Google DeepMind CEO Demis Hassabis called for a US-led standards body to oversee frontier AI, warning that an intense commercial and geopolitical race is exacerbating the technology’s risks.
AI could usher in “a new age for humanity” greater than the Industrial Revolution, Hassabis wrote, but only if deployed responsibly. As it stands, we are not “[giving] ourselves the time and space” to get it right.
Hassabis proposed that frontier AI labs should share new models with the self-regulatory organization 30 days before launch, and if major risks are uncovered, it could coordinate a slowdown in development.
His proposal follows some influential researchers’ blueprint for safe AI, “Plan A,” which calls for international coordination to delay the development of superintelligence.