Dan H

MLSN #18: Ad­ver­sar­ial Diffu­sion, Ac­ti­va­tion Or­a­cles, Weird Generalization

• Dan H 1 Dec 2025 17:18 UTC

  LW: 18 AF: 4

  2

  AF

  on: A Prag­matic Vi­sion for Interpretability

  Thank you to Neel for writing this. Most people pivot quietly.

  I’ve been most skeptical of mechanistic interpretability for years. I excluded interpretability in Unsolved Problems in ML Safety for this reason. Other fields like d/​acc (Systemic Safety) were included though, all the way back in 2021.

  Here’s are some earlier criticisms: https://​​www.lesswrong.com/​​posts/​​5HtDzRAk7ePWsiL2L/​​open-problems-in-ai-x-risk-pais-5#Transparency

  More recent commentary: https://​​ai-frontiers.org/​​articles/​​the-misguided-quest-for-mechanistic-ai-interpretability

  I think the community should reflect on its genius worship culture (in the case of Olah, a close friend of the inner circle) and epistemics: the approach was so dominant for years, and I think this outcome was entirely foreseeable.

MLSN #17: Mea­sur­ing Gen­eral AI Abil­ities and Miti­gat­ing Deception

AISN #65: Mea­sur­ing Au­toma­tion and Su­per­in­tel­li­gence Mo­ra­to­rium Letter

• Dan H 28 Oct 2025 5:09 UTC

  14 points

  4

  on: AIs should also re­fuse to work on ca­pa­bil­ities research

  This dynamic is captured in IABIED’s story and this paper from 2023: https://​​papers.ssrn.com/​​sol3/​​papers.cfm?abstract_id=4445706

AISN#64: New AGI Defi­ni­tion and Se­nate Bill Would Estab­lish Li­a­bil­ity for AI Harms

AISN #63: Cal­ifor­nia’s SB-53 Passes the Legislature

• Dan H 17 Sep 2025 3:46 UTC

  4 points

  0

  in reply to: Thane Ruthenis’s comment on: Thane Ruthe­nis’s Shortform

  just rehearsed variations on the arguments Eliezer/​MIRI already deployed

  I think they’re improved and simplified.

  My favorite chapter is “Chapter 5: Its Favorite Things.”

AISN #61: OpenAI Re­leases GPT-5

AISN #60: The AI Ac­tion Plan

AISN #59: EU Pub­lishes Gen­eral-Pur­pose AI Code of Practice

AISN #58: Se­nate Re­moves State AI Reg­u­la­tion Moratorium

AISN #57: The RAISE Act

AISN #56: Google Re­leases Veo 3

AISN #55: Trump Ad­minis­tra­tion Re­scinds AI Diffu­sion Rule, Allows Chip Sales to Gulf States

• Dan H 15 May 2025 21:36 UTC

  28 points

  4

  on: Eliezer and I wrote a book: If Any­one Builds It, Every­one Dies

  It’s a great book: it’s simple, memorable, and unusually convincing.

AISN #54: OpenAI Up­dates Restruc­ture Plan

AISN #53: An Open Let­ter At­tempts to Block OpenAI Restructuring

AISN#52: An Ex­pert Virol­ogy Benchmark

AISN #51: AI Frontiers