Sols­tice 2023 Roundup

dspeyerOct 11, 2023, 11:09 PM
28 points
6 comments1 min readLW link

Un­der­stand­ing LLMs: Some ba­sic ob­ser­va­tions about words, syn­tax, and dis­course [w/​ a con­jec­ture about grokking]

Bill BenzonOct 11, 2023, 7:13 PM
6 points
0 comments5 min readLW link

[Linkpost] Gen­er­al­iza­tion in diffu­sion mod­els arises from ge­om­e­try-adap­tive har­monic representation

Bogdan Ionut CirsteaOct 11, 2023, 5:48 PM
4 points
3 comments1 min readLW link

What I’ve been read­ing, Oc­to­ber 2023: The stir­rup in Europe, 19th-cen­tury art deco, and more

jasoncrawfordOct 11, 2023, 4:11 PM
18 points
2 comments11 min readLW link
(rootsofprogress.org)

EA Madrid social

Pablo VillalobosOct 11, 2023, 3:34 PM
6 points
0 comments1 min readLW link

At­tribut­ing to in­ter­ac­tions with GCPD and GWPD

jennyOct 11, 2023, 3:06 PM
20 points
0 comments6 min readLW link

You’re Mea­sur­ing Model Com­plex­ity Wrong

Oct 11, 2023, 11:46 AM
93 points
17 comments13 min readLW link

Up­date on the UK AI Task­force & up­com­ing AI Safety Summit

Elliot MckernonOct 11, 2023, 11:37 AM
84 points
2 comments4 min readLW link

An ex­pla­na­tion for ev­ery to­ken: us­ing an LLM to sam­ple an­other LLM

Max HOct 11, 2023, 12:53 AM
35 points
5 comments11 min readLW link

[Question] Ex­am­ples of Low Sta­tus Fun

niplavOct 10, 2023, 11:19 PM
18 points
17 comments1 min readLW link

A New Model for Com­pute Cen­ter Verification

Damin CurtisOct 10, 2023, 7:22 PM
8 points
0 comments5 min readLW link

An­nounc­ing MIRI’s new CEO and lead­er­ship team

Gretta DulebaOct 10, 2023, 7:22 PM
222 points
52 comments3 min readLW link

18 Hetero­dox lenses to look the world through

Shaurya GuptaOct 10, 2023, 6:33 PM
−1 points
2 comments5 min readLW link

Doc­u­ment­ing Jour­ney Into AI Safety

jacobhaimesOct 10, 2023, 6:30 PM
17 points
4 comments6 min readLW link

Look­ing for AI Art Col­lab­o­ra­tors!

beatrice@foresight.orgOct 10, 2023, 6:24 PM
1 point
0 comments1 min readLW link

Child­hood Roundup #3

ZviOct 10, 2023, 2:30 PM
49 points
3 comments30 min readLW link
(thezvi.wordpress.com)

My sim­ple model for Align­ment vs Capability

ryan_bOct 10, 2023, 12:07 PM
7 points
0 comments7 min readLW link

Next year in Jerusalem: The brilli­ant ideas and ra­di­ant legacy of Miriam Lip­schutz Ye­vick [in re­la­tion to cur­rent AI de­bates]

Bill BenzonOct 10, 2023, 9:06 AM
1 point
0 comments1 min readLW link
(3quarksdaily.com)

I’m a Former Is­raeli Officer. AMA

Yovel RomOct 10, 2023, 8:33 AM
78 points
70 comments1 min readLW link

Be­come a PIBBSS Re­search Affiliate

Oct 10, 2023, 7:41 AM
24 points
6 comments6 min readLW link

My 1st month at a “neu­ro­di­ver­gent gifted school” called Min­erva University

exanovaOct 10, 2023, 3:34 AM
4 points
1 comment1 min readLW link
(inawe.substack.com)

Epistemic Mo­tif of Ab­stract-Con­crete Cy­cles & Do­main Expansion

DalcyOct 10, 2023, 3:28 AM
26 points
2 comments3 min readLW link

Sim­ple Ter­mi­nal Colors

jefftkOct 10, 2023, 12:40 AM
11 points
1 comment1 min readLW link
(www.jefftk.com)

The Hand­book of Ra­tion­al­ity (2021, MIT press) is now open access

romeostevensitOct 10, 2023, 12:30 AM
48 points
4 comments1 min readLW link

Non-su­per­in­tel­li­gent pa­per­clip max­i­miz­ers are normal

jessicataOct 10, 2023, 12:29 AM
67 points
4 comments9 min readLW link
(unstableontology.com)

The Witch­ing Hour

Richard_NgoOct 10, 2023, 12:19 AM
113 points
1 comment9 min readLW link
(www.narrativeark.xyz)

One: a story

Richard_NgoOct 10, 2023, 12:18 AM
30 points
0 comments4 min readLW link
(www.narrativeark.xyz)

Truth­seek­ing when your dis­agree­ments lie in moral philosophy

Oct 10, 2023, 12:00 AM
99 points
4 comments4 min readLW link
(acesounderglass.com)

NYT on the Man­i­fest fore­cast­ing conference

Austin ChenOct 9, 2023, 9:40 PM
45 points
14 commentsLW link
(www.nytimes.com)

Fore­cast­ing and pre­dic­tion markets

CarlJOct 9, 2023, 8:43 PM
3 points
0 comments1 min readLW link

Com­par­ing Two Fore­cast­ers in an Ideal World

nikosOct 9, 2023, 7:52 PM
5 points
0 comments6 min readLW link

The case for af­ter­mar­ket blind spot mirrors

Brendan LongOct 9, 2023, 7:30 PM
59 points
14 comments2 min readLW link
(www.brendanlong.com)

New con­trac­tor role: Web se­cu­rity task force con­trac­tor for AI safety announcements

Oct 9, 2023, 6:36 PM
11 points
0 comments2 min readLW link
(survivalandflourishing.com)

[Question] Any­one work­ing on D. Amodei’s Bartlett show tran­script?

LeopardOct 9, 2023, 6:17 PM
10 points
0 comments1 min readLW link

Knowl­edge Base 3: Shop­ping ad­vi­sor and other uses of knowl­edge base about products

iwisOct 9, 2023, 11:53 AM
0 points
0 comments4 min readLW link

Knowl­edge Base 2: The struc­ture and the method of building

iwisOct 9, 2023, 11:53 AM
2 points
4 comments7 min readLW link

We don’t un­der­stand what hap­pened with cul­ture enough

Jan_KulveitOct 9, 2023, 9:54 AM
87 points
22 comments6 min readLW link1 review

Lev­er­ag­ing Bayes’ The­o­rem to Su­per­charge Me­mory Techniques

disohaOct 9, 2023, 3:34 AM
−15 points
1 comment4 min readLW link

Paper: Iden­ti­fy­ing the Risks of LM Agents with an LM-Emu­lated Sand­box—Univer­sity of Toronto 2023 - Bench­mark con­sist­ing of 36 high-stakes tools and 144 test cases!

Singularian2501Oct 9, 2023, 12:00 AM
6 points
0 comments1 min readLW link

AI Align­ment Break­throughs this week (10/​08/​23)

Logan ZoellnerOct 8, 2023, 11:30 PM
30 points
14 comments6 min readLW link

“The Heart of Gam­ing is the Power Fan­tasy”, and Co­hab­itive Games

RaemonOct 8, 2023, 9:02 PM
81 points
49 comments4 min readLW link
(bottomfeeder.substack.com)

FAQ: What the heck is goal ag­nos­ti­cism?

porbyOct 8, 2023, 7:11 PM
66 points
38 comments28 min readLW link

Time is ho­mo­ge­neous se­quen­tially-com­pos­able determination

TsviBTOct 8, 2023, 2:58 PM
15 points
0 comments21 min readLW link

Linkpost: Are Emer­gent Abil­ities in Large Lan­guage Models just In-Con­text Learn­ing?

Erich_GrunewaldOct 8, 2023, 12:14 PM
12 points
7 comments2 min readLW link
(arxiv.org)

Bird-eye view vi­su­al­iza­tion of LLM activations

SergiiOct 8, 2023, 12:12 PM
11 points
2 comments1 min readLW link
(grgv.xyz)

Per­spec­tive Based Rea­son­ing Could Ab­solve CDT

dadadarrenOct 8, 2023, 11:22 AM
4 points
5 comments5 min readLW link

The Gra­di­ent – The Ar­tifi­cial­ity of Alignment

micOct 8, 2023, 4:06 AM
12 points
1 comment5 min readLW link
(thegradient.pub)

Com­par­ing An­thropic’s Dic­tionary Learn­ing to Ours

Robert_AIZIOct 7, 2023, 11:30 PM
137 points
8 comments4 min readLW link

A thought about the con­straints of debtless­ness in on­line communities

mako yassOct 7, 2023, 9:26 PM
58 points
23 comments1 min readLW link

Ar­gu­ments for util­i­tar­i­anism are im­pos­si­bil­ity ar­gu­ments un­der un­bounded prospects

MichaelStJulesOct 7, 2023, 9:08 PM
7 points
7 comments21 min readLW link