SPAR seeks ad­vi­sors and stu­dents for AI safety pro­jects (Se­cond Wave)

micSep 14, 2023, 11:09 PM
21 points
0 comments1 min readLW link

“Did you lock it?”

ymeskhoutSep 14, 2023, 9:10 PM
33 points
36 comments2 min readLW link
(ymeskhout.substack.com)

Can I take ducks home from the park?

dynomightSep 14, 2023, 9:03 PM
67 points
8 comments3 min readLW link
(dynomight.net)

In­line Plot­ting in iTerm2

jefftkSep 14, 2023, 8:30 PM
13 points
0 comments1 min readLW link
(www.jefftk.com)

De­stroy­ing the fabric of the uni­verse as an in­stru­men­tal goal.

AI-doomSep 14, 2023, 8:04 PM
−7 points
5 comments1 min readLW link

The PUSA Sys­tem- Repost

Jaivardhan NawaniSep 14, 2023, 6:40 PM
4 points
1 comment5 min readLW link

# **An­nounce­ment of AI-Plans.com Cri­tique-a-thon Septem­ber 2023**

Kabir KumarSep 14, 2023, 5:43 PM
3 points
0 comments2 min readLW link

Cruxes for overhang

Zach Stein-PerlmanSep 14, 2023, 5:00 PM
12 points
5 comments6 min readLW link
(blog.aiimpacts.org)

A The­ory of Laugh­ter—Fol­low-Up

Steven ByrnesSep 14, 2023, 3:35 PM
37 points
3 comments8 min readLW link

Elic­it­ing Credit Hack­ing Be­havi­ours in LLMs

omegastickSep 14, 2023, 3:07 PM
3 points
2 comments7 min readLW link
(github.com)

In­stru­men­tal Con­ver­gence Bounty

Logan ZoellnerSep 14, 2023, 2:02 PM
62 points
24 comments1 min readLW link

[Question] In the age of mod­ern AI (LLMs and be­yond), is data still the new oil?

MPSep 14, 2023, 1:28 PM
4 points
1 comment2 min readLW link

AI #29: Take a Deep Breath

ZviSep 14, 2023, 12:00 PM
65 points
21 comments21 min readLW link
(thezvi.wordpress.com)

The om­ni­zoid—Heighn FDT De­bate #3: Con­tra om­ni­zoid con­tra me con­tra om­ni­zoid con­tra FDT

HeighnSep 14, 2023, 11:52 AM
6 points
0 comments4 min readLW link

High­lights: Went­worth, Shah, and Mur­phy on “Re­tar­get­ing the Search”

RobertMSep 14, 2023, 2:18 AM
87 points
4 comments8 min readLW link

Un­cov­er­ing La­tent Hu­man Wel­lbe­ing in LLM Embeddings

Sep 14, 2023, 1:40 AM
32 points
7 comments8 min readLW link
(far.ai)

A Call For Com­mu­nity: Scien­tific Lan­guage Learn­ing is Still Lan­guage Learn­ing

keltanSep 14, 2023, 12:32 AM
0 points
0 comments2 min readLW link

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)