Up­date to Mys­ter­ies of mode col­lapse: text-davinci-002 not RLHF

janusNov 19, 2022, 11:51 PM
71 points
8 comments2 min readLW link

Make the Drought Eva­po­rate!

AnthonyRepettoNov 19, 2022, 11:41 PM
32 points
25 comments3 min readLW link

Elas­tic Pro­duc­tivity Tools

Simon BerensNov 19, 2022, 9:59 PM
76 points
8 comments2 min readLW link
(simonberens.me)

A Short Dialogue on the Mean­ing of Re­ward Functions

Nov 19, 2022, 9:04 PM
45 points
0 comments3 min readLW link

By De­fault, GPTs Think In Plain Sight

Fabien RogerNov 19, 2022, 7:15 PM
88 points
36 comments9 min readLW link

Re­view: Bayesian Statis­tics the Fun Way by Will Kurt

mattoNov 19, 2022, 6:52 PM
4 points
2 comments2 min readLW link

[Question] How does acausal trade work in a de­ter­minis­tic mul­ti­verse?

sisyphusNov 19, 2022, 1:50 AM
2 points
13 comments1 min readLW link

Choos­ing the right dish

Adam ZernerNov 19, 2022, 1:38 AM
38 points
7 comments8 min readLW link

Reflec­tive Consequentialism

Adam ZernerNov 18, 2022, 11:56 PM
21 points
14 comments4 min readLW link

Value Created vs. Value Extracted

SableNov 18, 2022, 9:34 PM
8 points
6 comments6 min readLW link
(affablyevil.substack.com)

The Disas­trously Con­fi­dent And Inac­cu­rate AI

Sharat Jacob JacobNov 18, 2022, 7:06 PM
13 points
0 comments13 min readLW link

How AI Fails Us: A non-tech­ni­cal view of the Align­ment Problem

testingthewatersNov 18, 2022, 7:02 PM
7 points
1 comment2 min readLW link
(ethics.harvard.edu)

[Question] Is there any policy for a fair treat­ment of AIs whose friendli­ness is in doubt?

nahojNov 18, 2022, 7:01 PM
15 points
10 comments1 min readLW link

Distil­la­tion of “How Likely Is De­cep­tive Align­ment?”

NickGabsNov 18, 2022, 4:31 PM
24 points
4 comments10 min readLW link

Con­tra Chords

jefftkNov 18, 2022, 4:20 PM
12 points
1 comment7 min readLW link
(www.jefftk.com)

[Question] Up­dates on scal­ing laws for foun­da­tion mod­els from ′ Tran­scend­ing Scal­ing Laws with 0.1% Ex­tra Com­pute’

Nick_GreigNov 18, 2022, 12:46 PM
15 points
2 comments1 min readLW link

Hal­i­fax, NS – Monthly Ra­tion­al­ist, EA, and ACX Meetup

IdeopunkNov 18, 2022, 11:45 AM
10 points
0 comments1 min readLW link

In­tro­duc­ing The Log­i­cal Foun­da­tion, A Plan to End Poverty With Guaran­teed Income

Michael SimmNov 18, 2022, 8:13 AM
9 points
23 comments1 min readLW link

My Deon­tol­ogy Says Nar­row-Mind­ed­ness is Always Wrong

LVSNNov 18, 2022, 6:11 AM
6 points
2 comments1 min readLW link

AI Ethics != Ai Safety

DentinNov 18, 2022, 3:02 AM
2 points
0 comments1 min readLW link

Don’t de­sign agents which ex­ploit ad­ver­sar­ial inputs

Nov 18, 2022, 1:48 AM
72 points
64 comments12 min readLW link

Eng­ineer­ing Monose­man­tic­ity in Toy Models

Nov 18, 2022, 1:43 AM
75 points
7 comments3 min readLW link
(arxiv.org)

AGIs may value in­trin­sic re­wards more than ex­trin­sic ones

catubcNov 17, 2022, 9:49 PM
8 points
6 comments4 min readLW link

LLMs may cap­ture key com­po­nents of hu­man agency

catubcNov 17, 2022, 8:14 PM
27 points
0 comments4 min readLW link

Mastodon Replies as Comments

jefftkNov 17, 2022, 8:10 PM
20 points
0 comments1 min readLW link
(www.jefftk.com)

An­nounc­ing the Progress Forum

jasoncrawfordNov 17, 2022, 7:26 PM
83 points
9 comments1 min readLW link

[Question] What kind of bias is this?

Daniel SamuelNov 17, 2022, 6:44 PM
3 points
2 comments1 min readLW link

AI Fore­cast­ing Re­search Ideas

JsevillamolNov 17, 2022, 5:37 PM
21 points
2 comments1 min readLW link

Re­sults from the in­ter­pretabil­ity hackathon

Nov 17, 2022, 2:51 PM
81 points
0 comments6 min readLW link
(alignmentjam.com)

Covid 11/​17/​22: Slow Recovery

ZviNov 17, 2022, 2:50 PM
33 points
3 comments4 min readLW link
(thezvi.wordpress.com)

Sadly, FTX

ZviNov 17, 2022, 2:30 PM
133 points
18 comments47 min readLW link
(thezvi.wordpress.com)

Deon­tol­ogy and virtue ethics as “effec­tive the­o­ries” of con­se­quen­tial­ist ethics

Jan_KulveitNov 17, 2022, 2:11 PM
68 points
9 comments1 min readLW link1 review

The Ground Truth Prob­lem (Or, Why Eval­u­at­ing In­ter­pretabil­ity Meth­ods Is Hard)

Jessica RumbelowNov 17, 2022, 11:06 AM
27 points
2 comments2 min readLW link

[Question] [Per­sonal Ques­tion] Can any­one help me nav­i­gate this po­ten­tially painful in­ter­per­sonal dy­namic ra­tio­nally?

SlainLadyMondegreenNov 17, 2022, 8:53 AM
9 points
3 comments4 min readLW link

Mas­sive Scal­ing Should be Frowned Upon

harsimonyNov 17, 2022, 8:43 AM
4 points
6 comments5 min readLW link

[Question] Why are prof­itable com­pa­nies lay­ing off staff?

Yair HalberstadtNov 17, 2022, 6:19 AM
15 points
10 comments1 min readLW link

Dis­cus­sion: Was SBF a naive util­i­tar­ian, or a so­ciopath?

Nicholas / Heather KrossNov 17, 2022, 2:52 AM
0 points
4 comments1 min readLW link

Kel­sey Piper’s re­cent in­ter­view of SBF

agucovaNov 16, 2022, 8:30 PM
51 points
29 comments1 min readLW link

The Echo Principle

Jonathan MoregårdNov 16, 2022, 8:09 PM
4 points
0 comments3 min readLW link
(honestliving.substack.com)

[Question] Is there some rea­son LLMs haven’t seen broader use?

tailcalledNov 16, 2022, 8:04 PM
25 points
27 comments1 min readLW link

When should we be sur­prised that an in­ven­tion took “so long”?

jasoncrawfordNov 16, 2022, 8:04 PM
32 points
11 comments4 min readLW link
(rootsofprogress.org)

Ques­tions about Value Lock-in, Pa­ter­nal­ism, and Empowerment

Sam F. BrownNov 16, 2022, 3:33 PM
13 points
2 comments12 min readLW link
(sambrown.eu)

If Pro­fes­sional In­vestors Missed This...

jefftkNov 16, 2022, 3:00 PM
37 points
18 comments3 min readLW link
(www.jefftk.com)

Disagree­ment with bio an­chors that lead to shorter timelines

Marius HobbhahnNov 16, 2022, 2:40 PM
75 points
17 comments7 min readLW link1 review

Cur­rent themes in mechanis­tic in­ter­pretabil­ity research

Nov 16, 2022, 2:14 PM
89 points
2 comments12 min readLW link

Un­pack­ing “Shard The­ory” as Hunch, Ques­tion, The­ory, and Insight

Jacy Reese AnthisNov 16, 2022, 1:54 PM
31 points
9 comments2 min readLW link

Mir­a­cles and why not to be­lieve them

mruwnikNov 16, 2022, 12:07 PM
4 points
0 comments2 min readLW link

[Question] How do peo­ple do re­mote re­search col­lab­o­ra­tions effec­tively?

KriegerNov 16, 2022, 11:51 AM
8 points
0 comments1 min readLW link

Method of state­ments: an al­ter­na­tive to taboo

Q HomeNov 16, 2022, 10:57 AM
7 points
0 comments41 min readLW link

The two con­cep­tions of Ac­tive In­fer­ence: an in­tel­li­gence ar­chi­tec­ture and a the­ory of agency

Roman LeventovNov 16, 2022, 9:30 AM
17 points
0 comments4 min readLW link