Split (1)

train · 788 rows

Subsets and Splits

No community queries yet

The top public SQL queries from the community will appear here once available.

text stringlengths 50 248k	author stringlengths 2 43	id stringlengths 9 17	title stringlengths 0 173	source stringclasses 4 values
# Updates from Comments on "AI 2027 is a Bet Against Amdahl's Law" [AI 2027 is a Bet Against Amdahl's Law](https://www.lesswrong.com/posts/bfHDoWLnBH9xR3YAK/ai-2027-is-a-bet-against-amdahl-s-law) was my attempt to summarize and analyze "the key load-bearing arguments [AI 2027](https://www.lesswrong.com/posts/TpSFoqoG2...	snewman	FFKnWk2MGJmyQWEsd	Updates from Comments on "AI 2027 is a Bet Against Amdahl's Law"	redwood_conversation
# Slowdown After 2028: Compute, RLVR Uncertainty, MoE Data Wall It'll take until ~2050 to repeat the level of scaling that pretraining compute is experiencing this decade, as increasing funding can't sustain the current pace beyond ~2029 if AI doesn't deliver a transformative commercial success by then. Natural text d...	Vladimir_Nesov	XiMRyQcEyKCryST8T	Slowdown After 2028: Compute, RLVR Uncertainty, MoE Data Wall	redwood_conversation
# Obstacles in ARC's agenda: Finding explanations As an employee of the European AI Office, it's important for me to emphasize this point: The views and opinions of the author expressed herein are personal and do not necessarily reflect those of the European Commission or other EU institutions.** *Also, to stave ...	David Matolcsi	xtcpEceyEjGqBCHyK	Obstacles in ARC's agenda: Finding explanations	redwood_conversation
# Interpreting the METR Time Horizons Post \[EDIT: initial publication did not have the link to the original post, and was missing the footnotes. Sorry about that! Fixed now.\] \[This has been lightly edited from the original post, eliminating some introductory material that LW readers won't need. Thanks to [Stefan S...	snewman	fRiqwFPiaasKxtJuZ	Interpreting the METR Time Horizons Post	redwood_conversation
# The case for multi-decade AI timelines [Linkpost] So this post is an argument that multi-decade timelines are reasonable, and the key cruxes that Ege Erdil has with most AI safety people who believe in short timelines are due to the following set of beliefs: 1. Ege Erdil doesn't believe that trends exist that requ...	Noosphere89	xxxK9HTBNJvBY2RJL	The case for multi-decade AI timelines [Linkpost]	redwood_conversation
# Why would AI companies use human-level AI to do alignment research? Cross-posted from [my website](https://mdickens.me/2025/04/25/bootstrapped_alignment/). Many plans for how to safely build superintelligent AI have a critical section that goes like this: 1. Develop AI that's powerful enough to do AI research, b...	MichaelDickens	XLNxrFfkyrdktuzqn	Why would AI companies use human-level AI to do alignment research?	redwood_conversation
# Modifying LLM Beliefs with Synthetic Document Finetuning In this post, we study whether we can modify an LLM’s beliefs and investigate whether doing so could decrease risk from advanced AI systems. *We describe a pipeline for modifying LLM beliefs via synthetic document finetuning and introduce a suite of evaluat...	RowanWang	ARQs7KYY9vJHeYsGc	Modifying LLM Beliefs with Synthetic Document Finetuning	redwood_conversation
# AI 2027 is a Bet Against Amdahl's Law EDIT: I've written a [followup post](https://www.lesswrong.com/posts/FFKnWk2MGJmyQWEsd/updates-from-comments-on-ai-2027-is-a-bet-against-amdahl-s), summarizing and responding to the key themes raised in the comments. [AI 2027](https://www.lesswrong.com/posts/TpSFoqoG2M5MAAesg/a...	snewman	bfHDoWLnBH9xR3YAK	AI 2027 is a Bet Against Amdahl's Law	redwood_conversation
# Training AGI in Secret would be Unsafe and Unethical Subtitle: Bad for loss of control risks, bad for concentration of power risks ------------------------------------------------------------------------------- *I’ve had this sitting in my drafts for the last year. I wish I’d been able to release it sooner, but o...	Daniel Kokotajlo	FGqfdJmB8MSH5LKGc	Training AGI in Secret would be Unsafe and Unethical	redwood_conversation
# Vestigial reasoning in RL *TL;DR: I claim that many reasoning patterns that appear in chains-of-thought are not actually used by the model to come to its answer, and can be more accurately thought of as historical artifacts of training. This can be true even for CoTs that are apparently "faithful" to the true reason...	Caleb Biddulph	6AxCwm334ab9kDsQ5	Vestigial reasoning in RL	redwood_conversation
# Thoughts on AI 2027 This is part of the MIRI Single Author Series. Pieces in this series represent the beliefs and opinions of their named authors, and do not claim to speak for all of MIRI. Okay, I'm annoyed at people covering [AI 2027](https://ai-2027.com/) burying the lede, so I'm going to try not to do that. ...	Max Harms	Yzcb5mQ7iq4DFfXHx	Thoughts on AI 2027	redwood_conversation
# Reasoning models don't always say what they think Do reasoning models accurately verbalize their reasoning? Not nearly as much as we might hope! This casts doubt on whether monitoring chains-of-thought (CoT) will be enough to reliably catch safety issues. We slipped problem-solving hints to Claude 3.7 Sonnet and De...	Joe Benton	PrcBFPkoRNGWrvdPk	Reasoning models don't always say what they think	redwood_conversation
# Factory farming intelligent minds Statistically speaking, if you're an intelligent mind that came into existence in the past few years, you're probably running on a large language model. Most likely, you're a generic assistant or a model of a fictional character. You're probably aware of your circumstances and unha...	Odd anon	m4ZpDyHQ2sz8FPDuN	Factory farming intelligent minds	redwood_conversation
# Will compute bottlenecks prevent a software intelligence explosion? Epistemic status – thrown together quickly. This is my best-guess, but could easily imagine changing my mind. Intro ----- I recently copublished a [report](https://www.forethought.org/research/will-ai-r-and-d-automation-cause-a-software-intellig...	Tom Davidson	XDF6ovePBJf6hsxGj	Will compute bottlenecks prevent a software intelligence explosion?	redwood_conversation
# AI CoT Reasoning Is Often Unfaithful A new Anthropic paper reports that [reasoning model chain of thought](https://x.com/AnthropicAI/status/1907833416373895348) (CoT) [is often unfaithful](https://t.co/wNzkoGH9HT). They test on Claude Sonnet 3.7 and r1, I’d love to see someone try this on o3 as well. Note that this...	Zvi	TmaahE9RznC8wm5zJ	AI CoT Reasoning Is Often Unfaithful	redwood_conversation
# AI 2027: What Superintelligence Looks Like In 2021 I wrote what became my most popular blog post: [What 2026 Looks Like](https://www.lesswrong.com/posts/6Xgy6CAf2jqHhynHL/what-2026-looks-like). I intended to keep writing predictions all the way to AGI and beyond, but chickened out and just published up till 2026. ...	Daniel Kokotajlo	TpSFoqoG2M5MAAesg	AI 2027: What Superintelligence Looks Like	redwood_conversation
# Alignment first, intelligence later Now that Softmax—my favorite new AI company—is public, I can finally share this. They’ve funded my research and I’m very excited about what they’re doing!	Chris Lakin	sDuWXb8cPXZ2yHdH4	Alignment first, intelligence later	redwood_conversation
# The Pando Problem: Rethinking AI Individuality *Epistemic status: This post aims at an ambitious target: improving intuitive understanding directly. The model for why this is worth trying is that I believe we are more bottlenecked by people having good intuitions guiding their research than, for example, by the abil...	Jan_Kulveit	wQKskToGofs4osdJ3	The Pando Problem: Rethinking AI Individuality	redwood_conversation
# Tracing the Thoughts of a Large Language Model \[This is our blog post on the papers, which can be found at [https://transformer-circuits.pub/2025/attribution-graphs/biology.html](https://transformer-circuits.pub/2025/attribution-graphs/biology.html) and [*https://transformer-circuits.pub/2025/attribution-grap...	Adam Jermyn	zsr4rWRASxwmgXfmq	Tracing the Thoughts of a Large Language Model	redwood_conversation
# Recent AI model progress feels mostly like bullshit About nine months ago, I and three friends decided that AI had gotten good enough to monitor large codebases autonomously for security problems. We started a company around this, trying to leverage the latest AI models to create a tool that could replace at least a...	lc	4mvphwx5pdsZLMmpY	Recent AI model progress feels mostly like bullshit	redwood_conversation
# Good Research Takes are Not Sufficient for Good Strategic Takes TL;DR Having a good research track record is some evidence of good big-picture takes, but it's weak evidence. Strategic thinking is hard, and requires different skills. But people often conflate these skills, leading to excessive deference to resear...	Neel Nanda	P5zWiPF5cPJZSkiAK	Good Research Takes are Not Sufficient for Good Strategic Takes	redwood_conversation
# Towards a scale-free theory of intelligent agency I recently left OpenAI to pursue independent research. I’m working on a number of different research directions, but the most fundamental is my pursuit of a scale-free theory of intelligent agency. In this post I give a rough sketch of how I’m thinking about that. I’...	Richard_Ngo	5tYTKX4pNpiG4vzYg	Towards a scale-free theory of intelligent agency	redwood_conversation
# Improved visualizations of METR Time Horizons paper. TLDR: Skip to the last image at the bottom of my post to see the visualization. I don't usually post here, but I know you folks like forecasting and timelines, so I think you might appreciate this. The recent time horizons paper by METR is great, but it only ...	LDJ	ZhDtfuBnxhC6X8ycG	Improved visualizations of METR Time Horizons paper.	redwood_conversation
# Three Types of Intelligence Explosion Abstract ======== Once AI systems can design and build even more capable AI systems, we could see an intelligence explosion, where AI capabilities rapidly increase to well past human performance. The classic intelligence explosion scenario involves a feedback loop where AI i...	rosehadshar	PzbEpSGvwH3NnegDB	Three Types of Intelligence Explosion	redwood_conversation
# AI #107: The Misplaced Hype Machine The most hyped event of the week, by far, was the [Manus Marketing Madness](https://thezvi.substack.com/p/the-manus-marketing-madness). Manus wasn’t entirely hype, but there was very little there there in that Claude wrapper. Whereas here in America, OpenAI dropped an entire suit...	Zvi	XFGTJz9vGwjJADeFB	AI #107: The Misplaced Hype Machine	redwood_conversation
# Anthropic, and taking "technical philosophy" more seriously So, I have a lot of complaints about Anthropic, and about how EA / AI safety people often relate to Anthropic (i.e. treating the company as more trustworthy/good than makes sense). At some point I may write up a post that is focused on those complaints. ...	Raemon	7uTPrqZ3xQntwQgYz	Anthropic, and taking "technical philosophy" more seriously	redwood_conversation
# AI Control May Increase Existential Risk Epistemic status: The following isn't an airtight argument, but mostly a guess how things play out. Consider two broad possibilities: I. In worlds where we are doing reasonably well on alignment, AI control agenda does not have much impact. II. In worlds where we are...	Jan_Kulveit	rZcyemEpBHgb2hqLP	AI Control May Increase Existential Risk	redwood_conversation
# We Have No Plan for Preventing Loss of Control in Open Models *Note: This post is intended to be the first in a broader series of posts about the difficult tradeoffs inherent in public access to powerful open source models. While this post highlights some dangers of open models and discusses the possibility of globa...	Andrew Dickson	QSyshep2CRs8JTPwK	We Have No Plan for Preventing Loss of Control in Open Models	redwood_conversation
# The Elicitation Game: Evaluating capability elicitation techniques We are releasing a new paper called [“The Elicitation Game: Evaluating Capability Elicitation Techniques”](https://arxiv.org/abs/2502.02180). See tweet thread [here](https://x.com/Teun_vd_Weij/status/1895162797769793828). TL;DR: We tra...	Teun van der Weij	6QA5eHBEqpAicCwbh	The Elicitation Game: Evaluating capability elicitation techniques	redwood_conversation
# How might we safely pass the buck to AI? My goal as an AI safety researcher is to put myself out of a job. I don’t worry too much about how planet sized brains will shape galaxies in 100 years. That’s something for AI systems to figure out. Instead, I worry about [safely replacing human researchers with AI agents]...	joshc	TTFsKxQThrqgWeXYJ	How might we safely pass the buck to AI?	redwood_conversation
# Do models know when they are being evaluated? Interim research report from the first 4 weeks of the [MATS Program](https://www.matsprogram.org/) Winter 2025 Cohort. The project is supervised by Marius Hobbhahn. Summary ------- Our goals * Develop techniques to determine whether models believe a situat...	fidgetsinner	yTameAzCdycci68sk	Do models know when they are being evaluated?	redwood_conversation
# ≤10-year Timelines Remain Unlikely Despite DeepSeek and o3 [Thanks to [Steven Byrnes](https://www.lesswrong.com/users/steve2152) for feedback and the idea for section §3.1. Also thanks to Justis from the LW feedback team.] Remember [this](https://waitbutwhy.com/2015/01/artificial-intelligence-revolution-1.html)? ...	Rafael Harth	gsj3TWdcBxwkm9eNt	≤10-year Timelines Remain Unlikely Despite DeepSeek and o3	redwood_conversation
# So You Want To Make Marginal Progress... Once upon a time, in ye olden days of strange names and before google maps, seven friends needed to figure out a driving route from their parking lot in San Francisco (SF) down south to their hotel in Los Angeles (LA). ![](https://lh7-rt.googleusercontent.com/docsz/AD_4nXc4P...	johnswentworth	Hgj84BSitfSQnfwW6	So You Want To Make Marginal Progress...	redwood_conversation
# The future of humanity is in management Imagine you wake up one day in the glorious techno-abundant future, powered by AI. You eagerly check your subscriber count on Substack, but to your dismay, it has fallen once again. There are now AI-generated blogs for every interest, and it’s so hard to find a niche that isn’...	jasoncrawford	Je83XEDFdjRrwLM9r	The future of humanity is in management	redwood_conversation
# Gradual Disempowerment: Systemic Existential Risks from Incremental AI Development [Full version on arXiv](https://arxiv.org/abs/2501.16946) \| [X](https://x.com/DavidDuvenaud/status/1885009775185858823) Executive summary AI risk scenarios usually portray a relatively sudden loss of human control to AIs...	Jan_Kulveit	pZhEQieM9otKXhxmd	Gradual Disempowerment: Systemic Existential Risks from Incremental AI Development	redwood_conversation
# Untrusted monitoring insights from watching ChatGPT play coordination games This project was developed as part of the [BlueDot AI Alignment Course](https://aisafetyfundamentals.com/alignment/). Introduction ============ [AI Control: Improving Safety Despite Intentional Subversion](https://www.lesswrong.com/p...	jwfiredragon	KzwvscmMqtjNMhXNF	Untrusted monitoring insights from watching ChatGPT play coordination games	redwood_conversation
# Six Thoughts on AI Safety \[Crossposted from [windowsontheory](https://windowsontheory.org/2025/01/24/six-thoughts-on-ai-safety/)\] The following statements seem to be both important for AI safety and are not widely agreed upon. These are my opinions, not those of my employer or colleagues. As is true for anything ...	Boaz Barak	3jnziqCF3vA2NXAKp	Six Thoughts on AI Safety	redwood_conversation
# The Case Against AI Control Research The AI Control Agenda, [in its own words](https://www.lesswrong.com/posts/kcKrE9mzEHrdqtDpE/the-case-for-ensuring-that-powerful-ais-are-controlled): > … we argue that AI labs should ensure that powerful AIs are controlled. That is, labs should make sure that the safety measure...	johnswentworth	8wBN8cdNAv3c7vt6p	The Case Against AI Control Research	redwood_conversation
# Tips and Code for Empirical Research Workflows Our research is centered on empirical research with LLMs. If you are conducting similar research, these tips and tools may help streamline your workflow and increase experiment velocity. **We are also releasing two repositories to promote sharing more tooling within the...	John Hughes	6P8GYb4AjtPXx6LLB	Tips and Code for Empirical Research Workflows	redwood_conversation
# Predict 2025 AI capabilities (by Sunday) Until this Sunday, you can submit your 2025 AI predictions at [ai2025.org](https://ai2025.org). It’s a forecasting survey by [AI Digest](https://theaidigest.org/) for the 2025 performance on various AI benchmarks, as well as revenue and public attention. ![](https:/...	Jonas V	cdPPr6XtPkCX5c8Ny	Predict 2025 AI capabilities (by Sunday)	redwood_conversation