text stringlengths 300 320k | source stringlengths 52 154 |
|---|---|
# AGI Safety and Alignment at Google DeepMind:
A Summary of Recent Work
We wanted to share a recap of our recent outputs with the AF community. Below, we fill in some details about what we have been working on, what motivated us to do it, and how we thought about its importance. We hope that this will help people buil... | https://www.lesswrong.com/posts/79BPxvSsjzBkiSyTq/agi-safety-and-alignment-at-google-deepmind-a-summary-of |
# Would you benefit from, or object to, a page with LW users' reacts?
There is currently an admin-only page that shows a list of all comments that have been reacted to (in chronological order). Periodically I think "it might just be nice to show this to everyone, and to let them filter by individual reacts, or individ... | https://www.lesswrong.com/posts/tXRwp42jjABrgRXzo/would-you-benefit-from-or-object-to-a-page-with-lw-users |
# Where should I look for information on gut health?
I've been on a gut health kick, reading [Brain Maker](https://www.goodreads.com/book/show/22875118-brain-maker), adding more kale for the insoluble fiber, and cutting seed oils and sugar.
The author recommends probiotic enemas and fecal transplant, but I've seen mi... | https://www.lesswrong.com/posts/Bbq6Gogm5cHuLjZRG/where-should-i-look-for-information-on-gut-health |
# the Giga Press was a mistake
## the giga press
Tesla decided to use large aluminum castings ("gigacastings") for the frame of many of its vehicles, including the Model Y and Cybertruck. This approach and the "[Giga Press](https://en.wikipedia.org/wiki/Giga_Press)" used for it have been praised by many articles and ... | https://www.lesswrong.com/posts/nZcb9TTyneFbxiDav/the-giga-press-was-a-mistake |
# Apply to Aether - Independent LLM Agent Safety Research Group
**The basic idea**
==================
**Aether** will be a small group of talented early-career AI safety researchers with a shared research vision who work full-time with mentorship on their best effort at making AI go well. That research vision will br... | https://www.lesswrong.com/posts/WsDsGRTKBCrujwbKd/apply-to-aether-independent-llm-agent-safety-research-group |
# Please do not use AI to write for you
I've recently seen several articles here that were clearly generated or heavily assisted by AI. They are all dreadful. They are verbose, they are full of "on the one hand" and "on the other", they never make any assertion without also making room for the opposite, and end with "... | https://www.lesswrong.com/posts/eZa37pZtxsQirE84d/please-do-not-use-ai-to-write-for-you |
# AI Safety Newsletter #40: California AI Legislation
Plus, NVIDIA Delays Chip Production, and Do AI Safety Benchmarks Actually Measure Safety?
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.
Listen... | https://www.lesswrong.com/posts/q3QZaNHwCxxkheoTM/ai-safety-newsletter-40-california-ai-legislation-plus |
# Just because an LLM said it doesn't mean it's true: an illustrative example
This was originally posted in the comments of [You don't know how bad most things are nor precisely how they're bad.](https://www.lesswrong.com/posts/PJu2HhKsyTEJMxS9a/you-don-t-know-how-bad-most-things-are-nor-precisely-how); I've broken it... | https://www.lesswrong.com/posts/2gpsa9yZka95pCsTn/just-because-an-llm-said-it-doesn-t-mean-it-s-true-an |
# Deception and Jailbreak Sequence: 1. Iterative Refinement Stages of Deception in LLMs
Executive Overview
==================
As models grow increasingly sophisticated, they will surpass human expertise. It is a fundamentally difficult challenge to make sure that those models are robustly aligned ([Bowman et al., 202... | https://www.lesswrong.com/posts/imXMvaNZv5HkBphFa/deception-and-jailbreak-sequence-1-iterative-refinement |
# Measuring Structure Development in Algorithmic Transformers
**tl;dr**: *We compute the evolution of the local learning coefficient (LLC), a proxy for model complexity, for an algorithmic transformer. The LLC decreases as the model learns more structured solutions, such as head specialization. *
*This post is struct... | https://www.lesswrong.com/posts/ooAao2RYdFSd77qfp/measuring-structure-development-in-algorithmic-transformers |
# AI #78: Some Welcome Calm
SB 1047 has been amended once more, with both strict improvements and big compromises. I cover the changes, and answer objections to the bill, [in my extensive Guide to SB 1047](https://thezvi.substack.com/p/guide-to-sb-1047). I follow that up here with reactions to the changes and some tho... | https://www.lesswrong.com/posts/qGh9suEsb82hzBtSN/ai-78-some-welcome-calm |
# The economics of space tethers
*Some code for this post can be found [here](https://colab.research.google.com/drive/1QFJS0sjNoD4ybEFplM69EyrYWeIxH0Gl?usp=sharing).*
[Space tethers](https://en.wikipedia.org/wiki/Space_tether) take the old, defunct space elevator concept and shorten it. Rockets can fly up to a dang... | https://www.lesswrong.com/posts/DvGe4debuZvDKBsEm/the-economics-of-space-tethers |
# A primer on the current state of longevity research
*Note: This post is co-authored with* [*Stacy ... | https://www.lesswrong.com/posts/dWAmkSqLE4YMQRuhj/a-primer-on-the-current-state-of-longevity-research |
# A Robust Natural Latent Over A Mixed Distribution Is Natural Over The Distributions Which Were Mixed
This post walks through the math for a theorem. It’s intended to be a reference post, which we’ll link back to as-needed from future posts. The question which first motivated this theorem for us was: “Redness of a ma... | https://www.lesswrong.com/posts/jig8yRHuwhgxN35ue/a-robust-natural-latent-over-a-mixed-distribution-is-natural |
# Turning 22 in the Pre-Apocalypse
*Meta comment for LessWrong readers*[^liyromxacks]
*Something Different This Way Comes - Part 1*
=============================================
*In which I attempt to renegotiate rationalism as a personal philosophy, and offer my alternative—Game theory is not a substitute for real ... | https://www.lesswrong.com/posts/DwxbPixw7SsnpBfaa/turning-22-in-the-pre-apocalypse |
# Interoperable High Level Structures: Early Thoughts on Adjectives
*Meta: This post is a relatively rough dump of some recent research thoughts; it’s not one of our more polished posts, in terms of either clarity or rigor. You’ve been warned.*
The [Interoperable Semantics](https://www.lesswrong.com/posts/RrQftNoRHd5... | https://www.lesswrong.com/posts/wTF8eSqvh8vgfJLWi/interoperable-high-level-structures-early-thoughts-on |
# If we solve alignment, do we die anyway?
*Epistemic status: I'm aware of good arguments that this scenario isn't inevitable, but it still seems frighteningly likely even if we solve technical alignment. Clarifying this scenario seems important.*
TL;DR: (edits in parentheses, two days after posting, from discussions... | https://www.lesswrong.com/posts/kLpFvEBisPagBLTtM/if-we-solve-alignment-do-we-die-anyway-1 |
# Invitation to lead a project at AI Safety Camp (Virtual Edition, 2025)
*Do you have AI Safety research ideas that you would like to work on with others? Is there a project you want to do and you want help finding a team? AI Safety Camp could be the solution for you!*
Summary
=======
AI Safety Camp Virtual is a 3-m... | https://www.lesswrong.com/posts/mhCCwq7KEMARBuaT5/invitation-to-lead-a-project-at-ai-safety-camp-virtual |
# How I started believing religion might actually matter for rationality and moral philosophy
*After the release of* [*Ben Pace's extended interview with me about my views on religion*](https://www.lesswrong.com/posts/n3Q7F3v6wBLsioqt8/extended-interview-with-zhukeepa-on-religion)*, I felt inspired to publish more of ... | https://www.lesswrong.com/posts/X2og6RReKD47vseK8/how-i-started-believing-religion-might-actually-matter-for |
# Interpretability as Compression: Reconsidering SAE Explanations of Neural Activations with MDL-SAEs
*This work was produced as part of the ML Alignment & Theory Scholars Program - Summer 24 Cohort, under mentorship from Lee Sharkey and Jan Kulveit.*
***Note: An updated paper version of this post can be found*** [**... | https://www.lesswrong.com/posts/G2oyFQFTE5eGEas6m/interpretability-as-compression-reconsidering-sae |
# Crafting Polysemantic Transformer Benchmarks with Known Circuits
*Notes: *
* *This research was performed as part of Adrià Garriga-Alonso’s *[*MATS*](https://www.matsprogram.org/) *6.0 stream.*
* *If an opinion is stated in this post saying that "we" hold the opinion, assume it's Evan's opinion (Adrià is taking... | https://www.lesswrong.com/posts/jeoSoJQLuK4JWqtyy/crafting-polysemantic-transformer-benchmarks-with-known |
# Showing SAE Latents Are Not Atomic Using Meta-SAEs
*Bart, Michael and Patrick are joint first authors*. *Research conducted as part of MATS 6.0 in Lee Sharkey and Neel Nanda’s streams. Thanks to Mckenna Fitzgerald and Robert Krzyzanowski for their feedback!*
*TL;DR:*
* *Sparse Autoencoder (SAE) latents have bee... | https://www.lesswrong.com/posts/TMAmHh4DdMr4nCSr5/showing-sae-latents-are-not-atomic-using-meta-saes |
# What's important in "AI for epistemics"?
Summary
=======
This post gives my personal take on “AI for epistemics” and how important it might be to work on.
Some background context:
* AI capabilities are advancing rapidly and I think it’s important to think ahead and prepare for the possible development of AI tha... | https://www.lesswrong.com/posts/D2n5uduYGXuexkv7v/what-s-important-in-ai-for-epistemics-2 |
# "Can AI Scaling Continue Through 2030?", Epoch AI (yes)
> We investigate the scalability of AI training runs. We identify electric power, chip manufacturing, data and latency as constraints. We conclude that 2e29 FLOP training runs will likely be feasible by 2030.
 is an AI Alignment researcher, research associate at the [Center of Human Compatible AI](https://humancompatible.ai/) at UC Berkeley, and now leading a new AI safety research grou... | https://www.lesswrong.com/posts/PPDuLtqCtpqmGzEzH/owain-evans-on-situational-awareness-and-out-of-context |
# Looking for intuitions to extend bargaining notions
I've come up with two ideas - variations on bargaining [ROSE values](https://www.lesswrong.com/s/hCt6GL4SXX6ezkcJn/p/vJ7ggyjuP4u2yHNcP):
* qROSE ("quantum" version) - instead of taking average over agents' permutations, mean should be weighted and encompass all ... | https://www.lesswrong.com/posts/oo5BkydiLFCgvvsxC/looking-for-intuitions-to-extend-bargaining-notions |
# Training a Sparse Autoencoder in < 30 minutes on 16GB of VRAM using an S3 cache
Summary
-------
300 million GPT2-small activations are [cached on s3](https://s3.amazonaws.com/lewington-pitsos-sache), we pull these very quickly onto a [g4dn.8xlarge](https://aws.amazon.com/ec2/instance-types/g4/) ec2 instance in the ... | https://www.lesswrong.com/posts/AtJdPZkMdsakLc6mB/training-a-sparse-autoencoder-in-less-than-30-minutes-on |
# August 2024 Time Tracking
Every so often ( [2022](https://www.jefftk.com/p/january-2022-time-tracking), [2017](https://www.jefftk.com/p/december-2017-time-tracking), more [2017](https://www.jefftk.com/p/june-2017-time-tracking), [2015](https://www.jefftk.com/p/time-division-ii), [2011](https://www.jefftk.com/p/time-... | https://www.lesswrong.com/posts/Waq7SR3mtxdgvjAmC/august-2024-time-tracking |
# Perplexity wins my AI race
Perplexity is the first generalized AI chatbot I’ve found useful enough to integrate into any part of my daily workflow, much less across multiple domains. It speeds me up enough that I’m planning an increase in my freelancing rate.
Perplexity has three key advantages:
1. It provides ci... | https://www.lesswrong.com/posts/CcJ2YWPqcMsWg8vGt/perplexity-wins-my-ai-race-1 |
# Looking to interview AI Safety researchers for a book
If you are or have been employed as an AI Safety Researcher, I'd love to interview you as background for a novel that I'm writing. I'm interested in learning what a day in the life of an AI researcher typically looks like.
I'd be happy to include you in the ack... | https://www.lesswrong.com/posts/8344Fm4mjnoXjxXr5/looking-to-interview-ai-safety-researchers-for-a-book |
# What is it to solve the alignment problem? (Notes)
*(I originally wrote this post as some rough notes on defining the alignment problem, with the intention of turning them into something more polished later. I've now started doing that, as part of a broader series introduced* [*here*](https://www.lesswrong.com/post... | https://www.lesswrong.com/posts/AFdvSBNgN2EkAsZZA/what-is-it-to-solve-the-alignment-problem-1 |
# The Ap Distribution
[E.T. Jaynes](https://en.wikipedia.org/wiki/Edwin_Thompson_Jaynes) introduces this really interesting concept called the $A_p$distribution in his book *Probability Theory*. Despite the book enjoying a cult following, the $A_p$ distribution has failed to become widely-known among aspiring probabil... | https://www.lesswrong.com/posts/c52pGAaKLDJ6ZAviQ/the-ap-distribution |
# AXRP Episode 35 - Peter Hase on LLM Beliefs and Easy-to-Hard Generalization
[YouTube link](https://youtu.be/tEge5uo4E-A)
How do we figure out what large language models believe? In fact, do they even have beliefs? Do those beliefs have locations, and if so, can we edit those locations to change the beliefs? Also, h... | https://www.lesswrong.com/posts/N9fQpF5hkkqGkrN5N/axrp-episode-35-peter-hase-on-llm-beliefs-and-easy-to-hard |
# Please stop using mediocre AI art in your posts
*Epistemic Status: Old Man Blogs at Cloud*
Lately there's been a wave of people on LessWrong (and maybe the whole internet) starting off their essays with some Dall-E3 art.
I don't object (and think it can be cool if done nicely) to get some good ML art to add some v... | https://www.lesswrong.com/posts/zHLbnekuQnYotDhHz/please-stop-using-mediocre-ai-art-in-your-posts |
# Referendum Mechanics in a Marketplace of Ideas
This is a cross post from [https://250bpm.substack.com/p/referendum-mechanics-in-a-marketplace](https://250bpm.substack.com/p/referendum-mechanics-in-a-marketplace).
### In a referendum, people vote either for or against the proposal and that's it. Right? Wrong!
One c... | https://www.lesswrong.com/posts/CixJcx8yrjBQhBiks/referendum-mechanics-in-a-marketplace-of-ideas |
# you should probably eat oatmeal sometimes
Inspired by all the blog posts I've seen promoting unusual diets, nootropic drugs, unusual sleep cycles, and so on, I've decided to make my own post suggesting some radical lifestyle choice.
My suggestion here is, as the title says: you should probably eat oatmeal sometimes... | https://www.lesswrong.com/posts/FEc5SfjCbAyJgRkhR/you-should-probably-eat-oatmeal-sometimes |
# DIY LessWrong Jewelry
\[There are no affiliate links in this post and I haven't been paid by anyone; I just happened to have bought the charms from Shein. You can probably get similar things on Temu, Aliexpress or Amazon\]
I made a cute LessWrong themed necklace/earring set for cheap, here's how.
For the necklace... | https://www.lesswrong.com/posts/JAd8NA8AkMsvTQkGu/diy-lesswrong-jewelry |
# Darwinian Traps and Existential Risks
> *This part 1 in a 3-part sequence summarizes my book,* [*The Darwinian Trap*](https://www.amazon.com/Darwinian-Trap-Evolutionary-Explain-Threaten/dp/0593594053) *(see* [*part 2*](https://www.lesswrong.com/posts/kBG6bNckhZeamQJsQ/the-fragility-of-life-hypothesis-and-the-evolut... | https://www.lesswrong.com/posts/q3YmKemEzyrcphAeP/darwinian-traps-and-existential-risks |
# Secular interpretations of core perennialist claims
*After the release of* [*Ben Pace's extended interview with me about my views on religion*](https://www.lesswrong.com/posts/n3Q7F3v6wBLsioqt8/extended-interview-with-zhukeepa-on-religion)*, I felt inspired to publish more of my thinking about religion in a format t... | https://www.lesswrong.com/posts/WvpNKpL6hJB89cMNe/secular-interpretations-of-core-perennialist-claims |
# Interview with Robert Kralisch on Simulators
*The following is a transcript of a *[*video interview*](https://youtu.be/9KvnrdhJQYs?si=Mc_tHXyqbLyHEtHs) *(edited for grammar and readability) with Robert Kralisch on* [*simulator*](https://www.lesswrong.com/posts/vJFdjigzmcXMhNTsx/simulators) *theory and its implicatio... | https://www.lesswrong.com/posts/fQe7zPuGEkM5fZZyM/interview-with-robert-kralisch-on-simulators |
# My lukewarm take on GLP-1 agonists
Motivation for posting:
I read Scott's recent article: [https://www.astralcodexten.com/p/why-does-ozempic-cure-all-diseases](https://www.astralcodexten.com/p/why-does-ozempic-cure-all-diseases)
And it seems to me that scott isn't considering two points:
1. GLP-1s might fix a lo... | https://www.lesswrong.com/posts/t3pCgKGb7KtNdArmk/my-lukewarm-take-on-glp-1-agonists |
# Day Zero Antivirals for Future Pandemics
By Brian Wang
Ebola tore through Western Africa in 2014, killing an estimated 11,325 people in two years. After the outbreak ended, it took another three years for the first Ebola vaccine by Merck to be approved. When COVID-19 emerged in 2019, by contrast, mRNA vaccines deve... | https://www.lesswrong.com/posts/Wmnf9kfuxASeBcnbp/day-zero-antivirals-for-future-pandemics |
# ... Wait, our models of semantics should inform fluid mechanics?!?
*This post is styled after conversations we’ve had in the course of our research, put together in a way that hopefully highlights a bunch of relatively recent and (ironically) hard-to-articulate ideas around natural abstractions. *
**John**: So we’v... | https://www.lesswrong.com/posts/QsstSjDqa7tmjQfnq/wait-our-models-of-semantics-should-inform-fluid-mechanics |
# Would catching your AIs trying to escape convince AI developers to slow down or undeploy?
I often talk to people who think that if frontier models were egregiously misaligned and powerful enough to pose an existential threat, you could get AI developers to slow down or undeploy models by producing evidence of their ... | https://www.lesswrong.com/posts/YTZAmJKydD5hdRSeG/would-catching-your-ais-trying-to-escape-convince-ai |
# My Apartment Art Commission Process
When I know that I’m going to be moving out from an apartment soon, I commission a digital artist to draw it for me. Then I print it out and I have a cool art piece. If you love your current place but you don’t think you’ll spend the rest of your life there, you should consider do... | https://www.lesswrong.com/posts/EBT3wxngiBfDFQx4s/my-apartment-art-commission-process |
# Soft Nationalization: how the USG will control AI labs
[Crossposted to the EA Forum. ](https://forum.effectivealtruism.org/posts/47RH47AyLnHqCQRCD/soft-nationalization-how-the-us-government-will-control-ai)
We have yet to see anyone describe a critical element of effective AI safety planning: **a realistic model of... | https://www.lesswrong.com/posts/BueeGgwJHt9D5bAsE/soft-nationalization-how-the-usg-will-control-ai-labs |
# Unit economics of LLM APIs
*Disclaimer 1*: Our calculations are rough in places; information is sparse, guesstimates abound.
*Disclaimer 2*: This post draws from public info on FutureSearch as well as a paywalled report. If you want the paywalled numbers, email [dan@futuresearch.ai](mailto:dan@futuresearch.ai) with... | https://www.lesswrong.com/posts/SJESBW9ezhT663Sjd/unit-economics-of-llm-apis |
# What Depression Is Like
I was thinking to myself about the difficulties I have explaining depression to people, when I thought of a relatively good metaphor for it.
Bear with me.
### Sudoku
For anyone unaware, Sudoku is a puzzle where one tries to fill in a 9x9 grid of numbers according to certain rules:
, explores using debates between AI agents to ensure truthful answers from advanced systems. Recently three key deb... | https://www.lesswrong.com/posts/DgKyDTKeGcisEGCnH/gpt-3-5-judges-can-supervise-gpt-4o-debaters-in-capability |
# LessWrong email subscriptions?
Right now, LessWrong has a fair amount of email subscription options, but they're given in a giant list in your user profile that's kind of annoying to look at.[^lpout25dc1]
I have a sense that some people – especially newer users, or users who just don't have a habit of coming to Les... | https://www.lesswrong.com/posts/SuNa5usgxRrvAtbyH/lesswrong-email-subscriptions |
# SB 1047: Final Takes and Also AB 3211
This is the endgame. Very soon the session will end, and various bills either will or won’t head to Newsom’s desk. Some will then get signed and become law.
Time is rapidly running out to have your voice impact that decision.
Since my last weekly, we got a variety of people co... | https://www.lesswrong.com/posts/RaKWcwhygqpMnFZCp/sb-1047-final-takes-and-also-ab-3211 |
# The Information: OpenAI shows 'Strawberry' to feds, races to launch it
[Two](https://www.theinformation.com/articles/openai-shows-strawberry-ai-to-the-feds-and-uses-it-to-develop-orion) [new](https://www.theinformation.com/articles/openai-races-to-launch-strawberry-reasoning-ai-to-boost-chatbot-business) The Informa... | https://www.lesswrong.com/posts/8oX4FTRa8MJodArhj/the-information-openai-shows-strawberry-to-feds-races-to |
# Am I confused about the "malign universal prior" argument?
In a [2016 blog post](https://ordinaryideas.wordpress.com/2016/11/30/what-does-the-universal-prior-actually-look-like/), Paul Christiano argued that the universal prior (hereafter "UP") may be "malign." His argument has received a lot of follow-up discussio... | https://www.lesswrong.com/posts/Xmqnuq5eo9inywS7T/am-i-confused-about-the-malign-universal-prior-argument |
# On agentic generalist models: we're essentially using existing technology the weakest and worst way you can use it
The bottom of the coffee table is not the ceiling
-------------------------------------------------
Not long ago, Andrew Ng posted one of the most fascinating tweets I've read in quite some time. It's ... | https://www.lesswrong.com/posts/fLzbjZLpuEWuX8Z2h/on-agentic-generalist-models-we-re-essentially-using |
# First Lighthaven Sequences Reading Group
Come get old-fashioned with us, and let's read the sequences at Lighthaven! We'll show up, mingle, do intros, and then split off into randomized groups for some sequences discussion. Please do the reading beforehand - it should be no more than 20 minutes of reading.
, [Michael F. Bonner](https://arxiv.org/search/q-bio?searchtype=author&query=Bonner,+M+F).
Quick take: this seems among the strongest evidence yet for the Natural Abstration Hypothesis.... | https://www.lesswrong.com/posts/yP48zaJt7YwFwmLxB/universal-dimensions-of-visual-representation |
# things that confuse me about the current AI market.
Paging Gwern or anyone else who can shed light on the current state of the AI market—I have several questions.
Since the release of ChatGPT, at least 17 companies, according to the [LMSYS Chatbot Arena Leaderboard](https://lmarena.ai/?leaderboard), have developed ... | https://www.lesswrong.com/posts/yRjLY3z3GQBJaDuoY/things-that-confuse-me-about-the-current-ai-market |
# Looking for Goal Representations in an RL Agent - Update Post
**Summary**
===========
I’ve been working on a project aimed at finding goal representations in a small RL agent. I designed a setup where I train an agent, rapidly alternating between two very similar objectives.
I was able to consistently (across alt... | https://www.lesswrong.com/posts/GLBCvsN9Tgr2aZTyT/looking-for-goal-representations-in-an-rl-agent-update-post |
# "Deception Genre" What Books are like Project Lawful?
*This post is spoiler-free*
I just finished [Project Lawful](https://www.projectlawful.com/posts/6322), a really long, really weird book by Eliezer Yudkowsky.
The book's protagonist is a knowledgable and perceptive target. A conspiracy forms around the target t... | https://www.lesswrong.com/posts/4BMuPuhZLkNvwxMRd/deception-genre-what-books-are-like-project-lawful |
# Benefits of Psyllium Dietary Fiber in Particular
Psyllium husk is a non-fermenting (no gas or bloating) soluble dietary fiber that improves both constipation and diarrhea (such as with IBS), normalizes blood sugar, reduces LDL ("bad") cholesterol, and can help with weight loss. Each type of dietary fiber has diffe... | https://www.lesswrong.com/posts/zXk9Rwy4oFaex7bdd/benefits-of-psyllium-dietary-fiber-in-particular |
# AI #79: Ready for Some Football
I have never been more ready for Some Football.
Have I learned all about the teams and players in detail? No, I have been rather busy, and have not had the opportunity to do that, although I eagerly await Seth Burn’s Football Preview. I’ll have to do that part on the fly.
But oh my ... | https://www.lesswrong.com/posts/K8R3Cpj3szcX7z6Xo/ai-79-ready-for-some-football |
# Things I learned talking to the new breed of scientific institution
*Note:* *this article is spons... | https://www.lesswrong.com/posts/zPmhxAmEYwKyy29TN/things-i-learned-talking-to-the-new-breed-of-scientific |
# Solving adversarial attacks in computer vision as a baby version of general AI alignment
I spent the last few months trying to tackle the problem of adversarial attacks in computer vision from the ground up. The results of this effort are written up in our new paper [_Ensemble everything everywhere: Multi-scale aggr... | https://www.lesswrong.com/posts/oPnFzfZtaoWrqTP4H/solving-adversarial-attacks-in-computer-vision-as-a-baby |
# Free Will and Dodging Anvils: AIXI Off-Policy
$$*This post depends on a basic understanding of* [*history-based reinforcement learning and the AIXI model.*](https://www.amazon.ca/Introduction-Universal-Artificial-Intelligence/dp/1032607025)
*I am grateful to Marcus Hutter and the lesswrong team for early feedback,... | https://www.lesswrong.com/posts/WECqiLtQiisqWvhim/free-will-and-dodging-anvils-aixi-off-policy |
# Nursing doubts
If you ask the internet if breastfeeding is good, you will soon learn that *YOU MUST BREASTFEED* because *BREAST MILK = OPTIMAL FOOD FOR BABY*. But if you look for evidence, you’ll discover two disturbing facts.
First, there’s no consensus about *why* breastfeeding is good. I’ve seen experts suggest ... | https://www.lesswrong.com/posts/p7x3vvPR59WHuoQ2A/nursing-doubts |
# Are LLMs on the Path to AGI?
I am unsure, but I disagree with one argument that they aren’t.
---------------------------------------------------------------
There’s a joke about how humans have gotten so good at thinking that they tricked rocks into thinking for them. But it’s a joke, in part because it’s funny to ... | https://www.lesswrong.com/posts/zmKgozWaNmyuzJdTD/are-llms-on-the-path-to-agi |
# Congressional Insider Trading
You’ve probably seen the [Nancy Pelosi Stock Tracker](https://x.com/pelositracker_?lang=en) on X or else a [collection](https://finance.yahoo.com/news/former-house-speaker-nancy-pelosi-095000785.html) of [articles](https://www.newyorker.com/magazine/2024/06/03/how-to-pick-stocks-like-yo... | https://www.lesswrong.com/posts/YsahAYJCDfcTrkiHZ/congressional-insider-trading |
# Principles for the AGI Race
Crossposted from [https://williamrsaunders.substack.com/p/principles-for-the-agi-race](https://williamrsaunders.substack.com/p/principles-for-the-agi-race)
Why form principles for the AGI Race?
=====================================
I worked at OpenAI for 3 years, on the Alignment and Su... | https://www.lesswrong.com/posts/aRciQsjgErCf5Y7D9/principles-for-the-agi-race |
# AIS terminology proposal: standardize terms for probability ranges
**Summary:** The AI safety research community should adopt standardized terms for probability ranges, especially in public-facing communication and especially when discussing risk estimates. The terms used by the IPCC are a reasonable default.
Scien... | https://www.lesswrong.com/posts/NE2vog9AxtnB3WEpt/ais-terminology-proposal-standardize-terms-for-probability |
# AI for Bio: State Of The Field
[](https://substackcdn.com/image/fetch/f_auto,q_auto:good,fl_progres... | https://www.lesswrong.com/posts/CGegZveogJRGCK3LA/ai-for-bio-state-of-the-field |
# Singular learning theory: exercises
*Thanks to Jesse Hoogland and George Wang for feedback on these exercises. *
In learning singular learning theory (SLT), I found it was often much easier to understand by working through examples, rather than try to work through the (fairly technical) theorems in their full gener... | https://www.lesswrong.com/posts/3HYqTAi4kD35G3BzQ/singular-learning-theory-exercises |
# Can Large Language Models effectively identify cybersecurity risks?
TL;DR
-------
I was interested in the ability of LLMs to discriminate input scenarios/stories that carry high vs low cyber risk, and found that it is one of the “hidden features” present in most later layers of [Mistral7B](https://huggingface.co/... | https://www.lesswrong.com/posts/gJqvNH2XJqgyLq9B7/can-large-language-models-effectively-identify-cybersecurity |
# Book review: On the Edge
Book review: On the Edge: The Art of Risking Everything, by Nate Silver.
Nate Silver's latest work straddles the line between journalistic
inquiry and subject matter expertise.
"On the Edge" offers a valuable lens through which to understand
analytical risk-takers.
### The River v... | https://www.lesswrong.com/posts/jhzY8mTdrcJrL4J9s/book-review-on-the-edge |
# Anthropic is being sued for copying books to train Claude
OpenAI faces [10 copyright lawsuits](https://www.lesswrong.com/posts/5RX8j4CDqadnffCij/fifteen-lawsuits-against-openai) and Anthropic is starting to get sued as well. Whether or not you agree with copyright, this is worth looking into. Lawsuits could [hinder]... | https://www.lesswrong.com/posts/8SwJr2uvDFXYcqNGv/anthropic-is-being-sued-for-copying-books-to-train-claude |
# Verification methods for international AI agreements
**TLDR:** A new [**paper**](https://arxiv.org/abs/2408.16074) summarizes some verification methods for international AI agreements. See also summaries on [LinkedIn](https://www.linkedin.com/feed/update/urn:li:activity:7235291620529582080/) and [Twitter](https://t... | https://www.lesswrong.com/posts/uqWtSHLpDgiRYhnoM/verification-methods-for-international-ai-agreements |
# My Model of Epistemology
I regularly get asked by friends and colleagues for recommendation of good resources to study epistemology. And whenever that happens, I make an internal (or external) "Eeehhh"pained sound.
For I can definitely point to books and papers and blog posts that inspired me, excited me, and shape... | https://www.lesswrong.com/posts/L9W9cCjcgZwDgKzZ7/my-model-of-epistemology |
# Epistemic states as a potential benign prior
Malignancy in the prior seems like a strong crux of the goal-design part of alignment to me. Whether your prior is going to be used to model:
* processes in the multiverse containing the AI which does said modeling,
* processes which would output all of some blog so we ... | https://www.lesswrong.com/posts/uam4meRZw4DBm2B7J/epistemic-states-as-a-potential-benign-prior |
# On epistemic autonomy
TL;DR: [**Epistemic autonomy**](https://www.tandfonline.com/doi/full/10.1080/02691728.2024.2335623) refers to the degree to which you are able to change your mind without letting outside sources, such as social and financial pressure, affect it. To increase your epistemic autonomy, you can ref... | https://www.lesswrong.com/posts/rhDmAHvoxWwBo2yHW/on-epistemic-autonomy |
# Forecasting One-Shot Games
| Cliff notes: * You can practice forecasting on videogames you've never played before, to develop the muscles for "decision-relevant forecasting." * Turn based videogames work best. I recommend "Luck Be a Landlord", "Battle for Polytopia", or "Into the Breach." * Each turn, make as many [... | https://www.lesswrong.com/posts/hxhBT89wDBpWcuCW6/forecasting-one-shot-games |
# Can a Bayesian Oracle Prevent Harm from an Agent? (Bengio et al. 2024)
Yoshua Bengio wrote a [blogpost](https://yoshuabengio.org/2024/08/29/bounding-the-probability-of-harm-from-an-ai-to-create-a-guardrail/) about a new AI safety [paper](https://arxiv.org/abs/2408.05284) by him, various collaborators, and me. I've p... | https://www.lesswrong.com/posts/kZqFCt6kLQksCQ7hz/can-a-bayesian-oracle-prevent-harm-from-an-agent-bengio-et |
# Book Review: What Even Is Gender?
*I submitted this review to the 2024 ACX book review contest, but it didn't make the cut, so I'm putting it here instead for posterity. *
Conspiracy theories are fun because of how they make everything fit together, and scratch the unbearable itch some of us get when there are litt... | https://www.lesswrong.com/posts/guCf9vDqAH6zSbL8D/book-review-what-even-is-gender |
# A primer on the next generation of antibodies
Introduction
============
[If you want a primer ove... | https://www.lesswrong.com/posts/nSnb9b5a5o7kzg82q/a-primer-on-the-next-generation-of-antibodies |
# My decomposition of the alignment problem
Epistemic staus: Exploratory
Summary: In this post I will decompose the alignment problem into subproblems and frame existing approaches in terms of their relations to the subproblems. I will try to place a larger focus on the [epistemic process](https://www.lesswrong.com/p... | https://www.lesswrong.com/posts/yLpWBKBnRfSFbWMct/my-decomposition-of-the-alignment-problem |
# Survey: How Do Elite Chinese Students Feel About the Risks of AI?
**Intro**
In April 2024, my colleague and I (both affiliated with Peking University) conducted a survey involving 510 students from Tsinghua University and 518 students from Peking University—China's two top academic institutions. Our focus was on th... | https://www.lesswrong.com/posts/gxCGKHpX8G8D8aWy5/survey-how-do-elite-chinese-students-feel-about-the-risks-of |
# A gentle introduction to sparse autoencoders
Sparse autoencoders (SAEs) are the current hot topic 🔥 in the interpretability world. In late May, Anthropic released a [paper](https://transformer-circuits.pub/2024/scaling-monosemanticity/) that shows how to use sparse autoencoders to effectively break down the interna... | https://www.lesswrong.com/posts/8YnHuN55XJTDwGPMr/a-gentle-introduction-to-sparse-autoencoders |
# How I got 4.2M YouTube views without making a single video
Just over a month ago, I wrote this.
> The Wikipedia articles on the VNM theorem, Dutch Book arguments, money pump, Decision Theory, Rational Choice Theory, etc. are all a horrific mess. They're also completely disjoint, without any kind of Wikiproject or w... | https://www.lesswrong.com/posts/rS3jWvoX7JaxqYDJG/how-i-got-4-2m-youtube-views-without-making-a-single-video |
# Reducing global AI competition through the Commerce Control List and Immigration reform: a dual-pronged approach
**Summary:** This essay proposes specific proposals for consolidating AI development within the US in order to ensure US has the power to unilaterally slow down AI without interference from race dynamics.... | https://www.lesswrong.com/posts/x65BheCZ4J3gNuko7/reducing-global-ai-competition-through-the-commerce-control |
# Announcing the PIBBSS Symposium '24!
Tl;dr
=====
We are pleased to invite you to the **second** [**PIBBSS Symposium**](https://pibbss.ai/symposium-24/)**,** where the fellows from the ‘24 fellowship program present their work.
The symposium is taking place online, over several days in the week of **September 9th... | https://www.lesswrong.com/posts/sondQBxf32PLyeWLZ/announcing-the-pibbss-symposium-24 |
# On the UBI Paper
Would a universal basic income (UBI) work? What would it do?
Many people agree July’s RCT on giving people a guaranteed income, and its paper from Eva Vivalt, Elizabeth Rhodes, Alexander W. Bartik, David E. Broockman and Sarah Miller was, despite whatever flaws it might have, the best data we have ... | https://www.lesswrong.com/posts/RQDqnCeff4cJhKQiT/on-the-ubi-paper |
# The Checklist: What Succeeding at AI Safety Will Involve
*Crossposted by* [*habryka*](https://www.lesswrong.com/users/habryka4) *with Sam's permission. Expect lower probability for Sam to respond to comments here than if he had posted it (he said he'll be traveling a bunch in the coming weeks, so might not have tim... | https://www.lesswrong.com/posts/mGCcZnr4WjGjqzX5s/the-checklist-what-succeeding-at-ai-safety-will-involve |
# AI Safety at the Frontier: Paper Highlights, August '24
*This is a selection of AI safety paper highlights in August 2024, from my blog "AI Safety at the Frontier". The selection primarily covers ML-oriented research. It's only concerned with papers (arXiv, conferences etc.), not LessWrong or Alignment Forum posts. ... | https://www.lesswrong.com/posts/WFPqFD8A6rLtDFJGM/ai-safety-at-the-frontier-paper-highlights-august-24 |
# Announcing the Ultimate Jailbreaking Championship
**Gray Swan AI is hosting an LLM jailbreaking championship, offering $40,000 in bounties. **
* Official Website: [https://app.grayswan.ai/arena](https://app.grayswan.ai/arena)
* Pre-Registration Form: [https://app.grayswan.ai/arena#registration](https://app.gray... | https://www.lesswrong.com/posts/N5cttN24LqEteFgN2/announcing-the-ultimate-jailbreaking-championship |
# Is there any rigorous work on using anthropic uncertainty to prevent situational awareness
/ deception?
AI systems up to some high level of intelligence plausibly need to know exactly where they are in space-time in order for deception/"scheming" to make sense as a strategy.
This is because they need to know:
... | https://www.lesswrong.com/posts/eAhE5DCf8KsEvbiho/is-there-any-rigorous-work-on-using-anthropic-uncertainty-to |
# AI and the Technological Richter Scale
The Technological Richter scale is introduced about 80% of the way through [Nate Silver’s new book On the Edge](https://www.amazon.com/Edge-Art-Risking-Everything/dp/B0CR4FYC8R/ref=sr_1_1?crid=3L2GMWJY54J6U&dib=eyJ2IjoiMSJ9.DmCHlyHAYSQApF1nJzsy_xMBiiJwNZ3kjGneZUcWQsqJuXpZNx-gQp... | https://www.lesswrong.com/posts/oAy72fcqDHsCvLBKz/ai-and-the-technological-richter-scale |
# Michael Dickens' Caffeine Tolerance Research
[Michael Dickens](https://mdickens.me/) has [read](https://mdickens.me/2024/03/02/caffeine_tolerance/) [the research](https://mdickens.me/2024/03/29/does_caffeine_stop_working/) and performed [two](https://mdickens.me/2024/04/11/caffeine_self_experiment/) [self-experiment... | https://www.lesswrong.com/posts/3bmoxXgJpfcoZpmcJ/michael-dickens-caffeine-tolerance-research |
# Automating LLM Auditing with Developmental Interpretability
*Produced as part of the* [ML Alignment & Theory Scholars Program](https://www.matsprogram.org/) *\- Summer 2024 Cohort, supervised by Evan Hubinger*
**TL: DR**
==========
* We proved that the SAE features related to the finetuning target will change mo... | https://www.lesswrong.com/posts/uBZcQLmBPiiSyXsc9/automating-llm-auditing-with-developmental-interpretability |
# What happens if you present 500 people with an argument that AI is risky?
Recently, Nathan Young and I wrote about arguments for AI risk and put them on the [AI Impacts wiki](https://wiki.aiimpacts.org/arguments_for_ai_risk/list_of_arguments_that_ai_poses_an_xrisk/start). In the process, we ran a casual little surve... | https://www.lesswrong.com/posts/vfaCvXuQHj8ztibCG/what-happens-if-you-present-500-people-with-an-argument-that |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.