RSS

Large Lan­guage Models

TagLast edit: Nov 24, 2023, 4:38 PM by Toby Tremlett🔹

This topic is for posts discussing Large Language Models (LLMs) -- for example, the GPT models produced by OpenAI.

Related Entries


AI safety | Artificial intelligence | AI governance | AI forecasting

LLMs are weirder than you think

Derek ShillerNov 20, 2024, 1:39 PM
64 points
3 comments22 min readEA link

In­tro­duc­ing Senti—An­i­mal Ethics AI Assistant

Animal_EthicsMay 9, 2024, 7:33 AM
40 points
2 comments2 min readEA link

In­tro­duc­ing Squig­gle AI

Ozzie GooenJan 3, 2025, 5:53 PM
84 points
13 comments8 min readEA link

My Cur­rent Claims and Cruxes on LLM Fore­cast­ing & Epistemics

Ozzie GooenJun 26, 2024, 12:40 AM
46 points
7 comments24 min readEA link

Im­pact of Quan­ti­za­tion on Small Lan­guage Models (SLMs) for Mul­tilin­gual Math­e­mat­i­cal Rea­son­ing Tasks

Angie Paola GiraldoMay 7, 2025, 9:48 PM
1 point
0 comments14 min readEA link

The An­i­mal Welfare Case for Open Ac­cess: Break­ing Bar­ri­ers to Scien­tific Knowl­edge and En­hanc­ing LLM Training

Wladimir J. AlonsoNov 23, 2024, 1:07 PM
32 points
2 comments3 min readEA link

Ten­ta­tive prac­ti­cal tips for us­ing chat­bots in research

Erich_Grunewald 🔸Mar 29, 2023, 3:01 PM
48 points
7 comments5 min readEA link

Sleeper Agents: Train­ing De­cep­tive LLMs that Per­sist Through Safety Training

evhubJan 12, 2024, 7:51 PM
65 points
0 comments1 min readEA link
(arxiv.org)

Briefly how I’ve up­dated since ChatGPT

rimeApr 25, 2023, 7:39 PM
29 points
8 comments2 min readEA link
(www.lesswrong.com)

A short con­ver­sa­tion I had with Google Gem­ini on the dan­gers of un­reg­u­lated LLM API use, while mildly drunk in an air­port.

EvanMcCormickDec 17, 2024, 12:25 PM
1 point
0 comments8 min readEA link

The case for more am­bi­tious lan­guage model evals

JozdienJan 30, 2024, 9:24 AM
7 points
0 comments5 min readEA link

Prob­lem-solv­ing tasks in Graph The­ory for lan­guage mod­els

Bruno López OrozcoOct 1, 2024, 12:36 PM
21 points
1 comment9 min readEA link

Claude 3.5 Sonnet

Zach Stein-PerlmanJun 20, 2024, 6:00 PM
31 points
0 comments1 min readEA link
(www.anthropic.com)

ChatGPT not so clever or not so ar­tifi­cial as hyped to be?

Haris ShekerisMar 2, 2023, 6:16 AM
−7 points
2 comments1 min readEA link

How to quickly set up Claude as a chat bot for on­line fel­low­ships and courses

Jamie_HarrisJul 22, 2023, 7:53 AM
38 points
10 comments4 min readEA link

Pos­si­ble OpenAI’s Q* break­through and Deep­Mind’s AlphaGo-type sys­tems plus LLMs

BurnydelicNov 23, 2023, 7:02 AM
13 points
4 comments2 min readEA link

New Ar­tifi­cial In­tel­li­gence quiz: can you beat ChatGPT?

AndreFerrettiMar 3, 2023, 3:46 PM
29 points
3 comments1 min readEA link

Life of GPT

Odd anonNov 8, 2023, 10:31 PM
−1 points
0 comments5 min readEA link

Pros and Cons of boy­cotting paid Chat GPT

NickLaingMar 18, 2023, 8:50 AM
14 points
11 comments2 min readEA link

[Question] How would a lan­guage model be­come goal-di­rected?

David MJul 16, 2022, 2:50 PM
113 points
20 comments1 min readEA link

In fa­vor of an AI-pow­ered trans­la­tion but­ton on the EA Forum

Alix PhamJun 6, 2024, 8:29 PM
49 points
4 comments1 min readEA link

Dis­cussing AI-Hu­man Col­lab­o­ra­tion Through Fic­tion: The Story of Laika and GPT-∞

LaikaJul 27, 2023, 6:04 AM
1 point
0 comments1 min readEA link

The Prospect of an AI Winter

Erich_Grunewald 🔸Mar 27, 2023, 8:55 PM
56 points
13 comments1 min readEA link

Open Phil re­leases RFPs on LLM Bench­marks and Forecasting

Lawrence ChanNov 11, 2023, 3:01 AM
12 points
0 comments1 min readEA link
(www.openphilanthropy.org)

EA Ex­plorer GPT: A New Tool to Ex­plore Effec­tive Altruism

Vlad_TislenkoNov 12, 2023, 3:36 PM
12 points
1 comment1 min readEA link

LLMs won’t lead to AGI—Fran­cois Chollet

tobycrisford 🔸Jun 11, 2024, 8:19 PM
37 points
23 comments1 min readEA link
(www.youtube.com)

[Question] What am I miss­ing re. open-source LLM’s?

another-anon-do-gooderDec 4, 2023, 4:48 AM
1 point
2 comments1 min readEA link

On the Dwarkesh/​Chol­let Pod­cast, and the cruxes of scal­ing to AGI

JWS 🔸Jun 15, 2024, 8:24 PM
72 points
49 comments17 min readEA link

LLM-Se­cured Sys­tems: A Gen­eral-Pur­pose Tool For Struc­tured Transparency

Ozzie GooenJun 18, 2024, 12:20 AM
36 points
1 comment21 min readEA link

Fore­cast­ing With LLMs—An Open and Promis­ing Re­search Direction

Marcel DMar 12, 2024, 4:23 AM
13 points
0 comments4 min readEA link

On the fu­ture of lan­guage models

Owen Cotton-BarrattDec 20, 2023, 4:58 PM
125 points
3 comments36 min readEA link

AI scal­ing myths

Noah Varley🔸Jun 27, 2024, 8:29 PM
30 points
0 comments1 min readEA link
(open.substack.com)

LLMs can­not use­fully be moral patients

LGSJul 2, 2024, 4:43 AM
35 points
24 comments4 min readEA link

LLM Eval­u­a­tors Rec­og­nize and Fa­vor Their Own Generations

Arjun PanicksseryApr 17, 2024, 9:09 PM
21 points
4 comments1 min readEA link
(tiny.cc)

‘Chat with im­pact­ful re­search & eval­u­a­tions’ (Un­jour­nal Note­bookLMs)

david_reinsteinSep 24, 2024, 8:19 PM
8 points
1 comment2 min readEA link

Scal­ing of AI train­ing runs will slow down af­ter GPT-5

Maxime Riché 🔸Apr 26, 2024, 4:06 PM
10 points
2 comments3 min readEA link

An­i­mal ethics in ChatGPT and Claude

Elijah WhippleJan 16, 2024, 9:38 PM
47 points
2 comments9 min readEA link

The In­ten­tional Stance, LLMs Edition

Eleni_AMay 1, 2024, 3:22 PM
8 points
2 comments8 min readEA link

LLMs as a Plan­ning Overhang

LarksJul 14, 2024, 4:57 AM
49 points
3 comments1 min readEA link

RAND re­port finds no effect of cur­rent LLMs on vi­a­bil­ity of bioter­ror­ism attacks

LizkaJan 26, 2024, 8:10 PM
108 points
17 comments3 min readEA link
(www.rand.org)

Utility Eng­ineer­ing: An­a­lyz­ing and Con­trol­ling Emer­gent Value Sys­tems in AIs

Matrice JacobineFeb 12, 2025, 9:15 AM
13 points
0 comments1 min readEA link
(www.emergent-values.ai)

[Question] Find­ing ‘pivotal ques­tions’ from 80k pod­cast tran­scripts, sug­ges­tions, LLM ap­proaches/​ Is there already an “80k chat­bot”?

david_reinsteinJan 8, 2025, 5:16 PM
10 points
2 comments1 min readEA link

Wor­ri­some Trends for Digi­tal Mind Evaluations

Derek ShillerFeb 20, 2025, 3:35 PM
79 points
10 comments8 min readEA link

Knowl­edge, Rea­son­ing, and Superintelligence

Owen Cotton-BarrattMar 26, 2025, 11:28 PM
21 points
3 comments1 min readEA link
(strangecities.substack.com)

Ab­solute Zero: Re­in­forced Self-play Rea­son­ing with Zero Data

Matrice JacobineMay 12, 2025, 3:20 PM
14 points
1 comment1 min readEA link
(www.arxiv.org)

How to Catch a ChatGPT Cheat: 7 Prac­ti­cal Tips

MarshallDec 27, 2022, 4:09 PM
8 points
2 comments4 min readEA link

[Question] How in­de­pen­dent is the re­search com­ing out of OpenAI’s pre­pared­ness team?

EarthlingFeb 10, 2024, 4:59 PM
18 points
0 comments1 min readEA link

o3

Zach Stein-PerlmanDec 20, 2024, 9:00 PM
84 points
9 comments1 min readEA link

LLMs Out­perform Ex­perts on Challeng­ing Biol­ogy Benchmarks

ljustenMay 14, 2025, 4:09 PM
21 points
1 comment1 min readEA link
(substack.com)

We are in a New Paradigm of AI Progress—OpenAI’s o3 model makes huge gains on the tough­est AI bench­marks in the world

GarrisonDec 22, 2024, 9:45 PM
26 points
0 comments4 min readEA link
(garrisonlovely.substack.com)

Beyond Meta: Large Con­cept Models Will Win

Anthony RepettoDec 30, 2024, 12:57 AM
3 points
0 comments3 min readEA link

How much is 1.8 mil­lion years of work?

rosehadsharAug 16, 2024, 12:35 PM
21 points
3 comments2 min readEA link

Straight­for­wardly elic­it­ing prob­a­bil­ities from GPT-3

NunoSempereFeb 9, 2023, 7:25 PM
41 points
5 comments4 min readEA link

En­hanc­ing Math­e­mat­i­cal Model­ing with LLMs: Goals, Challenges, and Evaluations

Ozzie GooenOct 28, 2024, 9:37 PM
11 points
3 comments15 min readEA link

Claude vs GPT

Maxwell TabarrokMar 14, 2024, 12:44 PM
14 points
1 comment2 min readEA link
(www.maximum-progress.com)

[Linkpost] Vague Ver­biage in Forecasting

trevor1Mar 22, 2024, 6:05 PM
5 points
0 comments1 min readEA link
(goodjudgment.com)

Share your re­quests for ChatGPT

Kate TranDec 5, 2022, 6:43 PM
8 points
5 comments1 min readEA link

“This might be the first large-scale ap­pli­ca­tion of AI tech­nol­ogy to geopoli­tics.. 4o, o3 high, Gem­ini 2.5 pro, Claude 3.7, Grok all give the same an­swer to the ques­tion on how to im­pose tar­iffs eas­ily.”

Matrice JacobineApr 3, 2025, 10:50 AM
3 points
0 comments1 min readEA link
(x.com)

Was Re­leas­ing Claude-3 Net-Negative

Logan RiggsMar 27, 2024, 5:41 PM
12 points
1 comment4 min readEA link

Large Lan­guage Models Pass the Tur­ing Test

Matrice JacobineApr 2, 2025, 5:41 AM
11 points
6 comments1 min readEA link
(arxiv.org)

Is Text Water­mark­ing a lost cause?

Egor TimatkovOct 1, 2024, 1:07 PM
4 points
0 comments10 min readEA link

[Question] Could AI-gen­er­ated con­tent help think-tanks & re­search orgs be­come more effec­tive?

Justin OliveJan 10, 2023, 10:58 PM
13 points
0 comments2 min readEA link

LLM chat­bots have ~half of the kinds of “con­scious­ness” that hu­mans be­lieve in. Hu­mans should avoid go­ing crazy about that.

Andrew CritchNov 22, 2024, 3:26 AM
11 points
3 comments1 min readEA link

Sum­mary: In­tro­spec­tive Ca­pa­bil­ities in LLMs (Robert Long)

rileyharrisJul 2, 2024, 6:08 PM
11 points
1 comment4 min readEA link

BenchMo­ral: A bench­mark­ing to as­sess the moral sen­si­tivity of large lan­guage mod­els (LLMs) in Span­ish.

Flor Betzabeth Ampa FloresApr 30, 2025, 9:26 PM
1 point
0 comments18 min readEA link

Have your timelines changed as a re­sult of ChatGPT?

Chris LeongDec 5, 2022, 3:03 PM
30 points
18 comments1 min readEA link

How LLMs Work, in the Style of The Economist

utilistrutilApr 22, 2024, 7:06 PM
17 points
0 comments1 min readEA link

Perfor­mance com­par­i­son of Large Lan­guage Models (LLMs) in code gen­er­a­tion and ap­pli­ca­tion of best prac­tices in fron­tend web development

Diana V. Guaiña A.May 1, 2025, 2:57 PM
1 point
0 comments24 min readEA link

Digest: three pa­pers that have shaped my un­der­stand­ing of the po­ten­tial for con­scious­ness in AI systems

rileyharrisAug 21, 2024, 3:09 PM
5 points
0 comments1 min readEA link

What is scaf­fold­ing?

Vishakha AgrawalMar 27, 2025, 9:40 AM
3 points
0 comments2 min readEA link
(aisafety.info)

Fea­si­bil­ity of train­ing and in­fer­ring ad­vanced large lan­guage mod­els (LLMs) in data cen­ters in Mex­ico and Brazil.

Tatiana SandovalMay 2, 2025, 1:42 PM
1 point
0 comments24 min readEA link

AIs Are Ex­pert-Level at Many Virol­ogy Skills

Center for AI SafetyMay 2, 2025, 4:07 PM
22 points
0 comments1 min readEA link

[Question] Is Deep­Seek-R1 already bet­ter than o3 when in­fer­ence costs are held con­stant?

Magnus VindingJan 24, 2025, 3:29 PM
33 points
2 comments1 min readEA link

Com­par­i­son of LLM scal­a­bil­ity and perfor­mance be­tween the U.S. and China based on benchmark

Ivanna_alvaradoOct 12, 2024, 9:51 PM
8 points
0 comments34 min readEA link

Risk Align­ment in Agen­tic AI Systems

Hayley ClatterbuckOct 1, 2024, 10:51 PM
31 points
1 comment3 min readEA link
(static1.squarespace.com)

ChatGPT un­der­stands, but largely does not gen­er­ate Span­glish (and other code-mixed) text

Milan Weibel🔹Jan 4, 2023, 10:10 PM
6 points
0 comments4 min readEA link
(www.lesswrong.com)

AISN #35: Lob­by­ing on AI Reg­u­la­tion Plus, New Models from OpenAI and Google, and Le­gal Regimes for Train­ing on Copy­righted Data

Center for AI SafetyMay 16, 2024, 2:26 PM
14 points
0 comments6 min readEA link
(newsletter.safe.ai)

Wor­ri­some mi­s­un­der­stand­ing of the core is­sues with AI transition

Roman LeventovJan 18, 2024, 10:05 AM
4 points
3 comments1 min readEA link

Can­cel­ling GPT subscription

adekczMay 20, 2024, 4:19 PM
26 points
14 comments3 min readEA link

Perfor­mance of Large Lan­guage Models (LLMs) in Com­plex Anal­y­sis: A Bench­mark of Math­e­mat­i­cal Com­pe­tence and its Role in De­ci­sion Mak­ing.

Jaime Esteban Montenegro BarónMay 6, 2025, 9:08 PM
1 point
0 comments23 min readEA link

Does Re­in­force­ment Learn­ing Really In­cen­tivize Rea­son­ing Ca­pac­ity in LLMs Beyond the Base Model?

Matrice JacobineApr 24, 2025, 2:11 PM
10 points
0 comments1 min readEA link
(limit-of-rlvr.github.io)

“Suc­cess­ful lan­guage model evals” by Ja­son Wei

Arjun PanicksseryMay 25, 2024, 9:34 AM
11 points
0 comments1 min readEA link
(www.jasonwei.net)

Abil­ity to solve long-hori­zon tasks cor­re­lates with want­ing things in the be­hav­iorist sense

So8resNov 24, 2023, 5:37 PM
38 points
1 comment1 min readEA link

Dona­tion offsets for ChatGPT Plus subscriptions

Jeffrey LadishMar 16, 2023, 11:11 PM
76 points
10 comments3 min readEA link

INTELLECT-1 Re­lease: The First Globally Trained 10B Pa­ram­e­ter Model

Matrice JacobineNov 29, 2024, 11:03 PM
2 points
1 comment1 min readEA link
(www.primeintellect.ai)

[Question] I’m in­ter­view­ing the au­thor of ‘Not Born Yes­ter­day’ — Hugo Mercier. He ar­gues peo­ple are less gullible and more savvy than you think. What should I ask him?

Robert_WiblinNov 17, 2023, 5:43 PM
12 points
3 comments1 min readEA link

Still no strong ev­i­dence that LLMs in­crease bioter­ror­ism risk

freedomandutilityNov 2, 2023, 9:23 PM
58 points
9 comments1 min readEA link

Ideas for Next-Gen­er­a­tion Writ­ing Plat­forms, us­ing LLMs

Ozzie GooenJun 4, 2024, 6:40 PM
17 points
0 comments2 min readEA link

The Dis­solu­tion of AI Safety

RokoDec 12, 2024, 10:46 AM
−7 points
0 comments1 min readEA link
(www.transhumanaxiology.com)

Pro­ject Pro­posal Look­ing for Feed­back: Mak­ing Policy Im­pacts Trans­par­ent — A Rea­son­ing Model for Trade, Jobs, and Prices

Echo HuangMay 7, 2025, 5:14 PM
13 points
4 comments7 min readEA link

François Chol­let on why LLMs won’t scale to AGI

Yarrow🔸Apr 15, 2025, 11:01 PM
6 points
2 comments1 min readEA link
(www.youtube.com)

Who owns AI-gen­er­ated con­tent?

Johan S DanielDec 7, 2022, 3:03 AM
−2 points
0 comments2 min readEA link

Open Prob­lems and Fun­da­men­tal Limi­ta­tions of RLHF

stecasAug 17, 2023, 4:50 PM
5 points
0 comments1 min readEA link
(arxiv.org)

ChatGPT bug leaked users’ con­ver­sa­tion histories

Ian TurnerMar 27, 2023, 12:17 AM
15 points
2 comments1 min readEA link
(www.bbc.com)

Ex­plor­ing Tacit Linked Premises with GPT

RomeoStevensMar 24, 2023, 10:50 PM
5 points
0 comments1 min readEA link

GPT5 won’t be what kills us all

DPiepgrassSep 28, 2024, 5:11 PM
3 points
3 comments1 min readEA link
(dpiepgrass.medium.com)

[Question] Can we ever en­sure AI al­ign­ment if we can only test AI per­sonas?

Karl von WendtMar 16, 2025, 8:06 AM
8 points
0 comments1 min readEA link

ChatGPT is ca­pa­ble of cog­ni­tive em­pa­thy!

Miquel Banchs-Piqué (prev. mikbp)Mar 30, 2023, 8:42 PM
3 points
0 comments1 min readEA link
(nonzero.substack.com)

Γαμινγκ the Al­gorithms: Large Lan­guage Models as Mirrors

Haris ShekerisApr 1, 2023, 2:14 AM
5 points
3 comments4 min readEA link

GPTs are Pre­dic­tors, not Imitators

EliezerYudkowskyApr 8, 2023, 7:59 PM
74 points
12 comments1 min readEA link

Scale, schlep, and systems

AjeyaOct 10, 2023, 4:59 PM
59 points
3 comments6 min readEA link

Scal­able And Trans­fer­able Black-Box Jailbreaks For Lan­guage Models Via Per­sona Modulation

soroushjpNov 7, 2023, 6:00 PM
10 points
0 comments2 min readEA link
(arxiv.org)

Fa­vorite Re­cent LLM Prompts & Tips?

Ozzie GooenMar 18, 2025, 4:25 AM
33 points
12 comments1 min readEA link

What is “wire­head­ing”?

Vishakha AgrawalDec 17, 2024, 5:59 PM
1 point
0 comments1 min readEA link
(aisafety.info)

“Long-Ter­mism” vs. “Ex­is­ten­tial Risk”

Scott AlexanderApr 6, 2022, 9:41 PM
526 points
81 comments3 min readEA link

Si­mu­lat­ing a pos­si­ble al­ign­ment solu­tion in GPT2-medium us­ing Archety­pal Trans­fer Learning

MiguelMay 2, 2023, 4:23 PM
4 points
0 comments18 min readEA link

ChatGPT & The EthiSizer Game(s)

Velikovsky_of_NewcastleMay 24, 2023, 8:12 PM
1 point
0 comments40 min readEA link

Google Deep­Mind re­leases Gemini

Yarrow🔸Dec 6, 2023, 5:39 PM
21 points
7 comments1 min readEA link
(deepmind.google)

Align­ment Fak­ing in Large Lan­guage Models

Ryan GreenblattDec 18, 2024, 5:19 PM
142 points
9 comments1 min readEA link
No comments.