Jake Gold: warning people does not require a made-up probability
Jake Gold·medium signal
Gold's September 14 post attacks the practice of attaching precise numbers to AI existential risk, targeting Evan Hubinger's 'greater than 10% chance that AI kills all humans in the next decade.' His argument is about implied credentials, a stated probability signals to a general audience both domain expertise and empirical backing that nobody in this field has, and he points out Hubinger graduated college in 2019. He grants the underlying risks deserve discussion, then lands on the line 'Warning people does not require a made-up probability.'