Overview
- Bengio resigned as Mila’s scientific director in March to focus intensively on AI risk mitigation through his new nonprofit
- Scientist AI is designed to eschew autonomous goals and instead generate probabilistic assessments of correctness and harm potential
- Drawing its name from Isaac Asimov’s zeroth law of robotics, the system flags and blocks actions forecast to exceed a safety threshold
- Prominent philanthropic supporters such as Schmidt Sciences, Open Philanthropy, the Future of Life Institute and engineer Jaan Tallinn contributed to the initial funding
- LawZero plans to validate its approach on open-source models before persuading governments and major AI labs to adopt Scientist AI at scale