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

So8res14 May 2025 19:00 UTC
652 points
143 comments2 min readLW link

Ori­ent­ing Toward Wizard Power

johnswentworth8 May 2025 5:23 UTC
576 points
148 comments5 min readLW link

What We Learned from Briefing 70+ Law­mak­ers on the Threat from AI

leticiagarcia27 May 2025 18:23 UTC
495 points
17 comments16 min readLW link
(substack.com)

In­ter­pretabil­ity Will Not Reli­ably Find De­cep­tive AI

Neel Nanda4 May 2025 16:32 UTC
331 points
68 comments7 min readLW link

Truth or Dare

Duncan Sabien (Inactive)29 May 2025 0:07 UTC
263 points
61 comments69 min readLW link

If you’re not sure how to sort a list or grid—se­ri­ate it!

gwern28 May 2025 3:54 UTC
218 points
8 comments3 min readLW link
(www.jstatsoft.org)

Too Soon

Gordon Seidoh Worley13 May 2025 15:01 UTC
218 points
19 comments4 min readLW link

Win­ning the power to lose

KatjaGrace20 May 2025 6:40 UTC
218 points
88 comments2 min readLW link
(worldspiritsockpuppet.com)

PSA: The LessWrong Feed­back Service

JustisMills12 May 2025 16:34 UTC
216 points
12 comments2 min readLW link

Slow­down After 2028: Com­pute, RLVR Uncer­tainty, MoE Data Wall

Vladimir_Nesov1 May 2025 13:54 UTC
200 points
35 comments5 min readLW link

Gem­ini Diffu­sion: watch this space

Yair Halberstadt20 May 2025 19:29 UTC
194 points
39 comments1 min readLW link
(deepmind.google)

The Best Refer­ence Works for Every Subject

Parker Conley14 May 2025 0:58 UTC
158 points
31 comments6 min readLW link
(parconley.com)

It’s hard to make schem­ing evals look re­al­is­tic for LLMs

24 May 2025 19:17 UTC
150 points
29 comments5 min readLW link

It’s Okay to Feel Bad for a Bit

moridinamael10 May 2025 23:24 UTC
149 points
34 comments3 min readLW link

So­cial Anx­iety Isn’t About Be­ing Liked

Chris Lakin16 May 2025 22:26 UTC
148 points
21 comments2 min readLW link
(chrislakin.blog)

Con­sider not donat­ing un­der $100 to poli­ti­cal candidates

DanielFilan11 May 2025 3:20 UTC
141 points
33 comments1 min readLW link
(danielfilan.com)

AI Doomerism in 1879

David Gross13 May 2025 2:48 UTC
139 points
36 comments8 min readLW link

Sea­son Re­cap of the Village: Agents raise $2,000

Shoshannah Tekofsky27 May 2025 12:34 UTC
135 points
14 comments6 min readLW link
(theaidigest.org)

Med­i­ta­tions on Doge

Martin Sustrik25 May 2025 12:00 UTC
131 points
44 comments9 min readLW link
(250bpm.substack.com)

Five Hinge‑Ques­tions That De­cide Whether AGI Is Five Years Away or Twenty

charlieoneill6 May 2025 2:48 UTC
127 points
17 comments5 min readLW link

Please Donate to CAIP (Post 1 of 7 on AI Gover­nance)

Mass_Driver7 May 2025 17:13 UTC
123 points
20 comments33 min readLW link

Notes on the Long Tasks METR pa­per, from a HCAST task contributor

abstractapplic4 May 2025 23:17 UTC
115 points
8 comments2 min readLW link

UK AISI’s Align­ment Team: Re­search Agenda

7 May 2025 16:33 UTC
113 points
3 comments11 min readLW link

[linkpost] One Year in DC

tlevin19 May 2025 19:46 UTC
112 points
6 comments1 min readLW link
(www.greentape.pub)

We’re Not Ad­ver­tis­ing Enough (Post 3 of 7 on AI Gover­nance)

Mass_Driver22 May 2025 17:05 UTC
110 points
10 comments28 min readLW link

AI Gover­nance to Avoid Ex­tinc­tion: The Strate­gic Land­scape and Ac­tion­able Re­search Questions

1 May 2025 22:46 UTC
109 points
7 comments8 min readLW link
(techgov.intelligence.org)

Pri­ori­tiz­ing Work

jefftk1 May 2025 2:00 UTC
109 points
11 comments1 min readLW link
(www.jefftk.com)

What OpenAI Told Cal­ifor­nia’s At­tor­ney General

garrison17 May 2025 23:14 UTC
108 points
3 comments8 min readLW link
(www.obsolete.pub)

Do you even have a sys­tem prompt? (PSA /​ repo)

Croissanthology29 May 2025 18:49 UTC
108 points
77 comments2 min readLW link

As­so­ci­a­tion taxes are col­lu­sion subsidies

KatjaGrace27 May 2025 6:50 UTC
106 points
7 comments1 min readLW link
(worldspiritsockpuppet.com)

Gen­er­at­ing the Fun­niest Joke with RL (ac­cord­ing to GPT-4.1)

agg16 May 2025 5:09 UTC
103 points
22 comments4 min readLW link

RA x Con­trolAI video: What if AI just keeps get­ting smarter?

Writer2 May 2025 14:19 UTC
100 points
18 comments9 min readLW link

Grad­ual Disem­pow­er­ment: Con­crete Re­search Projects

Raymond Douglas29 May 2025 18:55 UTC
100 points
10 comments10 min readLW link

The Ukraine War and the Kill Market

Martin Sustrik4 May 2025 7:50 UTC
98 points
14 comments5 min readLW link
(250bpm.substack.com)

The Sweet Les­son: AI Safety Should Scale With Compute

Jesse Hoogland5 May 2025 19:03 UTC
97 points
3 comments3 min readLW link

Re­quiem for the hopes of a pre-AI world

Mitchell_Porter27 May 2025 14:47 UTC
97 points
0 comments3 min readLW link

a con­fu­sion about prefer­ence orderings

nostalgebraist11 May 2025 19:30 UTC
93 points
39 comments11 min readLW link

Slow cor­po­ra­tions as an in­tu­ition pump for AI R&D automation

9 May 2025 14:49 UTC
91 points
25 comments9 min readLW link

Dreams of Ideas

Joseph Miller19 May 2025 2:15 UTC
90 points
3 comments4 min readLW link

Sleep need re­duc­tion therapies

harsimony21 May 2025 15:22 UTC
87 points
19 comments10 min readLW link
(splittinginfinity.substack.com)

Claude 4 You: Safety and Alignment

Zvi25 May 2025 14:00 UTC
86 points
8 comments63 min readLW link
(thezvi.wordpress.com)

Let­ting Kids Be Kids

Zvi30 May 2025 10:50 UTC
86 points
15 comments20 min readLW link
(thezvi.wordpress.com)

AIs at the cur­rent ca­pa­bil­ity level may be im­por­tant for fu­ture safety work

ryan_greenblatt12 May 2025 14:06 UTC
82 points
2 comments4 min readLW link

An­thropic is Quietly Backpedal­ling on its Safety Commitments

garrison23 May 2025 2:26 UTC
81 points
7 comments5 min readLW link
(www.obsolete.pub)

Bet­ter Air Purifiers

jefftk11 May 2025 16:50 UTC
81 points
21 comments3 min readLW link
(www.jefftk.com)

Misal­ign­ment and Strate­gic Un­der­perfor­mance: An Anal­y­sis of Sand­bag­ging and Ex­plo­ra­tion Hacking

8 May 2025 19:06 UTC
80 points
3 comments15 min readLW link

Highly Opinionated Ad­vice on How to Write ML Papers

Neel Nanda12 May 2025 1:59 UTC
80 points
4 comments32 min readLW link

The stakes of AI moral status

Joe Carlsmith21 May 2025 18:20 UTC
79 points
65 comments14 min readLW link
(joecarlsmith.substack.com)

Nav­i­gat­ing burnout

gw3 May 2025 22:07 UTC
77 points
2 comments9 min readLW link
(www.georgeyw.com)

Policy recom­men­da­tions re­gard­ing re­pro­duc­tive technology

TsviBT22 May 2025 14:49 UTC
76 points
2 comments3 min readLW link