All 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 20232024

AllJanFeb Mar Apr May

All1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31

There is way too much serendipity

Malmesbury19 Jan 2024 19:37 UTC

350 points

56 comments7 min readLW link

Sleeper Agents: Training Deceptive LLMs that Persist Through Safety Training

evhub, Carson Denison, Meg, Monte M, David Duvenaud, Nicholas Schiefer and Ethan Perez

12 Jan 2024 19:51 UTC

291 points

94 comments3 min readLW link

(arxiv.org)

Gentleness and the artificial Other

Joe Carlsmith2 Jan 2024 18:21 UTC

267 points

33 comments11 min readLW link

The case for ensuring that powerful AIs are controlled

ryan_greenblatt and Buck

24 Jan 2024 16:11 UTC

245 points

66 comments28 min readLW link

MIRI 2024 Mission and Strategy Update

Malo5 Jan 2024 0:20 UTC

216 points

44 comments8 min readLW link

Toward A Mathematical Framework for Computation in Superposition

Dmitry Vaintrob, jake_mendel and Kaarel

18 Jan 2024 21:06 UTC

182 points

17 comments73 min readLW link

This might be the last AI Safety Camp

Remmelt and Linda Linsefors

24 Jan 2024 9:33 UTC

181 points

33 comments1 min readLW link

The impossible problem of due process

mingyuan16 Jan 2024 5:18 UTC

180 points

63 comments14 min readLW link

Introducing Alignment Stress-Testing at Anthropic

evhub12 Jan 2024 23:51 UTC

179 points

23 comments2 min readLW link

Without fundamental advances, misalignment and catastrophe are the default outcomes of training powerful AI

Jeremy Gillen and peterbarnett

26 Jan 2024 7:22 UTC

160 points

60 comments57 min readLW link

What’s up with LLMs representing XORs of arbitrary features?

Sam Marks3 Jan 2024 19:44 UTC

154 points

61 comments16 min readLW link

Apologizing is a Core Rationalist Skill

johnswentworth2 Jan 2024 17:47 UTC

152 points

42 comments5 min readLW link

Making every researcher seek grants is a broken model

jasoncrawford26 Jan 2024 16:06 UTC

151 points

41 comments4 min readLW link

(rootsofprogress.org)

What good is G-factor if you’re dumped in the woods? A field report from a camp counselor.

Hastings12 Jan 2024 13:17 UTC

133 points

22 comments1 min readLW link

Deep atheism and AI risk

Joe Carlsmith4 Jan 2024 18:58 UTC

131 points

22 comments27 min readLW link

Processor clock speeds are not how fast AIs think

Ege Erdil29 Jan 2024 14:39 UTC

128 points

55 comments2 min readLW link

The case for training frontier AIs on Sumerian-only corpus

Alexandre Variengien, Charbel-Raphaël and Jonathan Claybrough

15 Jan 2024 16:40 UTC

127 points

14 comments3 min readLW link

Notice When People Are Directionally Correct

Chris_Leong14 Jan 2024 14:12 UTC

127 points

7 comments2 min readLW link

An even deeper atheism

Joe Carlsmith11 Jan 2024 17:28 UTC

125 points

47 comments15 min readLW link

A Shutdown Problem Proposal

johnswentworth and David Lorell

21 Jan 2024 18:12 UTC

122 points

61 comments6 min readLW link

Steering Llama-2 with contrastive activation additions

Nina Rimsky, Wuschel Schulz, NickGabs, Meg, evhub and TurnTrout

2 Jan 2024 0:47 UTC

121 points

29 comments8 min readLW link

(arxiv.org)

Why I take short timelines seriously

NicholasKees28 Jan 2024 22:27 UTC

117 points

29 comments4 min readLW link

Four visions of Transformative AI success

Steven Byrnes17 Jan 2024 20:45 UTC

112 points

22 comments15 min readLW link

Practically A Book Review: Appendix to “Nonlinear’s Evidence: Debunking False and Misleading Claims” (ThingOfThings)

tailcalled3 Jan 2024 17:07 UTC

111 points

25 comments2 min readLW link

(thingofthings.substack.com)

Gender Exploration

sapphire14 Jan 2024 18:57 UTC

110 points

23 comments5 min readLW link

(open.substack.com)

Being nicer than Clippy

Joe Carlsmith16 Jan 2024 19:44 UTC

109 points

23 comments27 min readLW link

′ petertodd’’s last stand: The final days of open GPT-3 research

mwatkins22 Jan 2024 18:47 UTC

107 points

15 comments45 min readLW link

The case for more ambitious language model evals

Jozdien30 Jan 2024 0:01 UTC

105 points

25 comments5 min readLW link

2023 in AI predictions

jessicata1 Jan 2024 5:23 UTC

105 points

34 comments5 min readLW link

Deceptive AI ≠ Deceptively-aligned AI

Steven Byrnes7 Jan 2024 16:55 UTC

97 points

19 comments6 min readLW link

Almost everyone I’ve met would be well-served thinking more about what to focus on

Henrik Karlsson5 Jan 2024 21:01 UTC

95 points

8 comments11 min readLW link

(www.henrikkarlsson.xyz)

RAND report finds no effect of current LLMs on viability of bioterrorism attacks

StellaAthena25 Jan 2024 19:17 UTC

94 points

14 comments1 min readLW link

(www.rand.org)

On the abolition of man

Joe Carlsmith18 Jan 2024 18:17 UTC

88 points

18 comments41 min readLW link

The Aspiring Rationalist Congregation

maia10 Jan 2024 22:52 UTC

86 points

23 comments10 min readLW link

Sparse Autoencoders Work on Attention Layer Outputs

Connor Kissane, robertzk, Arthur Conmy and Neel Nanda

16 Jan 2024 0:26 UTC

82 points

5 comments19 min readLW link

Catching AIs red-handed

ryan_greenblatt and Buck

5 Jan 2024 17:43 UTC

82 points

18 comments17 min readLW link

An Introduction To The Mandelbrot Set That Doesn’t Mention Complex Numbers

Yitz17 Jan 2024 9:48 UTC

81 points

11 comments9 min readLW link

Survey of 2,778 AI authors: six parts in pictures

KatjaGrace6 Jan 2024 4:43 UTC

80 points

1 comment2 min readLW link

Palworld development blog post

bhauth28 Jan 2024 5:56 UTC

79 points

12 comments1 min readLW link

(note.com)

Some Vacation Photos

johnswentworth4 Jan 2024 17:15 UTC

78 points

0 comments1 min readLW link

Universal Love Integration Test: Hitler

Raemon10 Jan 2024 23:55 UTC

76 points

65 comments9 min readLW link

When “yang” goes wrong

Joe Carlsmith8 Jan 2024 16:35 UTC

72 points

6 comments13 min readLW link

The True Story of How GPT-2 Became Maximally Lewd

Writer and Jai

18 Jan 2024 21:03 UTC

70 points

7 comments6 min readLW link

(youtu.be)

Epistemic Hell

rogersbacon27 Jan 2024 17:13 UTC

69 points

20 comments14 min readLW link

[Repost] The Copenhagen Interpretation of Ethics

mesaoptimizer25 Jan 2024 15:20 UTC

69 points

3 comments5 min readLW link

(web.archive.org)

InterLab – a toolkit for experiments with multi-agent interactions

Tomáš Gavenčiak, Ada Böhm and Jan_Kulveit

22 Jan 2024 18:23 UTC

68 points

0 comments8 min readLW link

(acsresearch.org)

OpenAI’s Preparedness Framework: Praise & Recommendations

Akash2 Jan 2024 16:20 UTC

66 points

1 comment7 min readLW link

The Perceptron Controversy

Yuxi_Liu10 Jan 2024 23:07 UTC

65 points

18 comments1 min readLW link

(yuxi-liu-wired.github.io)

We need a Science of Evals

Marius Hobbhahn and Jérémy Scheurer

22 Jan 2024 20:30 UTC

63 points

13 comments9 min readLW link

[Question] Will quantum randomness affect the 2028 election?

Thomas Kwa and habryka

24 Jan 2024 22:54 UTC

63 points

48 comments1 min readLW link