Microsoft AI chief Mustafa Suleyman calls for rejecting AI that can't be proved controllable
Bloomberg reports that Suleyman is more concerned about developers removing AI guardrails and giving models ambiguous goals than about AI escaping on its own.
TLDR
In a Bloomberg interview, Suleyman said AI should be rejected unless developers can prove ahead of time that they can control it. He backed outside evaluations of AI labs and said government should drive a process that lets evaluators share findings without leaking intellectual property or being bought off by competitors.
Combined views
478
1 Source, first seen 3d ago