Why I don’t think AGI is right around the corner
- Source
- dwarkesh.com
- Author
- Dwarkesh Patel
- Date
Context
An AI becoming helpful during one conversation is different from an assistant that keeps improving over months of work. In his June 3, 2025 essay, Dwarkesh Patel describes models adapting to his writing feedback within a session, while losing subtle preferences afterward. He contrasts this with human editors who gradually learn his audience and working style. That experience explains his skepticism about quickly replacing whole roles: useful individual outputs still leave someone managing repeated mistakes and rebuilding . His argument concerns durable learning from work, rather than denying that models can adapt within an ongoing interaction.
- context window — The maximum amount of text a model can consider at once — its working memory for the current conversation or task.
“This lack of continual learning is a huge, huge problem.”
Dwarkesh Patel
“If AI progress totally stalls today, I think less than 25% of white-collar employment goes away.”
Dwarkesh Patel
“I think we're in the GPT-2 era for computer use, but we have no free training corpus”
Dwarkesh Patel
“most proximal, concise, and accurate explanation is simply that it's powered by baby artificial intelligence.”
Dwarkesh Patel
“AGI timelines are very log-normal. It's either this decade or bust.”
Dwarkesh Patel
Checking sign-in…
Loading comments…