Mech In­terp Challenge: Septem­ber—De­ci­pher­ing the Ad­di­tion Model

CallumMcDougallSep 13, 2023, 10:23 PM
35 points
0 comments4 min readLW link

Linkpost for Jan Leike on Self-Exfiltration

Daniel KokotajloSep 13, 2023, 9:23 PM
59 points
1 comment2 min readLW link
(aligned.substack.com)

MLSN: #10 Ad­ver­sar­ial At­tacks Against Lan­guage and Vi­sion Models, Im­prov­ing LLM Hon­esty, and Trac­ing the In­fluence of LLM Train­ing Data

Sep 13, 2023, 6:03 PM
15 points
1 comment5 min readLW link
(newsletter.mlsafety.org)

Ex­pand­ing the Scope of Superposition

Derek LarsonSep 13, 2023, 5:38 PM
10 points
0 comments4 min readLW link

Con­tra Yud­kowsky on Epistemic Con­duct for Author Criticism

Zack_M_DavisSep 13, 2023, 3:33 PM
70 points
38 comments7 min readLW link

Ap­ply to lead a pro­ject dur­ing the next vir­tual AI Safety Camp

Sep 13, 2023, 1:29 PM
19 points
0 comments5 min readLW link
(aisafety.camp)

Is AI Safety drop­ping the ball on pri­vacy?

markovSep 13, 2023, 1:07 PM
50 points
17 comments7 min readLW link

UDT shows that de­ci­sion the­ory is more puz­zling than ever

Wei DaiSep 13, 2023, 12:26 PM
219 points
56 comments1 min readLW link

[Question] Align­ment & Ca­pa­bil­ities: what’s the differ­ence?

johnhalsteadSep 13, 2023, 11:48 AM
6 points
3 comments1 min readLW link

Duty to res­cue /​ Non-as­sis­tance à per­sonne en danger

Thomas SepulchreSep 13, 2023, 9:49 AM
15 points
5 comments3 min readLW link

The Flow-Through Fallacy

Chris_LeongSep 13, 2023, 4:28 AM
21 points
7 comments1 min readLW link

Book re­view: The Im­por­tance of What We Care About (Harry G. Frank­furt)

David GrossSep 13, 2023, 4:17 AM
7 points
0 comments4 min readLW link

Padding the Corner

jefftkSep 13, 2023, 1:30 AM
32 points
4 comments1 min readLW link
(www.jefftk.com)

[Question] Should an un­der­grad avoid a ca­pa­bil­ities pro­ject?

DoubleSep 12, 2023, 11:16 PM
4 points
2 comments1 min readLW link

[Linkpost] Con­tra four-wheeled suit­cases, sort of

Gunnar_ZarnckeSep 12, 2023, 8:36 PM
18 points
4 comments1 min readLW link
(dynomight.substack.com)

Seek­ing Feed­back on My Mechanis­tic In­ter­pretabil­ity Re­search Agenda

RGRGRGSep 12, 2023, 6:45 PM
3 points
1 comment3 min readLW link

Au­to­mat­i­cally find­ing fea­ture vec­tors in the OV cir­cuits of Trans­form­ers with­out us­ing probing

Jacob DunefskySep 12, 2023, 5:38 PM
16 points
2 comments29 min readLW link

Startup Roundup #1: Happy Demo Day

ZviSep 12, 2023, 1:20 PM
38 points
5 comments15 min readLW link
(thezvi.wordpress.com)

[Question] Is there some­thing fun­da­men­tally wrong with the Uni­verse?

Caerulea-LawrenceSep 12, 2023, 12:02 PM
6 points
80 comments2 min readLW link

Stu­pidity is also hard

walkthroughwallsSep 12, 2023, 2:45 AM
−8 points
4 comments2 min readLW link

Ap­ple Cider Baklava

jefftkSep 12, 2023, 2:10 AM
15 points
0 comments1 min readLW link
(www.jefftk.com)

How use­ful is Cor­rigi­bil­ity?

martinkunevSep 12, 2023, 12:05 AM
11 points
4 comments5 min readLW link

Con­tra Heighn Con­tra Me Con­tra Func­tional De­ci­sion The­ory

Bentham's BulldogSep 11, 2023, 7:49 PM
−10 points
14 comments6 min readLW link

Ma­chine Evolution

Sep 11, 2023, 7:29 PM
11 points
2 comments22 min readLW link

[Question] Is there a hard copy of the se­quences available any­where?

Cole WyethSep 11, 2023, 7:01 PM
3 points
1 comment1 min readLW link

Ama­zon KDP AI con­tent guidelines

ChristianKlSep 11, 2023, 6:36 PM
12 points
0 comments1 min readLW link

A Case for AI Safety via Law

JWJohnstonSep 11, 2023, 6:26 PM
20 points
12 comments4 min readLW link

Erdős Prob­lems in Al­gorith­mic Probability

Aidan RockeSep 11, 2023, 4:44 PM
13 points
4 comments2 min readLW link

PSA: The com­mu­nity is in Berkeley/​Oak­land, not “the Bay Area”

maiaSep 11, 2023, 3:59 PM
105 points
7 comments1 min readLW link

A Bat and Ball made me Sad

Darren McKeeSep 11, 2023, 1:48 PM
14 points
26 comments1 min readLW link

Fo­cus on the Hardest Part First

Johannes C. MayerSep 11, 2023, 7:53 AM
42 points
13 comments1 min readLW link

The Promises and Pit­falls of Long-Term Forecasting

GeoVaneSep 11, 2023, 5:04 AM
1 point
0 comments5 min readLW link

Log­i­cal Share Splitting

DaemonicSigilSep 11, 2023, 4:08 AM
93 points
16 comments9 min readLW link
(pbement.com)

[Question] High school advice

bohaskaSep 11, 2023, 1:26 AM
11 points
16 comments1 min readLW link

Seat­tle As­tral Codex Ten Monthly Social

a7xSep 10, 2023, 7:00 PM
1 point
0 comments1 min readLW link

[Question] What are some good lan­guage mod­els to ex­per­i­ment with?

tailcalledSep 10, 2023, 6:31 PM
16 points
3 comments1 min readLW link

Play­ing the game vs. find­ing a cheat code

MetacelsusSep 10, 2023, 6:11 PM
34 points
1 comment3 min readLW link
(open.substack.com)

Cruxes on US lead for some do­mes­tic AI regulation

Zach Stein-PerlmanSep 10, 2023, 6:00 PM
26 points
3 comments2 min readLW link

Us­ing Nega­tive Hal­lu­ci­na­tions to Man­age Sex­ual Desire

Johannes C. MayerSep 10, 2023, 11:56 AM
−2 points
24 comments1 min readLW link

Fea­ture pro­posal: Ex­port ACX meetups

ViliamSep 10, 2023, 10:50 AM
11 points
7 comments1 min readLW link

Bet­ting and forecasting

CarlJSep 9, 2023, 8:03 PM
2 points
0 comments1 min readLW link

AI pres­i­dents dis­cuss AI al­ign­ment agendas

Sep 9, 2023, 6:55 PM
217 points
23 comments1 min readLW link
(www.youtube.com)

Prob­a­bil­is­tic ar­gu­ment re­la­tion­ships and an in­vi­ta­tion to the ar­gu­ment map­ping community

lunatic_at_largeSep 9, 2023, 6:45 PM
13 points
4 comments10 min readLW link

How teams went about their re­search at AI Safety Camp edi­tion 8

Sep 9, 2023, 4:34 PM
28 points
0 comments13 min readLW link

Panel dis­cus­sion on AI con­scious­ness with Rob Long and Jeff Sebo

Aaron BergmanSep 9, 2023, 3:38 AM
10 points
0 commentsLW link
(www.youtube.com)

Pos­si­ble Diver­gence in AGI Risk Tol­er­ance be­tween Selfish and Altru­is­tic agents

Brad West Sep 9, 2023, 12:23 AM
1 point
1 comment2 min readLW link

Cap­ture the Flag Mechanis­tic In­ter­pretabil­ity Challenges

Sep 8, 2023, 11:00 PM
24 points
0 comments7 min readLW link

[Question] What is to be done? (About the profit mo­tive)

Connor BarberSep 8, 2023, 7:27 PM
1 point
21 comments1 min readLW link

What is the op­ti­mal fron­tier for due dili­gence?

Sep 8, 2023, 6:20 PM
41 points
1 comment1 min readLW link

Progress links di­gest, 2023-09-08: The Con­ser­va­tive Fu­tur­ist, cargo air­ships, and more

jasoncrawfordSep 8, 2023, 5:48 PM
14 points
7 comments5 min readLW link
(rootsofprogress.org)