text stringlengths 300 320k | source stringlengths 52 154 |
|---|---|
# Electrostatic Airships?
Airships are pretty dang cool. Airplanes need a continuous expenditure of energy to stay in the air, but if you just fill a bag with a light gas, you can stay up in the air with no energy expenditure at all. The two lightest gases are hydrogen and helium. Though the hydrogen atom is 4 times l... | https://www.lesswrong.com/posts/6bX3xwf9dvhvuB6an/electrostatic-airships |
# Video lectures on the learning-theoretic agenda
This is a YouTube playlist of recorded lectures on the [learning-theoretic AI alignment agenda](https://www.alignmentforum.org/posts/ZwshvqiqCvXPsZEct/the-learning-theoretic-agenda-status-2023) (LTA) I gave for my [MATS](https://www.matsprogram.org/) scholars of the Wi... | https://www.lesswrong.com/posts/NWKk2eQwfuGzRXusJ/video-lectures-on-the-learning-theoretic-agenda |
# The hostile telepaths problem
*Epistemic status: model-building based on observation, with a few successful unusual predictions. Anecdotal evidence has so far been consistent with the model. This puts it at risk of seeming more compelling than the evidence justifies just yet. Caveat emptor.*
* * *
Imagine you're a... | https://www.lesswrong.com/posts/5FAnfAStc7birapMx/the-hostile-telepaths-problem |
# On Shifgrethor
A small number of terms are elevated from the pages of literature, up to the Mount Olympus of blog post vernacular. [Moloch](https://slatestarcodex.com/2014/07/30/meditations-on-moloch/), as the dark god of failed coordination problems. [The Dark Forest](https://www.lesswrong.com/posts/xDNyXGCDephBuNF... | https://www.lesswrong.com/posts/ngFa92YFdCxrqAWqe/on-shifgrethor |
# Hiring a writer to co-author with me (Spencer Greenberg for ClearerThinking.org)
Hi all! I'm looking to hire a writer for 3-5 hours per week (who can put in that time reliably nearly every week) to co-author writing projects with me (on topics like psychology, philosophy, and society). I'm looking for someone who:
... | https://www.lesswrong.com/posts/SNAjLYEXrbQ2aWBkW/hiring-a-writer-to-co-author-with-me-spencer-greenberg-for |
# Open Source Replication of Anthropic’s Crosscoder paper for model-diffing
Intro
-----
Anthropic recently released an exciting mini-paper on crosscoders ([Lindsey et al.](https://transformer-circuits.pub/2024/crosscoders/index.html)). In this post, we open source a [model-diffing crosscoder](https://huggingface.co/c... | https://www.lesswrong.com/posts/srt6JXsRMtmqAJavD/open-source-replication-of-anthropic-s-crosscoder-paper-for |
# Substituting Talkbox for Breath Controller
One of the inputs of my [rhythm stage setup](https://www.jefftk.com/p/rhythm-stage-setup-v4) has been a breath controller, which let's me gives me a continuous controller. I use it for a few different things, especially when playing music that's farther in the electronic di... | https://www.lesswrong.com/posts/6DrxzuSKjsauJak64/substituting-talkbox-for-breath-controller |
# Lighthaven Sequences Reading Group #8 (Tuesday 10/29)
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 readi... | https://www.lesswrong.com/events/CcrNegGiPmEaq4jQW/lighthaven-sequences-reading-group-8-tuesday-10-29 |
# Care Doesn't Scale
> If you wanted to design a social system to care for children who have lost their parents, I don’t know if you could do much better. With four children, each kid can get individual care and attention, but there were four social workers each had three 24-hour blocks per week, so they had time to h... | https://www.lesswrong.com/posts/nhtKJMbgbgCFznytf/care-doesn-t-scale |
# How Likely Are Various Precursors of Existential Risk?
Seven [Samotsvety](https://samotsvety.org/) forecasters give their probabilities for different possible catastrophes to lead to more than **1M direct deaths in any one year, in any of the next 10 years**. This post presents results, starting with a table which p... | https://www.lesswrong.com/posts/8PcjymbWGJCgT7bSf/how-likely-are-various-precursors-of-existential-risk |
# Bridging the VLM and mech interp communities for multimodal interpretability
*Cross-posted* [*here*](https://www.soniajoseph.ai/multimodal-interpretability-in-2024/)*.*
A note to the LW community
--------------------------
I wrote this post after s... | https://www.lesswrong.com/posts/aa5fzGr8JA3pqvhYC/bridging-the-vlm-and-mech-interp-communities-for-multimodal |
# SAEs you can See: Applying Sparse Autoencoders to Clustering
TL;DR
-----
* We train sparse autoencoders (SAEs) on artificial datasets of 2D points, which are arranged to fall into pre-defined, visua... | https://www.lesswrong.com/posts/f7gd7riceJPaPKkNS/saes-you-can-see-applying-sparse-autoencoders-to-clustering |
# Death notes - 7 thoughts on death
Both my grandmother and my great aunt passed away recently, which has caused me to muse upon death.
Slowly, then all at once
------------------------
What *is* death? I guess it's the ending of the process of consciousness. Not that we really know what *consciousness* is?
In that... | https://www.lesswrong.com/posts/jkDqiDKdumjGDyANo/death-notes-7-thoughts-on-death |
# ~80 Interesting Questions about Foundation Model Agent Safety
*Many people helped us a great deal in developing the questions and ideas in this post, including people at CHAI, MATS, various other places in Berkeley, and*[* Aether*](https://forum.effectivealtruism.org/posts/BEuJdpTPcyKYaD9X7/apply-to-aether-independ... | https://www.lesswrong.com/posts/ZJzyDdKsDhFvqQhdQ/80-interesting-questions-about-foundation-model-agent-safety |
# Miles Brundage: Finding Ways to Credibly Signal the Benignness of AI Development and Deployment is an Urgent Priority
Miles Brundage has a new [substack](https://milesbrundage.substack.com) and I like this post. Here's the introduction.
> A simplified view of AI policy is that there are two “arenas” with distinct c... | https://www.lesswrong.com/posts/rFGqh4wx2YnudvLGK/miles-brundage-finding-ways-to-credibly-signal-the |
# Winners of the Essay competition on the Automation of Wisdom and Philosophy
We’re delighted to announce the winners of the [*Essay competition on the Automation of Wisdom and Philosophy*](https://blog.aiimpacts.org/p/essay-competition-on-the-automation).
[*.*
* * *
Summary
-------
Philosophy and wisdom, and the processes under... | https://www.lesswrong.com/posts/LbfWAq4F8tHaQsTyT/towards-the-operationalization-of-philosophy-and-wisdom |
# Finishing The SB-1047 Documentary In 6 Weeks
*(Crossposted from* [*Manifund*](https://manifund.org/projects/finishing-the-sb-1047-documentary-in-6-weeks)*)*
Summary of the proposal
============... | https://www.lesswrong.com/posts/YGn5ryzAS4L2mkmtd/finishing-the-sb-1047-documentary-in-6-weeks |
# AI & wisdom 1: wisdom, amortised optimisation, and AI
*Written for the* [*AI Impacts essay competition on the automation of wisdom and philosophy*](https://blog.aiimpacts.org/p/essay-competition-on-the-automation)
At least in name, AI is about intelligence. How... | https://www.lesswrong.com/posts/4WLQvcgDWxeuFRymz/ai-and-wisdom-1-wisdom-amortised-optimisation-and-ai |
# AI & wisdom 2: growth and amortised optimisation
*Written for the* [*AI Impacts essay competition on the automation of wisdom and philosophy*](https://blog.aiimpacts.org/p/essay-competition-on-the-automation)
Growth modes and direct v amortised optimisation
---... | https://www.lesswrong.com/posts/SoGpdNMsWpzPEwjNJ/ai-and-wisdom-2-growth-and-amortised-optimisation |
# AI & wisdom 3: AI effects on amortised optimisation
*Written for the* [*AI Impacts essay competition on the automation of wisdom and philosophy*](https://blog.aiimpacts.org/p/essay-competition-on-the-automation)
Having waxed philosophical about wisdom and amor... | https://www.lesswrong.com/posts/QrJAcpapGyG4LMpDi/ai-and-wisdom-3-ai-effects-on-amortised-optimisation |
# How might we solve the alignment problem? (Part 1: Intro, summary, ontology)
***(Update Dec 2025: This series has been almost entirely superseded by a newer and more polished series, available*** [***here***](https://www.lesswrong.com/posts/fMqgLGoeZFFQqAGyC/how-do-we-solve-the-alignment-problem)***.)***
Introducti... | https://www.lesswrong.com/posts/iDRxuJyte6xvppCa3/how-might-we-solve-the-alignment-problem-part-1-intro |
# 5 homegrown EA projects, seeking small donors
What do I mean by “homegrown”? These projects are:
* **Local**: Creators have a good track record in the EA or AI Safety community
* **Modest**: The amount requested is not large; $5k would be meaningful
* **Overlooked**: Not already backed by large institutional ... | https://www.lesswrong.com/posts/nAgciBjxDh5MSeoGs/5-homegrown-ea-projects-seeking-small-donors |
# Gwern: Why So Few Matt Levines?
> [Matt Levine](https://en.wikipedia.org/wiki/Matt_Levine_(columnist)) is the most well-known newslettrist ([“Money Stuff”](https://www.bloomberg.com/opinion/authors/ARbTQlRLRjE/matthew-s-levine)) in the financial industry, having blogged or written since 2011, finding his niche in po... | https://www.lesswrong.com/posts/qMKvGDCXCBpbtwojc/gwern-why-so-few-matt-levines |
# D&D.Sci Coliseum: Arena of Data Evaluation and Ruleset
This is a follow-up to [last week's D&D.Sci scenario](https://www.lesswrong.com/posts/WjGcbR3RcRfGtmFfJ/d-and-d-sci-coliseum-arena-of-data): if you intend to play that, and haven't done so yet, you should do so now before spoiling yourself.
There is a web inter... | https://www.lesswrong.com/posts/MwYLnBzn4YpyJZ7Z6/d-and-d-sci-coliseum-arena-of-data-evaluation-and-ruleset |
# A path to human autonomy
> "Each one of us, and also us as the current implementation of humanity are going to be replaced. Persistence in current form is impossible. It's impossible in biology; every species will either die out or it will change and adapt, in which case it is again not the same species. So the next... | https://www.lesswrong.com/posts/NRZfxAJztvx2ES5LG/a-path-to-human-autonomy |
# AI #87: Staying in Character
The big news of the week was the release of a new version of Claude Sonnet 3.5, complete with its ability (for now only through the API) to outright use your computer, if you let it. It’s too early to tell how big an upgrade this is otherwise. ChatGPT got some interface tweaks that, whil... | https://www.lesswrong.com/posts/3AcK7Pcp9D2LPoyR2/ai-87-staying-in-character |
# Searching for phenomenal consciousness in LLMs: Perceptual reality monitoring and introspective confidence
To update our credence on whethe... | https://www.lesswrong.com/posts/BEZzG56iL5K7y6SXA/searching-for-phenomenal-consciousness-in-llms-perceptual |
# A Poem Is All You Need: Jailbreaking ChatGPT, Meta & More
> ***This project report was created in September 2024 as part of the BlueDot AI Safety Fundamentals Course, with the guidance of my facilitator, Alexandra Abbas. Work on this project originated as part of an ideation at a Apart Research hackathon.***
This r... | https://www.lesswrong.com/posts/fKboizBPnPDa9suLF/a-poem-is-all-you-need-jailbreaking-chatgpt-meta-and-more |
# Review: “The Case Against Reality”
This is not a red stop sign:
*
7.1 Post summary / Table of contents
====================================
*Part of the* [*Intuitive Self-Models series*](https://www.lesswrong.com/s/qhdHbCJ3PYesL9dde).
The main thrust of th... | https://www.lesswrong.com/posts/k8uMmw45k3qp8LPNc/intuitive-self-models-7-hearing-voices-and-other |
# Housing Roundup #10
There’s more campaign talk about housing. The talk of needing more housing is highly welcome, as one prominent person after another (including Jerome Powell!) talking like a YIMBY.
A lot of the concrete proposals are of course terrible, but not all of them. I’ll start off covering all that along... | https://www.lesswrong.com/posts/jJqPfzhhCyK5XjtTH/housing-roundup-10 |
# The Alignment Trap: AI Safety as Path to Power
Recent discussions about artificial intelligence safety have focused heavily on ensuring AI systems remain under human control. While this goal seems laudable on its surface, we should carefully examine whether some proposed safety measures could paradoxically enable ra... | https://www.lesswrong.com/posts/zWJTcaJCkYiJwCmgx/the-alignment-trap-ai-safety-as-path-to-power |
# Three Notions of "Power"
We begin with three stories about three people.
First, Zhu Di, emperor of China from 1402 to 1424. In that period, it was traditional for foreign envoys to present gifts to the emperor and make a show of submission, reinforcing the emperor’s authority and China’s image as the center of civi... | https://www.lesswrong.com/posts/ancAQBkfKPdcu2JDN/three-notions-of-power |
# Occupational Licensing Roundup #1
We’re coming out firmly against it.
Our attitude:
... | https://www.lesswrong.com/posts/bac4wxb9F4sciuAh6/occupational-licensing-roundup-1 |
# Updating the NAO Simulator
*Cross-posted from my [NAO Notebook](https://data.securebio.org/jefftk-notebook/simulator-update).*
In April we released [released](https://naobservatory.org/blog/simulating-approaches-to-metagenomic-pandemic-identification) a [tool](https://data.securebio.org/simulator/) to model the eff... | https://www.lesswrong.com/posts/ZP5tmbh6TXmQmQmuw/updating-the-nao-simulator |
# Motivation control
***(Update Dec 2025: This series has been almost entirely superseded by a newer and more polished series, available*** [***here***](https://www.lesswrong.com/posts/fMqgLGoeZFFQqAGyC/how-do-we-solve-the-alignment-problem)***.***
*This is the second in a* [*series*](https://www.lesswrong.com/s/btmY... | https://www.lesswrong.com/posts/vYfeCbvtZT35XgA2B/motivation-control |
# AI as a powerful meme, via CGP Grey
In [episode 158 of the Cortex podcast](https://www.relay.fm/cortex/158), CGP Grey gives their high-level reason why they are worried about AI.
My one line summary: AI should not be compared to nuclear weapons but instead to biological weapons or memes, which evolve under the impl... | https://www.lesswrong.com/posts/PSL88nJjMghpMNTfe/ai-as-a-powerful-meme-via-cgp-grey |
# I turned decision theory problems into memes about trolleys
I hope it has some educational, memetic or at least humorous potential.
Newcomb's problem
 linked here from some advice. Unfortunately, that advice has some caveats. See below:
* There exist some people who should not do the advice.
* Moreover, [people are different](https://everythingstudies.com/2017/04/24/people-are-different/?ref=brasstacks.blog).
* Mor... | https://www.lesswrong.com/posts/fzGTp9nbSraxqHF47/generic-advice-caveats |
# Standard SAEs Might Be Incoherent: A Choosing Problem & A “Concise” Solution
*This work was produced as part of the ML Alignment & Theory Scholars Program - Summer 24 Cohort.*
TL;DR
-----
The current approach to explaining model internals is to (1) disentangle the language model representations into feature activa... | https://www.lesswrong.com/posts/vNCAQLcJSzTgjPaWS/standard-saes-might-be-incoherent-a-choosing-problem-and-a |
# What TMS is like
There are two nuclear options for treating depression: Ketamine and TMS; This post is about the latter.
TMS stands for [Transcranial Magnetic Stimulation](https://www.mayoclinic.org/tests-procedures/transcranial-magnetic-stimulation/about/pac-20384625#:~:text=Transcranial%20magnetic%20stimulation%2... | https://www.lesswrong.com/posts/g3iKYS8wDapxS757x/what-tms-is-like |
# The Compendium, A full argument about extinction risk from AGI
We (Connor Leahy, Gabriel Alfour, Chris Scammell, Andrea Miotti, Adam Shimi) have just published [The Compendium](https://www.thecompendium.ai/), which brings together in a single place the most important arguments that drive our models of the AGI race, ... | https://www.lesswrong.com/posts/prm7jJMZzToZ4QxoK/the-compendium-a-full-argument-about-extinction-risk-from |
# AI #88: Thanks for the Memos
Following up on [the Biden Executive Order](https://thezvi.substack.com/p/on-the-executive-order?utm_source=publication-search) on AI, the White House has now issued an extensive memo outlining its AI strategy. The main focus is on government adaptation and encouraging innovation and com... | https://www.lesswrong.com/posts/HHkYEyFaigRpczhHy/ai-88-thanks-for-the-memos |
# Toward Safety Cases For AI Scheming
Developers of frontier AI systems will face increasingly challenging decisions about whether their AI systems are safe enough to develop and deploy. One reason why systems may not be safe is if they engage in *scheming*. In our new report ["Towards evaluations-based safety cases f... | https://www.lesswrong.com/posts/FXoEFdTq5uL8NPDGe/toward-safety-cases-for-ai-scheming-1 |
# Lighthaven Sequences Reading Group #9 (Tuesday 11/05)
**Edit2: We're in the Bayes Attic! Ask staff for directions.**
**If you're still lost, call me at five one oh, nine nine eight, five seven seven one.**
**—**
**Normally we would invite you to come get old fashioned with us at a Sequences reading group, but thi... | https://www.lesswrong.com/events/aoKELteMoGSqTHP7x/lighthaven-sequences-reading-group-9-tuesday-11-05 |
# Spooky Recommendation System Scaling
Crosspost of [https://phoropter.substack.com/p/spooky-recommendation-system-scaling](https://phoropter.substack.com/p/spooky-recommendation-system-scaling)
Today I wish to focus on one reason that recommendation systems should be regulated: the benefits of scaling that famously ... | https://www.lesswrong.com/posts/QCYDNbD2zmwxuzMcr/spooky-recommendation-system-scaling |
# Toward Safety Case Inspired Basic Research
Abstract
========
AI safety is a young science. In its early history, deep and speculative questions regarding the risks of artificial superintelligence attracted the attention of basic science and philosophy, but the near-horizon focus of industry and governance, coupled ... | https://www.lesswrong.com/posts/EzrFeLRuww7zTSXhy/toward-safety-case-inspired-basic-research |
# The slingshot helps with learning
The [*slingshot effect*](https://machinelearning.apple.com/research/slingshot-effect) is a late-stage training anomaly found in various adaptive gradient optimization methods. In particular, slingshots are present with [AdamW](https://arxiv.org/abs/1711.05101), the optimizer most wi... | https://www.lesswrong.com/posts/LncYobrn3vRr7qkZW/the-slingshot-helps-with-learning |
# GPT-4o Guardrails Gone: Data Poisoning & Jailbreak-Tuning
Imagine your once reliable, trusty AI assistant suddenly suggesting dangerous actions or spreading misinformation. This is a growing threat as large language models (LLMs) become more capable and pervasive. The culprit? Data poisoning, where LLMs are trained ... | https://www.lesswrong.com/posts/9S8vnBjLQg6pkuQNo/gpt-4o-guardrails-gone-data-poisoning-and-jailbreak-tuning |
# JargonBot Beta Test
We've just launched a new experimental feature: "Automated Jargon Glossaries." If it goes well, it may pave the way for things like LaTeX hoverovers and other nice things.
Whenever an author with 100+ karma ~~saves a draft of a post~~[^viz5jugo1rr] or presses the "Generate Terms" button, our dat... | https://www.lesswrong.com/posts/sZvMLWrWomN28uWA9/jargonbot-beta-test |
# Trading Candy
There are a lot of fun things about halloween, with costumes, neighbors, and sweets, but maybe the part I like best is the trading. Two kids sit down, each with a bucket full of candy. After a while they get back up, each with a better bucket than they started with. This feels like it shouldn't be poss... | https://www.lesswrong.com/posts/T5vntejHAZ2jWXMHW/trading-candy |
# When engaging with a large amount of resources during a literature review, how do you prevent yourself from becoming overwhelmed?
I was originally going to email Gwern directly, but figured being in a public space would benefit others who have the same questions and also put more eyes on it.
---
BLUF: I'm writing t... | https://www.lesswrong.com/posts/9Le7HuuFyhvFeMcAF/when-engaging-with-a-large-amount-of-resources-during-a |
# 2024 Unofficial LW Community Census, Request for Comments
Overview
--------
The LessWrong Community Census is an entertaining site tradition that doubles as a useful way to answer various questions about what the userbase looks like. This is a request for comments, constructive criticism, careful consideration, and... | https://www.lesswrong.com/posts/T9DAHucovCqhLqCdq/2024-unofficial-lw-community-census-request-for-comments |
# Levers for Biological Progress - A Response to "Machines of Loving Grace"
By Niko McCarty for [Asimov Press](https://www.asimov.press/p/levers)
Dario Amodei, the CEO of [Anthropic](https://www.anthropic.com/), recently published an essay called “[Machines of Loving Grace](https://darioamodei.com/machines-of-loving-... | https://www.lesswrong.com/posts/2zmxYKKsSaWWjALg2/levers-for-biological-progress-a-response-to-machines-of |
# Complete Feedback
A simple, weak notion of corrigibility is *having a "complete" feedback interface.* In logical induction terms, I mean the AI trainer can insert any trader into the market. I want to contrast this with "partial" feedback, in which only some propositions get feedback and others ("latent" proposition... | https://www.lesswrong.com/posts/3ag99iJEgFFwyj64Z/complete-feedback |
# Seeking Collaborators
I've been accepted as a mentor for the next [AI Safety Camp](https://www.aisafety.camp/). You can apply to work with me on the tiling problem. The goal will be to develop reflectively consistent UDT-inspired decision theories, and try to prove tiling theorems for them.
The deadline for applica... | https://www.lesswrong.com/posts/7AzexLYpXKMqevttN/seeking-collaborators |
# Live Machinery: An Interface Design Philosophy for Wholesome AI Futures
**Fluid interfaces for sensemaking at pace with AI development.**
* [***Register interest here for the workshop in November 2024 (or to be invited to future workshops)***](https://forms.gle/CGJ3N3NHsqTYCEGD9)**.** *To know some practical deta... | https://www.lesswrong.com/posts/9KamjXbTaQpPnNsxp/live-machinery-an-interface-design-philosophy-for-wholesome |
# Dentistry, Oral Surgeons, and the Inefficiency of Small Markets
There are million dollar bills lying on the ground for those with eyes to see them.
I'm in Miami today, visiting a friend who is opening a dental practice. Chatting with someone very knowledgeable about a niche industry is an interesting experience bec... | https://www.lesswrong.com/posts/xAYbPv33gvZZJZsgo/dentistry-oral-surgeons-and-the-inefficiency-of-small |
# Prediction markets and Taxes
(Epistemic status: This is super trivial information)
(Note 2: there is a timely Culture war issue that triggered this post but I will *not* be mentioning any Culture war issues in this post)
Imagine you worked for the US mint, and somebody was betting that a coin you manufactured ... | https://www.lesswrong.com/posts/Pr9AmRbKu9Evztrjc/prediction-markets-and-taxes |
# SAE Probing: What is it good for?
*Subhash and Josh are co-first authors. Work done as part of the two week research sprint in Neel Nanda’s MATS stream*
Update February 2025:
=====================
We have recently expanded this post into a full paper: [https://arxiv.org/abs/2502.16681](https://arxiv.org/abs/2502.... | https://www.lesswrong.com/posts/NMLq8yoTecAF44KX9/sae-probing-what-is-it-good-for |
# Science advances one funeral at a time
Major scientific institutions talk a big game about innovation, but the reality is that many of the mechanisms designed to ensure quality—peer review, funding decisions, the academic hierarchy—[explicitly](https://www.nber.org/papers/w31409) [incentivize](https://arxiv.org/abs/... | https://www.lesswrong.com/posts/tHncgnNSZN6oPgfZX/science-advances-one-funeral-at-a-time |
# What can we learn from insecure domains?
Cryptocurrency is terrible. With a single click of a button, it is possible to accidentally lose *all of your funds*. 99.9% of all cryptocurrency projects are complete scams (conservative estimate). Crypto is also tailor-made for ransomware attacks, since it makes it possi... | https://www.lesswrong.com/posts/qW9uW2NQvEcKjNvxn/what-can-we-learn-from-insecure-domains |
# Two arguments against longtermist thought experiments
*Epistemic status: shower thoughts.*
I am currently going through the [EA Introductory Course](https://docs.google.com/document/d/1mekz3QWN9GePZfeU_tIUUF8mHhB6XcwyANKBskuwfys/edit?tab=t.0#heading=h.uzvudrnhy12k) and we discussed two arguments against longtermism... | https://www.lesswrong.com/posts/zFXLSWhceutjNneMx/two-arguments-against-longtermist-thought-experiments |
# Fragile, Robust, and Antifragile Preference Satisfaction
What do I want to do?
This sounds like a question with an obvious answer, but the more I reflect on it, the more I notice complexity and subtleties.
First, there's a difficult tradeoff between what I want locally and globally: I regularly wish to do thing in... | https://www.lesswrong.com/posts/mDDFnqoLSNqN7xfAp/fragile-robust-and-antifragile-preference-satisfaction |
# The Median Researcher Problem
Claim: memeticity in a scientific field is mostly determined, not by the most competent researchers in the field, but instead by roughly-median researchers. We’ll call this the “median researcher problem”.
Prototypical example: imagine a scientific field in which the large majority of ... | https://www.lesswrong.com/posts/vZcXAc6txvJDanQ4F/the-median-researcher-problem-1 |
# electric turbofans
It seems I've been typecast as a person to ask about "hard tech startups" with big claims that are getting attention. This time it's been Astro Mechanica, a Y Combinator startup making aircraft engines. [Here's a video](https://www.youtube.com/watch?v=zwM2f6irFs4), possibly set up by a PR agency. ... | https://www.lesswrong.com/posts/5LwWEcC5FjdLykA3r/electric-turbofans |
# Why our politicians aren't Median
[Scott asks](https://www.astralcodexten.com/p/secrets-of-the-median-voter-theorem) why, if the Median Voter Theorem is true, American politicians aren't all middle of the road, and barely distinguishable from each other.
> Elegant as this proof may be, it fails to describe the real... | https://www.lesswrong.com/posts/oxyGrTczrgjGodWAh/why-our-politicians-aren-t-median |
# Current safety training techniques do not fully transfer to the agent setting
TL;DR: We are presenting three recent papers which all share a similar finding, i.e. the safety training techniques for chat models don’t transfer well from chat models to the agents built from them. In other words, models won’t tell you h... | https://www.lesswrong.com/posts/ZoFxTqWRBkyanonyb/current-safety-training-techniques-do-not-fully-transfer-to |
# Goal: Understand Intelligence
Save the world by understanding intelligence.
Instead of having SGD "grow" intelligence, design the algorithms of intelligence directly to get a system we can reason about. Align this system to a narrow but pivotal task, e.g. upload a human.
The key to intelligence is finding the algo... | https://www.lesswrong.com/posts/Y83twCDAdGkWLN8Jp/goal-understand-intelligence |
# Drug development costs can range over two orders of magnitude
_This is a cross-post from [my new newsletter](https://essays.soletta.vc/p/100x), where I intend to post about clinical trials and biotech, and from [my personal blog](https://blog.rossry.net/100x), where I intend to go on posting about all other sorts of... | https://www.lesswrong.com/posts/hiyHmq5jxCSdBLaau/drug-development-costs-can-range-over-two-orders-of |
# Survival without dignity
I open my eyes and find myself lying on a bed in a hospital room. I blink.
"Hello", says a middle-aged man with glasses, sitting on a chair by my bed. "You've been out for quite a long while."
"Oh no ... is it Friday already? I had that report due -"
"It's Thursday", the man says.
"Oh gr... | https://www.lesswrong.com/posts/BarHSeciXJqzRuLzw/survival-without-dignity |
# Context-dependent consequentialism
This dialogue is still in progress, but due to other commitments we don't have much time to continue it. We think the content is interesting, so we decided to publish it unfinished. We will maybe very slowly continue adding to it in the future, but can't commit to doing so.
Contex... | https://www.lesswrong.com/posts/gz3SEm77Yvqs2RbKK/context-dependent-consequentialism |
# [Linkpost] Building Altruistic and Moral AI Agent with
Brain-inspired Affective Empathy Mechanisms
Abstract
> As AI closely interacts with human society, it is crucial to ensure that its decision-making is safe, altruistic, and aligned with human ethical and moral values. However, existing research on embedding eth... | https://www.lesswrong.com/posts/pztgKjcSeiXXDCKNp/linkpost-building-altruistic-and-moral-ai-agent-with-brain |
# Abstractions are not Natural
*(This was inspired by a conversation with* [*Alex Altair*](https://www.lesswrong.com/users/alex_altair) *and other fellows as part of the* [*agent foundations fellowship*](https://www.lesswrong.com/posts/WiMnhMXu9CM5ytu3t/work-with-me-on-agent-foundations-independent-fellowship)*, funde... | https://www.lesswrong.com/posts/F4nzox6oh5oAdX9D3/abstractions-are-not-natural |
# A brief history of the automated corporation
Looking back from 2041
----------------------
When people in the early 21st Century imagined an AI-empowered economy, they tended to project person-like AI entities doing the work. “There will be demand for agent-like systems,” they argued, “so we’ll see AI labs making a... | https://www.lesswrong.com/posts/G8FWk2e2hPJv3xkgC/a-brief-history-of-the-automated-corporation |
# Does the "ancient wisdom" argument have any validity? If a particular teaching or tradition is old, to what extent does this make it more trustworthy?
Proponents of spirituality and alternative medicine often use the argument "this has been practiced for 2000 years", with the subtext "therefore it must work". Does t... | https://www.lesswrong.com/posts/bkcBDkuSMHdwEjNbH/does-the-ancient-wisdom-argument-have-any-validity-if-a |
# The current state of RSPs
*This is a reference post. It contains no novel facts and almost no novel analysis.*
The idea of [responsible scaling policies](https://metr.org/blog/2023-09-26-rsp/) is now over a year old. [Anthropic](https://www.anthropic.com/news/announcing-our-updated-responsible-scaling-policy), [Ope... | https://www.lesswrong.com/posts/jjMNCxYnnJZKoPAN2/the-current-state-of-rsps |
# Option control
Introduction and summary
========================
*(**Update Dec 2025: This series has been almost entirely superseded by a newer and more polished series, available*** [***here***](https://www.lesswrong.com/posts/fMqgLGoeZFFQqAGyC/how-do-we-solve-the-alignment-problem)***.***
*This is the third in ... | https://www.lesswrong.com/posts/YSJCPC5Tu2vRRryto/option-control |
# [Intuitive self-models] 8. Rooting Out Free Will Intuitions
*(Last revised Jan. 2026—see Changelog at the bottom.)*
8.1 Post summary / Table of contents
====================================
*This is the final post of the* [*Intuitive Self-Models series*](https://www.lesswrong.com/s/qhdHbCJ3PYesL9dde)*.*
**One-par... | https://www.lesswrong.com/posts/JLZnSnJptzmPtSRTc/intuitive-self-models-8-rooting-out-free-will-intuitions |
# Update on the Mysterious Trump Buyers on Polymarket
I've written a follow-up post on the mysterious Trump buyers on Polymarket. While mainstream media has extensively covered this story, it has overlooked some critical details—most notably, that this trader's bet on Trump is closer to $75 million USDC, making it the... | https://www.lesswrong.com/posts/WCtnc2YHtrjjJvJrx/update-on-the-mysterious-trump-buyers-on-polymarket |
# Empathy/Systemizing Quotient is a poor/biased model for the autism/sex link
*Thank you to Justis Millis for providing feedback and proofreading on this post. This post is also available* [*on my Substack*](https://tailcalled.substack.com/p/empathysystemizing-quotient-is-a)*.*
*TL/DR: Contrary to the theory that neu... | https://www.lesswrong.com/posts/eTHvMqxaLsrxtRLq3/empathy-systemizing-quotient-is-a-poor-biased-model-for-the |
# Bay Winter Solstice 2024: Speech Auditions
Hello! I'm looking for community members to read speeches at the Bay Winter Solstice event this year. If you're interested, please email me at ozybrennan@gmail.com by the end of **November 17** with the speeches you're potentially interested in. I'll ask you to record yours... | https://www.lesswrong.com/posts/DoT6G6raWWuSqpdR7/bay-winter-solstice-2024-speech-auditions |
# Metastatic Cancer Treatment Since 2010: The Success Stories
[
](https://substackcdn.com/image/... | https://www.lesswrong.com/posts/5mgAd5kY6Jag7d5o9/metastatic-cancer-treatment-since-2010-the-success-stories |
# Could orcas be (trained to be) smarter than humans?
*(Btw everything I write here about orcas also applies to a slightly lesser extent to pilot whales (especially long finned ones)*[^v66waa5g5hk]*.)*
*(I'm very very far from an orca expert - basically everything I know about them I learned today.)*
I always thoug... | https://www.lesswrong.com/posts/gRShuaWgKjizM4xPM/could-orcas-be-trained-to-be-smarter-than-humans |
# ML4Good (AI Safety Bootcamp) - Experience report
Introduction
============
This is a short summary of my experience attending the ML4Good UK bootcamp in September 2024. There are 2 previous experience reports I link to at the bottom, but because the program is refined each time, I wanted to describe my experience a... | https://www.lesswrong.com/posts/qovWG7EmYBzcea9Mh/ml4good-ai-safety-bootcamp-experience-report |
# The Shallow Bench
*Cross posting from my personal blog:* [*https://spiralprogress.com/2024/10/28/the-shallow-bench/*](https://spiralprogress.com/2024/10/28/the-shallow-bench/)
Spoilers for "Project Hail Mary", stop reading here if you don't want to be spoiled.
Project Hail Mary follows Ryland Grace, a disgraced ac... | https://www.lesswrong.com/posts/9XtMrLzuf6Pmymckv/the-shallow-bench |
# Anthropic - The case for targeted regulation
The first two sections are below:
>Increasingly powerful AI systems have the potential to accelerate scientific progress, unlock new medical treatments, and grow the economy. But along with the remarkable new capabilities of these AIs come significant risks. Governments ... | https://www.lesswrong.com/posts/DqNxuLH3kaiwwEmWZ/anthropic-the-case-for-targeted-regulation |
# Winning isn't enough
In our jobs as AI safety researchers, we think a lot about what it means to have reasonable beliefs and to make good decisions. This matters because we want to understand how powerful AI systems might behave. It also matters because we ourselves need to know how to make good decisions in light o... | https://www.lesswrong.com/posts/QxoGM89f8zr3JmNrz/winning-isn-t-enough |
# Why Recursion Pharmaceuticals abandoned cell painting for brightfield imaging
*Note: thank you to* [*Brita Belli*](https://www.linkedin.com/in/brita-belli/)*, senior communications manager at Recursion, for connecting me to* [*Charles Baker,*](https://www.linkedin.com/in/charlesjbaker/) *a VP at Recursion, who led a... | https://www.lesswrong.com/posts/5SsSZx5dMkRksrT85/why-recursion-pharmaceuticals-abandoned-cell-painting-for |
# Intent alignment as a stepping-stone to value alignment
I think [Instruction-following AGI is easier and more likely than value aligned AGI](https://www.lesswrong.com/posts/7NvKrqoQgJkZJmcuD/instruction-following-agi-is-easier-and-more-likely-than), and that this accounts for one major [crux of disagreement on align... | https://www.lesswrong.com/posts/587AsXewhzcFBDesH/intent-alignment-as-a-stepping-stone-to-value-alignment |
# An alternative approach to superbabies
*(Note: This post might be slightly funnily written but it is not a joke but serious and important.)*
Me: I think it's possible that we might not need to figure out biotechnological advances for how we can create superbabies through embryo selection or so, but that actually t... | https://www.lesswrong.com/posts/vKM4CTjz5fPB7vznb/an-alternative-approach-to-superbabies |
# Graceful Degradation
There’s a concept I think about when teaching, which I call Graceful Degradation. The basic idea is, how well does this lesson work if someone doesn’t remember it very well or if I teach it badly?
I picked up Graceful Degradation as an engineer, and you might be familiar with it from [that mili... | https://www.lesswrong.com/posts/N65uQ5RP6dYEC6CHW/graceful-degradation |
# Lighthaven Sequences Reading Group #10 (Tuesday 11/12)
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 read... | https://www.lesswrong.com/events/C2XMJwWyydgBx7meq/lighthaven-sequences-reading-group-10-tuesday-11-12 |
# LDT (and everything else) can be irrational
> you should not reject the 'offer' of a field that yields an 'unfair' amount of grain! - [*Ultimatum Game (Arbital)*](https://arbital.com/p/ultimatum_game/)
In this post, I demonstrate a problem in which there is an agent that outperforms [Logical Decision Theory](https:... | https://www.lesswrong.com/posts/2LvMxknC8g9Aq3S5j/ldt-and-everything-else-can-be-irrational |
# How to put California and Texas on the campaign trail!
In the USA, the president isn't determined by a straight vote. Instead, each state gets a certain number of Electoral College (EC) votes, and the candidate with 270 EC votes wins.
It's up to each state to decide how to allocate its EC votes. Most do “winner-tak... | https://www.lesswrong.com/posts/wK6G4Bdjs3mAaNR4w/how-to-put-california-and-texas-on-the-campaign-trail |
# How to cite LessWrong as an academic source?
Hi there.
Quick question. I am using a few articles from LessWrong for a dissertation. Are there any mainstream articles/sources that reference LessWrong as being the catalyst/partial source for AI alignment, researchers, and other academic literature? I think it's snobb... | https://www.lesswrong.com/posts/Zhcd4Ap487u8sWAaE/how-to-cite-lesswrong-as-an-academic-source |
# Scissors Statements for President?
(Epistemic status: I spoke simply / without "appears to" hedges, but I'm not sure of this at all.)
I’m confused why we keep getting [scissors](https://en.m.wiktionary.org/wiki/scissor_statement) [statements](https://slatestarcodex.com/2018/10/30/sort-by-controversial/) as our Pres... | https://www.lesswrong.com/posts/FkYAYQQig4FDTN6r5/scissors-statements-for-president |
# Advisors for Smaller Major Donors?
Open Philanthropy (OP) is the largest grantmaker who is moving money to the things I think are most valuable, including (disclosure!) [my work at the NAO](https://naobservatory.org/). There's been a lot of discussion in the effective altruism community about where this leaves small... | https://www.lesswrong.com/posts/RZDTDJwR7taQGgmhm/advisors-for-smaller-major-donors |
# Meme Talking Points
I’m going to describe a phenomenon that’s likely very obvious, but nevertheless I think it’s worth documenting because I’ve noticed it more and more. I’ll refer to it as Meme Talking Points, and the first instance I noticed this was based on my [conversation on the Ray Epps conspiracy theory](htt... | https://www.lesswrong.com/posts/KfzX5SQmCGxJuWn5X/meme-talking-points |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.