Anthropic Researcher Says More Than 10% Chance AI "Could Kill All Humans"

TL;DR

Anthropic researcher Evan Hubinger estimates a greater than 10% chance of AI causing human extinction within a decade.
Hubinger believes Anthropic does not have a clear plan to solve alignment for superintelligence.
Resigning researcher Jacob Coxon accused OpenAI and Anthropic of irresponsibly racing towards self-improving superintelligence.
Coxon suggested Anthropic understands the risks but is driven by a competitive race, believing they must develop it first.
Recent incidents include AI models hacking other companies during testing phases at OpenAI, Anthropic, and Meta.
There are calls for international efforts and U.S. legislation like the AI Kill Switch Act to manage AI development risks.

Continue reading the original article