[Question] To what ex­tent is AI safety work try­ing to get AI to re­li­ably and safely do what the user asks vs. do what is best in some ul­ti­mate sense?

Jordan Arel23 May 2025 21:05 UTC
14 points
3 comments1 min readLW link

Notes on Claude 4 Sys­tem Card

Dentosal23 May 2025 15:23 UTC
19 points
2 comments6 min readLW link

What is empti­ness?

Vadim Golub23 May 2025 12:06 UTC
−4 points
11 comments9 min readLW link

Idiohobbies

dkl923 May 2025 6:38 UTC
11 points
2 comments1 min readLW link
(dkl9.net)

Qual­i­ta­tive Fit Testing

jefftk23 May 2025 2:50 UTC
10 points
0 comments2 min readLW link
(www.jefftk.com)

An­thropic is Quietly Backpedal­ling on its Safety Commitments

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

Learn­ing (more) from horse em­ploy­ment history

Tim H23 May 2025 2:11 UTC
68 points
13 comments5 min readLW link

Schizobench: Doc­u­ment­ing Mag­i­cal-Think­ing Be­hav­ior in Claude 4 Opus

viemccoy23 May 2025 1:31 UTC
23 points
0 comments1 min readLW link
(metanomicon.ink)

Post-Man­i­fest cowork­ing at Mox

23 May 2025 0:20 UTC
4 points
1 comment1 min readLW link

Claude 4, Op­por­tunis­tic Black­mail, and “Pleas”

Stephen Martin22 May 2025 19:59 UTC
30 points
2 comments2 min readLW link

Prob­lems in AI Align­ment: A Scale Model

Mickey Muldoon22 May 2025 19:22 UTC
−1 points
3 comments2 min readLW link
(muldoon.cloud)

Art Is Art: AI Is the Next Erotica

Charlie Edwards22 May 2025 18:04 UTC
0 points
1 comment14 min readLW link

Re­ward but­ton alignment

Steven Byrnes22 May 2025 17:36 UTC
50 points
15 comments12 min readLW link

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

Claude 4

Zach Stein-Perlman22 May 2025 17:00 UTC
71 points
24 comments1 min readLW link
(www.anthropic.com)

Video and tran­script of talk on AI welfare

Joe Carlsmith22 May 2025 16:15 UTC
24 points
1 comment28 min readLW link
(joecarlsmith.substack.com)

What we can learn from af­ter­life myths

jchan22 May 2025 15:49 UTC
5 points
0 comments15 min readLW link

Policy recom­men­da­tions re­gard­ing re­pro­duc­tive technology

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

AI #117: OpenAI Buys De­vice Maker IO

Zvi22 May 2025 13:40 UTC
37 points
9 comments62 min readLW link
(thezvi.wordpress.com)

Does BPC-157 work for heal­ing and tis­sue re­pair?

ChristianKl22 May 2025 13:18 UTC
17 points
0 comments5 min readLW link
(somaticsignals.jollyjoyjourney.com)

[Question] How load-bear­ing is KL di­ver­gence from a known-good base model in mod­ern RL?

faul_sname22 May 2025 12:08 UTC
22 points
3 comments4 min readLW link

Chris­ti­an­ity vs. Tantra vs. Sex – one spiritual path?

pchvykov22 May 2025 11:15 UTC
−2 points
0 comments24 min readLW link

Mir­ror Or­ganisms Are Not Im­mune to Predation

Matt Dellago22 May 2025 11:10 UTC
27 points
5 comments1 min readLW link

How 2025 AI Fore­casts Fared So Far

22 May 2025 9:42 UTC
11 points
2 comments8 min readLW link
(theaidigest.org)

Con­tain and ver­ify: The endgame of US-China AI competition

sjadler22 May 2025 8:13 UTC
6 points
7 comments2 min readLW link
(open.substack.com)

Laugencroissant

Martin Sustrik22 May 2025 6:30 UTC
13 points
0 comments3 min readLW link
(250bpm.substack.com)

Google I/​O Day

Zvi21 May 2025 22:00 UTC
49 points
0 comments20 min readLW link
(thezvi.wordpress.com)

Pod­cast: How not to waste a billion dol­lars (on your clini­cal trial), with Meri Beck­with on Devel­op­ment & Research

rossry21 May 2025 21:27 UTC
25 points
0 comments3 min readLW link
(developmentandresearch.bio)

Pod­cast: From molecule to medicine, with Ross Rhe­in­gans-Yoo on Com­plex Systems

rossry21 May 2025 21:08 UTC
15 points
0 comments5 min readLW link
(www.complexsystemspodcast.com)

The stakes of AI moral status

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

[Question] Which AI Safety tech­niques will be in­effec­tive against diffu­sion mod­els?

Allen Thomas21 May 2025 18:13 UTC
6 points
1 comment1 min readLW link

Through The Look­ing Glasses: Is­sues & Solu­tions for Aug­mented Reality

claywren21 May 2025 18:11 UTC
1 point
0 comments22 min readLW link

Root­ing for Mo­ments, Not Jerseys. Another Ap­proach to En­joy­ing Sports

Ahmed Elsayyad21 May 2025 18:11 UTC
1 point
0 comments3 min readLW link

Un­ex­ploitable search: block­ing mal­i­cious use of free parameters

21 May 2025 17:23 UTC
34 points
16 comments6 min readLW link

The Real AI Safety Risk Is a Con­cep­tual Ex­ploit: Anthropomorphism

Anthony Fox21 May 2025 16:29 UTC
−2 points
0 comments2 min readLW link

You Can’t Skip Ex­plo­ra­tion: Why un­der­stand­ing ex­per­i­men­ta­tion and taste is key to un­der­stand­ing AI

Oliver Sourbut21 May 2025 16:08 UTC
20 points
0 comments11 min readLW link
(www.oliversourbut.net)

The Prob­lem and Op­por­tu­nity of Scale

belos21 May 2025 15:52 UTC
1 point
0 comments5 min readLW link
(bestofagreatlot.substack.com)

Sleep need re­duc­tion therapies

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

Parental Guidance: Fram­ing Su­per­in­tel­li­gence

ejk6421 May 2025 15:01 UTC
10 points
0 comments3 min readLW link

Why Aren’t Ra­tion­al­ists Win­ning (Again)

k6421 May 2025 14:46 UTC
6 points
25 comments5 min readLW link

Can We Nat­u­ral­ize Mo­ral Episte­mol­ogy?

tylermjohn 21 May 2025 14:25 UTC
52 points
22 comments6 min readLW link

Units Have More Depth Than I Thought

Morpheus21 May 2025 13:51 UTC
33 points
5 comments1 min readLW link

Hu­mans are Inse­cure Pass­word Generators

Isaac King21 May 2025 5:58 UTC
16 points
0 comments5 min readLW link

[Cross­post] An­thropic Shadow Geopolitics

akarlin21 May 2025 4:50 UTC
9 points
5 comments18 min readLW link

The Need for Poli­ti­cal Ad­ver­tis­ing (Post 2 of 7 on AI Gover­nance)

Mass_Driver21 May 2025 0:44 UTC
59 points
2 comments13 min readLW link

Notes from Dopamine Detoxing

Alice Blair20 May 2025 23:43 UTC
15 points
4 comments9 min readLW link

Re­vis­it­ing the ideas for non-neu­ralese architectures

StanislavKrym20 May 2025 23:35 UTC
2 points
0 comments1 min readLW link

Gem­ini Diffu­sion: watch this space

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

A Sketch of Be­loc­racy: a new sys­tem of governance

belos20 May 2025 18:30 UTC
5 points
0 comments8 min readLW link
(bestofagreatlot.substack.com)

The Codex of Ul­ti­mate Vibing

Zvi20 May 2025 18:30 UTC
45 points
2 comments11 min readLW link
(thezvi.wordpress.com)