Yudkowsky and Aschenbrenner Clash Over AI Timelines

An online debate sparked by writer Zvi Mowshowitz has illuminated a historical clash between AI theorist Eliezer Yudkowsky and researcher Leopold Aschenbrenner over epistemic modesty. The disagreement centered on whether independent thinkers must defer to established consensus. Yudkowsky recalled an encounter where Aschenbrenner criticized him for not updating his short AI timelines toward the Open Philanthropy doctrine of a 2050 arrival. Yudkowsky refused to defer, stating he did not consider those consensus-builders to be his epistemic peers, which provoked contempt from Aschenbrenner.

Mowshowitz argues that Yudkowsky was ultimately vindicated. Since that initial exchange, Aschenbrenner and many others have radically shortened their timelines. In August 2021, Aschenbrenner estimated only a 25 percent chance of artificial general intelligence arriving by 2070. However, following recent technological leaps, he shifted his public prediction to AGI arriving by 2027, advocating for an urgent technological arms race.

This shift away from modesty also manifested in Aschenbrenner's financial endeavors. He founded the hedge fund Situational Awareness, which took highly leveraged market positions. Although the fund suffered a massive decline, with assets dropping from $45 billion to $10 billion due to margin calls, Mowshowitz defends the underlying philosophy of backing one's own analysis over market consensus.

For AI practitioners and forecasters, Mowshowitz reaffirms his February 9, 2021 stance rejecting the idea that smart individuals cannot improve upon expert consensus. He outlines several reasons to reject epistemic modesty, such as identifying motivated reasoning or information cascades. Ultimately, Mowshowitz argues that progress relies on individuals contributing their own independent assessments rather than passively deferring to the crowd.