Having been established in July this year, the regulation-facing forum is looking to promote research into possible AI safety guardrails to mitigate harms such as disinformation and security risks, as well as appointing Chris Meserole as its first executive director.
In the lead-up to a global summit exploring the safety of artificial intelligence taking place in the UK next week, the new safety fund initiative comes in line with commitments to facilitate third-party auditing of vulnerabilities within AI models, says a Forum announcement.
Primarily, focus is said to be on supporting development of adversarial tests on models — a kind of ethical system hacking known as ‘red teaming‘.
Through funding such projects, the tech consortium hopes to raise safety and security standards, as well as provide insights into measures required of government and industry stakeholders to effectively mitigate flaws.
UK government calls for AI infrastructure access ahead of global summit — As big tech AI innovation continues, government officials push for under-the-hood access to key start-ups’ technology ahead of the world’s first AI safety summit.
The fund is being delivered with support with onboarded ‘philanthropic’ partners, the Patrick J. McGovern Foundation; The David and Lucile Packard Foundation; ex-Google CEO Eric Schmidt; and computer programmer and investor Jaan Tallinn.
New executive director Meserole joins following a six year tenure serving in directoral and fellowship roles at the social sciences research and education think tank, The Brookings Institution.
Meserole said of the new AI Safety Fund, which he will oversee: “The most powerful AI models hold enormous promise for society, but to realise their potential we need to better understand how to safely develop and evaluate them. I’m excited to take on that challenge with the Frontier Model Forum.”
Proposals for funding under the new AI safety scheme are expected to be sent out within a matter of months.
Geoffrey Hinton realised mankind was history when AI got the joke — So-called ‘godfather of AI’ Geoffrey Hinton says he realised that humanity had created its own successor when Google’s PaLM explained why a joke was funny.