Announced by Technology Secretary Michelle Donelan, the UK government’s AI Safety Institute is gearing up to establish its first overseas office in San Francisco this summer. This move signifies a strategic development in the institute’s growth, aiming to tap into the tech talent pool of the SF Bay Area, home to many leading tech and AI firms worldwide.
The AI Safety Institute (AISI) will be recruiting an initial team of technical staff, led by a research director, in its new San Francisco office. This team will complement the existing pool of over 30 experts at the Institute’s headquarters in London, where continuous efforts are made to scale risk assessments of advanced AI systems.
British AI Safety Institute taking global strides
Notably, the UK and the US have both committed to similar AI safety agreements, exemplified by signing the Bletchley Declaration at a British-organized summit alongside 25 other countries and the European Union. This shared commitment sets the stage for enhanced collaboration between the two nations through the new San Francisco office of the AISI.
Additionally, a partnership between the UK and Canada has been forged to bolster AI safety research efforts, further solidifying the global impact of initiatives in this field.
Reflecting on the expansion, Donelan emphasized the significance of a global perspective in understanding the risks and potential of AI, asserting the UK’s leadership role in AI safety. He mentioned that the expansion marks a pivotal moment for the UK as it strengthens partnerships with the US and opens avenues for other countries to leverage their expertise.
An important milestone for the Institute, the announcement coincides with the release of AI safety test results, shedding light on technical limitations of large language models. The study unveiled vulnerabilities in these models, highlighting the need for robust evaluations to mitigate risks, especially concerning national security.
Looking ahead, Ian Hogarth, the Chair of AISI, expressed the institute’s ambition to push boundaries in the emerging field of AI safety. By focusing on cutting-edge evaluations and addressing national security risks, AISI aims to continue pioneering advancements in this critical area.
The upcoming AI Seoul Summit 2024, succeeding the Bletchley Park summit held in 2023, is expected to bring together global stakeholders from governments, AI companies, academia, and civil society to advance discussions on AI safety.
