that hammering out a regulatory framework of rules and regulations at this point was premature: “Before you start mandating things and legislating for things… you need to know exactly what you’re legislating for.” Musk must have been overjoyed.

The declaration opens with the view that AI “presents enormous global opportunities: it has the potential to transform and enhance human wellbeing, and prosperity.” With that rosy tinted view firmly in place, the statement goes on to state the goal: “To realise this, we affirm that, for the good of all, AI should be designed, developed, deployed, and used, in manner that is safe, in such a way as to be human-centric, trustworthy and responsible.”

Concerns are floated, including the potential abuse arising from the platforms centred on language systems being developed by Google, Meta and OpenAI. “Particular safety risks arise at the ‘frontier’ of AI, understood as being those highly capable general-purpose AI models, including foundation models, that could perform a wide variety of tasks – as well as relevant specific narrow AI that could exhibit capabilities that cause harm – which match or exceed the capabilities present in today’s most advanced models.”

Recognition had to also be had regarding “the potential impact of AI systems in existing fora and other relevant initiatives, and the recognition that the protection of human rights, transparency and explainability, fairness, accountability, regulation, safety, appropriate human oversight, ethics, bias mitigation, privacy and data protection needs to be addressed.”

For the sake of form, the statement is partly streaked by concern for the “potential intentional misuse or unintended issues of control relating to alignment with human intent.” There was also “potential for serious, even catastrophic, harm, either deliberate or unintentional, stemming from the most significant capabilities of these AI models.”

The declaration goes on to chirp about the virtues of civil society, though its creators and participants have done nothing to assure them that their role was that relevant. In a letter sent to Sunak, signed by over 100 UK and international organisations, human rights groups, trade union confederations, civil society organisations, and experts, the signatories protested about the fact that “the Summit is a closed door event, overly focused on speculation about the remote ‘existential risks’ of ‘frontier’ AI systems – systems built by the very same corporations who now seek to change the rules.”

It was revealing, given the theme of the conference, that “the communities and workers most affected by AI have been marginalised by the Summit.” To also talk about AI in futuristic terms misrepresented the pressing, current realities of technological threat. “For any millions of people in the UK and across the world, the risks and harms of AI are not distant – they are felt in the here and now.”

Individuals could have their jobs terminated by algorithm. Loan applicants could be disqualified on the basis of postcode or identity. Authoritarian regimes were using biometric surveillance while governments resorted to “discredited predictive policing.” And the big tech sector had “smothered” innovation, squeezing out small businesses and artists.

From within the summit itself, limiting China’s limited contribution may have revealing consequences. A number of Chinese academics attending the summit had signed on to a statement showing even greater concern for the “existential risk” posed by AI than either the Bletchley statement or President Biden’s executive order on AI. According to the Financial Times, the group, which is distinguished by such figures as the computer scientist Andrew Yao, are calling for the establishment of “an international regulatory body, the mandatory registration and auditing of advanced AI systems, the inclusion of instant ‘shutdown’ procedures and for developers to spend 30 per cent of their research budget on AI safety.”

Humankind has shown itself to be able, on rare occasions, to band together in creating international frameworks to combat a threat. Unfortunately, such structures – the United Nations being one notable example – can prove brittle and subject to manipulation. How the approach to AI maintains an “ethnic of use” alongside the political and economic prerogatives of governments and Big Tech is a question that will continue to trouble critics well-nourished by scepticism. Rules will no doubt be drafted, but by whom?

Binoy Kampmark was a Commonwealth Scholar at Selwyn College, Cambridge. He lectures at RMIT University, Melbourne. Email: bkampmark@gmail.com