jsteinhardt(Jacob Steinhardt)

Ap­proach­ing Hu­man-Level Fore­cast­ing with Lan­guage Models

An­a­lyz­ing the His­tor­i­cal Rate of Catastrophes

Fore­cast­ing AI (Overview)

GPT-2030 and Catas­trophic Drives: Four Vignettes

In­trin­sic Drives and Ex­trin­sic Mi­suse: Two In­ter­twined Risks of AI

AI Pause Will Likely Back­fire (Guest Post)

AI Fore­cast­ing: Two Years In

What will GPT-2030 look like?

Com­plex Sys­tems are Hard to Control

Prin­ci­ples for Pro­duc­tive Group Meetings

Emer­gent De­cep­tion and Emer­gent Optimization

Some Les­sons Learned from Study­ing Indi­rect Ob­ject Iden­ti­fi­ca­tion in GPT-2 small

Fore­cast­ing ML Bench­marks in 2023

AI Fore­cast­ing: One Year In

How fast can we perform a for­ward pass?

Early 2022 Paper Round-up (Part 2)

Early 2022 Paper Round-up

Ap­pendix: More Is Differ­ent In Other Domains

Em­piri­cal Find­ings Gen­er­al­ize Sur­pris­ingly Far

ML Sys­tems Will Have Weird Failure Modes