LessWrong
PulseAugur coverage of LessWrong — every cluster mentioning LessWrong across labs, papers, and developer communities, ranked by signal.
-
Anthropic adopts alignment pretraining for AI safety
Anthropic is now employing an alignment pretraining technique, which involves training AI models on data demonstrating desired behavior in challenging ethical scenarios. This method, also referred to as safety pretraini…
-
AI's lack of introspection doesn't mean it's uncooperative, argues LessWrong
This article argues that a lack of introspective ability in AI does not equate to a lack of corrigibility. It draws an analogy to human capabilities like face recognition, which are complex and not fully understood by t…
-
Inner work testimonials focus on experience, not results, analysis finds
A recent analysis of testimonials from prominent "inner work" practitioners suggests that the field may be prioritizing experiences over tangible life improvements. The author reviewed numerous testimonials and found th…
-
Impact Accelerator Program opens applications for career transition
High Impact Professionals (HIP) has opened applications for its 6-week Impact Accelerator Program (IAP). This free program aims to equip experienced professionals with the skills to pursue high-impact careers. To date, …
-
Community organizer term creates false assumptions, author argues
The author argues that the term "community organizer" is a problematic oxymoron, suggesting that its continued use creates false assumptions. Specifically, it implies that a community must have an organizer and that suc…
-
Essay likens civilization to gaming exploits and nature's 'bio-hacking'
This essay explores the concept of exploiting system loopholes, drawing parallels between gaming "munchkinry" and real-world security exploits. It posits that nature itself is the original "bio-hacker," having exploited…
-
LessWrong post offers Guesstimate model for prediction market returns
A LessWrong post introduces a Guesstimate model designed to calculate the expected growth rate for investments in real-money prediction markets. The model takes inputs such as share cost, holding rewards, win probabilit…
-
Author defines 'Hot Take' as novel, personal, normative claim
The author explores the concept of a "Hot Take," defining it as a simple, novel, and personal normative claim that challenges conventional wisdom. They argue that while many opinions are not truly "hot takes," crafting …
-
Sales interactions involve buyer and seller deception and persuasion tactics
The dynamics of sales involve a complex interplay of deception and persuasion, where both buyers and sellers may employ fallacies and untruths. Buyers often use the phrase "let me think about it" as a polite way to avoi…
-
AI alignment faces challenge distinguishing guidance from manipulation
This post explores the difficulty in distinguishing between beneficial guidance and harmful manipulation when conceptualizing AI alignment. The author argues that human desires are inherently manipulable, making it chal…
-
Developmental psychology: Uneven progress across life stages explained
Developmental psychology suggests that individuals progress through distinct stages, but this progression is often uneven across different life domains. A person might exhibit advanced behaviors in one area, like profes…
-
LessWrong user asks about deleting content from moderation log
This post on LessWrong asks a technical question regarding the platform's content moderation system. Specifically, the user is inquiring about the possibility of removing deleted content from the moderation log. The que…
-
AI safety advocates propose prize money to spur innovation
The author proposes a shift from traditional "push funding" in AI safety to "pull funding" mechanisms, which reward specific outcomes rather than processes. This approach, inspired by initiatives like DARPA Grand Challe…
-
AI's 'Anti-Singularity' Future: Task-Specific Models Over Universal Intelligence
A recent blog post proposes a new paradigm in machine learning, moving away from abstract theories towards using large language models to tirelessly iterate on complex designs for specific tasks. This approach, termed t…
-
Author questions historical progress via 'Darwinian honeymoon' concept
The author argues that historical human progress, often cited as evidence for optimism, is less convincing than it appears. They propose a "Darwinian honeymoon" concept, suggesting that initial optimization processes be…
-
New alignment strategy targets LLM context rot
A recent analysis on LessWrong proposes a novel approach to address the
-
LessWrong user seeks accessible AI existential risk explainers
A user on LessWrong is seeking recommendations for the best introductory resources on AI existential risk, aiming for materials that are simple, concise (around 15 minutes), and accessible to a broad audience, including…