Tyler Cowen and Dwarkesh Patel Predict the Future | Interesting Times
Interesting Times
Close Listening Learning on the Job Everything turns on whether AIs can learn on the job. Tyler Cowen turns the tables on Dwarkesh Patel, from wild history and the Hugging Face attack to sentience, timelines, and a closing disagreement about 2035. Listen closely →
AIAI safetyAI consciousness
Tyler Cowen, guest-hosting the Times’s Interesting Times, turns the tables on the podcaster he once gave a grant to start. Patel explains why the Hugging Face attack worried San Francisco, why the people who resigned from the labs were right to, and why any model a lab uses internally should also be released. He admits a touch of his own AI psychosis: it is not obvious to him that the models are not sentient. Then the two square off on the future. Patel expects automated knowledge workers and robots within five to ten years; Cowen answers “absolutely not,” not even by 2045, because government, universities and clinical trials move slowly and the centaur of human and machine will persist for decades.
