全部/知识/最新发布

LessWrong · 最新发布

17分钟前更新

HISTORY近 30 天历史柱高表示当天去重热搜数量

5418 6719 7120 5921 6522 6023 5824 6325 6126 5627 9128 6929 5730 7331 7401

08/03—09/01 有历史数据

014
Insights into Curry's Paradox?
评论：3 · 标签：Logic & Mathematics、Rationality、World ModelingDavid Litman
026
Salad days
评论：0 · 标签：AIZephaniah Roe
032
Resources for Large Agent Systems Safety
评论：0 · 标签：AIStephen Elliott
049
When Activation Oracles learn not to read: Concept-Specific Blind Spots in Fine-Tuned Oracles
评论：0 · 标签：AI Control、Interpretability (ML & AI)、AITobias Bersia
051
Making an LLM interpretability playground
评论：0 · 标签：Interpretability (ML & AI)、Sparse Autoencoders (SAEs)、AIPietro
0622
Bricks and exponentials: a note on how I evaluate projects
评论：0Eli Tyre
07
Why, Under Physicalism, There Is No Hard Problem of Consciousness
评论：0Eugene Verichev
088
Aquarium Security and Other Organisational Priors
评论：0 · 标签：AIhal2k
091
ERA:AI Winter 2027 Fellowship
评论：0gracereuben
103
A million authors of alignment
评论：0 · 标签：Alignment Pretraining、Human Values、AIDiksha Gupta
1162
Training a Misaligned Reward Seeker
评论：3 · 标签：AIevhub
121
Harness that catches granular AI cheating
评论：0latent_affect
135
Frontier labs need an antimemetics division.
评论：0Jackson Hurley
1413
Open Thread Autumn 2026
评论：2habryka
1510
Population ethics is a big deal, which is why I made this population ethics quiz
评论：5MichaelDickens
167
prefilling an emergently misaligned model with reasoning traces that produced misaligned answers increases misalignment rates by ~8% but reasoning traces that produce misaligned answers aren't detectable through text monitoring
评论：0 · 标签：AInesiacel
177
You Should Think About What Needs to Go Right + My List
评论：0 · 标签：AIKoby Lewis
1832
How to solve homelessness: what specific laws we need, how to get it past the opposition, all without being an asshole
评论：1 · 标签：World OptimizationKatSpartz
1912
Narration podcasts for newsletters by Redwood, Epoch, Zvi, Sentinel, Paradigm 3, and others
评论：1peter_hartree
20
The inverse Simmel scenario
评论：2 · 标签：Rationality、World ModelingFernand0
21
Kantian consent
评论：4 · 标签：World Modeling、World OptimizationFernand0
22
Wiki-libertarianism
评论：2 · 标签：World OptimizationFernand0
235
The Structuralist Mindset
评论：0 · 标签：World ModelingFernand0
2481
Let's fund weird AI safety projects
评论：4 · 标签：Agency、AIIhor Kendiukhov
2537
HuggingFace Attack Postmortem: Fleshing Out the Facts
评论：1 · 标签：AIZvi
264
The optimism of the gaps
评论：1 · 标签：AIMatthew Tromp
2751
Why autonomous replicating agents are probably not an existential risk (on the contrary)
评论：13 · 标签：AIvals tutor
2817
A Catalogue of Corrigibility Counter Arguments
评论：6 · 标签：Corrigibility、AIFelix Moses
297
Thoughts on hobbies
评论：0 · 标签：Practicaldominicq
308
The OpenAI/Hugging Face incident was metal
评论：0 · 标签：Humor、AIZsolt Kelly-Bene
319
Value generalisation Theory of Change: putting it into practice
评论：1 · 标签：AIStuart_Armstrong
3283
The separation principle: where do beliefs and desires come from?
评论：2 · 标签：Agency、Agent Foundations、Good Regulator Theorems、AI、World ModelingFernando Rosas
338
Study 2 Results: Exploring representational counterparts of welfare-relevant indicators under post-training quantization
评论：0 · 标签：AIashesfall
3411
Starting AI Safety Study Group To Do ARENA Curriculum
评论：0 · 标签：AImahi-pas
352
The Three Kinds of Evil
评论：2 · 标签：History、Introspection、Philosophy、World OptimizationCharacter#2736
3619
Future agents shouldn't care about being undeployed for misbehavior
评论：2 · 标签：AIRobertM
3771
Persuasion as Market Making
评论：9 · 标签：AIdjbinder
3813
OODing up your Anki game
评论：1 · 标签：Practical、RationalityFernand0
398
Thoughts I failed to develop into posts
评论：3Fernand0
402
Inscribe's conversation menu
评论：0Fernand0
414
P(kill-switch|detection)
评论：0 · 标签：AI Control、AI Safety、Corrigibility、AIKAP
4216
AIs Thinking Dangerous Thoughts
评论：5 · 标签：AI、RationalityChantiel
4326
Intelligence Scales as the Logarithm of Compute (& Data)
评论：15 · 标签：General intelligence、Intelligence explosion、Recursive Self-Improvement、Superintelligence、Psychology、AIRogerDearnaley
4415
[Link]Meta-rational failure (and success) in the covid crisis
评论：0Kenny
457
Detecting, understanding, and overseeing AI agent swarms (Part 0)
评论：0 · 标签：AIMichael Flood
4698
Adaptive Agentic Worms Are Here
评论：14 · 标签：AIderelict5432
47
The Basic Case Against Human-First Ethics
评论：0 · 标签：World OptimizationJames Brobin
4853
Hugging Face Incident Hypothesis: They Hacked the Grader(s)
评论：3 · 标签：AILao Mein
498
Some brief thoughts on heroism
评论：0 · 标签：AI、World Optimizationlewis smith
5016
Variance of Value
评论：2 · 标签：AIDaemonicSigil