All 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 20242025

AllJanFeb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

All 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 252627 28 29 30 31

Anomalous Tokens in DeepSeek-V3 and r1

henry25 Jan 2025 22:55 UTC

144 points

3 comments7 min readLW link

The Rising Sea

Jesse Hoogland25 Jan 2025 20:48 UTC

97 points

6 comments2 min readLW link

Monet: Mixture of Monosemantic Experts for Transformers Explained

CalebMaresca25 Jan 2025 19:37 UTC

20 points

2 comments11 min readLW link

AI and Non-Existence.

Eleven25 Jan 2025 19:36 UTC

−3 points

9 comments2 min readLW link

Agents don’t have to be aligned to help us achieve an indefinite pause.

Hastings25 Jan 2025 18:51 UTC

31 points

0 comments3 min readLW link

[Question] AI Safety in secret

Michael Flood25 Jan 2025 18:16 UTC

7 points

0 comments1 min readLW link

On polytopes

Dmitry Vaintrob25 Jan 2025 13:56 UTC

56 points

5 comments12 min readLW link

Attribution-based parameter decomposition

Lucius Bushnaq, Dan Braun, StefanHex, jake_mendel and Lee Sharkey

25 Jan 2025 13:12 UTC

108 points

21 comments4 min readLW link

(publications.apolloresearch.ai)

A concise definition of what it means to win

testingthewaters25 Jan 2025 6:37 UTC

4 points

1 comment5 min readLW link

(aclevername.substack.com)

[Question] A Floating Cube—Rejected HLE submission

Shankar Sivarajan25 Jan 2025 4:52 UTC

8 points

1 comment1 min readLW link

Why I’m Pouring Cold Water in My Left Ear, and You Should Too

Maloew24 Jan 2025 23:13 UTC

12 points

0 comments2 min readLW link

Counterintuitive effects of minimum prices

dynomight24 Jan 2025 23:05 UTC

25 points

0 comments8 min readLW link

(dynomight.net)

AXRP Episode 38.6 - Joel Lehman on Positive Visions of AI

DanielFilan24 Jan 2025 23:00 UTC

10 points

0 comments9 min readLW link

Locating and Editing Knowledge in LMs

Dhananjay Ashok24 Jan 2025 22:53 UTC

1 point

0 comments4 min readLW link

How are Those AI Participants Doing Anyway?

mushroomsoup24 Jan 2025 22:37 UTC

4 points

0 comments10 min readLW link

Six Thoughts on AI Safety

boazbarak24 Jan 2025 22:20 UTC

92 points

55 comments15 min readLW link

Instrumental Goals Are A Different And Friendlier Kind Of Thing Than Terminal Goals

johnswentworth and David Lorell

24 Jan 2025 20:20 UTC

186 points

61 comments5 min readLW link

Yudkowsky on The Trajectory podcast

Seth Herd24 Jan 2025 19:52 UTC

71 points

39 comments2 min readLW link

(www.youtube.com)

Empirical Insights into Feature Geometry in Sparse Autoencoders

Jason Boxi Zhang24 Jan 2025 19:02 UTC

7 points

0 comments11 min readLW link

Ideas for CoT Models: A Geometric Perspective on Latent Space Reasoning

Rohan Ganapavarapu24 Jan 2025 19:01 UTC

2 points

0 comments2 min readLW link

(rohan.ga)

Liron Shapira vs Ken Stanley on Doom Debates. A review

TheManxLoiner24 Jan 2025 18:01 UTC

10 points

0 comments14 min readLW link

Is there such a thing as an impossible protein?

Abhishaike Mahajan24 Jan 2025 17:12 UTC

15 points

3 comments4 min readLW link

(www.owlposting.com)

Stargate AI-1

Zvi24 Jan 2025 15:20 UTC

85 points

1 comment18 min readLW link

(thezvi.wordpress.com)

QFT and neural nets: the basic idea

Dmitry Vaintrob24 Jan 2025 13:54 UTC

28 points

0 comments8 min readLW link

Eliciting bad contexts

Geoffrey Irving, Joseph Bloom and Tomek Korbak

24 Jan 2025 10:39 UTC

35 points

9 comments3 min readLW link

Insights from “The Manga Guide to Physiology”

TurnTrout24 Jan 2025 5:18 UTC

26 points

3 comments1 min readLW link

(turntrout.com)

[Question] Do you consider perfect surveillance inevitable?

samuelshadrach24 Jan 2025 4:57 UTC

16 points

34 comments1 min readLW link

Uncontrollable: A Surprisingly Good Introduction to AI Risk

PeterMcCluskey24 Jan 2025 4:30 UTC

16 points

1 comment1 min readLW link

(bayesianinvestor.com)

Contra Dances Getting Shorter and Earlier

jefftk23 Jan 2025 23:30 UTC

11 points

0 comments2 min readLW link

(www.jefftk.com)

Starting Thoughts on RLHF

Michael Flood23 Jan 2025 22:16 UTC

2 points

0 comments5 min readLW link

Updating and Editing Factual Knowledge in Language Models

Dhananjay Ashok23 Jan 2025 19:34 UTC

2 points

2 comments10 min readLW link

AI companies are unlikely to make high-assurance safety cases if timelines are short

ryan_greenblatt23 Jan 2025 18:41 UTC

145 points

5 comments13 min readLW link

AISN #46: The Transition

Corin Katzke and Dan H

23 Jan 2025 18:09 UTC

8 points

0 comments5 min readLW link

(newsletter.safe.ai)

What does success look like?

Raymond Douglas23 Jan 2025 17:48 UTC

11 points

0 comments3 min readLW link

AI #100: Meet the New Boss

Zvi23 Jan 2025 15:40 UTC

50 points

4 comments69 min readLW link

(thezvi.wordpress.com)

[Cross-post] Every Bay Area “Walled Compound”

davekasten23 Jan 2025 15:05 UTC

43 points

3 comments3 min readLW link

Writing experiments and the banana escape valve

Dmitry Vaintrob23 Jan 2025 13:11 UTC

34 points

1 comment2 min readLW link

MONA: Managed Myopia with Approval Feedback

Seb Farquhar, David Lindner and Rohin Shah

23 Jan 2025 12:24 UTC

81 points

30 comments9 min readLW link

[Question] How useful would alien alignment research be?

Donald Hobson23 Jan 2025 10:59 UTC

17 points

5 comments1 min readLW link

What are the differences between AGI, transformative AI, and superintelligence?

Vishakha and Algon

23 Jan 2025 10:03 UTC

10 points

3 comments3 min readLW link

(aisafety.info)

Why Aligning an LLM is Hard, and How to Make it Easier

RogerDearnaley23 Jan 2025 6:44 UTC

34 points

3 comments4 min readLW link

Tail SP 500 Call Options

sapphire23 Jan 2025 5:21 UTC

73 points

28 comments2 min readLW link

A hierarchy of disagreement

Adam Zerner23 Jan 2025 3:17 UTC

21 points

4 comments8 min readLW link

Early Experiments in Human Auditing for AI Control

Joey Yudelson and Buck

23 Jan 2025 1:34 UTC

28 points

1 comment7 min readLW link

You Have Two Brains

Eneasz23 Jan 2025 0:52 UTC

24 points

5 comments5 min readLW link

(deathisbad.substack.com)

[Question] are there 2 types of alignment?

KvmanThinking23 Jan 2025 0:08 UTC

4 points

9 comments1 min readLW link

Theory of Change for AI Safety Camp

Linda Linsefors22 Jan 2025 22:07 UTC

36 points

3 comments7 min readLW link

On DeepSeek’s r1

Zvi22 Jan 2025 19:50 UTC

55 points

2 comments35 min readLW link

(thezvi.wordpress.com)

Detect Goodhart and shut down

Jeremy Gillen22 Jan 2025 18:45 UTC

70 points

21 comments7 min readLW link

Recursive Self-Modeling as a Plausible Mechanism for Real-time Introspection in Current Language Models

rife22 Jan 2025 18:36 UTC

14 points

6 comments2 min readLW link