Clarifying and predicting AGI by Richard Ngo
The Inside View
00:00
The Generality of AGI
Daniel Koko Tajlo, who also works at OpenAI, believes otherwise. He claims that AIs would end up generalising from tasks that take humans a short amount of time to tasks that takes humans a long time. In the comments section, Richard gives a rough ballpark estimate. If right now we're at one second AGI and we assume that we could gain an order of magnitude every one and a half years, then we would reach something like a one-day AGI in roughly six to seven years.
Transcript
Play full episode
