UN AI Experts Call for Calm Talk on Risks, Not Doom

Several experts from a United Nations AI committee warned against doomsday-style claims about AI risks during the UN General Assembly in New York.

Joelle Barral, a member of the International Scientific Panel on AI and an executive at Google DeepMind, said researchers should focus on science to clarify what is settled and what remains unknown. She added that spreading fear is not helpful. Earlier, a former employee of Anthropic and OpenAI said AI could destroy all of humanity before the end of this decade.

Nobel Peace Prize winner Maria Ressa said the debate often swings between two extremes: either claiming an existential threat or treating AI risks as trivial. She said the truth is far more nuanced and complex.

Yoshua Bengio, one of the founders of AI and co-chair of the panel, said the key is to draw a line between facts scientists have established about AI risks and reasonable extrapolations from existing evidence. He called for a global, rational conversation that faces reality, accepts uncertainty, and carefully assesses the severity of various risks.