Avoid­ing AI De­cep­tion: Lie De­tec­tors can ei­ther In­duce Hon­esty or Evasion

5 Jun 2025 23:07 UTC
22 points
2 comments5 min readLW link
(far.ai)

In­tro­duc­ing: Meri­dian Cam­bridge’s new on­line lec­ture se­ries cov­er­ing fron­tier AI and AI safety

Meridian Cambridge5 Jun 2025 21:55 UTC
1 point
0 comments1 min readLW link

cheaper sodium electrolysis

bhauth5 Jun 2025 21:49 UTC
23 points
3 comments4 min readLW link
(www.bhauth.com)

His­tograms are to CDFs as cal­ibra­tion plots are to...

Optimization Process5 Jun 2025 20:20 UTC
35 points
9 comments1 min readLW link
(optimizationprocess.com)

In­te­gra­tion Band­width: The Mechanism Be­hind In­tel­li­gence and Puberty

Dortex5 Jun 2025 19:37 UTC
−1 points
4 comments1 min readLW link
(osf.io)

Levels of Doom: Eu­topia, Disem­pow­er­ment, Extinction

Vladimir_Nesov5 Jun 2025 19:08 UTC
34 points
1 comment2 min readLW link

LLM in-con­text learn­ing as (ap­prox­i­mat­ing) Solomonoff induction

Cole Wyeth5 Jun 2025 17:45 UTC
31 points
3 comments4 min readLW link

Fun­da­men­tal Uncer­tainty: Chap­ter 2 - How do words get their mean­ing?

Gordon Seidoh Worley5 Jun 2025 16:32 UTC
11 points
2 comments11 min readLW link

AI Might Kill Everyone

Bentham's Bulldog5 Jun 2025 15:37 UTC
6 points
0 comments4 min readLW link

AI #119: Good­bye AISI?

Zvi5 Jun 2025 14:00 UTC
42 points
8 comments60 min readLW link
(thezvi.wordpress.com)

Pow­er­ful Predictions

Alvin Ånestrand5 Jun 2025 10:44 UTC
2 points
0 comments6 min readLW link
(forecastingaifutures.substack.com)

Po­ten­tially Use­ful Pro­jects in Wise AI

Chris_Leong5 Jun 2025 8:13 UTC
12 points
0 comments5 min readLW link

Build­ing as gardening

Itay Dreyfus5 Jun 2025 6:41 UTC
3 points
1 comment4 min readLW link
(productidentity.co)

Semi­con­duc­tor Fabs I: The Equipment

nomagicpill4 Jun 2025 22:09 UTC
19 points
0 comments19 min readLW link
(nomagicpill.github.io)

The Stereo­type of the Stereotype

Ike4 Jun 2025 21:06 UTC
58 points
17 comments9 min readLW link

2. Why in­tu­itive com­par­i­sons of large-scale im­pact are unjustified

Anthony DiGiovanni4 Jun 2025 20:30 UTC
25 points
0 comments16 min readLW link

Dat­ing Roundup #6

Zvi4 Jun 2025 20:00 UTC
36 points
2 comments55 min readLW link
(thezvi.wordpress.com)

Ra­tional Prime Calendar

RickHull4 Jun 2025 19:30 UTC
−1 points
0 comments3 min readLW link

A Tech­nique of Pure Reason

Adam Newgas4 Jun 2025 19:07 UTC
11 points
3 comments2 min readLW link

“Flaky break­throughs” per­vade in­ner work — but al­most no one tracks them

Chris Lakin4 Jun 2025 19:02 UTC
207 points
44 comments2 min readLW link
(chrislakin.blog)

[Question] LessOn­line saved my life. Now how do I let go of this house?

RedMan4 Jun 2025 18:47 UTC
22 points
7 comments1 min readLW link

Linkpost: Pre­dict­ing Em­piri­cal AI Re­search Out­comes with Lan­guage Models

quetzal_rainbow4 Jun 2025 18:14 UTC
10 points
1 comment1 min readLW link
(arxiv.org)

Self-Co­or­di­nated De­cep­tion in Cur­rent AI Models

Avi Brach-Neufeld4 Jun 2025 17:59 UTC
8 points
5 comments4 min readLW link

To MAIM or Not to MAIM. In­tro­duc­ing MARS: The Nu­clear Deter­rent case for Har­dened Datacenters

kinsman4 Jun 2025 17:56 UTC
1 point
0 comments7 min readLW link

The Be­lo­crat: a ser­vant leader

belos4 Jun 2025 17:25 UTC
1 point
0 comments10 min readLW link
(bestofagreatlot.substack.com)

A list of books which are ad­ja­cent to EA

marco moldo4 Jun 2025 12:31 UTC
−1 points
0 comments3 min readLW link

Philo­soph­i­cal Jailbreaks: Demo of LLM Nihilism

Artem Karpov4 Jun 2025 12:03 UTC
3 points
0 comments5 min readLW link

Notes from a mini-repli­ca­tion of the al­ign­ment fak­ing paper

Ben_Snodin4 Jun 2025 11:01 UTC
13 points
5 comments9 min readLW link
(www.bensnodin.com)

ARENA 6.0 - Call for Applicants

4 Jun 2025 10:19 UTC
26 points
3 comments6 min readLW link

Quickly Assess­ing Re­ward Hack­ing-like Be­hav­ior in LLMs and its Sen­si­tivity to Prompt Variations

AndresCampero4 Jun 2025 7:22 UTC
26 points
1 comment17 min readLW link

Draft: A con­cise the­ory of agen­tic consciousness

Martin Vlach4 Jun 2025 5:00 UTC
2 points
4 comments1 min readLW link

In­di­vi­d­ual AI rep­re­sen­ta­tives don’t solve Grad­ual Disempowerement

Jan_Kulveit4 Jun 2025 1:26 UTC
60 points
3 comments3 min readLW link

Lec­tures on AI for high school stu­dents (and oth­ers)

Radford Neal3 Jun 2025 23:54 UTC
6 points
0 comments1 min readLW link
(radfordneal.wordpress.com)

Does the Taiwan in­va­sion pre­vent mankind from ob­tain­ing the al­igned ASI?

StanislavKrym3 Jun 2025 23:35 UTC
−14 points
1 comment5 min readLW link

Self-inquiry

Vadim Golub3 Jun 2025 22:15 UTC
−3 points
0 comments5 min readLW link

Ques­tion to LW devs: does LessWrong tries to be face­booky?

Roman Malov3 Jun 2025 22:08 UTC
5 points
1 comment1 min readLW link

Your Strat­egy Roadmap: Ex­pert Tips + Live Training

Deena Englander3 Jun 2025 21:10 UTC
−4 points
0 comments4 min readLW link

Steer­ing Vec­tors Can Help LLM Judges De­tect Sub­tle Dishonesty

3 Jun 2025 20:33 UTC
12 points
1 comment5 min readLW link

Schel­ling Co­or­di­na­tion via Agen­tic Loops

Callum-Luis Kindred3 Jun 2025 20:13 UTC
10 points
0 comments9 min readLW link

Vi­sual Prompt In­jec­tions: Re­sults on test­ing AI spam-defense and AI vuln­er­a­bil­ity to de­cep­tive web ads.

Seon Gunness3 Jun 2025 20:10 UTC
4 points
0 comments12 min readLW link

Broad-Spec­trum Cancer Treatments

sarahconstantin3 Jun 2025 19:40 UTC
148 points
10 comments7 min readLW link
(sarahconstantin.substack.com)

How to work through the ARENA pro­gram on your own

Leon Lang3 Jun 2025 17:38 UTC
37 points
3 comments6 min readLW link

How the veil of ig­no­rance grounds sentientism

HoVY3 Jun 2025 17:29 UTC
−2 points
21 comments6 min readLW link
(forum.effectivealtruism.org)

In Which I Make the Mis­take of Fully Cover­ing an Epi­sode of the All-In Podcast

Zvi3 Jun 2025 15:50 UTC
42 points
2 comments28 min readLW link
(thezvi.wordpress.com)

Trans­former Mo­du­lar Ad­di­tion Through A Sig­nal Pro­cess­ing Lens

Benjamin Kelley3 Jun 2025 15:32 UTC
1 point
0 comments1 min readLW link

AXRP Epi­sode 41 - Lee Sharkey on At­tri­bu­tion-based Pa­ram­e­ter Decomposition

DanielFilan3 Jun 2025 3:40 UTC
28 points
1 comment61 min readLW link

Notes on dy­namism, power, & virtue

Lizka3 Jun 2025 1:40 UTC
19 points
0 comments12 min readLW link

Trends – Ar­tifi­cial Intelligence

Archimedes3 Jun 2025 0:48 UTC
1 point
1 comment1 min readLW link
(www.bondcap.com)

LLMs might have sub­jec­tive ex­pe­riences, but no con­cepts for them

No77e2 Jun 2025 21:18 UTC
10 points
5 comments2 min readLW link

In defense of memes (and thought-ter­mi­nat­ing clichés)

Harjas Sandhu2 Jun 2025 20:18 UTC
11 points
4 comments10 min readLW link