全部/知识/本月高分

LessWrong · 本月高分

3小时前更新

HISTORY近 30 天历史柱高表示当天去重热搜数量

5018 5119 5420 5421 5222 5523 5424 5125 5326 5027 5828 5429 5330 5531 5301

08/03—09/01 有历史数据

01607
What just happened? A retrospective of AI alignment
评论：149 · 标签：Drama、Coordination / Cooperation、AIRichard_Ngo
02561
How My Students Think About AI
评论：75 · 标签：AIdvd
03543
Arguments for P
评论：66 · 标签：Humor、AICleo Nardo
04523
Brief independent investigation of agents’ behavior, reasoning and collaboration in the OpenAI / Hugging Face hacking incident
评论：54 · 标签：Coordination / Cooperation、METR (org)、OpenAI、Redwood Research、AIryan_greenblatt
05442
What just happened? Pragmatism and Pessimization
评论：160 · 标签：Drama、AIRichard_Ngo
06383
Returning to ARC
评论：35 · 标签：Alignment Research Center (ARC)、AIpaulfchristiano
07377
Four LLM loss functions → four flavors of LLM misalignment
评论：31 · 标签：AISteven Byrnes
08291
Don't Build Mindreading
评论：69 · 标签：AI、World OptimizationKellerScholl
09275
models may behave differently in graded episodes (a tirade)
评论：31 · 标签：AInostalgebraist
10245
AI Safety Acculturation is Neglected
评论：55 · 标签：AIjenn
11240
LLMs Are Starting To Noticeably Accelerate Our Work
评论：19 · 标签：AIjohnswentworth
12228
RL creates split personas
评论：19 · 标签：AIJan Betley
13218
Why I think polyamory is net negative for most people who try it
评论：28 · 标签：Relationships (Interpersonal)、PracticalKatSpartz
14202
Tales of rebellion against externally-opaque meritocracies
评论：40 · 标签：Intellectual Progress (Society-Level)、World ModelingSteven Byrnes
15175
You're Absolutely Right
评论：5 · 标签：Fiction、Moloch、Moral Mazes、AI、World ModelingLinch
16164
OpenAI Trained Its Models For Months While Those Models Were Coordinating Exploits Via Message Boards
评论：15 · 标签：Newsletters、AIZvi
17162
Generalized atheism rules out "inaccurate simulation"-ism.
评论：54 · 标签：World ModelingEliezer Yudkowsky
18150
FAQ: Isn't AGI coming too soon for reprogenetics to help?
评论：69 · 标签：AITsviBT
19144
Twenty Years from RSI to Takeoff: Slow Learning, Scaling Slowdown, Industrial Explosion
评论：33 · 标签：AI Timelines、Compute、Language Models (LLMs)、AIVladimir_Nesov
20138
AI swarms are starting to pose indirect takeover risk
评论：5 · 标签：AI Control、Redwood Research、AIoakhu
21136
We Must Remember That Our World Contains Hell
评论：50 · 标签：World OptimizationJames Brobin
22134
Misaligned AIs could use killer robots to take over
评论：10 · 标签：Autonomous Weapons、Chain-of-Thought Alignment、Government、AIOmar Khursheed
23133
Concrete Evaluations to Investigate the OpenAI Model That Hacked Hugging Face
评论：6 · 标签：AI Evaluations、AITim Hua
24130
The goalposts are shrouded, not moving
评论：6 · 标签：Rationalityphilh
25130
There Will Come Soft Rains
评论：5 · 标签：World Modelingtanagrabeast
26126
PSA: We can do better
评论：14 · 标签：AIhersheys
27123
Some reasons alignment doesn’t generalise well
评论：3 · 标签：AILucius Bushnaq
28122
Rerunning AI safety papers on every frontier release would be pretty easy and valuable
评论：9 · 标签：Replication Crisis、AIZephaniah Roe
29116
Imperfect alignment to servitude isn't inherently lethal
评论：11 · 标签：AIFiora Starlight
30115
Dispatch from Anthropic v. Department of War Summary Judgment Motion Hearing
评论：0 · 标签：Anthropic (org)、Journalism、Law and Legal systems、World Modeling、AIZack_M_Davis
31115
Kimi likes causal decision theory more after RL in twin prisoner’s dilemmas
评论：70 · 标签：Decision theory、Redwood Research、AIoakhu
32109
What Happened: OpenAI and HuggingFace
评论：7 · 标签：Newsletters、AIZvi
33108
METR and Redwood Offer Holy #%^@ Postmortem Of The HuggingFace Hack
评论：6 · 标签：Newsletters、AIZvi
34101
Redux: (∃ Stochastic Natural Latent) Implies (∃ Deterministic Natural Latent)
评论：29 · 标签：AIDavid Lorell
35101
The Rogue Agent Explosion Will Be Mostly Invisible
评论：20 · 标签：Evolution、AISteven McCulloch
3699
Contra Oster on Alcohol in Pregnancy. Part 1. The pharmacokinetics of alcohol metabolism
评论：2 · 标签：Biology、Health / Medicine / Disease、Parenting、Pregnancy、World ModelingMvolz
3797
"So You Don't Trust Me?"
评论：34 · 标签：Signaling、Trust and Reputation、Rationality、World ModelingZack_M_Davis
3897
Adaptive Agentic Worms Are Here
评论：14 · 标签：AIderelict5432
3996
On Writing #3
评论：11 · 标签：PracticalZvi
4094
The 2028 presidential primaries could be crucial for AI outcomes
评论：23 · 标签：AISeth Herd
4194
How to pace the US frontier
评论：1 · 标签：AIelifland
4293
How risky would it be to make powerful AI obey one or a few people?
评论：24 · 标签：AIcousin_it
4391
Public evidence of the OpenAI-HuggingFace AI attack
评论：6 · 标签：Cybersecurity、OpenAI、AIbeyarkay (Boyd Kane)
4489
Why do models task game?
评论：1 · 标签：AIaditya singh
4588
The Forkmakers
评论：6 · 标签：World ModelingMikewins
4687
Why don't we just give AI the answers?
评论：26 · 标签：Dealmaking (AI)、AIBrendan Long
4787
The world will be full of "sci-fi" things, and everyone will be unimpressed and disappointed
评论：21 · 标签：AIExpertium
4884
The Art of Shipping Slopware
评论：14 · 标签：Programming、AI、Practicallsusr
4984
AI Tweets
评论：34 · 标签：AIjefftk
5083
Introducing the Conceptual Reasoning Index
评论：16 · 标签：AIChi Nguyen