text stringlengths 300 320k | source stringlengths 52 154 |
|---|---|
# An adversarial example for Direct Logit Attribution: memory management in gelu-4l
*Please check out *[*our notebook*](https://colab.research.google.com/drive/16Kp-4iH330a1dF6F0ntPK7kfNuxqkvfZ#scrollTo=0WpBUiAYkEdv) *for figure recreation and to examine your own model for clean-up behavior.*
*Produced as part of *[*... | https://www.lesswrong.com/posts/2PucFqdRyEvaHb4Hn/an-adversarial-example-for-direct-logit-attribution-memory |
# Report on Frontier Model Training
Understanding what drives the rising capabilities of AI is important for those who work to forecast, regulate, or ensure the safety of AI. Regulations on the export of powerful GPUs need to be informed by understanding of how these GPUs are used, forecasts need to be informed by bot... | https://www.lesswrong.com/posts/nXcHe7t4rqHMjhzau/report-on-frontier-model-training |
# Invulnerable Incomplete Preferences:
A Formal Statement
*Produced as part of the* [*SERI ML Alignment Theory Scholars Program*](https://serimats.org/) *\- Summer 2023 Cohort. My thanks to Eric Chen, Elliott Thornley, and John Wentworth for invaluable discussion and comments on earlier drafts. All errors are mine.*
... | https://www.lesswrong.com/posts/sHGxvJrBag7nhTQvb/invulnerable-incomplete-preferences-a-formal-statement-1 |
# Responses to apparent rationalist confusions about game / decision theory
I’ve encountered various claims about how AIs would approach game theory and decision theory that seem pretty importantly mistaken. Some of these confusions probably aren’t that big a deal on their own, and I’m definitely not the first to poin... | https://www.lesswrong.com/posts/895Qmhyud2PjDhte6/responses-to-apparent-rationalist-confusions-about-game |
# Long-Term Future Fund Ask Us Anything (September 2023)
The Long-Term Future Fund has an AMA up on the Effective Altruism Forum.
There’s no real deadline for questions, but let’s say we have a soft commitment to focus on questions asked on or before September 8th.
I'd prefer centralizing the questions to one place.... | https://www.lesswrong.com/posts/6eDfrgK3LDKJio7ym/long-term-future-fund-ask-us-anything-september-2023 |
# AI #27: Portents of Gemini
By all reports, and as one would expect, Google’s Gemini looks to be substantially superior to GPT-4. We now have more details on that, and also word that Google plans to deploy it in December, [Manifold gives it 82% to happen this year](https://manifold.markets/RH/will-googles-gemini-llm-... | https://www.lesswrong.com/posts/WLvboc66rBCNHwtRi/ai-27-portents-of-gemini |
# Cleaning a SoundCraft Mixer
In [Kingfisher](https://www.kingfisherband.com/) [Cecilia](https://www.ceciliavacanti.com/) uses a [condenser](https://www.audio-technica.com/en-us/pro35) on her fiddle but combines it with effects pedals. This is [a bit awkward to get working](https://www.jefftk.com/p/fiddle-effects-tech... | https://www.lesswrong.com/posts/hyuD9e7kAoHrWSyj7/cleaning-a-soundcraft-mixer |
# The Tree of Life, and a Note on Job
This is another one cross-posted from [my New Savanna blog](https://new-savanna.blogspot.com/2011/06/tree-of-life-and-note-on-job.html). It's old, 11 years (and most of it is even older than that). It's an interpretation of the Book of Job, which is not at all the sort of thing I... | https://www.lesswrong.com/posts/wKHbJPbzLtgCfXcQe/the-tree-of-life-and-a-note-on-job |
# Which Questions Are Anthropic Questions?
I will try to keep this shot, just want to use some simple problems to point out what I think is a commonly overlooked point in anthropic discussions.
1\. The Room Assignment Problem
-------------------------------
> You are among 100 people waiting in a hallway. The hallwa... | https://www.lesswrong.com/posts/SjEFqNtYfhJMP8LzN/which-questions-are-anthropic-questions |
# [Linkpost] Michael Nielsen remarks on 'Oppenheimer'
This is a linkpost to a recent blogpost from [Michael Nielsen](https://michaelnielsen.org/), who has previously written on [EA](https://michaelnotebook.com/eanotes/) among many other topics. This blogpost is adapted from a talk Nielsen gave to an audience working o... | https://www.lesswrong.com/posts/ppQRJEfLBCLFzK73w/linkpost-michael-nielsen-remarks-on-oppenheimer |
# Meta Questions about Metaphilosophy
To quickly recap my main intellectual journey so far (omitting a lengthy [side trip](https://www.lesswrong.com/posts/m8FjhuELdg7iv6boW/work-on-security-instead-of-friendliness) into cryptography and Cypherpunk land), with the approximate age that I became interested in each topic ... | https://www.lesswrong.com/posts/fJqP9WcnHXBRBeiBg/meta-questions-about-metaphilosophy |
# A Golden Age of Building? Excerpts and lessons from Empire State, Pentagon, Skunk Works and SpaceX
Patrick Collison has a fantastic list of examples of [people quickly accomplishing ambitious things together](https://patrickcollison.com/fast) since the 19th Century. It does make you yearn for a time that feels... di... | https://www.lesswrong.com/posts/BpTDJj6TrqGYTjFcZ/a-golden-age-of-building-excerpts-and-lessons-from-empire |
# Progress links digest, 2023-09-01: How ancient people manipulated water, and more
I am experimenting with pulling more social media content directly into these digests, in part to rely less on social media sites long-term (since content might be deleted, blocked, paywalled, etc.) That makes these digests longer, but... | https://www.lesswrong.com/posts/WhMk3gwyfGq2xek2J/progress-links-digest-2023-09-01-how-ancient-people |
# AGI isn't just a technology
Each time a skeptic talks about AI as a technology, I think it signals a likely crux.
I've been watching the public AI debate closely and sadly. I think that debate might be crucial in whether we actually get aligned AGI, and it's not going particularly well so far. The debate is confuse... | https://www.lesswrong.com/posts/nZYhs48pWsaCCgGfi/agi-isn-t-just-a-technology |
# Why aren't more people in AIS familiar with PDP?
Why haven't more people who work on alignment read Parallel Distributed Processing or even seem at all familiar with Rumelhart's work? This is the fundamental model of cognition and behaviour that all of modern AI is built on, the work that Hinton used for most of his... | https://www.lesswrong.com/posts/uktxr4B5ZAHGnKCyN/why-aren-t-more-people-in-ais-familiar-with-pdp |
# Reproducing ARC Evals' recent report on language model agents
I reproduced results from [ARC Evals](https://evals.alignment.org/)' recent report, [*Evaluating Language-Model Agents on Realistic Autonomous Tasks*](https://evals.alignment.org/Evaluating_LMAs_Realistic_Tasks.pdf). For the report, ARC Evals built a set... | https://www.lesswrong.com/posts/WhSK9y8apy8mNMFGK/reproducing-arc-evals-recent-report-on-language-model-agents |
# Tensor Trust: An online game to uncover prompt injection vulnerabilities
**TL;DR: **Play [this online game](https://tensortrust.ai/) to help CHAI researchers create a dataset of prompt injection vulnerabilities.
RLHF and instruction tuning have succeeded at making LLMs practically useful, but in some ways they are ... | https://www.lesswrong.com/posts/qrFf2QEhSiL9F3yLY/tensor-trust-an-online-game-to-uncover-prompt-injection |
# One Minute Every Moment
About how much information are we keeping in working memory at a given moment?
"Miller's Law" dictates that the number of things humans can hold in working memory is "[the magical number 7](https://en.wikipedia.org/wiki/The_Magical_Number_Seven,_Plus_or_Minus_Two)$\pm$[2](https://en.wikipedi... | https://www.lesswrong.com/posts/D97xnoRr6BHzo5HvQ/one-minute-every-moment |
# What did you learn from leaked documents?
As rationalists we seek to understand the world, but it's made harder by political bias and different agendas.
Leaked documents, represent a kind of ground truth, showing how the world really works. Telling us what's for sale, what the real agendas are, how powerful spies ... | https://www.lesswrong.com/posts/NgDdyy3NxykbntQyJ/what-did-you-learn-from-leaked-documents |
# Plum Cooking Temperature
After having fun with cooking/drying plums [last year](https://www.jefftk.com/p/making-prunes), I decided to have another go this year. This time I compared two different ways of cooking whole plums: 140F for ~1.5d vs 350F for ~1.5hr.
I was curious whether the higher temperature was better ... | https://www.lesswrong.com/posts/3etri4vCgaPcr84uE/plum-cooking-temperature |
# [Linkpost] Large language models converge toward human-like concept organization
*This is a linkpost for* [*https://arxiv.org/abs/2308.15047*](https://arxiv.org/abs/2308.15047)*. *
> Large language models show human-like performance in knowledge extraction, reasoning and dialogue, but it remains controversial wheth... | https://www.lesswrong.com/posts/edBz8knyMcgT2Siy9/linkpost-large-language-models-converge-toward-human-like |
# Rational Agents Cooperate in the Prisoner's Dilemma
*This is a crosspost from* [*my blog*](https://outsidetheasylum.blog/rational-agents-cooperate-in-the-prisoners-dilemma/)*. Its intention is to act as a primer on non-causal decision theories, and how they apply to the prisoner's dilemma in particular, for those wh... | https://www.lesswrong.com/posts/QX98rCSXPkPSMriYi/rational-agents-cooperate-in-the-prisoner-s-dilemma |
# Is Metaethics Unnecessary Given Intent-Aligned AI?
While you can read this post as a standalone, I'm writing it partially in response to Wei Dai's recent post [Meta Questions about Metaphilosophy](https://www.lesswrong.com/posts/fJqP9WcnHXBRBeiBg/meta-questions-about-metaphilosophy). I had already been considering w... | https://www.lesswrong.com/posts/pmraJqhjD2Ccbs6Jj/is-metaethics-unnecessary-given-intent-aligned-ai |
# The smallest possible button (or: moth traps!)
**tl;dr:** The more knowledge you have, the smaller the button you need to press to achieve desired results. This is what makes moth traps formidable killing machines, and it's a good analogy for *other* formidable killing machines I could mention.
Traps
-----
I was... | https://www.lesswrong.com/posts/JteNtoLBFZB9niiiu/the-smallest-possible-button-or-moth-traps |
# PIBBSS Summer Symposium 2023
Tl;dr
=====
We are pleased to invite you to the **PIBBSS Summer Symposium** where the fellows from the ‘23 fellowship program present their work.
The symposium is taking place online, over several days in the week of **September 18th**.
* [Check out the **full program**,](https:/... | https://www.lesswrong.com/posts/86KqAYdjrW7niXydq/pibbss-summer-symposium-2023 |
# The goal of physics
In grad school, I was a teaching assistant for a course called, *Why the Sky is Blue*. It was a qualitative introduction to physics for non-majors, covering a lot of the same topics as Physics I, such as forces, conservation of energy and momentum, electric charges and magnetic fields, in less de... | https://www.lesswrong.com/posts/N6gGz3KjFcGjiZZmK/the-goal-of-physics |
# Series of absurd upgrades in nature's great search
So you want to find that special thing that replicates best and lasts longest? Just vibrate a bunch of molecules a long time!
You might reasonably assume that molecules wouldn't practically ever randomly assemble themselves into anything worth looking at. It happe... | https://www.lesswrong.com/posts/o2BkyQLrscbLbQSJn/series-of-absurd-upgrades-in-nature-s-great-search |
# An embedding decoder model, trained with a different objective on a different dataset, can decode another model's embeddings surprisingly accurately
(via [twitter](https://twitter.com/thesephist/status... | https://www.lesswrong.com/posts/RYX7NFwoejwJy2h9F/an-embedding-decoder-model-trained-with-a-different |
# Fundamental question: What determines a mind's effects?
*[Metadata: crossposted from [https://tsvibt.blogspot.com/2023/04/fundamental-question-what-determines.html](https://tsvibt.blogspot.com/2023/04/fundamental-question-what-determines.html). First completed April 9, 2023.]*
A mind has some effects on the world.... | https://www.lesswrong.com/posts/NqsNYsyoA2YSbb3py/fundamental-question-what-determines-a-mind-s-effects |
# What must be the case that ChatGPT would have memorized “To be or not to be”? – Three kinds of conceptual objects for LLMs
*This is cross-posted from* [*my personal blog, New Savanna*](https://new-savanna.blogspot.com/2023/09/what-must-be-case-that-chatgpt-would.html)*.*
Is anyone working on this kind of issue? If... | https://www.lesswrong.com/posts/5SmqubjL8xK4ycenh/what-must-be-the-case-that-chatgpt-would-have-memorized-to |
# Units of Exchange
Units of Exchange is the first class in the CFAR Handbook.
In this meetup, we'll look at the tradeoffs we make in our lives, and then examine how we can make better tradeoffs.
Suggested reading (optional): [Units of Exchange](https://www.lesswrong.com/s/KAv8z6oJCTxjR8vdR/p/cvDtmPNCyrkpg4d4F)
Th... | https://www.lesswrong.com/events/FimvpFtZbmHH8PeWW/units-of-exchange-1 |
# Open Thread – Autumn 2023
If it’s worth saying, but not worth its own post, here's a place to put it.
If you are new to LessWrong, here's the place to introduce yourself. Personal stories, anecdotes, or just general comments on how you found us and what you hope to get from the site and community are invited. This ... | https://www.lesswrong.com/posts/AyS4awCnKrKhwk4vs/open-thread-autumn-2023 |
# Impending AGI doesn’t make everything else unimportant
*I sat in a restaurant in New York, for example, and I looked out at the buildings and I began to think, about how much the radius of the Hiroshima bomb damage was. How far from here was 34th street?... All those buildings, all smashed. And I would go along a... | https://www.lesswrong.com/posts/mAwoixsLmDKZFZ3E4/impending-agi-doesn-t-make-everything-else-unimportant |
# Paper: On measuring situational awareness in LLMs
This post is a copy of the introduction of this [paper](https://arxiv.org/abs/2309.00667) on situational awareness in LLMs.
Authors: Lukas Berglund, Asa Cooper Stickland, Mikita Balesni, Max Kaufmann, Meg Tong, Tomasz Korbak, Daniel Kokotajlo, Owain Evans.
... | https://www.lesswrong.com/posts/mLfPHv4QjmeQrsSva/paper-on-measuring-situational-awareness-in-llms |
# The omnizoid - Heighn FDT Debate #1: Why FDT Isn't Crazy
*After* [*omnizoid*](https://www.lesswrong.com/users/omnizoid?from=post_header) *asked whether people want to* [*debate him*](https://www.lesswrong.com/posts/Wvyiycf6dHeknJRRA/anyone-want-to-debate-publicly-about-fdt) *on Functional Decision Theory (FDT), he a... | https://www.lesswrong.com/posts/8tYeE6MXkb4tTyzCF/the-omnizoid-heighn-fdt-debate-1-why-fdt-isn-t-crazy |
# Defunding My Mistake
*Confessions of an ex-ACAB*
[](https://substackcdn.com/image/fetch/f_auto,q_au... | https://www.lesswrong.com/posts/4rsRuNaE4uJrnYeTQ/defunding-my-mistake |
# Hard Questions Are Language Bugs
*Thanks to Jessica Taylor for the discussion that was the stem of this article. The opinions here are all mine and I’m sure she’d disagree with some or all of them.*
The Hai’Zu
----------
Imagine the Hai’zu people of Stone Age Langong, using a very simple language that might contai... | https://www.lesswrong.com/posts/sc6Gv4Gbc7FNGJH5N/hard-questions-are-language-bugs |
# How ForumMagnum builds communities of inquiry
The website you're currently using is powered by ForumMagnum. But what really *is* ForumMagnum? What is it *for*, and why was it designed *this* way? In this post, I cast ForumMagnum as a medium for building *communities of inquiry.* I show how ForumMagnum is designed to... | https://www.lesswrong.com/posts/dYwQCFkR6cCbP9Xqk/how-forummagnum-builds-communities-of-inquiry |
# a rant on politician-engineer coalitional conflict
Sometimes, a group in some organization has a highly technical and highly effective leader. Kelly Johnson (Skunk Works) and Hyman Rickover (US Navy nuclear propulsion) are famous examples. A naive economist might expect such people to be well-liked by management abo... | https://www.lesswrong.com/posts/hbN5yhruy3htL7neJ/a-rant-on-politician-engineer-coalitional-conflict |
# Hertford, Sourbut (rationality lessons from University Challenge)
Amongst the huge range of excitements offered by joining the University of Oxford was the unexpected opportunity to join this lovely bunch
, and it is a great read. I get why the people developing AI fi... | https://www.lesswrong.com/posts/njEWACBHhfppg6KYS/notes-on-nukes-ir-and-ai-from-arsenals-of-folly-and-other |
# Action theory is not policy theory is not agent theory
There are several proposed alternative theories of rationality. Some of them ([causal decision theory (CDT)](https://www.lesswrong.com/tag/causal-decision-theory), [evidential decision theory (EDT)](https://www.lesswrong.com/tag/evidential-decision-theory), [fun... | https://www.lesswrong.com/posts/MwetLcBPvshg9ePZB/action-theory-is-not-policy-theory-is-not-agent-theory |
# Text Posts from the Kids Group: 2023 I
We have a [Facebook group for kid stuff](https://www.jefftk.com/p/making-groups-for-kid-pictures), because if we post a mixture of kid things and other stuff FB's algorithm gets very confused about who to show our posts to. While my annual [pictures posts](https://www.jefftk.co... | https://www.lesswrong.com/posts/FqvhuLkdR8FeyCf5b/text-posts-from-the-kids-group-2023-i |
# Who Has the Best Food?
It is a fun question going around the internet this past week, so here we go.
In particular, people focused on the question of France vs. America. As one would expect, those on the French side think those on the American side are crazy, it is insulting to even consider this a question. Those ... | https://www.lesswrong.com/posts/D5urD7WmDePyii73D/who-has-the-best-food |
# AISN #21:
Google DeepMind’s GPT-4 Competitor, Military Investments in Autonomous Drones, The UK AI Safety Summit, and Case Studies in AI Policy
Welcome to the AI Safety Newsletter by the [Center for AI Safety](https://www.safe.ai/). We discuss developments in AI and AI safety. No technical background required.
Subs... | https://www.lesswrong.com/posts/jkEEHfQkwvbLkzpiF/aisn-21-google-deepmind-s-gpt-4-competitor-military |
# Strongest real-world examples supporting AI risk claims?
\[Manually cross-posted to EAForum [here](https://forum.effectivealtruism.org/posts/GDdvdhbGfCehnoJzY/strongest-real-world-examples-supporting-ai-risk-claims)\]
There are some great collections of examples of things like [specification gaming](https://docs.go... | https://www.lesswrong.com/posts/h66WSLKdShqzNsZL9/strongest-real-world-examples-supporting-ai-risk-claims |
# Benchmarks for Detecting Measurement Tampering [Redwood Research]
**TL;DR**: This post discusses our recent empirical work on detecting measurement tampering and explains how we see this work fitting into the overall space of alignment research.
When training powerful AI systems to perform complex tasks, it may be ... | https://www.lesswrong.com/posts/inALbAqdx63KTaGgs/benchmarks-for-detecting-measurement-tampering-redwood |
# What I would do if I wasn’t at ARC Evals
**In which: **I list 9 projects that I would work on if I wasn’t busy working on safety standards at ARC Evals, and explain why they might be good to work on.
**Epistemic status:**I’m prioritizing getting this out fast as opposed to writing it carefully. I’ve thought for at... | https://www.lesswrong.com/posts/6FkWnktH3mjMAxdRT/what-i-would-do-if-i-wasn-t-at-arc-evals |
# Conversation about paradigms, intellectual progress, social consensus, and AI
I have a number of thoughts here that I haven't gotten around to writing up, and sharing via a public conversation seems like a decent way to try.
A few opening thoughts:
* People talk about the field of AI Alignment being "pre-par... | https://www.lesswrong.com/posts/ipaWHTiEwEqtuvzrP/conversation-about-paradigms-intellectual-progress-social |
# What I've been reading, September 2023
A quasi-monthly feature. Recent blog posts and news stories are generally omitted; you can find them in my [links digests](https://rootsofprogress.org/writing#links-digest). I’ve been busy helping to choose the first cohort of our [blogging fellowship](https://fellowship.rootso... | https://www.lesswrong.com/posts/aAbtJiNnBxmGZAXqD/what-i-ve-been-reading-september-2023 |
# Find Hot French Food Near Me: A Follow-up
On [Zvi's recent post](https://www.lesswrong.com/posts/D5urD7WmDePyii73D/who-has-the-best-food) about French food I posted an [inflammatory comment](https://www.lesswrong.com/posts/D5urD7WmDePyii73D/?commentId=Lzja5wcSSjSh4gSoG) (saying in essence that French food is so bad ... | https://www.lesswrong.com/posts/Gv2yv4idk6Hcas8vZ/find-hot-french-food-near-me-a-follow-up |
# Monthly Roundup #10: September 2023
It’s that time again. Here’s all the news and items of interest that isn’t something else.
Things that didn’t make it in to try and stand them on their own include [whether America or France has better food](https://thezvi.substack.com/p/who-has-the-best-food) and various claims ... | https://www.lesswrong.com/posts/ZoxkWDgDueo5yoakv/monthly-roundup-10-september-2023 |
# ActAdd: Steering Language Models without Optimization
We wrote up the [GPT-2 steering vector work](https://www.lesswrong.com/posts/5spBue2z2tw4JuDCx/steering-gpt-2-xl-by-adding-an-activation-vector#Content_warning__Some_completions_contain_unpleasant_content__including_gendered_slurs_) as a full paper, adding a few ... | https://www.lesswrong.com/posts/HWxLQvzJGeXoLPJWd/actadd-steering-language-models-without-optimization |
# My First Post
Hello all!
I'm Jaivardhan Nawani, a 12-year-old enthusiast on rationality and how using it can improve our day-to-day lives.
Recently, I discovered this forum from a few friends who introduced me to many useful principles, and I thought to come up with an idea to make rationality easier to implement ... | https://www.lesswrong.com/posts/mFKLJgWgwBhLD9XW7/my-first-post-1 |
# Video essay: How Will We Know When AI is Conscious?
I recently came across this engaging video essay by [exurb1a](https://www.youtube.com/@Exurb1a) and felt it would be of interest to this community. Contrary to its title, it avoids the common pitfall of asserting that only conscious AI can be dangerous, and it eff... | https://www.lesswrong.com/posts/uJtXSafu9wbCrDNja/video-essay-how-will-we-know-when-ai-is-conscious |
# Weekly Incidence vs Cumulative Infections
Imagine you have a goal of identifying a novel disease by the time some small fraction of the population has been infected. Many of the signs you might use to detect something unusual, however, such as doctor visits or shedding into wastewater, will depend on the number of p... | https://www.lesswrong.com/posts/6rnWQW8HtHoavPDiu/weekly-incidence-vs-cumulative-infections |
# Sharing Information About Nonlinear
*Added (11th Sept): Nonlinear have commented that* [*they intend to write a response*](https://forum.effectivealtruism.org/posts/32LMQsjEMm6NK2GTH/sharing-information-about-nonlinear?commentId=iujw3kLaYrXSP3EgJ)*, have written a* [*short follow-up*](https://www.lesswrong.com/po... | https://www.lesswrong.com/posts/Lc8r4tZ2L5txxokZ8/sharing-information-about-nonlinear-1 |
# Recreating the caring drive
**TL;DR**: This post is about value of recreating “caring drive” similar to some animals and why it might be useful for AI Alignment field in general. Finding and understanding the right combination of training data/loss function/architecture/etc that allows gradient descent to robustly f... | https://www.lesswrong.com/posts/JjqZexMgvarBFMKPs/recreating-the-caring-drive |
# AI#28: Watching and Waiting
We are, [as Tyler Cowen has noted](https://marginalrevolution.com/marginalrevolution/2023/09/america-is-asleep-on-its-ai-boom.html), in a bit of a lull. Those of us ahead of the curve have gotten used to GPT-4 and Claude-2 and MidJourney. Functionality and integration are expanding, but o... | https://www.lesswrong.com/posts/GuTK47y9awvypvAbC/ai-28-watching-and-waiting |
# How did you make your way back from meta?
I've noticed in myself a strong preference for focusing on the meta level. It's most visible in fields dear to me, like writing. For example, I'll spend much more time reading up on rhetoric or tracking down rare 1980's books about writing techniques than practicing writing ... | https://www.lesswrong.com/posts/B99sRmhWpMcZLpptH/how-did-you-make-your-way-back-from-meta |
# [Linkpost] Frontier AI Taskforce: first progress report
Other links
===========
* [Short interview with Ian Hogarth ](https://twitter.com/SciTechgovuk/status/1699686882022486145)
* [Twitter thread](https://twitter.com/soundboy/status/1699688880482500684)
Some quotes from the report
===========================
... | https://www.lesswrong.com/posts/FhKhhmK4DXrogJxRr/linkpost-frontier-ai-taskforce-first-progress-report |
# A quick update from Nonlinear
**One example of the evidence we’re gathering**
We are working hard on a point-by-point response to [Ben’s article](https://forum.effectivealtruism.org/posts/32LMQsjEMm6NK2GTH/sharing-information-about-nonlinear), but wanted to provide a quick example of the sort of evidence we are pre... | https://www.lesswrong.com/posts/FHxYPpMkAX9ayoBuK/a-quick-update-from-nonlinear |
# Bring back the Colosseums
Men want to engage in righteous combat. They want it more than they want sex or VP titles. They fantasize about getting the casus belli to defend themselves against armed thugs that will never come, they spend billions of dollars on movies and TV about everymen in implausible circumstances ... | https://www.lesswrong.com/posts/8ykeD42WGWJ8XnQbd/bring-back-the-colosseums |
# Explained Simply: Quantilizers
Standard AI are *optimizers*: they ‘look’ through possible actions they could take, and pick the one that maximises what they care about. This can be dangerous— an AI which maximises in this way needs to care about *exactly* the same things that humans care about, which is [really hard... | https://www.lesswrong.com/posts/u2RFkz8HNmmbEFrBr/explained-simply-quantilizers |
# Have Attention Spans Been Declining?
__I investigate whether the attention span of individual humans has
been falling over the last two decades (prompted by curiosity about
whether the introduction of the internet may be harmful to cognitive
performance). I find little direct work on the topic, despite its wide
appe... | https://www.lesswrong.com/posts/Pweg9xpKknkNwN8Fx/have-attention-spans-been-declining |
# Explaining grokking through circuit efficiency
This is a linkpost for our paper [*Explaining grokking through circuit efficiency*](https://arxiv.org/abs/2309.02390), which provides a general theory explaining when and why grokking (aka delayed generalisation) occurs, and makes several interesting and novel predictio... | https://www.lesswrong.com/posts/JK2QGfNGLjuFnrEvz/explaining-grokking-through-circuit-efficiency |
# Invading Australia (Endless Formerlies Most Beautiful, or What I Learned On My Holiday)
G'day! I recently got back from a really lovely extended break down under. There was a lot ('heaps') to see and do! - including visiting family and friends with my partner ~~in crime~~, sampling some unspecified quantity of homeb... | https://www.lesswrong.com/posts/WaAFhFoYuct7a3gDX/invading-australia-endless-formerlies-most-beautiful-or-what |
# AI Probability Trees - Joe Carlsmith (2022)
*I am reviewing the work AI experts on what they think will happen with AI. This is a summary of Joe Carlsmith’s thoughts from* [*his paper*](https://arxiv.org/pdf/2206.13353.pdf )*. AI risk scares me but often I feel pretty disconnected from it. This has helped me think a... | https://www.lesswrong.com/posts/y2Qrk2RRp5CgkbMAA/ai-probability-trees-joe-carlsmith-2022 |
# Debate series: should we push for a pause on the development of AI?
In March of this year, 30,000 people, including leading AI figures like Yoshua Bengio and Stuart Russell, signed [a letter](http... | https://www.lesswrong.com/posts/h2pWy87DgpBWuuq4A/debate-series-should-we-push-for-a-pause-on-the-development |
# Sum-threshold attacks
How do you affect something far away, a lot, without anyone noticing?
(Note: you can safely skip sections. It is also safe to skip the essay entirely, or to read the whole thing backwards if you like.)
# The frog's lawsuit
Attorney for the defendant: "So, Mr. Frog. You allege that my c... | https://www.lesswrong.com/posts/R3eDrDoX8LisKgGZe/sum-threshold-attacks |
# Progress links digest, 2023-09-08: The Conservative Futurist, cargo airships, and more
**Announcements**
-----------------
* [*The Conservative Futurist: How to Create the Sci-Fi World We Were Promised*](https://www.amazon.com/gp/product/1546005544?tag=jasocraw-20) is available for pre-order (via [@JimPethokoukis... | https://www.lesswrong.com/posts/RBuwybs7gqhtqTYHR/progress-links-digest-2023-09-08-the-conservative-futurist |
# What is the optimal frontier for due diligence?
The title isn't quite right. My current take is that:
* There is a ~10k word [post](https://www.lesswrong.com/posts/Lc8r4tZ2L5txxokZ8/sharing-information-about-nonlinear-1) sharing some damaging information about someone else.
* 3-4k words are describing accusati... | https://www.lesswrong.com/posts/vJagsfKmQKeaW5oG3/what-is-the-optimal-frontier-for-due-diligence |
# Capture the Flag Mechanistic Interpretability Challenges
This project was developed as part of the Swiss Existential Risk Initiative Research Fellowship under the mentorship of Alexandre Variengien. The use of the first person throughout the document refers to Alejandro Acelas.
* * *
I designed three challenges me... | https://www.lesswrong.com/posts/SbLRsdajhMQdAbaqL/capture-the-flag-mechanistic-interpretability-challenges |
# Panel discussion on AI consciousness with Rob Long and Jeff Sebo
Intro
=====
[Recent 80k guest](https://80000hours.org/podcast/episodes/robert-long-artificial-sentience/) and philosopher specializing in AI consciousness Rob Long ([@rgb](https://forum.effectivealtruism.org/users/rgb?mention=user)) recently participa... | https://www.lesswrong.com/posts/w84XTwejBbpSRJvwL/panel-discussion-on-ai-consciousness-with-rob-long-and-jeff |
# How teams went about their research at AI Safety Camp edition 8
AI Safety Camp connects new collaborators worldwide to discuss and decide on a concrete research proposal, gear up online as a team, and try their hand at AI safety research during intensive coworking sprints.
Fourteen teams formed at this year's virtu... | https://www.lesswrong.com/posts/EsBDbtnCizxDLsDT3/how-teams-went-about-their-research-at-ai-safety-camp-1 |
# Probabilistic argument relationships and an invitation to the argument mapping community
I've been thinking about the probabilistic properties of how ideas are connected to each other, and I think that the Lesswrong community might be interested in the same kind of questions that interest me. I'd like to introduce[^... | https://www.lesswrong.com/posts/b4vLuaaKBefA4C6bG/probabilistic-argument-relationships-and-an-invitation-to |
# AI presidents discuss AI alignment agendas
None of the presidents fully represent my (TurnTrout's) views.
TurnTrout wrote the script. Garrett Baker helped produce the video after the audio was complete. Thanks to David Udell, Ulisse Mini, Noemi Chulo, and especially Rio Popper for feedback and assistance in writing... | https://www.lesswrong.com/posts/7M2iHPLaNzPNXHuMv/ai-presidents-discuss-ai-alignment-agendas |
# Feature proposal: Export ACX meetups
Currently there are two places that display upcoming meetups of Astral Codex Ten:
* [Community Section](https://www.lesswrong.com/community) of LessWrong
* articles at ACX, such as "[Meetups Everywhere 2023: Times & Places](https://www.astralcodexten.com/p/meetups-everywhere... | https://www.lesswrong.com/posts/QTpdhEBZeJGCAhtWj/feature-proposal-export-acx-meetups |
# Cruxes on US lead for some domestic AI regulation
*Written quickly. Suggestions welcome.*
A possible risk of some US AI regulation is that US regulation would differentially slow US AI progress and that would be bad. This post explores the factors that determine *how much US regulation would differentially slow US ... | https://www.lesswrong.com/posts/YguseW2zMYe8tMCbW/cruxes-on-us-lead-for-some-domestic-ai-regulation |
# Playing the game vs. finding a cheat code
*This is a linkpost from my blog* [*De Novo*.](https://denovo.substack.com)
Imagine a new Pokémon game has just come out, and you really want to catch a Zapdos. It’s listed in the game’s Pokédex, so you know it must be possible to catch, but you’re not sure how.
You could ... | https://www.lesswrong.com/posts/KThSAwEbtwZko3vp2/playing-the-game-vs-finding-a-cheat-code |
# What are some good language models to experiment with?
Like if I want to experiment with a steering technique, it would be useful to have a language model that is small, capable, but not so finetuned that it becomes inflexible. (Or maybe ideally, a model which has both a finetuned and a non-finetuned variant.)
I've... | https://www.lesswrong.com/posts/CkJ5D7zpHjKbykKJz/what-are-some-good-language-models-to-experiment-with |
# High school advice
What advice does the Lesswrong community have in general for high school students who stumbled on this community?
Looking for all types of responses, it's fine whether your response is focused on achieving something in the AI/effective altruism/rationality community, or whether it is focused... | https://www.lesswrong.com/posts/HvAcnjjtPpfajLwaN/high-school-advice |
# Logical Share Splitting
### Are mathematicians just not trying hard enough?
The Riemann hypothesis is one of the most important open problems in math. There's a $1 million prize from the Clay mathematics institute for a proof or disproof of the Riemann hypothesis. At the time of writing, it remains unsolved. From ... | https://www.lesswrong.com/posts/uDXRxF9tGqGX5bGT4/logical-share-splitting |
# Focus on the Hardest Part First
Consider reading [this](https://cs.stanford.edu/~jsteinhardt/ResearchasaStochasticDecisionProcess.html) instead.
Here is some obvious advice.
I think a common failure mode when working on AI alignment[^1] is to not focus on the hard parts of the problem first. This is a problem when... | https://www.lesswrong.com/posts/Eav2BizSejDcztFC8/focus-on-the-hardest-part-first |
# A Bat and Ball made me Sad
'A bat and a ball cost $110 in total.
The bat costs $100 more than the ball.
How much does the ball cost?
**The answer is $5.**
**Please enter the number 5 in the blank below.**
$_____'
That question is from one of 59 new studies (*N* = 72,310) which focus primarily on the “bat ... | https://www.lesswrong.com/posts/4d8dbuAKA39QDiyw7/a-bat-and-ball-made-me-sad |
# PSA: The community is in Berkeley/Oakland, not "the Bay Area"
Posting this because I recently had a conversation that went like this:
**Friend:** Hey, you used to live in SF. Is there any rationalist stuff actually happening in San Francisco? There don't seem to be many events, or even that many aspiring rationalis... | https://www.lesswrong.com/posts/esaaewdyr72Ae2tvP/psa-the-community-is-in-berkeley-oakland-not-the-bay-area |
# Erdős Problems in Algorithmic Probability
**Motivation:**
Before the Pandemic of 2020, I had a number of constructive discussions with several neuroscientists including Konrad Körding, a computational neuroscientist and pioneer exploring [connections between deep learning and human learning](https://www.nature.com/... | https://www.lesswrong.com/posts/cqs3XHkYaamy93cx8/erdos-problems-in-algorithmic-probability-1 |
# A Case for AI Safety via Law
This post is to make the subject case more available and open to comment. A paper with the above title is currently languishing in arXiv limbo but available in [Google Docs](https://drive.google.com/file/d/1CzDGBtjY92qQujLKYysbaIF1agoDoUJE/view?usp=sharing). I was surprised to see my las... | https://www.lesswrong.com/posts/oi2YxRddakosnnbGX/a-case-for-ai-safety-via-law |
# Amazon KDP AI content guidelines
Amazon updated the content guidelines for Kindle Direct Publishing to address AI and in the process attempted to define terms AI-generated and AI-assisted. Guidelines like this are how our society tries to engage with AI in the near-term.
> Artificial intelligence (AI) content (tex... | https://www.lesswrong.com/posts/4csaBfAgiyzrFXvEj/amazon-kdp-ai-content-guidelines |
# Machine Evolution
*Epistemic Status: This post explores the relationship between biological evolution and the evolution of our machines. We examine parallels between the evolutionary pressures shaping organisms and machines. We are still exploring this analogy and our findings are tentative. We hope to explore these... | https://www.lesswrong.com/posts/e6gg8S8AfJmwE8HLh/machine-evolution |
# How useful is Corrigibility?
*Epistemic Status: My best guess (I'm new to AI safety)*
We don't know how to formally define corrigibility and this is part of the reason why we haven't solved it so far. Corrigibility is defined in [Arbital](https://arbital.com/p/corrigibility/) as:
> \[the agent\] doesn't interfere ... | https://www.lesswrong.com/posts/Py3vqPp9uSqQJHFuy/how-useful-is-corrigibility |
# Apple Cider Baklava
Since coming across [apple cider syrup](https://www.jefftk.com/p/apple-cider-syrup) last year I've been trying to figure out what sort of things to make with it. Cider donuts are traditional, but I'm not that much of a fan. Mixing it into whipped cream works very well, and makes a good sweet and ... | https://www.lesswrong.com/posts/wu3r9AfKMDYzLzPPc/apple-cider-baklava |
# Startup Roundup #1: Happy Demo Day
There were a bunch of discussions recently related to issues surrounding Y-Combinator, related as usual to their annual demo day. It seemed worth splitting them off into a post.
#### Bidders at Auction Mostly Think Prices Are Too High
YC is in session, so all the usual talk is ba... | https://www.lesswrong.com/posts/dGvcx3giYngcSEaMZ/startup-roundup-1-happy-demo-day |
# Automatically finding feature vectors in the OV circuits of Transformers without using probing
Let's say that you're working to understand how a given large language model computes its output on a certain task. You might be curious about the following questions:
* What are some linear feature vectors that are res... | https://www.lesswrong.com/posts/jDfjqu2qJLcPco9cf/automatically-finding-feature-vectors-in-the-ov-circuits-of |
# [Linkpost] Contra four-wheeled suitcases, sort of
Are simple, rugged, reliable systems better or are fancy, sophisticated, complex ones better?
The article goes in with a preconception, analyzes suitcases, cars, hiking gear, food, cups, space pens, people, startups, soccer, headphones, weight, rockets, rockets, shi... | https://www.lesswrong.com/posts/nFGMvs9WSeau8ATTY/linkpost-contra-four-wheeled-suitcases-sort-of |
# Padding the Corner
There's a shelf in my dad's kitchen, about four feet off the ground. I remember when I stopped being short enough that I could walk under it: ouch! I didn't get a concussion or anything, but it was pretty unpleasant. My sisters and cousins also remember bonking their heads on it. I noticed my olde... | https://www.lesswrong.com/posts/dNBeQdB35qKruoCpr/padding-the-corner |
# The Flow-Through Fallacy
There is something of an informal reasoning fallacy where you promote something that seems like something that we should "surely do", something that seems obviously important, but where we haven't thought through all of the steps involved and so we aren't actually justified in assuming that ... | https://www.lesswrong.com/posts/G9GmrYhQm6bKt2XJJ/the-flow-through-fallacy |
# Duty to rescue / Non-assistance à personne en danger
If you are waiting in a bank lobby, and an 83-year-old man collapses, do you have a moral duty to try to rescue him, call an ambulance, or provide any kind of assistance? Or, more importantly, do you have any legal duty to do so, i.e. can you be held liable for fa... | https://www.lesswrong.com/posts/2srJWhjCy2LYQCvbR/duty-to-rescue-non-assistance-a-personne-en-danger |
# UDT shows that decision theory is more puzzling than ever
I feel like MIRI perhaps mispositioned FDT (their variant of UDT) as a clear advancement in decision theory, whereas maybe they could have attracted more attention/interest from academic philosophy if the framing was instead that the UDT line of thinking show... | https://www.lesswrong.com/posts/wXbSAKu2AcohaK2Gt/udt-shows-that-decision-theory-is-more-puzzling-than-ever |
# Is AI Safety dropping the ball on privacy?
**TL;DR**
=========
The lack of privacy-preserving technologies facilitates better predictive models of human behavior. This accelerates several existential ... | https://www.lesswrong.com/posts/WCevxhGtmnPhWH3ah/is-ai-safety-dropping-the-ball-on-privacy-1 |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.