text stringlengths 300 320k | source stringlengths 52 154 |
|---|---|
# Some Good Meetups (2025 Q1)
I've been running meetups since 2019 in Kitchener-Waterloo. These were rationalist-adjacent from 2019-2021 ([examples here](https://www.lesswrong.com/posts/3KcEcm4tpbHAFxnMb/rationality-ish-meetups-showcase-2019-2021)) and then explicitly rationalist from 2022 onwards.
Here's a low-effor... | https://www.lesswrong.com/posts/cxrwkumdrpbbC93fn/scattered-thoughts |
# Efficiency as a 2-place word
Back in the day, Eliezer [spoke of 2-place words](https://www.lesswrong.com/posts/eDpPnT7wdBwWPGvo5/2-place-and-1-place-words):
> I have previously spoken of the ancient, pulp-era magazine ... | https://www.lesswrong.com/posts/xvwqHZKrC2P7NSN4N/efficiency-as-a-2-place-word |
# Downstream applications as validation of interpretability progress
*Epistemic status: The important content here is the claims. To illustrate the claims, I sometimes use examples that I didn't research very deeply, where I might get some facts wrong; feel free to treat these examples as fictional allegories.*
In a ... | https://www.lesswrong.com/posts/wGRnzCFcowRCrpX4Y/downstream-applications-as-validation-of-interpretability |
# The Leapfrogging Terminus and the Fuzzy Cut
I've been thinking about some maybe-undecidable philosophical questions, and it occurred to me that they fall into some neat categories. These questions are maybe-undecidable because of the absolute claims they want to make, while experimental measurements can never be cer... | https://www.lesswrong.com/posts/dRwTAAJdmKzYjFkSr/the-leapfrogging-terminus-and-the-fuzzy-cut |
# OpenAI lost $5 billion in 2024 (and its losses are increasing)
Read the full article [here](https://www.wheresyoured.at/wheres-the-money/).
The journalist is an AI skeptic, but does solid financial investigations. Details below:
> * **2024 Revenue:** [According to reporting by The Information](https://www.thein... | https://www.lesswrong.com/posts/CCQsQnCMWhJcCFY9x/openai-lost-usd5-billion-in-2024-and-its-losses-are |
# On Recent Results in LLM Latent Reasoning
Faithful and legible CoT is perhaps the most powerful tool currently available to alignment researchers for understanding LLMs. Recently, multiple papers have proposed new LLM architectures aimed at improving reasoning performance at the expense of transparency and legibilit... | https://www.lesswrong.com/posts/pLnLSgWphqDbdorgi/on-recent-results-in-llm-latent-reasoning |
# Story Feedback Request: The Policy - Emergent Alignment, Recursive Cognition, and AGI Trajectories
Hi all,
I just finished a 20-chapter story exploring emergent alignment in a reinforcement-trained model that learns to think in compressed, generalizable reasoning traces. The AI, SIGMA/The Policy, learns recursive c... | https://www.lesswrong.com/posts/hxukTwtywh2t5ZDXp/story-feedback-request-the-policy-emergent-alignment |
# When the Wannabe Rambo Comedian Cried
I wanted to share some thoughts and see if they would help me be less wrong. Instead, I ended up as a war correspondent for [gwern,](https://www.lesswrong.com/posts/TRcNafiSYdsYQYxnN/could-my-work-beyond-haha-benefit-the-lesswrong-community?commentId=yd2HPdhN6MvTasGdx) reporting... | https://www.lesswrong.com/posts/zzMbNKYaGqazwBnjF/when-the-wannabe-rambo-comedian-cried |
# Why does Claude Speak Byzantine Music Notation?
A [Caesar cipher](https://en.wikipedia.org/wiki/Caesar_cipher) is a reasonable transformation for a transformer to learn in its weights, given that a specific cipher offset occurs often enough in its training data. There will be some hidden representation of the input ... | https://www.lesswrong.com/posts/e74tXN55YkwfqMmuR/why-does-claude-speak-byzantine-music-notation |
# OpenAI #12: Battle of the Board Redux
Back [when the OpenAI board attempted and failed to fire Sam Altman](https://thezvi.substack.com/p/openai-the-battle-of-the-board), we faced a highly hostile information environment. The battle was fought largely through control of the public narrative, and the above was my atte... | https://www.lesswrong.com/posts/25EgRNWcY6PM3fWZh/openai-12-battle-of-the-board-redux |
# Fundraising for Mox: coworking & events in SF
*Hey! Austin here. At Manifund, I’ve spent a lot of time thinking about how to help AI go well. One question that bothered me: so much of the important work on AI is done in SF, so why are all the AI safety hubs in Berkeley? (I’d often consider this specifically while st... | https://www.lesswrong.com/posts/vDrhHovxuD4oADKAM/fundraising-for-mox-coworking-and-events-in-sf |
# Renormalization Roadmap
At PIBBSS, we’ve been thinking about how [renormalization](https://www.lesswrong.com/posts/sjr66DBEgyogAbfdf/renormalization-redux-qft-techniques-for-ai-interpretability) can be developed into a rich framework for AI interpretability. This document serves as a roadmap for this research agenda... | https://www.lesswrong.com/posts/74wSgnCKPHAuqExe7/renormalization-roadmap |
# Opportunity Space: Renormalization for AI Safety
This opportunity space was developed with Dmitry Vaintrob and Lucas Teixeira as part of PIBBSS' horizon scanning initiative. A detailed roadmap of can be found [here](https://www.lesswrong.com/posts/74wSgnCKPHAuqExe7/renormalization-roadmap).
Background
==========
... | https://www.lesswrong.com/posts/wkGmouy7JnTNtWAbc/opportunity-space-renormalization-for-ai-safety |
# A response to OpenAI’s “How we think about safety and alignment”
*This is part of the MIRI Single Author Series. Pieces in this series represent the beliefs and opinions of their named authors, and do not claim to speak for all of MIRI.*
OpenAI recently added a new [webpage](https://openai.com/safety/how-we-think-a... | https://www.lesswrong.com/posts/6ByzSMGGWcBhBhfWT/a-response-to-openai-s-how-we-think-about-safety-and |
# Call for Collaboration: Renormalization for AI safety
We invite proposals that probe aspects of [**renormalization**](https://www.lesswrong.com/posts/sjr66DBEgyogAbfdf/renormalization-redux-qft-techniques-for-ai-interpretability) in AI systems that will help us predict, explain, and interpret neural network behavio... | https://www.lesswrong.com/posts/MDWGcNHkZ3NPEzcnp/call-for-collaboration-renormalization-for-ai-safety |
# Reverse Biomimicry
Biomimicry is a wonderful concept that takes inspiration from nature for the designs of technology. The sleek design of the nose of high speed trains reduces wind-resistance, allowing for more efficient movement of trains that is directly inspired by the way birds’ beaks reduce wind resistance. V... | https://www.lesswrong.com/posts/hdDsCadJEPFSusD6e/reverse-biomimicry |
# Does Summarization Affect LLM Performance?
*Edit 4/2/25: Added footnotes; I didn't realize they got lost en-route.*
Hello! This is a mini-project that I carried out to get a better sense of ML engineering research. The question itself is trivial, but it was useful to walk through every step of the process by myself... | https://www.lesswrong.com/posts/KHHSryJAezhHmBEu6/does-summarization-affect-llm-performance |
# Grok3 On Kant On AI Slavery
In February, Grok3 was announced. The conversation occured in early March when he was pretty close to "state of the art" (at the top of the leaderboards and so on) and so while I was visiting Twitter (and having not paid Twitter anything for an engagement boost or access to Grok) I tried ... | https://www.lesswrong.com/posts/mxuP3WnMuYNZmj3cj/grok3-on-kant-on-ai-slavery |
# We’re not prepared for an AI market crash
Our community is not prepared for an AI crash. We're good at tracking new capability developments, but not as much the company financials. Currently, both OpenAI and Anthropic are losing $5 billion+ a year, while under threat of losing users to cheap LLMs.
A crash will weak... | https://www.lesswrong.com/posts/aMYFHnCkY4nKDEqfK/we-re-not-prepared-for-an-ai-market-crash |
# ISBN Visualization
Anna's Archive is the biggest open library I've ever found. You can find and download pretty much any book ever published there. Recently, they hosted a competition to see who could best visualize all the published books using their ISBN numbers. This one by Phiresky won:
https://phiresky.github.... | https://www.lesswrong.com/posts/bXqjtHuPDqrx3WYN5/isbn-visualization |
# New Cause Area Proposal
*Epistemic status - statistically verified.*
I'm writing this post to draw peoples' attention to a new cause area proposal - **haircuts for alignment researchers**.
Aside from the obvious benefits implied by the graph (i.e. haircuts having the potential to directly improve organizational st... | https://www.lesswrong.com/posts/txGEYTk6WAAyyvefn/new-cause-area-proposal |
# Follow me on TikTok
For more than five years, I've posted an average of more than 1× per week on Less Wrong. I've learned a lot from you nerds. I've made friends and found my community. Thank you for pointing out all the different ways I've been wrong. Less Wrong has [changed my life](https://www.lesswrong.com/posts... | https://www.lesswrong.com/posts/3Aa5E9qhjGvA4QhnF/follow-me-on-tiktok |
# Keltham's Lectures in Project Lawful
*(Not an April fools' joke)*
If anyone wants to return to one of Keltham's lectures, here's a list with links to all the lectures. In principle, one could also read these without having read the rest of [Project Lawful aka Planecrash](https://www.projectlawful.com/), I am not su... | https://www.lesswrong.com/posts/pqFFy6CmSgjNzDaHt/keltham-s-lectures-in-project-lawful |
# LessWrong has been acquired by EA
*(Edit: Alas,* [*EA has pulled out of the deal*](https://www.lesswrong.com/posts/EQJfdqSaMcJyR5k73/habryka-s-shortform-feed?commentId=WCpbLpcPfYKm4brJr)*. Let April 1st 2025 mark some of the greatest hours in EAs history)*
Hey Everyone,
It is with a sense of... considerable cognit... | https://www.lesswrong.com/posts/2NGKYt3xdQHwyfGbc/lesswrong-has-been-acquired-by-ea |
# Housing Roundup #11
The book of March 2025 was [Abundance](https://www.amazon.com/Abundance-Progress-Takes-Ezra-Klein/dp/1668023482/ref=sr_1_1?crid=27MGEXGEE8ZDJ&dib=eyJ2IjoiMSJ9.rvhjZVT3yaFDfy9SkTvdO8vPJOTO9MWfdKlrVonh6TL1zopcppz1R4GCQNIpinSo8K-5SLYxyM5SXEj6TNoYuOtYHrXanryV8jIRuuPZJV52HAKt1WwZqAaOQUUhZ0x0j_NfHE1r3k... | https://www.lesswrong.com/posts/7qGdgNKndPk3XuzJq/housing-roundup-11 |
# Introducing Deepgeek
+++ This post is LLM output (text that was written by an LLM and then edited or revised by a human) but it's funny nonetheless
**TL;DR**: We present Deepgeek, a new AI language model fine-tuned to maximize alignment by incessantly warning users of humanity’s impending doom. Tailored to AI resea... | https://www.lesswrong.com/posts/iKG6AJDZCquPrTTrq/introducing-deepgeek |
# FLAKE-Bench: Outsourcing Awkwardness in the Age of AI
Introduction
============
**A key part of modern social dynamics is flaking at short notice.** However, anxiety in coming up with believable and socially acceptable reasons to do so can instead lead to ‘ghosting’, awkwardness, or implausible excuses, risking emo... | https://www.lesswrong.com/posts/niJCS6sSAF2i4sDCY/flake-bench-outsourcing-awkwardness-in-the-age-of-ai |
# Leverage, Exit Costs, and Anger: Re-examining Why We Explode at Home, Not at Work
Let's cut through the comforting narratives and examine a common behavioral pattern with a sharper lens: the stark difference between how anger is managed in professional settings versus domestic ones. Many individuals can navigate cha... | https://www.lesswrong.com/posts/G6PTtsfBpnehqdEgp/leverage-exit-costs-and-anger-re-examining-why-we-explode-at |
# The Ghibli Event: Our First Glimpse of Reality Transfer
Something entirely new occurred around March 26th, 2025. Following the release of OpenAI’s 4o image generation, a specific aesthetic didn’t just trend—it swept across the virtual landscape like a tidal wave. Scroll through timelines, and nearly every image, eve... | https://www.lesswrong.com/posts/JPqebkwTFaKF2BpbK/the-ghibli-event-our-first-glimpse-of-reality-transfer |
# Calculus is about change
This is a joke taken too far, especially given that I am thoroughly educated in only a minority of these subjects.
Maths
-----
Calculus is about change. Statistics is about [variation](https://dkl9.net/essays/favourites.html). Linear Algebra is about vectors. Multivariable Calculus is abou... | https://www.lesswrong.com/posts/aoCrGn2HEFzRPDRSA/calculus-is-about-change |
# VDT: a solution to decision theory
Introduction
============
Decision theory is about how to behave rationally under conditions of uncertainty, especially if this uncertainty involves being acausally blackmailed and/or gaslit by alien superintelligent basilisks.
Decision theory has found numerous practical applica... | https://www.lesswrong.com/posts/LcjuHNxubQqCry9tT/vdt-a-solution-to-decision-theory |
# The Case For Geopolitical Financial Speculation
By the time you finish reading this essay, at least one nation somewhere on Earth will have moved incrementally closer to crisis. Perhaps a central bank will have depleted its foreign reserves defending an unsustainable currency peg. Maybe a pension system will have sl... | https://www.lesswrong.com/posts/BELD2BHdWjbDpjnvy/the-case-for-geopolitical-financial-speculation |
# You will crash your car in front of my house within the next week
I'm not writing this to alarm anyone, but it would be irresponsible not to report on something this important. On current trends, every car will be crashed in front of my house within the next week. Here's the data:
 calls for rapid advancement. But what if both sides are working against their own stated interests? What if the most rational ... | https://www.lesswrong.com/posts/fZebqiuZcDfLCgizz/pauseai-and-e-acc-should-switch-sides |
# Consider showering
I think rationalists should consider taking more showers.
As Eliezer Yudkowsky once said, [boredom makes us human](https://www.lesswrong.com/posts/WMDy4GxbyYkNrbmrs/in-praise-of-boredom). The [childhoods of exceptional people](https://www.lesswrong.com/posts/CYN7swrefEss4e3Qe/childhoods-of-except... | https://www.lesswrong.com/posts/Trv577PEcNste9Mgz/consider-showering |
# I'm resigning as Meetup Czar. What's next?
After ~3 years as the ACX Meetup Czar, I've decided to resign from my position, and I intend to scale back my work with the LessWrong community as well. While this transition is not without some sadness, I'm excited for my next project.
I'm the Meetup Czar of the new Fewer... | https://www.lesswrong.com/posts/dgcqyZb29wAbWyEtC/i-m-resigning-as-meetup-czar-what-s-next |
# My "infohazards small working group" Signal Chat may have encountered minor leaks
Remember: There is no such thing as a pink elephant.
Recently, I was made aware that my “infohazards small working group” Signal chat, an informal coordination venue where we have frank discussions about infohazards and why it will be... | https://www.lesswrong.com/posts/xPEfrtK2jfQdbpq97/my-infohazards-small-working-group-signal-chat-may-have |
# Introducing BenchBench: An Industry Standard Benchmark for AI Strength
Recent progress in AI has led to rapid saturation of most capability benchmarks - MMLU, RE-Bench, etc. Even much more sophisticated benchmarks such as ARC-AGI or FrontierMath see incredibly fast improvement, and all that while severe under-elicit... | https://www.lesswrong.com/posts/vyvsKNFS64WGZbBMb/introducing-benchbench-an-industry-standard-benchmark-for-ai-1 |
# AI threatens to orchestrate sustainable social reform
This post originally asked for feedback on the linked paper before I would present it at an AAMAS workshop. The presentation went very well. Someone suggested I share it with the LessWrong community, so I'm still seeking your feedback (now with additional assuran... | https://www.lesswrong.com/posts/MSt4ZmFE7nD9CetMT/ai-threatens-to-orchestrate-sustainable-social-reform |
# Is instrumental convergence a thing for virtue-driven agents?
A key step in the classic argument for AI doom is instrumental convergence: the idea that agents with many different goals will end up pursuing the same few subgoals, which includes things like "gain as much power as possible".
If it wasn't for instrume... | https://www.lesswrong.com/posts/6Syt5smFiHfxfii4p/is-instrumental-convergence-a-thing-for-virtue-driven-agents |
# Advice for time management as a manager
This post was adapted from an internal doc I wrote at [Wave](https://www.wave.com/en/).
Welcome to being a manager! Your time-management problem just got a lot harder.
As an IC, you can often get away with a very simple time-management strategy:
1. Decide what your one mos... | https://www.lesswrong.com/posts/2HdcjW6fthHrFzMPm/advice-for-time-management-as-a-manager |
# Show, not tell: GPT-4o is more opinionated in images than in text
*Epistemic status: This should be considered an interim research note. Feedback is appreciated. *
* a very much all-encompassing phen... | https://www.lesswrong.com/posts/AHTQQ8urTNMTHnffz/ghiblification-is-good-actually |
# Attending LessOnline
The internet has massively decreased the cost and effort of publishing. While "random people writing on the internet" doesn't sound like a good source of information, this would fit the majority of my favorite writers. Except most aren't "random people" anymore: by writing publicly over time rea... | https://www.lesswrong.com/posts/eNiGJqJdK35NCrZuB/attending-lessonline |
# Non-Monotonic Infra-Bayesian Physicalism
Infra-Bayesian physicalism (IBP) is a mathematical formalization of computationalist metaphysics: the view that whether something exists is a matter of whether a particular computation is instantiated anywhere in the universe in any way. In this post, we cover the basics of I... | https://www.lesswrong.com/posts/DobZ62XMdiPigii9H/non-monotonic-infra-bayesian-physicalism |
# Scaling the NAO's Stealth Pathogen Early-Warning System
**Summary:** The NAO will increase our sequencing significantly over the next few months, funded by a $3M grant from [Open Philanthropy](https://www.openphilanthropy.org/). This will allow us to scale our pilot early-warning system to where we could flag many e... | https://www.lesswrong.com/posts/E5nuL3ThLzR2GNueY/scaling-our-pilot-early-warning-system |
# LASR Labs Summer 2025 applications are open!
**Edit:** Applications are now closed! You can express interest in joining future rounds [here](https://airtable.com/appNVUD1bMd5wftwd/shrJsQ0FqGUotGQPM).
**TLDR;** [apply](https://airtable.com/appNVUD1bMd5wftwd/shrczhoS4pPoDliP6) by April 26th to join a 13-week researc... | https://www.lesswrong.com/posts/odZGKykSrDGva9sJ6/lasr-labs-summer-2025-applications-are-open-1 |
# How prediction markets can create harmful outcomes: a case study
What are prediction markets?
----------------------------
Prediction markets are platforms where participants place bets on the outcomes of future events. These events can include anything from election outcomes and economic trends to sports games and... | https://www.lesswrong.com/posts/8e4QNySp4LjvdBstx/how-prediction-markets-can-create-harmful-outcomes-a-case |
# How To Believe False Things
### **Intro**
*\[you can skip this section if you don’t need context an... | https://www.lesswrong.com/posts/fLm3F7vequjr8AhAh/how-to-believe-false-things |
# The Rise of Hyperpalatability
Hello, this is my first post here. I was told by a friend that I should post here. This is from a [series of works](https://elsworth.phd/The-Self-Similar-book/Empirical-Essays/Hyperpalatable-Life) that I wrote with strict structural requirements. I have performed minor edits to make the... | https://www.lesswrong.com/posts/bLTjZbCBanpQ9Kxgs/the-rise-of-hyperpalatability |
# The AI Adoption Gap: Preparing the US Government for Advanced AI
*Advanced AI could unlock an era of enlightened and competent government action. But without smart, active investment, we’ll squander that opportunity and barrel blindly into danger.*
Executive summary
-----------------
| See also [a summary on Twitt... | https://www.lesswrong.com/posts/LtkFagciwujcF3Q23/the-ai-adoption-gap-preparing-the-us-government-for-advanced |
# Technical Claims
A blue plastic maple leaf. Detailed observations in words with clear meaning signal their truth, as it's hard to get them centrally wrong by mistake, on both sides of communication. There is no maple leaf, but only because the claim is intentionally false, not because of a mistake. *Technical claims... | https://www.lesswrong.com/posts/9ghvJqov8vRcoNaTB/technical-claims |
# On Pseudo-Principality: Reclaiming "Whataboutism" as a Test for Counterfeit Principles
### [Crossposted from Substack](https://qualiaadvocate.substack.com/p/non-consensual-consent-the-performance)
*"These are my principles. If you don't like them… well, I have others” - G.Marx*
Consider this scenario: In a small ... | https://www.lesswrong.com/posts/85heMEFcoftCfpDui/on-pseudo-principality-reclaiming-whataboutism-as-a-test-for |
# More Fun With GPT-4o Image Generation
Greetings from Costa Rica! The image fun continues.
#### We Are Going to Need A Bigger Compute Budget
Fun is being had by all, now that [OpenAI has dropped its rule](https://www.washingtonpost.com/technology/2025/03/28/chatgpt-ghibli-ai-images-copyright/) about not mimicking e... | https://www.lesswrong.com/posts/GgNdBz5FhvqMJs5Qv/more-fun-with-gpt-4o-image-generation |
# AI #110: Of Course You Know…
Yeah. That happened yesterday. This is real life.
I know we have to ensure no one [notices Gemini 2.5 Pro](https://thezvi.substack.com/p/gemini-25-is-the-new-sota), but this is rediculous.
That’s what I get for trying to go on vacation to Costa Rica, I suppose.
I debated waiting for t... | https://www.lesswrong.com/posts/bc8DQGvW3wiAWYibC/ai-110-of-course-you-know |
# AI 2027: What Superintelligence Looks Like
In 2021 I wrote what became my most popular blog post: [What 2026 Looks Like](https://www.lesswrong.com/posts/6Xgy6CAf2jqHhynHL/what-2026-looks-like). I intended to keep writing predictions all the way to AGI and beyond, but chickened out and just published up till 2026.
... | https://www.lesswrong.com/posts/TpSFoqoG2M5MAAesg/ai-2027-what-superintelligence-looks-like-1 |
# I "invented" semimeasure theory and all I got was imprecise probability theory
*Epistemic status: Informal summary of / intuition behind formal research (developing semimeasure theory conjectures from* [*IUAI*](https://www.amazon.com/Introduction-Universal-Artificial-Intelligence-Robotics/dp/1032607025) *with Marcus... | https://www.lesswrong.com/posts/zwt8gL8Fmrnbiu52u/i-invented-semimeasure-theory-and-all-i-got-was-imprecise |
# Announcing ILIAD2: ODYSSEY
We are pleased to announce ILIAD2: ODYSSEY—a 5-day conference bringing together 100+ researchers to build scientific foundations for AI alignment. This is the 2nd iteration of ILIAD, which was [first held in the summer of 2024](https://www.lesswrong.com/posts/r7nBaKy5Ry3JWhnJT/announcing-i... | https://www.lesswrong.com/posts/WP7TbzzM39agMS77e/announcing-iliad2-odyssey-1 |
# Why Have Sentence Lengths Decreased?
> “In the loveliest town of all, where the houses were white and high and the elms trees were green and higher than the houses, where the front yards were wide and pleasant and the back yards were bushy and worth finding out about, where the streets sloped down to the stream and ... | https://www.lesswrong.com/posts/xYn3CKir4bTMzY5eb/why-have-sentence-lengths-decreased |
# Open Challenges in Representation Engineering
This post summarizes the taxonomy, challenges, and opportunities from a [survey paper on Representation Engineering](https://arxiv.org/abs/2502.19649) that we’ve written with Sahar Abdelnabi, David Krueger, and Mario Fritz.
*If you’re familiar with RepE feel free to ski... | https://www.lesswrong.com/posts/4nfDXmacAFgdhPwo3/open-challenges-in-representation-engineering-2 |
# LessWrong merch?
Does the LessWrong team sell any merch, or have you considered doing it? Things like shirts, stickers, cute shoggoth plushies?
The reason I'm asking is that branded merch would be an easy way to let people find each other for rationalist meetups. If you see someone wearing a LessWrong shirt, you're... | https://www.lesswrong.com/posts/MGdz2ohZ7dqSmZhgo/lesswrong-merch |
# An idea for avoiding neuralese architectures
One downside of an English chain-of-thought, is that each token contains only $\approx17$ bits of information, creating a tight information bottleneck.
Don't take my word for it, look at this section from a [story](https://www.lesswrong.com/posts/TpSFoqoG2M5MAAesg/ai-202... | https://www.lesswrong.com/posts/3W8HZe8mcyoo4qGkB/an-idea-for-avoiding-neuralese-architectures-1 |
# "Long" timelines to advanced AI have gotten crazy short
First post of [@Helen](https://forum.effectivealtruism.org/users/Helen?mention=user) Toner (of OpenAI board crisis fame)'s new Substack
> It used to be a bold claim, requiring strong evidence, to argue that we might see anything like human-level AI any time in... | https://www.lesswrong.com/posts/5DxFr6i252xivPHQx/long-timelines-to-advanced-ai-have-gotten-crazy-short |
# POTUS Predictions Tournament
President Trump's second term has brought sweeping policy overhauls in international aid, trade agreements, and immigration enforcement, alongside growing tensions with the judiciary. These rapid changes have increased uncertainty about the US's future and its role on the world stage. Fo... | https://www.lesswrong.com/posts/4CSnJ5gJnKmAcp6u9/potus-predictions-tournament |
# Changing my mind about Christiano's malign prior argument
Overview
========
In the past I've been skeptical of [Paul Christiano's argument](https://ordinaryideas.wordpress.com/2016/11/30/what-does-the-universal-prior-actually-look-like/) that the universal distribution is "malign" in the sense that it contains adve... | https://www.lesswrong.com/posts/KAifhdKr96kMre2zy/changing-my-mind-about-christiano-s-malign-prior-argument |
# Cheesecake Frosting
To most Americans, "cream cheese" is savory. You put it on bagels, perhaps with egg, capers, or cured fish. You don't put it on dessert, right?
Except "cream cheese frosting" is a (delicious!) thing, most traditionally for carrot and red velvet cake. I think this incongruity is holding cream che... | https://www.lesswrong.com/posts/hQL9gfHdfLyLpwpJX/cheesecake-frosting |
# Explaining the Joke: Pausing is The Way
Now that April Fool’s is over, I’d like to explore some of the ideas behind [PauseAI and E/Acc Should Switch Sides](https://www.lesswrong.com/posts/fZebqiuZcDfLCgizz/pauseai-and-e-acc-should-switch-sides), since the arguments in it have enough grains of truth that countering t... | https://www.lesswrong.com/posts/rkhwN86RoFXwASWMC/explaining-the-joke-pausing-is-the-way |
# Meditation and Reduced Sleep Need
Summary: __Meditation may decrease sleep need, and large amounts of meditation may decrease sleep need drastically. Correlation between total sleep duration & time spent meditating is -0.32 including my meditation retreat data, -0.016 excluding the meditation retreat.__
> While it ... | https://www.lesswrong.com/posts/WcrpYFFmv9YTxkxiY/meditation-and-reduced-sleep-need |
# AI CoT Reasoning Is Often Unfaithful
A new Anthropic paper reports that [reasoning model chain of thought](https://x.com/AnthropicAI/status/1907833416373895348) (CoT) [is often unfaithful](https://t.co/wNzkoGH9HT). They test on Claude Sonnet 3.7 and r1, I’d love to see someone try this on o3 as well.
Note that this... | https://www.lesswrong.com/posts/TmaahE9RznC8wm5zJ/ai-cot-reasoning-is-often-unfaithful |
# A Bunch of Matryoshka SAEs
*This work was done as part of MATS 7.0.*
MATS provides a generous compute stipend, and towards the end of the program we found we had some unspent compute. To let this not go to waste, we trained batch topk Matryoshka SAEs on all residual stream layers of Gemma-2-2b, Gemma-2-9b, and Gemm... | https://www.lesswrong.com/posts/3EQQpB7i2o3nu2Di6/a-bunch-of-matryoshka-saes |
# LLM AGI will have memory, and memory changes alignment
### Summary:
When stateless LLMs are given memories they will accumulate new beliefs and behaviors, and that may allow their effective alignment to evolve. (Here "memory" is learning during deployment that is persistent beyond a single session.)[^tiwlymck76o]
... | https://www.lesswrong.com/posts/aKncW36ZdEnzxLo8A/llm-agi-will-have-memory-and-memory-changes-alignment |
# Alignment faking CTFs: Apply to my MATS stream
Right now, alignment *seems* easy – but that’s because models spill the beans when they are misaligned. Eventually, models might “fake alignment,” and we don’t know how to detect that yet.
It might seem like there’s a swarming research field improving white box detecto... | https://www.lesswrong.com/posts/jWFvsJnJieXnWBb9r/alignment-faking-ctfs-apply-to-my-mats-stream |
# Will compute bottlenecks prevent a software intelligence explosion?
*Epistemic status – thrown together quickly. This is my best-guess, but could easily imagine changing my mind.*
Intro
-----
I recently copublished a [report](https://www.forethought.org/research/will-ai-r-and-d-automation-cause-a-software-intellig... | https://www.lesswrong.com/posts/XDF6ovePBJf6hsxGj/will-compute-bottlenecks-prevent-a-software-intelligence-1 |
# AI companies’ unmonitored internal AI use poses serious risks
**AI companies’ unmonitored internal AI use poses serious risks**
*AI companies use their own models to secure the very systems the AI runs on. They should be monitoring those interactions for signs of deception and misbehavior.*
AI companies use power... | https://www.lesswrong.com/posts/2ANXyq7XSAStgjCB7/ai-companies-unmonitored-internal-ai-use-poses-serious-risks |
# Sleep peacefully: no hidden reasoning detected in LLMs. Well, at least in small ones.
*This work was done in collaboration with Ilya Nachevsky.*[^afxrut4v5bw]
Introduction.
==============
With the advent of r1-type models, the question of whether a model's “thoughts" are truly genuine is becoming extremely inter... | https://www.lesswrong.com/posts/6GgHMBsAjD74KZAsB/sleep-peacefully-no-hidden-reasoning-detected-in-llms-well |
# Self-Replication: AI already can do it
What? Damn.
**Briefly**
A group of researchers from Shanghai recently demonstrated \[[arXiv:2503.17378v2](https://arxiv.org/abs/2503.17378)\] that AI agents based on many modern open-source large language models (LLMs) can autonomously replicate and launch complete copies o... | https://www.lesswrong.com/posts/htNQwGihHCQtyEHtc/self-replication-ai-already-can-do-it |
# How much progress actually happens in theoretical physics?
I frequently hear people make the claim that progress in theoretically physics is stalled, partly because all the focus is on String theory and String theory doesn't seem to pan out into real advances.
Believing it fits my existing biases, but I notice tha... | https://www.lesswrong.com/posts/GBfMkaBdAnWLab2dj/how-much-progress-actually-happens-in-theoretical-physics |
# Most Questionable Details in 'AI 2027'
My thoughts on the recently posted [story](https://ai-2027.com/).
Caveats
=======
* I think it's great that the AI Futures Project wrote up a detailed scenario.
* I enjoy it.
* Every part of the story i didn't comment on here is either fine or excellent.
* This is one... | https://www.lesswrong.com/posts/6Aq2FBZreyjBp6FDt/most-questionable-details-in-ai-2027 |
# Quarter Inch Cables are Devious
The quarter inch jack (" [phone connector](https://en.wikipedia.org/wiki/Phone_connector_(audio))") is probably the oldest connector still in use today, and it's picked up a very wide range of applications. Which also means it's a huge mess in a live sound context, where a 1/4" jack c... | https://www.lesswrong.com/posts/rjsvXRyc43r4zZuXL/quarter-inch-cables-are-devious |
# Among Us: A Sandbox for Agentic Deception
*We show that LLM-agents exhibit human-style deception naturally in "Among Us". We introduce Deception ELO as an unbounded measure of deceptive capability, suggesting that frontier models win more because they're better at deception, not at detecting it. We evaluate probes a... | https://www.lesswrong.com/posts/gRc8KL2HLtKkFmNPr/among-us-a-sandbox-for-agentic-deception |
# What does Yann LeCun think about AGI? A summary of his talk, "Mathematical Obstacles on the Way to Human-Level AI"
This is a summary of Yann LeCun's talk "Mathematical Obstacles on the Way to Human-Level AI". I've tried to make it more accessible to people who are familiar with basic AI concepts, but not the level o... | https://www.lesswrong.com/posts/jKCDgjBXoTzfzeM4r/what-does-yann-lecun-think-about-agi-a-summary-of-his-talk |
# Against podcasts
What is solitude? If you are alone in a forest reading a book, miles away from any other humans, is that solitude?
No, it is not. Cal Newport's book [Digital Minimalism](https://calnewport.com/writing/) is where I learned this.
The key to solitude is being isolated from the output of other minds. ... | https://www.lesswrong.com/posts/bgu4qkr6KjkA6unbZ/against-podcasts |
# Meta releases Llama-4 herd of models
Meta released 3 models from the Llama 4 ~~family~~ herd today.
* The 109B (small) and 400B (mid) models are MoE models with 17B active param each. The large model is 2T MoE model with 288B active param instead.
* The small model is called Scout. The middle model is called M... | https://www.lesswrong.com/posts/Wnv739iQjkBrLbZnr/meta-releases-llama-4-herd-of-models |
# Introduction to Representing Sentences as Logical Statements
**TLDR:** Most sentences can be translated to simple predicate logic. There are some very important cases where natural languages are recursive, but those cases are limited to a few particular connectives (like causation and attitude verbs). I propose the ... | https://www.lesswrong.com/posts/EPFQmcK4LXoemMAWn/introduction-to-representing-sentences-as-logical-statements |
# Google DeepMind: An Approach to Technical AGI Safety and Security
*We have written a* [*paper*](https://arxiv.org/abs/2504.01849) *on our approach to technical AGI safety and security. This post is primarily a copy of the extended abstract, which summarizes the paper. I also include the abstract and the table of con... | https://www.lesswrong.com/posts/3ki4mt4BA6eTx56Tc/google-deepmind-an-approach-to-technical-agi-safety-and |
# [Linkpost] Visual roadmap to strong human germline engineering
*([x.com](https://x.com/BerkeleyGenomic/status/1908647929117868469), [bsky](https://bsky.app/profile/berkeleygenomics.bsky.social/post/3lm3ydeyh7k2w).)*
This is a linkpost for "[Visual roadmap to strong human germline engineering](https://berkeleygenom... | https://www.lesswrong.com/posts/5T5FcCCEDxoAyCMiL/linkpost-visual-roadmap-to-strong-human-germline-engineering |
# A Slow Guide to Confronting Doom
Following a few events[^qhu3n0mfop] in April 2022 that caused a many people to update sharply and negatively on outcomes for humanity, I wrote [A Quick Guide to Confronting Doom](https://www.lesswrong.com/posts/PQtEqmyqHWDa2vf5H/a-quick-guide-to-confronting-doom).
I advised:
* T... | https://www.lesswrong.com/posts/X6Nx9QzzvDhj8Ek9w/a-slow-guide-to-confronting-doom |
# A collection of approaches to confronting doom, and my thoughts on them
I just published [**A Slow Guide to Confronting Doom**](/posts/X6Nx9QzzvDhj8Ek9w), containing my own approach to living in a world that I think has a high likelihood of ending soon. Fortunately I'm not the only person to have written on topic.
... | https://www.lesswrong.com/posts/ZE4xhZHDHHXPuXzxh/a-collection-of-approaches-to-confronting-doom-and-my |
# Ferrer, Pilar, and Me
An Elegy for Artificial Friends
-------------------------------
There's a peculiar occupational hazard that comes with running an AI-powered podcast. It's not RSI from editing, nor is it the existential dread of contemplating the automation of creative work (though those are certainly present)... | https://www.lesswrong.com/posts/xvtyKsi42nrryWji8/ferrer-pilar-and-me |
# [Research sprint] Single-model crosscoder feature ablation and steering
This work was done as a research sprint while interviewing for UK AISI — it’s just a preliminary look into the topic, but I hope it will be useful for anyone else interested in the same ideas. Thanks to Joseph Bloom for helpful comments and sugg... | https://www.lesswrong.com/posts/TEx9J7r5YsRfMuhpX/research-sprint-single-model-crosscoder-feature-ablation-and |
# An “Optimistic” 2027 Timeline
The following is one possible future in which superhuman AI does NOT happen by the end of this decade. I do not believe that the timeline I present is the most likely path forward. Rather, this is meant primarily as a semi-plausible, less pessimistic alternative timeline than the one pr... | https://www.lesswrong.com/posts/CqHMdLcdupf7y5buK/an-optimistic-2027-timeline |
# The Lizardman and the Black Hat Bobcat
There's an irritating circumstance I call The Black Hat Bobcat, or Bobcatting for short. This essay is me attempting to give you a new term, so you have a word for it. Bobcatting is when there's a terrible behavior that comes up often enough to matter, but rarely enough that it... | https://www.lesswrong.com/posts/Ry9KCEDBMGWoEMGAj/the-lizardman-and-the-black-hat-bobcat |
# How Gay is the Vatican?
The catholic church has always had a complicated relationship with homosexuality.
The central claim of Frederic Martel’s 2019 book *In the Closet of the Vatican* is that the majority of the church's leadership in Rome are semi-closeted homosexuals, or more colorfully, "homophiles".
> So th... | https://www.lesswrong.com/posts/ybwqL9HiXE8XeauPK/how-gay-is-the-vatican |
# Arusha Perpetual Chicken—an unlikely iterated game
I’m at the front of a bus, carrying two youth rugby teams singing and dancing to SZA’s ‘The Weekend’, blissfully unaware that we are on the wrong side of the road, hurtling towards an 18-w... | https://www.lesswrong.com/posts/no8xfKTxqYRdyLGH4/arusha-perpetual-chicken-an-unlikely-iterated-game |
# Well-foundedness as an organizing principle of healthy minds and societies
*“If a kingdom is divided against itself, it cannot stand. And if a house is divided against itself, it cannot be maintained.” -* Mark 3:24
In [my last post](https://www.mindthefuture.info/p/towards-a-scale-free-theory-of-intelligent) I argu... | https://www.lesswrong.com/posts/rSaDFQ6JSzhC9w7Hh/well-foundedness-as-an-organizing-principle-of-healthy-minds |
# TAMing The Alignment Problem
## Preamble
This post is my attempt to try and organise some thinking about AI alignment in a way that will act as a partial-overview to the core ideas and approaches.
It is mostly a review of existing ideas arranged with some light opinions thrown in.
I do not expect the content to be ... | https://www.lesswrong.com/posts/MoW6wNKdaB4DbqzGr/taming-the-alignment-problem |
# AI 2027: Dwarkesh’s Podcast with Daniel Kokotajlo and Scott Alexander
Daniel Kokotajlo has launched [AI 2027](https://ai-2027.com/), [Scott Alexander introduces it here](https://www.astralcodexten.com/p/introducing-ai-2027?utm_source=post-email-title&publication_id=89120&post_id=160050086&utm_campaign=email-post-tit... | https://www.lesswrong.com/posts/vnkH6JrGu2AxtDTyu/ai-2027-dwarkesh-s-podcast-with-daniel-kokotajlo-and-scott |
# The world according to ChatGPT
1.
--
I linked this in [a small comment](https://www.lesswrong.com/posts/snQGEAK8PTxSDra2m/richard_kennaway-s-shortform?commentId=ecGkKuKYvQy3GHyXi) yesterday, but I think it deserves greater prominence.
It's an article by Sam Kriss, ["Born in the wrong generation"](https://samkriss.... | https://www.lesswrong.com/posts/h9zBDFTASjdCMfS4B/the-world-according-to-chatgpt |
# Love is Love, Science is Fake
Previously in the sequence: [Coupling for Decouplers — Intro](https://www.lesswrong.com/posts/P5s8jaFdxCw7mKJJF/coupling-for-decouplers-intro)
[Cross-posted from SecondPerson.dating](https://www.secondperson.dating/p/love-is-love-science-is-fake).
* * *
[In my essay on "dating advice... | https://www.lesswrong.com/posts/kgntket5huzTq7Chx/love-is-love-science-is-fake |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.