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
# "AI Safety for Fleshy Humans" an AI Safety explainer by Nicky Case Nicky Case, of ["The Evolution of Trust"](https://ncase.me/trust/) and ["We Become What We Behold"](https://ncase.itch.io/wbwwb) fame (two quite popular online explainers/mini-games) has written an intro explainer to AI Safety! It looks pretty good ...	habryka	gprh2HD6PDK6AZDqP	"AI Safety for Fleshy Humans" an AI Safety explainer by Nicky Case	redwood_conversation
# Questions for labs Associated with [AI Lab Watch](https://www.lesswrong.com/posts/N2r9EayvsWJmLBZuF/introducing-ai-lab-watch), I sent questions to some labs a week ago (except I failed to reach Microsoft). I didn't really get any replies (one person replied in their personal capacity; this was very limited and they ...	Zach Stein-Perlman	jbJ7FynonxFXeoptf	Questions for labs	redwood_conversation
# Mechanistically Eliciting Latent Behaviors in Language Models Produced as part of the MATS Winter 2024 program, under the mentorship of Alex Turner (TurnTrout). TL,DR: I introduce a method for eliciting latent behaviors in language models by learning unsupervised perturbations of an early layer of an LLM. The...	Andrew Mack	ioPnHKFyy4Cw2Gr2x	Mechanistically Eliciting Latent Behaviors in Language Models	redwood_conversation
# Introducing AI Lab Watch I'm launching [AI Lab Watch](https://ailabwatch.org). I collected actions for frontier AI labs to improve AI safety, then evaluated some frontier labs accordingly. ![](https://39669.cdn.cke-cs.com/rQvD3VnunXZu34m86e5f/images/273a1f624bdcdf6455346ed2554d94d752eb96a2128875f3.png) It's a coll...	Zach Stein-Perlman	N2r9EayvsWJmLBZuF	Introducing AI Lab Watch	redwood_conversation
# We are headed into an extreme compute overhang If we achieve AGI-level performance using an LLM-like approach, the training hardware will be capable of running ~1,000,000s concurrent instances of the model. Definitions =========== Although there is [some debate](https://www.lesswrong.com/posts/icR53xeAkeuzgzsWP/ta...	devrandom	cRFtWjqoNrKmgLbFw	We are headed into an extreme compute overhang	redwood_conversation
# Simple probes can catch sleeper agents This is a link post for the Anthropic Alignment Science team's first "Alignment Note" blog post. We expect to use this format to showcase early-stage research and work-in-progress updates more in the future. [Twitter thread here.](https://twitter.com/AnthropicAI/status/17829...	Monte M	gknc6NWCNuTCe8ekp	Simple probes can catch sleeper agents	redwood_conversation
# Motivation gaps: Why so much EA criticism is hostile and lazy Disclaimer: While I criticize several EA critics in this article, I am myself on the EA-skeptical side of things (especially on AI risk). ### Introduction I am a proud critic of effective altruism, and in particular a critic of AI existential ris...	titotal	ojPJYci3AdgdHsKY8	Motivation gaps: Why so much EA criticism is hostile and lazy	redwood_conversation
# Inducing Unprompted Misalignment in LLMs Emergent Instrumental Reasoning Without Explicit Goals ====================================================== TL;DR: LLMs can act and scheme without being told to do so. This is bad. *Produced as part of Astra Fellowship - Winter 2024 program, mentored by Evan Hubing...	Sam Svenningsen	ukTLGe5CQq9w8FMne	Inducing Unprompted Misalignment in LLMs	redwood_conversation
# Discriminating Behaviorally Identical Classifiers: a model problem for applying interpretability to scalable oversight In a new preprint, [Sparse Feature Circuits: Discovering and Editing Interpretable Causal Graphs in Language Models](https://arxiv.org/abs/2403.19647), my coauthors and I introduce a technique, Spar...	Sam Marks	s7uD3tzHMvD868ehr	Discriminating Behaviorally Identical Classifiers: a model problem for applying interpretability to scalable oversight	redwood_conversation
# Staged release "Staged release" is regularly mentioned as a good thing for frontier AI labs to do. But I've only ever seen one analysis of staged release,[^uj9a8bk9zt] and the term's meaning has changed, becoming vaguer since the GPT-2 era. This post is kinda a reference post, kinda me sharing my understanding/ta...	Zach Stein-Perlman	RdsqJnmP8XxutLbso	Staged release	redwood_conversation
# A quick experiment on LMs’ inductive biases in performing search TL;DR: Based on a toy setting, GPT-3.5-turbo and GPT-4-turbo are best at search (by which I mean computing an argmax) when using chain-of-thought, but neither of them can do internal search when forced to work from memory over only a few token positi...	Alex Mallen	XaKLjyDejtXDoRAzL	A quick experiment on LMs’ inductive biases in performing search	redwood_conversation
# Is LLM Translation Without Rosetta Stone possible? Suppose astronomers detect a binary radio signal, an alien message, from a star system many light years away. The message contains a large text dump (conveniently, about GPT-4 training text data sized) composed in an alien language. Let's call it Alienese.[^1] Unfo...	cubefox	J3zA3T9RTLkKYNgjw	Is LLM Translation Without Rosetta Stone possible?	redwood_conversation
# Run evals on base models too! (Creating more visibility for a [comment thread with Rohin Shah](https://www.lesswrong.com/posts/5Dz3ZrwBzzMfaucrH/ai-57-all-the-ai-news-that-s-fit-to-print?commentId=AaapY2KK4fDvwWaRT).) Currently, [DeepMind's capabilities evals](https://twitter.com/rohinmshah/status/17707466641411732...	orthonormal	dgFC394qZHgj2cWAg	Run evals on base models too!	redwood_conversation
# The Case for Predictive Models Thanks to Johannes Treutlein and Paul Colognese for feedback on this post. Just over a year ago, the [Conditioning Predictive Models](https://arxiv.org/abs/2302.00805) paper was released. It laid out an argument and a plan for using powerful predictive models to reduce existential r...	Rubi J. Hudson	RHDB3BdnvM233bnhG	The Case for Predictive Models	redwood_conversation
# Sparsify: A mechanistic interpretability research agenda Over the last couple of years, mechanistic interpretability has seen substantial progress. Part of this progress has been enabled by the identification of superposition as a key barrier to understanding neural networks ([Elhage et al., 2022](https://transforme...	Lee Sharkey	64MizJXzyvrYpeKqm	Sparsify: A mechanistic interpretability research agenda	redwood_conversation
# Modern Transformers are AGI, and Human-Level *This is my personal opinion, and in particular, does not represent anything like a MIRI consensus; I've gotten push-back from almost everyone I've spoken with about this, although in most cases I believe I eventually convinced them of the narrow terminological point I'm ...	abramdemski	gP8tvspKG79RqACTn	Modern Transformers are AGI, and Human-Level	redwood_conversation
# Announcing Neuronpedia: Platform for accelerating research into Sparse Autoencoders This posts assumes basic familiarity with Sparse Autoencoders. For those unfamiliar with this technique, we highly recommend the introductory sections of [these](https://transformer-circuits.pub/2023/monosemantic-features)[*paper...	Johnny Lin	BaEQoxHhWPrkinmxd	Announcing Neuronpedia: Platform for accelerating research into Sparse Autoencoders	redwood_conversation
# On the Confusion between Inner and Outer Misalignment Here’s my take on why the distinction between inner and outer-alignment frame is weird/unclear/ambiguous in some circumstances: My understanding is that these terms were originally used when talking about AGI. So outer alignment involved writing down a reward or ...	Chris_Leong	hueNHXKc4xdn6cfB4	On the Confusion between Inner and Outer Misalignment	redwood_conversation
# AI #56: Blackwell That Ends Well Hopefully, anyway. Nvidia has a new chip. Also Altman has a new interview. And most of Inflection has new offices inside Microsoft. #### Table of Contents 1. Introduction. 2. Table of Contents. 3. Language Models Offer Mundane Utility. Open the book. 4. Clauding Along. Claude...	Zvi	iH5Sejb4dJGA2oTaP	AI #56: Blackwell That Ends Well	redwood_conversation
# DeepMind: Evaluating Frontier Models for Dangerous Capabilities > To understand the risks posed by a new AI system, we must understand what it can and cannot do. Building on prior work, we introduce a programme of new “dangerous capability” evaluations and pilot them on Gemini 1.0 models. Our evaluations cover four ...	Zach Stein-Perlman	CCBaLzpB2qvwyuEJ2	DeepMind: Evaluating Frontier Models for Dangerous Capabilities	redwood_conversation
# New report: Safety Cases for AI ArXiv paper: [https://arxiv.org/abs/2403.10462](https://arxiv.org/abs/2403.10462) ![](https://39669.cdn.cke-cs.com/rQvD3VnunXZu34m86e5f/images/d5a38a73c2d9a9802feb02ca158bf3791c40e96ced522daf.png) The idea for this paper occurred to me when I saw Buck Shlegeris' MATS stream on "...	joshc	HrtyZm2zPBtAmZFEs	New report: Safety Cases for AI	redwood_conversation
# On Devin #### Introducing Devin Is the era of AI agents writing complex code systems without humans in the loop upon us? [Cognition is calling Devin](https://www.cognition-labs.com/blog) ‘the first AI software engineer.’ [Here is a two minute demo](https://twitter.com/cognition_labs/status/1767548763134964000) of...	Zvi	wovJBkfZ8rTyLoEKv	On Devin	redwood_conversation
# What is the best argument that LLMs are shoggoths? Where can I find a post or article arguing that the internal cognitive model of contemporary LLMs is quite alien, strange, non-human, even though they are trained on human text and produce human-like answers, which are rendered "friendly" by RLHF? To be clear...	JoshuaFox	FyRDZDvgsFNLkeyHF	What is the best argument that LLMs are shoggoths?	redwood_conversation
# 'Empiricism!' as Anti-Epistemology (Crossposted by [habryka](https://www.lesswrong.com/users/habryka4) after asking Eliezer whether I could post it under his account) ### i. "Ignore all these elaborate, abstract, theoretical predictions," the Spokesperson for Ponzi Pyramid Incorporated said in a firm, re...	Eliezer Yudkowsky	LvKDMWQ3yLG9R3gHw	'Empiricism!' as Anti-Epistemology	redwood_conversation
# Highlights from Lex Fridman’s interview of Yann LeCun Introduction ============ Yann LeCun is perhaps the most prominent critic of the “LessWrong view” on AI safety, the only one of the three "godfathers of AI" to not acknowledge the risks of advanced AI. So, when he recently [appeared on the Lex Fridman podcast](h...	Joel Burget	bce63kvsAMcwxPipX	Highlights from Lex Fridman’s interview of Yann LeCun	redwood_conversation
# Laying the Foundations for Vision and Multimodal Mechanistic Interpretability & Open Problems ![](https://39669.cdn.cke-cs.com/rQvD3VnunXZu34m86e5f/images/6091501cda563a07e8ee0efd8663c57df34225caa955db04.png) Behold the dogit lens. Patch-level logit attribution is an emergent segmentation map. Join our Discord [...	Sonia Joseph	kobJymvvcvhbjWFKe	Laying the Foundations for Vision and Multimodal Mechanistic Interpretability & Open Problems	redwood_conversation
# What could a policy banning AGI look like? [Caveat lector: I know roughly nothing about policy!] Suppose that there were political support to really halt research that might lead to an unstoppable, unsteerable transfer of control over the lightcone from humans to AGIs. What government policy could exert that politi...	TsviBT	X9Z9vdG7kEFTBkA6h	What could a policy banning AGI look like?	redwood_conversation
# Results from an Adversarial Collaboration on AI Risk (FRI) Authors of linked report: Josh Rosenberg, Ezra Karger, Avital Morris, Molly Hickman, Rose Hadshar, Zachary Jacobs, Philip Tetlock[^crpys2viudt] Today, the Forecasting Research Institute (FRI) released “[Roots of Disagreement on AI Risk: Exploring the Pote...	Josh Rosenberg	94K6pskgqBmuxsJLx	Results from an Adversarial Collaboration on AI Risk (FRI)	redwood_conversation
# Simple versus Short: Higher-order degeneracy and error-correction > TLDR: The simplicity bias in Bayesian statistics is not just a bias towards short description length. The folklore relating the simplicity bias in Bayesian statistics to description length is incomplete: while it is true that the fewer p...	Daniel Murfet	nWRj6Ey8e5siAEXbK	Simple versus Short: Higher-order degeneracy and error-correction	redwood_conversation
# Deconstructing Bostrom's Classic Argument for AI Doom I had a pretty great discussion with social psychologist and philosopher Lance Bush recently about the orthogonality thesis, which ended up turning into a broader analysis of Nick Bostrom's argument for AI doom as presented in Superintelligence, and some relate...	Nora Belrose	RbynKk3evb6RiLryL	Deconstructing Bostrom's Classic Argument for AI Doom	redwood_conversation
# One-shot strategy games? I'm looking for computer games that involve strategy, resource management, hidden information, and management of "value of information" (i.e. figuring out when to explore or exploit), which: * \can\ be beaten in 30 – 120 minutes on your first try (or, there's a clear milestone that's ab...	Raemon	DvRBSzFjfaPYBhwmj	One-shot strategy games?	redwood_conversation
# Many arguments for AI x-risk are wrong _The following is a lightly edited version of a memo I wrote for a retreat. It was inspired by a draft of [Counting arguments provide no evidence for AI doom](https://www.lesswrong.com/posts/YsFZF3K9tuzbfrLxo/counting-arguments-provide-no-evidence-for-ai-doom). I think that my ...	TurnTrout	yQSmcfN4kA7rATHGK	Many arguments for AI x-risk are wrong	redwood_conversation
# Are we so good to simulate? If you believe that,— a) a civilization like ours is likely to survive into technological incredibleness, and b) a technologically incredible civilization is very likely to create ‘ancestor simulations’, —then the Simulation Argument says you should expect that you are currently in suc...	KatjaGrace	di4Dhho4xZ4x9ABna	Are we so good to simulate?	redwood_conversation
# Bengio's Alignment Proposal: "Towards a Cautious Scientist AI with Convergent Safety Bounds" Yoshua Bengio recently posted a high-level overview of his alignment research agenda on his blog. I'm pasting the full text below since it's fairly short. > What can’t we afford with a future superintelligent AI? Among othe...	mattmacdermott	edvyWfKdJHnoPkM2J	Bengio's Alignment Proposal: "Towards a Cautious Scientist AI with Convergent Safety Bounds"	redwood_conversation
# Counting arguments provide no evidence for AI doom Crossposted from the [AI Optimists blog](https://optimists.ai/2024/02/27/counting-arguments-provide-no-evidence-for-ai-doom/). AI doom scenarios often suppose that future AIs will engage in scheming— planning to escape, gain power, and pursue ulterior mot...	Nora Belrose	YsFZF3K9tuzbfrLxo	Counting arguments provide no evidence for AI doom	redwood_conversation
# The Shutdown Problem: Incomplete Preferences as a Solution Preamble ============ This post is an updated explanation of the POST-Agents Proposal (PAP): my proposed solution to the shutdown problem.[^g6wtgrwwuj4] The post is shorter than my AI Alignment Awards [contest entry](https://s3.amazonaws.com/pf-user-fil...	Elliott Thornley (EJT)	YbEbwYWkf8mv9jnmi	The Shutdown Problem: Incomplete Preferences as a Solution	redwood_conversation
# Extinction Risks from AI: Invisible to Science? Abstract: In an effort to inform the discussion surrounding existential risks from AI, we formulate Extinction-level Goodhart’s Law as "Virtually any goal specification, pursued to the extreme, will result in the extinction[^wwed95q0dgc]* of humanity*'', and we a...	VojtaKovarik	d5oqvgCR7SDf5m4k4	Extinction Risks from AI: Invisible to Science?	redwood_conversation
# Auditing LMs with counterfactual search: a tool for control and ELK Consider a scalable oversight setting where a super-human model knows of a subtle flaw[^1] in an input (plan, code, or argument) but also knows that humans, with their limited knowledge, would overlook this flaw. In this case, ordinary language mode...	Jacob Pfau	Y2nxEbDfcoWhC8C2D	Auditing LMs with counterfactual search: a tool for control and ELK	redwood_conversation
# Critiques of the AI control agenda Thanks to John Wentworth, Garrett Baker, Theo Chapman, and David Lorell for feedback and discussions on drafts of this post. In this post I’ll describe some of my thoughts on the [AI control research agenda](https://www.lesswrong.com/posts/kcKrE9mzEHrdqtDpE/the-case-for-ensuring...	Jozdien	j9Ndzm7fNL9hRAdCt	Critiques of the AI control agenda	redwood_conversation
# Dreams of AI alignment: The danger of suggestive names Let's not forget the old, well-read post: [Dreams of AI Design](https://www.lesswrong.com/posts/p7ftQ6acRkgo6hqHb/dreams-of-ai-design). In that essay, Eliezer correctly points out errors in imputing meaning to nonsense by using suggestive names to describe the n...	TurnTrout	yxWbbe9XcgLFCrwiL	Dreams of AI alignment: The danger of suggestive names	redwood_conversation