3 Comments

Post hoc observation: of course ChatGPT does better with political predictions. It is a wordcel after all

Expand full comment

Politicians are effectively large language models: they are rewarded for sounded correct, not for producing real outcomes. So maybe that’s why political outcomes are easier for it to predict?

Expand full comment

"Politics" is at the top and "International politics" is close behind but "American politics" is near the bottom. Why the discrepancy?

Expand full comment