UK minister: Include China in AI summit for ‘safety of species’
China needs to be involved in discussions at the AI safety summit, set to be held this week, the UK’s minister for artificial intelligence (AI) has said.
“We cannot possibly have a world in which there is a Chinese method of [ensuring safety] and a rest of the world method of doing that. That would not be safe,” Jonathan Berry, known in Downing Street as the fifth Viscount Camrose, told City A.M.
“We have to try and bring everybody into the same shared risks of frontier risks of AI. I very much think that’s possible,” he added.
Berry, who assumed the newly created role of minister for AI and intellectual property in March, is also a hereditary peer in the House of Lords.
Berry’s remarks follow former Prime Minister Liz Truss’s letter to her successor, Rishi Sunak, which urged a reconsideration of China’s invitation to the UK government’s AI safety summit.
Truss had warned of China’s “fundamentally different attitude” toward AI compared to Western nations.
But Berry clarified the focus of the summit.
“This is talking about AI safety, we’re absolutely not talking about innovations within AI or taking AI forward technologically, which we would not want to share with anybody but our allies,” he said. “This is about ensuring the safety of our species.”
This week, the UK government is hosting the summit at Bletchley Park in Milton Keynes, bringing together politicians, leaders, and top AI companies from around the world.
The aim is to discuss strategies for mitigating the risks associated with AI. Last week, No. 10 said it is concerned that AI could pose an “existential threat” to humanity.
Some critics of the summit have expressed concerns that big tech is poised to capitalise on their invites, using their influence to encourage regulations that benefit them, effectively marking their own homework.
“I would absolutely hate to see some sense of regulatory capture by the big labs,” Berry said.
He cautioned against a scenario where AI success hinges on the size of a company but stressed the importance of companies with frontier models, such as Chat GPT, participating in discussions.
“What nobody has yet really taken a grasp of is the frontier risks of AI. which is what we’re doing [at the summit],” he explained.
In response to recent actions taken by US President Joe Biden, who intends to compel big tech companies to disclose the safety of their AI models, Berry did not confirm if the UK would follow suit.
However, he expressed support for the move. “It looks like an extremely positive step that we’ll look at with great care and take pretty seriously,” he said.
His biggest worry right now revolves around the democratic impact of deepfakes (false but extremely convincing AI-generated images or videos, often used to imitate high profile individuals like politicians).
Berry said government taskforces are “looking very hard” at this issue.
“Obviously we worry about the big frontier risks of AI but I think right now it is much easier to produce targeted disinformation to affect a whole population of people than it has ever been. That’s a real worry for me.”