What can be done about the “profound risks to society and humanity” that Elon Musk and others claim AI poses? That’s one of the questions the Government hopes to answer at the AI Safety Summit, which takes place this week.
The summit is an opportunity for governments and industry to come together to discuss some of the most serious threats created by AI and how to deal with them. It will, the Government hopes, make the UK a leading player in AI regulation.
Here we explore the purpose of the AI Safety Summit, who’s coming, what it hopes to achieve, and whether any government can really control the rapid advancement of AI.
When is the AI Safety Summit 2023?
The AI Safety Summit is being held over two days, starting Wednesday, November 1.
However, there are a series of related events being held before and after the summit. In the run-up, workshops have been held by The Alan Turing Institute, the British Academy, techUK, and the Royal Society, each discussing different aspects of AI.
During and after the summit, there will also be “public engagement” events, such as a LinkedIn Q&A with Michelle Donelan, the secretary of state for science, innovation and technology, on November 16.
Elon Musk and other tech leaders wrote an open letter... through fear that AI was developing so fast it could 'pose profound risks to society and humanity'
Where is the AI Safety Summit being held?
The summit itself is being held at Bletchley Park, a location loaded with computing history. It was the home of the allied codebreaking effort during World War II. It was where Alan Turing and others devised the Bombe, an electromechanical machine that could break the German Enigma codes more effectively than anything that had come before it.
Who is attending the AI Safety Summit?
The aim of the summit is to bring together governments and representatives of the AI industry to discuss how to manage the risks posed by advancements in AI. The summit has a very limited capacity of around 100 people, which the Government says is necessary to focus discussions, but it’s also a reflection of the limited capacity at Bletchley Park.
The full list of attendees has not been revealed, but the Government did confirm that representatives of the Chinese government have been invited to join. That has reportedly provoked opposition from Japan, which say it’s too early to involve Beijing when democratic nations haven’t agreed the ground rules for AI.
Foreign secretary James Cleverly responded to such criticism by saying: “We cannot keep the UK public safe from the risks of AI if we exclude one of the leading nations in AI tech.”
Industry attendees will reportedly include Google, OpenAI (the owners of ChatGPT), and smaller British AI start-ups, such as Graphcore and Stability AI.
What are the priorities of the AI Safety Summit?
The AI Safety Summit is very much focused on the threats, rather than the opportunities, of AI.
It’s concentrating largely on what the Government refers to as “Frontier AI”; large-language models such as GPT-4 that power services such as ChatGPT.
The agenda will particularly focus on two specific categories of risk: misuse and loss of control.
Britain has a history of pioneering rules for governing emerging technologies
Why is AI dangerous?
In terms of misuse, the summit will explore how AI could be used to co-ordinate biological or cyber attacks, or whether it could be deployed to interfere with critical systems. When it comes to loss of control, this seems to be focused on the notion of artificial general intelligence (AGI), where AI becomes so smart that we are no longer able to contain it. Elon Musk and other tech leaders wrote an open letter earlier this year arguing we should “pause giant AI experiments” through fear that AI was developing so fast it could “pose profound risks to society and humanity”.
Some AI experts argue that this threat is over-hyped and that the Government is wrong to put so much emphasis on AI’s dangers at this summit. Dr Keegan McBride, a department research lecturer in AI at the University of Oxford’s Internet Institute, claims that the “AI systems based on technology that we have now and in the foreseeable future aren't able to rise to the level of sophistication and intelligence that governments – the UK, basically – and companies like OpenAI are discussing”.
In fact, he argues the industry is exaggerating the dangers of AI to force governments to act, thus shutting out would-be rivals. “If you're able to convince governments that you are building something that can destroy the world, and then they're going to regulate it so only three or four companies can work on it”, that leads to “centralisation of AI development,” according to Dr McBride.
Other experts are more positive about opening discussions on how to tame AI. Pia Hüsch, a research analyst at London-based defence think-tank the Royal United Services Institute (RUSI), wrote last week that the summit provides an “opportunity for the UK to clarify its position on AI regulation and to seek alliances”. It could even lead to the formation of an international institution focused on AI safety, she argued.
Can AI be regulated?
Is it even possible to rein in AI with regulation? It’s fair to say opinion is divided on that, too.
One of the big problems for the Government, which is aiming to “play a leadership role in the global effort to address the challenges and opportunities of AI”, is that the vast majority of the leading AI companies are not British. Google (which bought the British AI firm DeepMind), Microsoft, OpenAI, Anthropic and many of the other leading lights in AI are based in the US, not over here, raising issues around jurisdiction.
Although the US is currently taking something of a hands-off approach to AI regulation, it did introduce some initial guidelines and regulations surrounding AI during the Trump administration, before anyone had even heard of ChatGPT and its like. The EU is currently proposing an AI Act to sit alongside other digital regulations, such as GDPR and the Digital Services Act. China has also developed its own AI regulations. So, far from taking a global lead, the UK is arguably playing catch-up with the superpowers.
That doesn’t necessarily mean the summit is a waste of time, according to Olivia O’Sullivan, director of the UK in the World programme for the Chatham House policy institute. Although the “EU will ultimately be more influential in setting standards for AI and other digital technologies,” she wrote recently, “Britain has a history of pioneering rules for governing emerging technologies, for example in stem-cell research, where its ‘strict but permissive’ regime is world-renowned”.
However, the summit’s narrow focus on AI’s existential threats does limit its scope, according to Oxford University’s Dr McBride, because laws already exist to counter those. “A lot of the stuff that are risks that can come from generative AI, we already have regulatory frameworks in place for,” he said.
“You can't kill people, you can't commit fraud, you can't discriminate. So, it just feels a bit weird. I don't know how else to describe it.”
