UK Prime Minister Rishi Sunak has emphasized the need for better understanding and testing of artificial intelligence (AI) models before enforcing any mandatory regulations in the industry. Speaking at the AI Safety Summit at Bletchley Park, Sunak highlighted the importance of accelerating efforts in this area. He recently announced the establishment of the UK AI Safety Institute, which will collaborate with technology companies on a voluntary basis to ensure the safety of their AI models for public use. However, the institute will not possess regulatory powers, and companies will not be compelled to undergo specific testing protocols.
During a press conference at the summit, Sunak acknowledged that regulation will eventually be necessary but should be evidence-based. Several major technology companies involved in AI, including Meta, Google DeepMind, and OpenAI, have agreed to engage with the new organization. Sunak acknowledged the need to conduct testing before releasing AI models to the public, stating that companies should not be responsible for evaluating their own work.
While acknowledging that regulation takes time, Sunak emphasized the need to gather more information about AI before effectively implementing regulations. He highlighted the importance of empirical and scientifically grounded regulation, particularly considering the surprises AI developers often encounter regarding the capabilities of their models.
Sunak expressed his belief in the significant role of governments in shaping the future of AI, particularly in testing national security risks. He emphasized that sovereign governments, along with intelligence agencies, have the responsibility and knowledge to handle this task.
The AI Safety Summit brought together around 100 politicians, business leaders, and academics to discuss the potential dangers associated with highly advanced AI. Sunak previously stated that such risks could be on par with those of nuclear war.
However, the summit received criticism for lack of transparency, as the UK government published a list of attending governments and organizations but did not disclose all the guests’ names. Reporters present at the event were also not allowed to interact with the delegates.
Nevertheless, one significant outcome of the summit was the signing of the Bletchley Declaration by 28 countries, including the US, China, and the European Union. The declaration acknowledges the risks associated with AI and recommends further research in this area. It also plans for a smaller summit on the same topic to be held in South Korea within the next six months, followed by a larger conference next year.
Experts, however, criticized the progress made at the summit as vague and slow. Carissa Véliz from the University of Oxford expressed concerns about the slow pace of AI regulation and the lack of ambitious timelines given the high stakes and rapid development of AI.
In addition to the summit, Prime Minister Sunak was scheduled to have a live-streamed conversation with Elon Musk, owner of xAI (formerly known as Twitter), on 2 November.
