The UK's newly established AI Safety Institute (AISI) should focus on setting global standards for artificial intelligence testing rather than attempting to conduct all vetting internally, according to Marc Warner, CEO of Faculty AI, a company assisting the government's AI safety initiatives.
Warner emphasized the need for the AISI to take a leadership role in establishing test standards that can be adopted internationally, leveraging the UK's advanced work in AI safety. He suggested that the institute should prioritize setting standards that other governments and companies can follow, promoting approaches like "red teaming," where specialists simulate misuse of AI models, instead of shouldering the entire workload.
Acknowledging the rapid pace of technological advancements, Warner emphasized the importance of scalability and long-term vision in ensuring AI safety. He cautioned against potential bottlenecks and urged the institute to adopt a strategic approach to prioritize the testing of the most advanced AI systems.
While the AISI has made significant progress in its testing program, Warner underscored the need for the institute to collaborate with international stakeholders and promote information sharing to advance global understanding of AI safety.
The UK government, through the Department for Science, Innovation, and Technology, affirmed its commitment to driving forward AI safety efforts. A spokesperson highlighted the AISI's role in conducting evaluations, research, and information sharing to inform policymakers worldwide on AI safety.
Warner's remarks come amid increasing pressure from major AI companies, including Google, OpenAI (developer of ChatGPT), Microsoft, and Meta, urging the UK government to accelerate its AI safety testing processes. The US has also announced its own AI safety institute and consortium to support its AI safety goals.
As governments worldwide recognize the importance of AI safety, the UK's leadership in establishing global standards through the AISI could have far-reaching implications for the responsible development and deployment of AI technologies on a global scale.
