Paul Graham Says 1980 View Would Call Current Models AGI

Paul Graham posted that people from 1980 would regard today's AI models as AGI. Pedro Domingos answered that observers would not, pointing to the absence of household robots. Xuan added that 1980 viewers would change their minds after seeing weak performance in Pong and Tetris. Alfred Lin noted that AI research has run since the 1950s while definitions of success keep shifting. David Pfau and others observed that people still distinguish LLM output from human text in ordinary use. Trevor Blackwell replied that easy detection does not rule out passing the Turing test.