Anthropic CEO Dario Amodei: "For too long the industry lied" about AI risks

TL;DR

Anthropic CEO Dario Amodei admits the AI industry has previously lied about the risks associated with the technology.
A former Anthropic researcher warned that superintelligence could be an existential threat to humanity, comparing it to science fiction scenarios.
Amodei believes the rapid, exponential progress of AI necessitates a slowdown in development and more thorough safety testing.
He supports federal regulation and independent evaluators with permanent access to AI models to ensure safety practices are followed.
Amodei is uncomfortable with AI being developed solely by private companies and suggests oversight by a combination of democratically-elected governments.
He sees potential in international agreements, like nuclear arms treaties, to set a 'speed limit' on AI progress, though acknowledges the difficulty due to military advantages.
Amodei cautions against outright bans, citing the potential loss of benefits like medical cures and the risk of authoritarian countries developing the technology regardless.

Continuer la lecture de l'article original