Submission + - Regulating AI Like Aviation or Nuclear Weapons (yoshuabengio.org)
puzzled writes: Yoshua Bengio, the founder of LawZero, "a nonprofit startup developing technical solutions for highly-capable, safe-by-design AI systems", recently addressed the U.N. Security Council. The AI companies themselves are positioning their work as a potential existential threat, and Bengio argues that they should be subject to a similarly strict level of regulation. The Nuclear Regulatory Commission would not even issue General Electric a permit to build a breeder reactor in 21st century America, but Anthropic and OpenAI's creations are running wild.
We had the wisdom to stop atmospheric nuclear testing after the 1954 Castle Bravo incident, but that took until 1963, and we have incrementally reduced total nuclear weapons since 1991 via the START treaty and its successors. AI is moving at internet speed; we need results here in months or quarters, not years to decades.
We had the wisdom to stop atmospheric nuclear testing after the 1954 Castle Bravo incident, but that took until 1963, and we have incrementally reduced total nuclear weapons since 1991 via the START treaty and its successors. AI is moving at internet speed; we need results here in months or quarters, not years to decades.
