[Preprint] The Com­pu­ta­tional Limits of Deep Learning

Gordon Seidoh WorleyJul 21, 2020, 9:25 PM
9 points
4 comments1 min readLW link
(arxiv.org)

Fresh Bread

ZviJul 21, 2020, 8:40 PM
22 points
1 comment2 min readLW link
(thezvi.wordpress.com)

Com­pe­ti­tion: Am­plify Ro­hin’s Pre­dic­tion on AGI re­searchers & Safety Concerns

stuhlmuellerJul 21, 2020, 8:06 PM
83 points
41 comments3 min readLW link

Align­ment As A Bot­tle­neck To Use­ful­ness Of GPT-3

johnswentworthJul 21, 2020, 8:02 PM
111 points
57 comments3 min readLW link

How good is hu­man­ity at co­or­di­na­tion?

BuckJul 21, 2020, 8:01 PM
82 points
44 comments3 min readLW link

$1000 bounty for OpenAI to show whether GPT3 was “de­liber­ately” pre­tend­ing to be stupi­der than it is

Bird ConceptJul 21, 2020, 6:42 PM
56 points
39 comments2 min readLW link
(twitter.com)

[Question] What are the limits of self-ed­u­ca­tion?

nitropieJul 21, 2020, 6:01 PM
3 points
2 comments1 min readLW link

[Meta] anony­mous merit or pub­lic status

[anonymous]Jul 21, 2020, 6:01 PM
6 points
4 comments1 min readLW link

AI Benefits Post 5: Out­stand­ing Ques­tions on Govern­ing Benefits

CullenJul 21, 2020, 4:46 PM
4 points
0 comments4 min readLW link

The “AI Dun­geons” Dragon Model is heav­ily path de­pen­dent (test­ing GPT-3 on ethics)

Rafael HarthJul 21, 2020, 12:14 PM
44 points
9 comments6 min readLW link

Un­cal­ibrated quan­tum ex­per­i­ments act cla­si­cally

justinpombrioJul 21, 2020, 5:31 AM
18 points
12 comments8 min readLW link

The Redis­cov­ery of In­te­ri­or­ity in Ma­chine Learning

DanBJul 21, 2020, 5:02 AM
5 points
4 commentsLW link
(danburfoot.net)

Chains, Bot­tle­necks and Optimization

curiJul 21, 2020, 2:07 AM
14 points
12 comments4 min readLW link

“Can you keep this con­fi­den­tial? How do you know?”

RaemonJul 21, 2020, 12:33 AM
164 points
43 comments3 min readLW link2 reviews

Par­allels Between AI Safety by De­bate and Ev­i­dence Law

CullenJul 20, 2020, 10:52 PM
10 points
1 comment2 min readLW link
(cullenokeefe.com)

Thiel on Progress and Stagnation

Richard_NgoJul 20, 2020, 8:27 PM
173 points
32 comments11 min readLW link
(docs.google.com)

Learn­ing Values in Practice

Stuart_ArmstrongJul 20, 2020, 6:38 PM
24 points
0 comments5 min readLW link

Ineffi­cient doesn’t mean in­differ­ent, but it might mean wimpy.

DirectedEvolutionJul 20, 2020, 6:27 PM
14 points
3 comments5 min readLW link

[Question] To what ex­tent is GPT-3 ca­pa­ble of rea­son­ing?

TurnTroutJul 20, 2020, 5:10 PM
70 points
73 comments16 min readLW link

Sel­ling real es­tate: should you over­price or un­der­price?

Steven ByrnesJul 20, 2020, 3:54 PM
19 points
5 comments10 min readLW link

[Question] “Do Noth­ing” util­ity func­tion, 3½ years later?

niplavJul 20, 2020, 11:09 AM
5 points
3 comments1 min readLW link

Oper­a­tional­iz­ing Interpretability

lifelonglearnerJul 20, 2020, 5:22 AM
20 points
0 comments4 min readLW link

Use re­silience, in­stead of im­pre­ci­sion, to com­mu­ni­cate uncertainty

habrykaJul 20, 2020, 5:08 AM
3 points
1 comment1 min readLW link
(forum.effectivealtruism.org)

What Would I Do? Self-pre­dic­tion in Sim­ple Algorithms

Scott GarrabrantJul 20, 2020, 4:27 AM
65 points
12 comments5 min readLW link

“Should Black­mail Be Le­gal” Han­son/​Zvi De­bate (Sun July 26th, 3pm PDT)

Ben PaceJul 20, 2020, 4:06 AM
36 points
13 comments1 min readLW link

The 8 Tech­niques to Tol­er­ify the Dark World

adamShimiJul 20, 2020, 12:58 AM
2 points
5 comments2 min readLW link

Praise of some pop­u­lar LW articles

DirectedEvolutionJul 20, 2020, 12:32 AM
40 points
1 comment7 min readLW link

Types Of On­line Meetups

Dan BJul 19, 2020, 11:51 PM
4 points
2 comments2 min readLW link

Mu­si­cal Outgroups

eapacheJul 19, 2020, 10:55 PM
9 points
1 comment4 min readLW link

Fo­rum As­sisted Discussion

Dan BJul 19, 2020, 10:38 PM
9 points
0 comments3 min readLW link

Pulse and Glide Cycling

jefftkJul 19, 2020, 7:02 PM
11 points
5 comments2 min readLW link
(www.jefftk.com)

[Question] Math. proof of the su­pe­ri­or­ity of in­de­pen­dent guesses?

MiltonJul 19, 2020, 2:38 AM
−3 points
7 comments1 min readLW link

Crit­i­cism of some pop­u­lar LW articles

DirectedEvolutionJul 19, 2020, 1:16 AM
71 points
19 comments6 min readLW link

Swiss Poli­ti­cal Sys­tem: More than You ever Wanted to Know (I.)

Martin SustrikJul 19, 2020, 1:11 AM
172 points
39 comments24 min readLW link2 reviews

[Question] Why is pseudo-al­ign­ment “worse” than other ways ML can fail to gen­er­al­ize?

nostalgebraistJul 18, 2020, 10:54 PM
45 points
9 comments2 min readLW link

Against Reopen­ing Ottawa

eapacheJul 18, 2020, 8:08 PM
6 points
2 comments5 min readLW link

Col­lec­tion of GPT-3 results

Kaj_SotalaJul 18, 2020, 8:04 PM
89 points
24 comments1 min readLW link
(twitter.com)

[Question] Is there an easy way to turn a LW se­quence into an epub?

ChristianKlJul 18, 2020, 6:20 PM
17 points
9 comments1 min readLW link

Cal­ibrate words, not just probabilities

MikkWJul 18, 2020, 5:56 AM
11 points
3 comments2 min readLW link

[Question] Erv­ing Goff­man’s ‘pa­per’

SaffronJul 18, 2020, 1:12 AM
5 points
2 comments1 min readLW link

Les­sons on AI Takeover from the conquistadors

Jul 17, 2020, 10:35 PM
61 points
31 comments6 min readLW link

[Question] Can an agent use in­ter­ac­tive proofs to check the al­ign­ment of suc­ce­sors?

PabloAMCJul 17, 2020, 7:07 PM
7 points
2 comments1 min readLW link

An­thro­po­mor­phiz­ing Humans

johnswentworthJul 17, 2020, 5:49 PM
46 points
6 comments2 min readLW link

Tel­ling more ra­tio­nal stories

DirectedEvolutionJul 17, 2020, 5:47 PM
26 points
21 comments3 min readLW link

Solv­ing Math Prob­lems by Relay

Jul 17, 2020, 3:32 PM
103 points
26 comments7 min readLW link

[Question] What are the best tools you have seen to keep track of knowl­edge around testable state­ments?

migueltorrescostaJul 17, 2020, 3:02 PM
2 points
1 comment1 min readLW link

En­vi­ron­ments as a bot­tle­neck in AGI development

Richard_NgoJul 17, 2020, 5:02 AM
41 points
19 comments6 min readLW link

My Dat­ing Plan ala Ge­offrey Miller

snog toddgrassJul 17, 2020, 4:52 AM
2 points
57 comments3 min readLW link

Meta-prefer­ences are weird

Jul 16, 2020, 11:03 PM
13 points
2 comments5 min readLW link

Sun­day July 19, 1pm (PDT) — talks by Rae­mon, ricraz, mr-hire, Jame­son Quinn

Jul 16, 2020, 8:04 PM
26 points
6 comments1 min readLW link