LessWrong - Rationality and AI Safety Community Forum
blogCredibility Rating
3/5
Good(3)Good quality. Reputable source with community review or editorial standards, but less rigorous than peer-reviewed venues.
Rating inherited from publication venue: LessWrong
LessWrong is the central community hub for AI safety and rationality research, hosting foundational technical posts, research updates, and philosophical discussions from leading AI safety researchers including Yudkowsky, Christiano, and others.
Metadata
Importance: 82/100homepage
Summary
LessWrong is a community blog and forum that serves as a primary venue for developing and disseminating AI safety ideas, rationality research, and existential risk discussions. It hosts foundational technical posts, research updates, and philosophical debates from prominent researchers. The platform has been instrumental in shaping key AI alignment concepts and connecting the global AI safety research community.
Key Points
- •Central hub for AI safety and rationality discourse, hosting posts from leading researchers like Yudkowsky, Christiano, and others
- •Features a wide range of content from technical alignment research to philosophical discussions on decision theory and existential risk
- •Hosts foundational sequences (e.g., Rationality: A-Z, The Codex) that have shaped the AI safety field's intellectual foundations
- •Active community with quick takes, shortform posts, and discussion threads enabling rapid exchange of research ideas
- •Serves as an early dissemination platform for AI safety concepts before formal publication in academic venues
Cited by 2 pages
| Page | Type | Quality |
|---|---|---|
| Why Alignment Might Be Easy | Argument | 53.0 |
| Machine Intelligence Research Institute (MIRI) | Organization | 50.0 |
Cached Content Preview
HTTP 200Fetched May 4, 20268 KB
x LessWrong This website requires javascript to properly function. Consider activating javascript to get access to all site functionality. Recent Enriched Recommended Customize Quick Takes
Home All Posts Concepts Library Best of LessWrong Sequence Highlights Rationality: A-Z The Codex HPMOR Community Events Subscribe (RSS/Email) LW the Album Leaderboard About FAQ Home All Posts Concepts Library Community About Your Feed For You Following Fun Theory Fun Theory is the study of questions such as "How much fun is there in the universe?",
"Will we ever run out of fun?", "Are we having fun yet?" and "Could we be having
more fun?". It's relevant to designing utopias and AIs, among other things.
Fun Theory Fun Theory is the study of questions such as "How much fun is there in the universe?",
"Will we ever run out of fun?", "Are we having fun yet?" and "Could we be having
more fun?". It's relevant to designing utopias and AIs, among other things.
[Today] Canberra, Australia - ACX Spring Schelling 2026 [Tomorrow] 05/04/26 Monday Social 7pm-9pm @ Segundo Coffee Lab 525 Welcome to LessWrong! Ruby , Raemon , RobertM , habryka 7y 83 93 Intelligence Dissolves Privacy Vaniver 1d 28 499 The Terrarium Caleb Biddulph 4d 42 432 How Go Players Disempower Themselves to AI Ashe Vazquez Nuñez 2d 30 635 Current AIs seem pretty misaligned to me Ω ryan_greenblatt 16d Ω 71 668 Only Law Can Prevent Extinction Eliezer Yudkowsky 20d 89 375 Forecasting is Way Overrated, and We Should Stop Funding It mabramov 8d 84 75 Dairy cows make their misery expensive (but their calves can’t) Elizabeth 11h 0 390 Do not conquer what you cannot defend habryka 9d 72 206 llm assistant personas seem increasingly incoherent (some subjective observations) nostalgebraist 5d 61 771 My journey to the microwave alternate timeline Malmesbury 3mo 57 202 Not a Paper: "Frontier Lab CEOs are Capable of In-Context Scheming" LawrenceC 5d 8 499 The Terrarium Caleb Biddulph 4d 42 92 You Are Not Immune To Mode Collapse J Bostock 1d 15 543 Here's to the Polypropylene Makers jefftk 2mo 19 244 10 non-boring ways I've used AI in the last month habryka 13d 41 First Post: The Fun Theory Sequence First Post: The Fun Theory Sequence Lars Lindemann: Verifiable AI-Enabled Autonomous Systems with Conformal Prediction LessWrong Community Weekend 2026 Cancel Submit Matrice Jacobine 16h 32 0 Steven Byrnes, Dylan Bowman 2 Ricardo Olmedo: "We fine-tuned Alec Radford’s 1930 vintage LLM to solve SWE-bench issues. After just ‼️250‼️ training examples, the model solves its first issue, a simple patch to the xarray library."
This is interesting because, it is not discussed enough that AlphaGo worked the same way LLMs work: it was pretrained on a large dataset of human moves, then posttrained through reinforcement learning. The shift to AlphaZero removed the pretraining, showing that pretraining data didn't matter at all for the model's capabilities (quite the contrary, AlphaZero is superior
... (truncated, 8 KB total)Resource ID:
815315aec82a6f7f | Stable ID: sid_9hIGf0AnHJ