Stephen Fowler

Re­search Jan/​Feb 2024

[Question] To what ex­tent is the UK Govern­ment’s re­cent AI Safety push en­tirely due to Rishi Su­nak?

[Question] What are the best pub­lished pa­pers from out­side the al­ign­ment com­mu­nity that are rele­vant to Agent Foun­da­tions?

Ate­liers: But what is an Ate­lier?

Ate­liers: Motivation

Scaf­folded LLMs: Less Ob­vi­ous Concerns

[Question] What do benefi­cial TDT trades for hu­man­ity con­cretely look like?

Req­ui­site Variety

Ng and LeCun on the 6-Month Pause (Tran­script)

No Sum­mer Har­vest: Why AI Devel­op­ment Won’t Pause

100 Din­ners And A Work­shop: In­for­ma­tion Preser­va­tion And Goals

Align­ment Tar­gets and The Nat­u­ral Ab­strac­tion Hypothesis

Stephen Fowler’s Shortform

Swap and Scale

Search­ing for Mo­du­lar­ity in Large Lan­guage Models

What Makes an Idea Un­der­stand­able? On Ar­chi­tec­turally and Cul­turally Nat­u­ral Ideas.

How Do We Align an AGI Without Get­ting So­cially Eng­ineered? (Hint: Box It)

Broad Bas­ins and Data Compression

Race Along Rashomon Ridge

Iden­ti­fi­ca­tion of Nat­u­ral Modularity