Agreements at UK summit tip balance of humanity against AI threats: Sunak

British Prime Minister Rishi Sunak said Thursday that achievements at the first international AI Safety Summit would tip the balance in favour of humanity in the race to contain the risks from rapid advances in cutting-edge artificial intelligence.

Speaking after two days of talks at Bletchley Park, a former codebreaking spy base near London, Sunak said agreements struck at the meeting of politicians, researchers and business leaders show that we have both the political will and the capability to control this technology, and secure its benefits for the long term.

Sunak organised the summit as a forum for officials, experts and the tech industry to better understand cutting-edge, frontier AI that some scientists warn could pose a risk to humanity's very existence.

He hailed the gathering's achievements, including a Bletchley Declaration committing nations to tackle the biggest threats from artificial intelligence, a deal to vet tech firms' AI models before their release, and an agreement to call together a global expert panel on AI, inspired by the United Nations' climate change panel.

Some argue that governments must go further and faster on oversight. Britain has no plans for specific legislation to regulate AI, unlike the US and the European Union.

Vice President Kamala Harris attended the summit, stressing steps the Biden administration has taken to hold tech firms to account. She said Thursday that the United States' bold action should be inspiring and instructive to other nations.

United Nations Secretary General Antonio Guterres urged a coordinated global effort, comparing risks from AI to the Nazi threat that Britain's wartime codebreakers worked to combat.

Bletchley Park played a vital part in the computing breakthroughs that helped to defeat Nazism, he said The threat posed by AI is more insidious but could be just as dangerous.

The UN chief, like many others, warned about the need to act swiftly to keep pace with AI's breathtaking advances. General purpose AI chatbots like ChatGPT released over the past year stirred both amazement and fear with their ability to generate text, audio and images that closely resembled human work.

The speed and reach of today's AI technology are unprecedented, Guterres said. The paradox is that in the future, it will never move as slowly as today. The gap between AI and its governance is wide and growing.

Sunak hailed the summit as a success, despite its arguably modest achievements. He managed to get 28 nations including the US and China to sign up to working toward shared agreement and responsibility about AI risks, and to hold further meetings in South Korea and France over the next year.

China did not attend the second day, which focused on meetings among what the UK termed a small group of countries with shared values. Sunak held a roundtable with politicians from the EU, the UN, Italy, Germany, France and Australia.

Announcing the expert panel on Thursday, Sunak said pioneering computer scientist Yoshua Bengio, dubbed one of the godfathers of AI, had agreed to chair production of its first report on the state of AI science.

Sunak said likeminded governments and AI companies also had reached a landmark agreement to work together on testing the safety of AI models before they're released to the public. Leading AI companies at the meeting including OpenAI, Google's DeepMind, Anthropic and Inflection AI have agreed to deepen access to their frontier AI models, he said.

Binding regulation for AI was not among the summit's goals. Sunak said the UK's approach should not be to rush into regulation but to fully understand AI first.

Harris emphasised the US administration's more hands-on approach in a speech at the US embassy on Wednesday, saying the world needs to act right away to address the full spectrum of AI risks, not just existential threats such as massive cyberattacks or AI-formulated bioweapons.

One of the Biden administration's main concerns is that advances in AI are widening inequality within societies and between countries. As a step towards addressing that, Britain's Foreign Secretary James Cleverly announced a $100 million fund, supported by the UK, the US and others, to help ensure African countries get a share of AI's benefits and that 46 African languages are fed into its models.

Cleverly told reporters that it's crucial there is a diversity of voice informing AI.

If it was just Euro-Atlantic and China, we would miss stuff, potentially huge amounts of stuff, he said.

Sunak is scheduled to discuss AI with Tesla CEO Elon Musk on Thursday evening in a conversation that will be played on the social network X, which Musk owns.

Musk, who attended the summit, is among tech executives who have warned that AI could pose a risk to humanity's future.

Here we are for the first time, really in human history, with something that is going to be far more intelligent than us, Musk said at the summit. It's not clear to me if we can control such a thing.

Sunak said it was important not to be alarmist about the technology, which could bring huge benefits.

But there is a case to believe that it may pose a risk on a scale like pandemics and nuclear war, and that's why, as leaders, we have a responsibility to act to take the steps to protect people, and that's exactly what we're doing," he said.