Overview
- This week Hinton used a series of high-profile interviews to say advanced AI could carry a non-negligible chance of catastrophic outcomes and that “we might have been better off never developing it.”
- He estimated a roughly 10–20% probability of extreme long-term harm and said his main near-term fear is AI enabling attacks on critical systems such as power grids or global financial infrastructure.
- Hinton called for an independent, drug-regulator-style approval process for models and legal accountability when systems cause harm as a way to force safer, tested deployments.
- He sharply criticized the concentration of AI power in a few big firms and named leaders he says prioritize profit over safety, arguing market and national competition are accelerating risky releases.
- Hinton’s stature as a founding neural‑network researcher has intensified calls for enforceable oversight but concrete regulatory action remains limited and governments are split on measures such as emergency shutdowns.