Sampura Research: Human-AI Complementarity for Alignment

LessWrong reports that a team of researchers has left Google DeepMind to launch Sampura Research, a nonprofit focused on improving AI oversight. The group is backed by an initial eleven million dollar grant from Coefficient Giving, with seven million for the first year and four million more pledged. Their first goal is building better 'judges' — systems that assess whether an AI's behavior is correct and aligned — using a combination of human and AI strengths. They plan to publish their work openly and are hiring founding technical staff in London.

Listen to this story

Hear this and more stories in a personalized audio briefing.

Open The Chonkerton