New o1-like model (QwQ) beats Claude 3.5 Son­net with only 32B parameters

Jesse HooglandNov 27, 2024, 10:06 PM
68 points
4 comments1 min readLW link
(qwenlm.github.io)

Per­sonal AI Planning

jefftkNov 10, 2024, 2:00 PM
68 points
11 comments2 min readLW link
(www.jefftk.com)

SAEs are highly dataset de­pen­dent: a case study on the re­fusal direction

Nov 7, 2024, 5:22 AM
66 points
4 comments14 min readLW link

AI Craftsmanship

abramdemskiNov 11, 2024, 10:17 PM
66 points
7 comments4 min readLW link

The Third Fun­da­men­tal Question

ScrewtapeNov 15, 2024, 4:01 AM
66 points
7 comments6 min readLW link

Win/​con­tinue/​lose sce­nar­ios and ex­e­cute/​re­place/​au­dit protocols

BuckNov 15, 2024, 3:47 PM
64 points
2 comments7 min readLW link

Per­ils of Gen­er­al­iz­ing from One’s So­cial Group

localdeityNov 24, 2024, 3:31 PM
63 points
1 comment3 min readLW link

elec­tric turbofans

bhauthNov 2, 2024, 10:50 PM
63 points
2 comments5 min readLW link
(bhauth.com)

Why our poli­ti­ci­ans aren’t Median

Yair HalberstadtNov 3, 2024, 2:03 PM
62 points
15 comments3 min readLW link

Why im­perfect ad­ver­sar­ial ro­bust­ness doesn’t doom AI control

Nov 18, 2024, 4:05 PM
62 points
25 comments2 min readLW link

Seek­ing Collaborators

abramdemskiNov 1, 2024, 5:13 PM
62 points
15 comments7 min readLW link

Train­ing AI agents to solve hard prob­lems could lead to Scheming

Nov 19, 2024, 12:10 AM
61 points
12 comments28 min readLW link

Read­ing RFK Jr so that you don’t have to

bracesNov 22, 2024, 12:59 AM
59 points
1 comment8 min readLW link

[Question] Could or­cas be (trained to be) smarter than hu­mans? 

Towards_KeeperhoodNov 4, 2024, 11:29 PM
56 points
23 comments1 min readLW link

U.S.-China Eco­nomic and Se­cu­rity Re­view Com­mis­sion pushes Man­hat­tan Pro­ject-style AI initiative

worseNov 19, 2024, 6:42 PM
56 points
7 comments1 min readLW link

The Evals Gap

Marius HobbhahnNov 11, 2024, 4:42 PM
55 points
7 comments7 min readLW link
(www.apolloresearch.ai)

a space habitat design

bhauthNov 25, 2024, 5:28 PM
55 points
13 comments9 min readLW link
(bhauth.com)

Es­ti­mates of GPU or equiv­a­lent re­sources of large AI play­ers for 2024/​5

CharlesDNov 28, 2024, 11:01 PM
54 points
7 comments9 min readLW link

A Con­flicted Linkspost

ScrewtapeNov 21, 2024, 12:37 AM
52 points
0 comments3 min readLW link

Which evals re­sources would be good?

Marius HobbhahnNov 16, 2024, 2:24 PM
51 points
4 comments5 min readLW link

Epistemic sta­tus: po­etry (and other po­ems)

Richard_NgoNov 21, 2024, 6:13 PM
51 points
5 comments2 min readLW link
(www.narrativeark.xyz)

On Tar­geted Ma­nipu­la­tion and De­cep­tion when Op­ti­miz­ing LLMs for User Feedback

Nov 7, 2024, 3:39 PM
51 points
7 comments11 min readLW link

Me­tastatic Cancer Treat­ment Since 2010: The Suc­cess Stories

sarahconstantinNov 4, 2024, 10:50 PM
51 points
2 comments6 min readLW link
(sarahconstantin.substack.com)

Two in­ter­views with the founder of DeepSeek

Cosmia_NebulaNov 29, 2024, 3:18 AM
50 points
6 comments31 min readLW link
(rentry.co)

The Choice Transition

Nov 18, 2024, 12:30 PM
50 points
4 comments15 min readLW link
(strangecities.substack.com)

Dave Kas­ten’s AGI-by-2027 vignette

davekastenNov 26, 2024, 11:20 PM
49 points
8 comments5 min readLW link

Ac­tive Re­call and Spaced Rep­e­ti­tion are Differ­ent Things

Saul MunnNov 8, 2024, 8:14 PM
49 points
2 comments3 min readLW link
(www.brasstacks.blog)

Look­ing back on the Fu­ture of Hu­man­ity In­sti­tute—Asterisk

jakeeatonNov 19, 2024, 12:44 AM
48 points
0 comments1 min readLW link

An al­ter­na­tive ap­proach to superbabies

Towards_KeeperhoodNov 5, 2024, 10:56 PM
48 points
19 comments3 min readLW link

The Shal­low Bench

Karl FaulksNov 5, 2024, 5:07 AM
48 points
5 comments3 min readLW link

Live Machin­ery: An In­ter­face De­sign Philos­o­phy for Whole­some AI Futures

SahilNov 1, 2024, 5:24 PM
48 points
3 comments35 min readLW link

What Ke­tamine Ther­apy Is Like

SableNov 11, 2024, 11:09 AM
47 points
8 comments6 min readLW link
(affablyevil.substack.com)

AI #91: Deep Thinking

ZviNov 21, 2024, 2:30 PM
47 points
11 comments56 min readLW link
(thezvi.wordpress.com)

An­a­lyz­ing how SAE fea­tures evolve across a for­ward pass

Nov 7, 2024, 10:07 PM
47 points
0 comments1 min readLW link
(arxiv.org)

Monthly Roundup #24: Novem­ber 2024

ZviNov 18, 2024, 1:20 PM
44 points
14 comments50 min readLW link
(thezvi.wordpress.com)

Liter­acy Rates Haven’t Fallen By 20% Since the Depart­ment of Ed­u­ca­tion Was Created

Maxwell TabarrokNov 22, 2024, 8:53 PM
44 points
0 comments3 min readLW link
(www.maximum-progress.com)

Danger­ous ca­pa­bil­ity tests should be harder

LucaRighettiNov 21, 2024, 5:20 PM
44 points
3 comments5 min readLW link
(www.planned-obsolescence.org)

Em­pa­thy/​Sys­tem­iz­ing Quo­tient is a poor/​bi­ased model for the autism/​sex link

tailcalledNov 4, 2024, 9:11 PM
43 points
0 comments7 min readLW link

ARENA 4.0 Im­pact Report

Nov 27, 2024, 8:51 PM
43 points
3 comments13 min readLW link

AI #89: Trump Card

ZviNov 7, 2024, 4:30 PM
42 points
12 comments42 min readLW link
(thezvi.wordpress.com)

Causal in­fer­ence for the home gardener

bracesNov 27, 2024, 5:55 PM
42 points
1 comment5 min readLW link

Col­lege tech­ni­cal AI safety hackathon ret­ro­spec­tive—Ge­or­gia Tech

yixNov 15, 2024, 12:22 AM
41 points
2 comments5 min readLW link
(open.substack.com)

Lo­cally op­ti­mal psychology

ChipmonkNov 25, 2024, 6:35 PM
41 points
7 comments2 min readLW link
(twitter.com)

How to use bright light to im­prove your life.

Nat MartinNov 18, 2024, 7:32 PM
40 points
10 comments10 min readLW link

Sig­nal­ing with Small Orange Diamonds

jefftkNov 7, 2024, 8:20 PM
40 points
1 comment1 min readLW link
(www.jefftk.com)

Win­ning isn’t enough

Nov 5, 2024, 11:37 AM
40 points
18 comments9 min readLW link

In­trin­sic Power-Seek­ing: AI Might Seek Power for Power’s Sake

TurnTroutNov 19, 2024, 6:36 PM
40 points
5 comments1 min readLW link
(turntrout.com)

[Question] Are You More Real If You’re Really For­get­ful?

Thane RuthenisNov 24, 2024, 7:30 PM
39 points
25 comments5 min readLW link

A Sober Look at Steer­ing Vec­tors for LLMs

Nov 23, 2024, 5:30 PM
38 points
0 comments5 min readLW link

Do­ing Re­search Part-Time is Great

casualphysicsenjoyerNov 22, 2024, 7:01 PM
38 points
7 comments5 min readLW link