text stringlengths 300 320k | source stringlengths 52 154 |
|---|---|
# Against Almost Every Theory of Impact of Interpretability
*Epistemic Status: I believe I am well-versed in this subject. I erred on the side of making claims that were too strong and allowing readers to disagree and start a discussion about precise points rather than trying to edge-case every statement. I also think... | https://www.lesswrong.com/posts/LNA8mubrByG7SFacm/against-almost-every-theory-of-impact-of-interpretability-1 |
# Model of psychosis, take 2
*(I am not by any stretch of the imagination an expert on psychosis. This is more like “live-blogging my thinking as I go”. I’m hoping to spur discussion and get feedback and pointers.)*
1\. Introduction
================
I suggested a model of psychosis in my blog post [“Schizophrenia as... | https://www.lesswrong.com/posts/tgaD4YnpGBhGGbAy5/model-of-psychosis-take-2 |
# Progress links digest, 2023-08-17: Cloud seeding, robotic sculptors, and rogue planets
**Opportunities**
-----------------
* [Academic biotech founders: apply for Nucleate’s Activator program](http://apply.nucleate.xyz/) (via [@kulesatony](https://twitter.com/kulesatony/status/1689316164764065793))
**News & anno... | https://www.lesswrong.com/posts/CJNbF9hM8PtxkrHHT/progress-links-digest-2023-08-17-cloud-seeding-robotic |
# Memetic Judo #1: On Doomsday Prophets v.3
There is a popular tendency to dismiss people who are concerned about AI-safety as "doomsday prophets", carrying with it the suggestion that predicting an existential risk in the near future would automatically discredit them (because "you know; *they* have always been wrong... | https://www.lesswrong.com/posts/G3TjhYN8ZvFs5jASW/memetic-judo-1-on-doomsday-prophets-v-3 |
# Managing risks of our own work
*Note: This is not a personal post. I am sharing on behalf of the ARC Evals team.*
Potential risks of publication and our response
-----------------------------------------------
*This document expands on an appendix to ARC Evals’ paper, “*[*Evaluating Language-Model Agents on Realis... | https://www.lesswrong.com/posts/fARMR2tiyCem8DD35/managing-risks-of-our-own-work |
# Risks from AI Overview: Summary
We’ve recently published on our website a summary of our [paper on catastrophic risks from AI](https://arxiv.org/abs/2306.12001), which we are cross-posting here. We hope that this summary helps to make our research more accessible and to share our policy recommendations in a more con... | https://www.lesswrong.com/posts/9dNxz2kjNvPtiZjxj/risks-from-ai-overview-summary |
# 6 non-obvious mental health issues specific to AI safety
Intro
========
I am a psychotherapist, and I help people working on AI safety. I noticed patterns of mental health issues highly specific to this group. It's not just doomerism, there are way more of them that are less obvious.
If you struggle with a men... | https://www.lesswrong.com/posts/tpLzjWqG2iyEgMGfJ/6-non-obvious-mental-health-issues-specific-to-ai-safety |
# AI labs' requests for input
AI labs sometimes publicly ask for input on their actions, their products, and the future of AI. See OpenAI's [Democratic inputs to AI](https://openai.com/blog/democratic-inputs-to-ai) and [ChatGPT Feedback Contest](https://cdn.openai.com/chatgpt/ChatGPT_Feedback_Contest_Rules.pdf), and m... | https://www.lesswrong.com/posts/5e3bXsBQLusuArzmN/ai-labs-requests-for-input |
# Meetup Tip: Board Games
Summary
-------
Many meetups play board games or card games. Hanging out playing games can be a fun way to spend time together, and organizing a board game meetup is unusually easy since the games do most of the heavy lifting. Some games are better suited to this purpose than others; I belie... | https://www.lesswrong.com/posts/SZ2n5cjp6uZs8kfTT/meetup-tip-board-games |
# The U.S. is becoming less stable
We focus so much on arguing over who is at fault in this country that I think sometimes we fail to alert on what's actually happening. I would just like to point out, without attempting to assign blame, that American political institutions appear to be losing common knowledge of thei... | https://www.lesswrong.com/posts/r2vaM2MDvdiDSWicu/the-u-s-is-becoming-less-stable |
# Rationality-ish Meetups Showcase: 2019-2021
These events were run bi-weekly by myself as a tired and burnt out university student, and I didn't spend too much time on the event descriptions at the time so they're a bit sparse. At this point, the meetups were not yet explicitly rationalist-coded, and so it wasn't a p... | https://www.lesswrong.com/posts/3KcEcm4tpbHAFxnMb/rationality-ish-meetups-showcase-2019-2021 |
# Is Chinese total factor productivity lower today than it was in 1956?
*tl;dr: Multifactor productivity data from a famous economic dataset, used often to proxy for technological progress and innovation, might be significantly biased by poor estimates of the social returns to education, the capital share of income, a... | https://www.lesswrong.com/posts/4SsoZLYk6efXFWzY5/is-chinese-total-factor-productivity-lower-today-than-it-was |
# Could fabs own AI?
_Disclaimer: Very rough thoughts. I am nothing approaching an expert on this topic. Comments, corrections, and references appreciated._
Suppose there's an amazing general-purpose neural net architecture that learns as it runs and can be scaled indefinitely with returns that don't diminish too muc... | https://www.lesswrong.com/posts/brgBYhX58deu9fnmB/could-fabs-own-ai |
# Supervised Program for Alignment Research (SPAR) at UC Berkeley: Spring 2023 summary
In Spring 2023, the [Berkeley AI Safety Initiative for Students](https://berkeleyaisafety.com/) (BASIS) organized an alignment research program for students, drawing inspiration from similar programs by [Stanford AI Alignment](https... | https://www.lesswrong.com/posts/PXr38b64ECtFcn4Yq/supervised-program-for-alignment-research-spar-at-uc |
# We can do better than DoWhatIMean (inextricably kind AI)
_Edit February 2024: Now I think maybe we can't do better._
Lately, "alignment" means "follow admin rules and do what users mean". Admins can put in rules like "don't give instructions for bombs, hacking, or bioweapons" and "don't take sides in politics". As ... | https://www.lesswrong.com/posts/QQdmb3TjrvmB9Pfzp/we-can-do-better-than-dowhatimean-inextricably-kind-ai |
# Chess as a case study in hidden capabilities in ChatGPT
*(Edit: This post is out of date, and understates ChatGPT's chess abilities.* `gpt-3.5-turbo-instruct`*, when prompted correctly, plays consistently legal moves at around the 1800-2000 Elo level. Thanks to commenter GoteNoSente for pointing this out and to Clev... | https://www.lesswrong.com/posts/F6vH6fr8ngo7csDdf/chess-as-a-case-study-in-hidden-capabilities-in-chatgpt |
# Ten variations on red-pill-blue-pill
Here are some variations on the original red-pill-blue-pill question, recently discussed [here](https://www.lesswrong.com/posts/ZdEhEeg9qnxwFgPMf) and [here](https://www.lesswrong.com/posts/ogcEgWEgmzBw8ePMC).
In all cases, everyone makes their decision in ignorance of everyone ... | https://www.lesswrong.com/posts/bKz2DWWuyNoYHwz5m/ten-variations-on-red-pill-blue-pill |
# Will AI kill everyone? Here's what the godfathers of AI have to say [RA video]
This video is based on [this article](https://forum.effectivealtruism.org/posts/fGfXrbtBJJasA2EKj/most-leading-ai-experts-believe-that-advanced-ai-could-be). [@jai](https://forum.effectivealtruism.org/users/jai?mention=user) has written b... | https://www.lesswrong.com/posts/Fix3xmnERoqMuhKbd/will-ai-kill-everyone-here-s-what-the-godfathers-of-ai-have |
# Interpreting a dimensionality reduction of a collection of matrices as two positive semidefinite block diagonal matrices
Here are some empirical observations that I have made on August 14, 2023 to August 19, 2023 that are characteristics of the interpretability of my own matrix dimensionality reduction algorithm. Th... | https://www.lesswrong.com/posts/exNgLzdGja2P5wJ4j/interpreting-a-dimensionality-reduction-of-a-collection-of |
# Four management/leadership book summaries
Preface
=======
A while ago I read four books on leadership and management and wrote summaries of them. The summaries are slightly biased in that they are mostly what I found to be useful and how I interpreted the books. I don’t necessarily endorse the points made by these ... | https://www.lesswrong.com/posts/yN6PjeZk9ikxJcnbg/four-management-leadership-book-summaries |
# AI Forecasting: Two Years In
Two years ago, I commissioned forecasts for state-of-the-art performance on several popular ML benchmarks. Forecasters were asked to predict state-of-the-art performance on June 30th of 2022, 2023, 2024, and 2025. While there were four benchmarks total, the two most notable were [MATH](h... | https://www.lesswrong.com/posts/SdkexhiynayG2sQCC/ai-forecasting-two-years-in |
# How do I read things on the internet
> This is a linkpost - I recommend reading it at the [original URL](https://vlad.roam.garden/How-do-I-read-things-on-the-internet) for a better reading UX
I have a somewhat elaborate process for reading things that I find on the web. I've been inspired to share it because after ... | https://www.lesswrong.com/posts/y5ZWAZ6BJYnwbSrxK/how-do-i-read-things-on-the-internet |
# Call for Papers on Global AI Governance from the UN
Copied from their LinkedIn page:
📢 𝐂𝐚𝐥𝐥 𝐟𝐨𝐫 𝐏𝐚𝐩𝐞𝐫𝐬 𝐨𝐧 𝐆𝐥𝐨𝐛𝐚𝐥 𝐀𝐈 𝐆𝐨𝐯𝐞𝐫𝐧𝐚𝐧𝐜𝐞🌍
Exciting news! As we gear up for the High-level Advisory Body on AI, we're inviting thought leaders, researchers, and enthusiasts to contribute short pa... | https://www.lesswrong.com/posts/PK3GcrJcPeFiSo6Q9/call-for-papers-on-global-ai-governance-from-the-un |
# “Dirty concepts” in AI alignment discourses, and some guesses for how to deal with them
*Meta: This is a short summary & discussion post of* [*a talk on the same topic by Javier Gomez-Lavin*](https://youtu.be/B9Ujr6uKmpc)*, which he gave as part of the* [*PIBBSS*](https://pibbss.ai) *speaker series. The speaker seri... | https://www.lesswrong.com/posts/bBicgqvwjPbaQrJJA/dirty-concepts-in-ai-alignment-discourses-and-some-guesses |
# Jan Kulveit's Corrigibility Thoughts Distilled
*Many thanks to Jan for commenting on a draft of this post.*
There were a lot of great comments on "[Let's see you write that corrigibility tag](https://www.lesswrong.com/posts/AqsjZwxHNqH64C2b6/let-s-see-you-write-that-corrigibility-tag)". This is my attempt at expand... | https://www.lesswrong.com/posts/Lg4voqq4vTXiCJNQP/jan-kulveit-s-corrigibility-thoughts-distilled |
# Steven Wolfram on AI Alignment
Joe Walker has [a general conversation with Wolfram](https://josephnoelwalker.com/148-stephen-wolfram/) about his work and things and stuff, but there are some remarks about AI alignment at the very end:
> WALKER: Okay, interesting. So moving finally to AI, many people worry about una... | https://www.lesswrong.com/posts/osPFESHQKLa2X3adb/steven-wolfram-on-ai-alignment |
# Ruining an expected-log-money maximizer
Suppose you have a game where you can bet any amount of money. You have a 60% chance of doubling your stake and a 40% chance of losing it.
Consider agents Linda and Logan, and assume they both have £1^[1](#fn:gbp)^. Linda has a utility function that's linear in money (and has... | https://www.lesswrong.com/posts/JAmyTWoukk8xzhE9n/ruining-an-expected-log-money-maximizer |
# Efficiency and resource use scaling parity
An interesting pattern I've now noticed across many different domains is that if we try to do an attribution of improvements in outcomes or performance in a domain to the two categories "we're using more resources now than in the past" and "we're making more efficient use o... | https://www.lesswrong.com/posts/Q3XaZTExzDpCLr4wu/efficiency-and-resource-use-scaling-parity |
# DIY Deliberate Practice
In the spirit of growth and self-improvement, I recently attempted to apply Ericsson’s principles of deliberate practice to my own growth goal: speeding up my writing. If you're unfamiliar with the minutia of Ericsson's methods, don't worry, I was in the same boat — and hence my initial goal ... | https://www.lesswrong.com/posts/oZnPabN2CQQdnAo63/diy-deliberate-practice |
# Self-shutdown AI
*This post describes my proposal for the* [*AI Alignment Awards*](https://www.lesswrong.com/posts/zFoAAD7dfWdczxoLH/winners-of-ai-alignment-awards-research-contest) *contest that won a "first round prize" / an honorable mention. The original submission is* [*here*](https://github.com/johny-b/aaa_23_... | https://www.lesswrong.com/posts/piAnXc2a4k5bFsKjL/self-shutdown-ai |
# Which possible AI systems are relatively safe?
Presumably some kinds of AI systems, architectures, methods, and *ways of building complex systems out of ML models* are safer or more alignable than others. Holding capabilities constant, you'd be happier to see some kinds of systems than others.
For example, Paul Chr... | https://www.lesswrong.com/posts/Djgws2Moi7Zefkj5y/which-possible-ai-systems-are-relatively-safe |
# Large Language Models will be Great for Censorship
*Produced as part of the* [*SERI ML Alignment Theory Scholars Program*](https://serimats.org/) *\- Summer 2023 Cohort*
*Thanks to* [*ev_*](https://www.lesswrong.com/users/ev_) *and* [*Kei*](https://www.lesswrong.com/users/kei-1) *for suggestions on this post.*
LLM... | https://www.lesswrong.com/posts/oqvsR2LmHWamyKDcj/large-language-models-will-be-great-for-censorship |
# Ideas for improving epistemics in AI safety outreach
In 2022 and 2023, there has been a growing focus on recruiting talented individuals to work on mitigating the potential existential risks posed by artificial intelligence. For example, we’ve seen an increase in the number of university clubs, retreats, and worksho... | https://www.lesswrong.com/posts/SDpaZ7MdH5yRnobrZ/ideas-for-improving-epistemics-in-ai-safety-outreach |
# Causality and a Cost Semantics for Neural Networks
$\textbf{Epistemic status:}$ I time-boxed this idea to three days of effort. So any calculations are pretty sloppy, and I haven't looked into any related works. I probably could have done much better if I knew anything about circuit complexity. There are some TODOs ... | https://www.lesswrong.com/posts/zkfmhWQXsZweijmzi/causality-and-a-cost-semantics-for-neural-networks |
# Seth Explains Consciousness
The Real Problem
----------------
For as long as there have been philosophers, they loved philosophizing about what *life really is*. Plato focused on nutrition and reproduction as the core features of living organisms. Aristotle claimed that it was ultimately about resisting perturbatio... | https://www.lesswrong.com/posts/FQhtpHFiPacG3KrvD/seth-explains-consciousness |
# State of Generally Available Self-Driving
After a lot of discussions online where people try to argue about where self-driving tech is going but seem pretty confused about where the tech currently is, I wanted to give a bit of an overview of the current state.
There are two main approaches: taxis and personal vehic... | https://www.lesswrong.com/posts/3KdbnsGnzSyiujz7Z/state-of-generally-available-self-driving |
# Walk while you talk: don't balk at "no chalk"
Some days ago, after a long conversation with a friend, I took retrospective notes, recalling what we discussed. (I wrote over 50 items. This is not a small sample.) For parts of the time, we walked; for other parts, we stayed in roughly the same place. Writing the notes... | https://www.lesswrong.com/posts/trZSdr49bygabxQtG/walk-while-you-talk-don-t-balk-at-no-chalk |
# Implications of evidential cooperation in large worlds
I've written several posts about the plausible implications of "evidential cooperation in large worlds" (ECL), on my newly-revived blog. This is a cross-post of [the first](https://lukasfinnveden.substack.com/p/implications-of-ecl). If you want to see the rest o... | https://www.lesswrong.com/posts/EeXSjvyQge5FZPeuL/implications-of-evidential-cooperation-in-large-worlds |
# Do agents with (mutually known) identical utility functions but irreconcilable knowledge sometimes fight?
Been pondering; will conflict always exist? A major subquestion: Suppose we all merge utility functions and form an interstellar community devoted to optimizing the merger. It'll probably make sense for us to sp... | https://www.lesswrong.com/posts/FTdtHHBPDdzk4pJz2/do-agents-with-mutually-known-identical-utility-functions |
# Why Is No One Trying To Align Profit Incentives With Alignment Research?
A whole lot of Alignment work seems to be resource-constrained. Many funders have talked about how they were only able to give grants to a small percentage of projects and work they found promising. Many researchers also receive a small fractio... | https://www.lesswrong.com/posts/XnAZHCi5QH58WcrkX/why-is-no-one-trying-to-align-profit-incentives-with |
# A Theory of Laughter
*(Last revised: March 2026. See changelog at the bottom.)*
1\. tl;dr
=========
There should be parallel explanations for laughter at two levels.
* At the brain level, there should be some mechanism / algorithm that produces laughter, and it should fit the data of when people laugh in practi... | https://www.lesswrong.com/posts/7kdBqSFJnvJzYTfx9/a-theory-of-laughter |
# A problem with the most recently published version of CEV
This post outlines a thought experiment, designed to illustrate a problem with the Parliamentarian version of CEV (PCEV). It then shows that the problem is not inherent in the scenario of the thought experiment, by describing a large space of other designs, t... | https://www.lesswrong.com/posts/LRuaCRpTc2zMhDDbb/a-problem-with-the-most-recently-published-version-of-cev |
# The Low-Hanging Fruit Prior and sloped valleys in the loss landscape
*You can find code for the referenced experiments in* [*this GitHub repository*](https://github.com/nrimsky/mlexperiments)
Many have postulated that training large neural networks will enforce a simplicity, or[ Solomonoff prior](https://www.lesswr... | https://www.lesswrong.com/posts/SbzptgFYr272tMbgz/the-low-hanging-fruit-prior-and-sloped-valleys-in-the-loss |
# Diet Experiment Preregistration: Long-term water fasting + seed oil removal
I am intrigued by exfatloss's [recent theory](https://exfatloss.substack.com/p/seed-oils-explain-the-8-mysteries) that the primary cause of the obesity epidemic is slow, subtle damage from linoleic acid over many years. Specifically, I want... | https://www.lesswrong.com/posts/5KkyygHQL2XCw3kTs/diet-experiment-preregistration-long-term-water-fasting-seed |
# China's position on autonomous weapons
> While continuing to endorse the necessity to develop a legally binding instrument with respect to LAWS, in 2018 China also proposed a somewhat narrow definition of LAWS comprising five essential features:
>> The first is lethality, which means sufficient payload (charge) fo... | https://www.lesswrong.com/posts/QaaunwFaGGFd8cTpy/china-s-position-on-autonomous-weapons |
# Regreasing a KitchenAid Mixer
I recently got an old KitchenAid mixer from a friend. It's a k45ss, from maybe the early 80s, and the general advice I see is that with typical home use they should be regreased every ~10y. I ordered 4oz of food-safe synthetic [grease](https://www.amazon.com/dp/B08ZYCQVRD) and followed ... | https://www.lesswrong.com/posts/5pcunrT9fJJ3hmHvu/regreasing-a-kitchenaid-mixer |
# The lost millennium
*Note: This post is speculative, and you should take the claims in it with a grain of salt.*
Throughout human history, we know that economic growth, for which population growth was a good proxy under Malthusian conditions, has accelerated many times. Indeed, we can see this just by extrapolating... | https://www.lesswrong.com/posts/hgf6FB9jMB7wMLuKA/the-lost-millennium |
# I measure Google's MusicLM over 3 months as it appears to go from jaw-dropping to embarrassingly repeating itself
Google Research submitted a paper on January 26 2023 for MusicLM, a mind-bogglingly incredibly powerful AI model that converts user text prompts into music.
https://google-research.github.io/seanet/musi... | https://www.lesswrong.com/posts/DfqcyGXcFcukYbWZ5/i-measure-google-s-musiclm-over-3-months-as-it-appears-to-go |
# The God of Humanity, and the God of the Robot Utilitarians
My personal religion involves two gods – the god of humanity (who I sometimes call "Humo") and the god of the robot utilitarians (who I sometimes call "Robutil").
When I'm facing a moral crisis, I query my [shoulder](https://www.lesswrong.com/posts/X79Rc5c... | https://www.lesswrong.com/posts/SfZRWxktiFFJ5FNk8/the-god-of-humanity-and-the-god-of-the-robot-utilitarians |
# What wiki-editing features would make you use the LessWrong wiki more?
The LessWrong wiki does not seem as well used as it should be. I guess this is a lack of editors.
[@Vladimir_Nesov](https://www.lesswrong.com/users/vladimir_nesov?mention=user) made a good point that many standard wiki-editing features are missi... | https://www.lesswrong.com/posts/xzm6F6rLt7oPTash2/what-wiki-editing-features-would-make-you-use-the-lesswrong |
# AI Regulation May Be More Important Than AI Alignment For Existential Safety
***Summary**: Aligning a single powerful AI is not enough: we're only safe if no-one, ever, can build an unaligned powerful AI. Yudkowsky tried to solve this with the pivotal act: the first aligned AI does something (such as melting all GPU... | https://www.lesswrong.com/posts/2cxNvPtMrjwaJrtoR/ai-regulation-may-be-more-important-than-ai-alignment-for |
# AI Safety Bounties
Earlier this year, Vaniver recommended [Bug Bounties for Advanced ML Systems](https://www.lesswrong.com/posts/5dKDLv4knhXLvNHT5/recommendation-bug-bounties-and-responsible-disclosure-for).
I spent a while at Rethink Priorities considering and expanding on this idea, suggesting potential progr... | https://www.lesswrong.com/posts/iXECTEyC5PQuYM2aJ/ai-safety-bounties |
# AI #26: Fine Tuning Time
GPT-3.5 fine tuning is here. GPT-4 fine tuning is only a few months away. It is about to get a lot easier to get a powerful system that does what you want it to do, and knows what you want it to know, especially for the purposes of a business or a website.
As an experiment, I am putting in ... | https://www.lesswrong.com/posts/kLa3HmkesF5w3MFEY/ai-26-fine-tuning-time |
# Nuclear consciousness
A passage from Mark Twain's *Connecticut Yankee in King Arthur's Court* has been stuck in my head since I read it some 35 years ago:
> As for me, all that I think about in this plodding sad pilgrimage, this pathetic drift between the eternities, is to look out and humbly live a pure and high a... | https://www.lesswrong.com/posts/qXZrobGdAycNDBhET/nuclear-consciousness |
# Apply for the 2023 Developmental Interpretability Conference!
**What**: A conference on developmental interpretability and singular learning theory.
**When**: 5-12 November 2023
**Where**: Wytham Abbey, Oxford
**How:** [Apply now!](https://forms.gle/MB3yeiJA4QkY9K5o6)
**Edit**: We're full! If you're still intere... | https://www.lesswrong.com/posts/QpFiEbqMdhaLBPb7X/apply-for-the-2023-developmental-interpretability-conference |
# What AI Posts Do You Want Distilled?
I'd like to distill AI Safety posts and papers, and I'd like to see more distillations generally. Ideally, posts and papers would meet the following criteria:
* Potentially high-impact for more people to understand
* Uses a lot of jargon or is generally complex and difficult... | https://www.lesswrong.com/posts/wdvGgEGMohpZcgASP/what-ai-posts-do-you-want-distilled |
# A Model-based Approach to AI Existential Risk
Introduction
============
Polarisation hampers cooperation and progress towards understanding whether future AI poses an existential risk to humanity and how to reduce the risks of catastrophic outcomes. It is exceptionally challenging to pin down what these risks are a... | https://www.lesswrong.com/posts/sGkRDrpphsu6Jhega/a-model-based-approach-to-ai-existential-risk |
# Assume Bad Faith
I've been trying to avoid the terms "good faith" and "bad faith". I'm suspicious that most people who have picked up the phrase "bad faith" from hearing it used, don't actually know what it means—and maybe, that the thing it does mean doesn't [carve reality at the joints](https://www.lesswrong.com/p... | https://www.lesswrong.com/posts/e4GBj6jxRZcsHFSvP/assume-bad-faith |
# GPT-powered EA/LW weekly summary
Skip to the summaries: [LW](https://www.lesswrong.com/posts/kjsDpnfGrArHhbtKQ/gpt-powered-ea-lw-weekly-summary#LessWrong) | [EAF](https://www.lesswrong.com/posts/kjsDpnfGrArHhbtKQ/gpt-powered-ea-lw-weekly-summary#EA forum)
[Originally posted on the EA forum](https://forum.effectivea... | https://www.lesswrong.com/posts/kjsDpnfGrArHhbtKQ/gpt-powered-ea-lw-weekly-summary |
# Hoodwinked: Evaluating Deception Capabilities in Large Language Models
*This is a linkpost for* [*https://arxiv.org/abs/2308.01404*](https://arxiv.org/abs/2308.01404)
Evaluating the hazardous capabilities of AI systems is useful for prioritizing technical research and informing policy decisions. To better evaluate ... | https://www.lesswrong.com/posts/eXNyqj9AgrKYGDqDn/hoodwinked-evaluating-deception-capabilities-in-large |
# When Omnipotence is Not Enough
Some knowledge cannot be gifted or received. It can only be stolen.
---
In the east, there was a garden. In the garden there was a God, a man, a woman and a snake. They all had legs.
"Did God really tell you, 'You must not eat from any tree in the garden'?" asked the snake.
"We may... | https://www.lesswrong.com/posts/r8w2SxmYqgssdjWwd/when-omnipotence-is-not-enough |
# Using Points to Rate Different Kinds of Evidence
*Epistemic Status: Briefly written. The specific equation here captures my quick intuition - this is meant primarily as a demonstration. *
There’s a lot of discussion on the EA Forum and LessWrong about epistemics, evidence, and updating.
I don’t know of many attem... | https://www.lesswrong.com/posts/6DuAddMcrv2jfQ5CW/using-points-to-rate-different-kinds-of-evidence |
# Lemonade Stand Costs?
My kids were thinking of having a lemonade stand, and in the past I've said that if they're doing it as a fundraiser we'll cover the cost of ingredients, but if they're doing it for their own money then they need to reimburse us. But how much should that be?
The recipe they're using is approxi... | https://www.lesswrong.com/posts/zAeEZ2WCwSkmC62zw/lemonade-stand-costs |
# Will an Overconfident AGI Mistakenly Expect to Conquer the World?
I'm wondering how selection effects will influence the first serious
attempt by an AGI to take over the world.
My question here is inspired by thoughts about people who say AGI
couldn't conquer the world because it will depend on humans to provid... | https://www.lesswrong.com/posts/ibknYnn8tf2ppXvxJ/will-an-overconfident-agi-mistakenly-expect-to-conquer-the |
# ACX Meetups Everywhere 2023: Times & Places
Thanks to everyone who responded to my request for ACX meetup organizers. Volunteers have arranged meetups in 169 cities around the world, from Baghdad to Bangalore to Buenos Aires.
You can find the list below, in the following order:
Africa & Middle East
Asia-Pacific
... | https://www.lesswrong.com/posts/ynpC7oXhXxGPNuCgH/acx-meetups-everywhere-2023-times-and-places |
# Chariots of Philosophical Fire
Sharing because I feel that the contribution of analytical philosophy is sometimes under appreciated here, particularly compared to other attempts at philosophy:
“In grappling with these mysteries, Oxford philosophers developed and refined old and new techniques. Reasoning, deduction,... | https://www.lesswrong.com/posts/TLbjABw6qQHo46p8v/chariots-of-philosophical-fire |
# Digital brains beat biological ones because diffusion is too slow
I've spent quite a bit of time thinking about the possibility of genetically enhancing humans to be smarter, healthier, more likely to care about others, and just generally better in ways that most people would recognize as such.
As part of this rese... | https://www.lesswrong.com/posts/MR7vY4ztw5L2pcDuj/digital-brains-beat-biological-ones-because-diffusion-is-too |
# On Ilya Sutskever's "A Theory of Unsupervised Learning"
*I highly recommend watching Sutskever's entire lecture instead of reading the AI-generated transcript*[^5aj106k53w7]*, as he presents his ideas exceptionally well. I believe Sutskever's effort to formalize or demistify unsupervised learning is a solution to t... | https://www.lesswrong.com/posts/KqgujtM3vSAfZE2dR/on-ilya-sutskever-s-a-theory-of-unsupervised-learning |
# Red-teaming language models via activation engineering
*Produced as part of the* [*SERI ML Alignment Theory Scholars Program*](https://serimats.org/) *\- Summer 2023 Cohort, under the mentorship of Evan Hubinger.*
Evaluating powerful AI systems for hidden functionality and out-of-distribution behavior is hard. In t... | https://www.lesswrong.com/posts/iHmsJdxgMEWmAfNne/red-teaming-language-models-via-activation-engineering |
# A list of core AI safety problems and how I hope to solve them
*Context: I sometimes find myself referring back to* [*this tweet*](https://x.com/davidad/status/1660547794627641345?s=46&t=Lap-izdY_pwXQfbJXZHE8g) *and wanted to give it a more permanent home. While I'm at it, I thought I would try to give a concise sum... | https://www.lesswrong.com/posts/mnoc3cKY3gXMrTybs/a-list-of-core-ai-safety-problems-and-how-i-hope-to-solve |
# Aumann-agreement is common
*Thank you to Justis Mills for proofreading and feedback. This post is also available on* [*my substack*](https://tailcalled.substack.com/p/aumann-agreement-is-common)*.*
[Aumann's agreement theorem](https://www.lesswrong.com/tag/aumann-s-agreement-theorem) is a family of theorems which s... | https://www.lesswrong.com/posts/4S6zunFNFY3f5JYxt/aumann-agreement-is-common |
# Mesa-Optimization: Explain it like I'm 10 Edition
There are a [couple](https://www.lesswrong.com/posts/XWPJfgBymBbL3jdFd/an-58-mesa-optimization-what-it-is-and-why-we-should-care) [of](https://ui.stampy.ai?state=8160_) [explanations](https://www.lesswrong.com/tag/mesa-optimization) of mesa-optimization available. I ... | https://www.lesswrong.com/posts/fLAvmWHmpJEiw8KEp/mesa-optimization-explain-it-like-i-m-10-edition |
# The Game of Dominance
Concerns about AI safety rely on the assumption that a sufficiently powerful AI might take control of our future in an undesirable way. Meta’s head of AI, Yann LeCun, correctly points this out [in a recent tweet](https://twitter.com/ylecun/status/1695056787408400778), but then argues that this ... | https://www.lesswrong.com/posts/NCDakH4nZrS9qeuL6/the-game-of-dominance |
# Apply to a small iteration of MLAB to be run in Oxford
TLDR: We’re running a small iteration of MLAB (~10 participants) in Oxford towards the end of September. If you’re interested in participating, apply [here](https://forms.gle/QV9z6cxNGnS1UJzq6) by 7 September. If you’re interested in being a TA, please email us ... | https://www.lesswrong.com/posts/k5anbk2pBZPFkrCqh/apply-to-a-small-iteration-of-mlab-to-be-run-in-oxford |
# Trying a Wet Suit
I get cold very quickly in the water, enough that unless it's close to body temperature I get chilled through within ~15min. This mostly wasn't a problem, because I'd take a quick dip to cool off and then hang out on the beach, but now that I have kids they (and I) want lots of swimming together ti... | https://www.lesswrong.com/posts/ZrQcLpL59Frjr3vE5/trying-a-wet-suit |
# High level overview on how to go about estimating "p(doom)" or the like
This is from a comment I left on a [recent blog post](https://garymarcus.substack.com/p/d28) from Gary Marcus, but I thought it might be of general interest here. Feedback and critiques definitely welcome!
\-\-\-\-\-\-
For what it's worth, her... | https://www.lesswrong.com/posts/ojiegFzywtpsc29QB/high-level-overview-on-how-to-go-about-estimating-p-doom-or |
# AI pause/governance advocacy might be net-negative, especially without a focus on explaining x-risk
I found myself repeating the same words to multiple people, hence a short post.
I think some of the AI pause/governance advocacy might be net-negative. Three reasons:
* Most importantly, it's easy to get regulatio... | https://www.lesswrong.com/posts/jiYLFomPPePy85eN8/ai-pause-governance-advocacy-might-be-net-negative |
# Dear Self; we need to talk about ambition
I keep seeing advice on ambition, aimed at people in college or early in their career, that would have been really bad for me at similar ages. Rather than contribute ([more](https://forum.effectivealtruism.org/posts/aXiLprnMGKuxTXcpF/elizabeth-s-quick-takes?commentId=cG6Fqhe... | https://www.lesswrong.com/posts/uGDtroD26aLvHSoK2/dear-self-we-need-to-talk-about-ambition-1 |
# No More Freezer Pucks
A while ago I wrote about how I'd make [freezer pucks](https://www.jefftk.com/p/frozen-lunch-pucks):
[](https://www.jefftk.com/p/frozen-lunch-pucks)
These work ok, but there are a couple issues:
* A large batch requires a lot of muffin tins... | https://www.lesswrong.com/posts/hhoLhMqbhLF3oWaee/no-more-freezer-pucks |
# The Evidence for Question Decomposition is Weak
Question decomposition appears to be a relatively common method for forecasting, see [Allyn-Feuer & Sanders 2023](https://forum.effectivealtruism.org/posts/ARkbWch5RMsj6xP5p/transformative-agi-by-2043-is-less-than-1-likely), [Silver 2016](http://fivethirtyeight.com/fea... | https://www.lesswrong.com/posts/YjZ8sJmkGJQhNcjHj/the-evidence-for-question-decomposition-is-weak |
# Introducing the Center for AI Policy (& we're hiring!)
### **Summary**
* [The Center for AI Policy](https://www.aipolicy.us/) is a new organization developing and advocating for policy to mitigate catastrophic AI risks.
* We are hiring for a [Government Affairs Director](https://www.aipolicy.us/careers/governme... | https://www.lesswrong.com/posts/unwRBRQivd2LYRfuP/introducing-the-center-for-ai-policy-and-we-re-hiring |
# OpenAI API base models are not sycophantic, at any size
In [Discovering Language Model Behaviors with Model-Written Evaluations" (Perez et al 2022)](https://cdn2.assets-servd.host/anthropic-website/production/files/model-written-evals.pdf), the authors studied language model "sycophancy" - the tendency to agree with... | https://www.lesswrong.com/posts/3ou8DayvDXxufkjHD/openai-api-base-models-are-not-sycophantic-at-any-size |
# An Interpretability Illusion for Activation Patching of Arbitrary Subspaces
*Produced as part of the *[*SERI ML Alignment Theory Scholars Program*](https://serimats.org/) *\- Summer 2023 Cohort*
*We would like to thank* [*Atticus Geiger*](https://atticusg.github.io/) *for his valuable feedback and in-depth discussi... | https://www.lesswrong.com/posts/RFtkRXHebkwxygDe2/an-interpretability-illusion-for-activation-patching-of |
# AI Deception: A Survey of Examples, Risks, and Potential Solutions
$$By Peter S. Park, Simon Goldstein, Aidan O’Gara, Michael Chen, and Dan Hendrycks
\[This post summarizes our new report on AI deception, available [**here**](https://arxiv.org/abs/2308.14752)\]
**Abstract:** This paper argues that a range of curre... | https://www.lesswrong.com/posts/YgAKhkBdgeTCn6P53/ai-deception-a-survey-of-examples-risks-and-potential |
# Incentives affecting alignment-researcher encouragement
My hypothesis: I think the incentives for "cultivating more/better researchers in a preparadigmatic field" lean towards "don't discourage even less-promising researchers, because they could luck out and suddenly be good/useful to alignment in an unexpected way"... | https://www.lesswrong.com/posts/wBgjQKNfJnPMjKpFa/incentives-affecting-alignment-researcher-encouragement |
# Newcomb Variant
There are two boxes. If you open a box, you get to keep the contents.
Omega is an omniscient oracle. Omega predicts what will happen if you open the first box and observe that it contains $100.
- If Omega predicts that you will open the second box, then Omega puts $0 in both boxes.
- If Omega predic... | https://www.lesswrong.com/posts/jzfLjhgMrCb5sE2Go/newcomb-variant |
# Barriers to Mechanistic Interpretability for AGI Safety
I gave a talk at MIT in March earlier this year on barriers to mechanistic interpretability being helpful to AGI/ASI safety, and why by default it will likely be net dangerous. Several people seem to be coming to similar conclusions recently (e.g., [this recent... | https://www.lesswrong.com/posts/KRDo2afKJtD7bzSM8/barriers-to-mechanistic-interpretability-for-agi-safety |
# Dating Roundup #1: This is Why You’re Single
Developments around relationships and dating have a relatively small speed premium, so I figured I would wait until I had a full post worth of them.
Indeed I now present such a post, in which I present several theories as to why so many of you might still be single.
Whi... | https://www.lesswrong.com/posts/BAqvAvPC7GiZhTyR3/dating-roundup-1-this-is-why-you-re-single |
# Loft Bed Fan Guard
I really like ceiling fans (big fan!) and we have them in most of the rooms of our house. About a year ago we also [built loft beds](https://www.jefftk.com/p/building-the-loft-beds) for the older two kids bedrooms:
[](https://www.jefftk.com/p/... | https://www.lesswrong.com/posts/7v4wDLvjcNxJfX6Qr/loft-bed-fan-guard |
# AISN #20: LLM Proliferation, AI Deception, and Continuing Drivers of AI Capabilities
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
Subscribe [here](https://newsletter.safe.ai/subscribe?utm_medium... | https://www.lesswrong.com/posts/j2ZnY4sgJqu8FAqGd/aisn-20-llm-proliferation-ai-deception-and-continuing |
# Broken Benchmark: MMLU
Phillip over at the AI Explained channel has been running some experiments on his SmartGPT framework against the MMLU benchmark and discovered a not-insignificant amount of issues with the problem set.
Among them:
* Crucial context missing from questions (apparently copy-paste errors?)
* ... | https://www.lesswrong.com/posts/rQBaftqKMfG2uMiWb/broken-benchmark-mmlu |
# Democratic Fine-Tuning
*The project below, “Democratic Fine-tuning with a Moral Graph” (*$DFT_{mg}$*), is a winner of* [*the OpenAI democratic process grant*](https://openai.com/blog/democratic-inputs-to-ai)*. It is an alternative to* [*Constitutional AI*](https://arxiv.org/abs/2212.08073) *or* [*simple RLHF-based a... | https://www.lesswrong.com/posts/ncb2ycEB3ymNqzs93/democratic-fine-tuning |
# The Economics of the Asteroid Deflection Problem (Dominant Assurance Contracts)
**Update 2023-09-04:** [Partial Update: I'm fully funded!](https://www.lesswrong.com/posts/CwgHX9tbfASqxjpsc/the-economics-of-the-asteroid-deflection-problem-dominant?commentId=QGvcggzBpFtRhhYyt)
**Update 2023-09-21:** [Full Update. Fun... | https://www.lesswrong.com/posts/CwgHX9tbfASqxjpsc/the-economics-of-the-asteroid-deflection-problem-dominant |
# An OV-Coherent Toy Model of Attention Head Superposition
**Background**
This project was inspired by Anthropic’s [post](https://transformer-circuits.pub/2023/may-update/index.html) on attention head superposition, which constructed a toy model trained to learn a circuit to identify skip-trigrams that are OV-incoher... | https://www.lesswrong.com/posts/cqRGZisKbpSjgaJbc/an-ov-coherent-toy-model-of-attention-head-superposition-1 |
# Paper Walkthrough: Automated Circuit Discovery with Arthur Conmy
Arthur Conmy's [Automated Circuit Discovery](https://arxiv.org/pdf/2304.14997.pdf) is a great paper that makes initial forays into automating parts of mechanistic interpretability (specifically, automatically finding a sparse subgraph for a circuit). I... | https://www.lesswrong.com/posts/FzqKXpTDaouMF6Chj/paper-walkthrough-automated-circuit-discovery-with-arthur |
# LTFF and EAIF are unusually funding-constrained right now
### **UPDATE 2023/09/13: **
Including only money that has already landed in our bank account and extremely credible donor promises of funding, LTFF has raised ~1.1M and EAIF has raised ~500K. After Open Phil matching, this means LTFF now has ~3.3M additional... | https://www.lesswrong.com/posts/gRfy2Q2Pg25a2cHyY/ltff-and-eaif-are-unusually-funding-constrained-right-now |
# Open Call for Research Assistants in Developmental Interpretability
We are excited to announce multiple positions for Research Assistants to join our [six-month research project](https://manifund.org/projects/scoping-developmental-interpretability-xg55b33wsfc) assessing the viability of [Developmental Interpretabili... | https://www.lesswrong.com/posts/DZHmEmzujfuqfxbJY/open-call-for-research-assistants-in-developmental |
# "Wanting" and "liking"
*Written as a result of AI Safety Camp Virtual 2023. Thanks to the following people for feedback and helpful conversations: Oliver Bridge, Tim Gothard, Rasmus Jensen, Linda Linsefors.*^[Ordered alphabetically, by last name.]
This post reviews the literature on *"wanting"* and *"liking"*, two ... | https://www.lesswrong.com/posts/opJxxfrN33xQx3eXu/wanting-and-liking |
# Why I hang out at LessWrong and why you should check-in there every now and then
As the title indicates, this post, which in [cross-posting from my home blog, New Savanna](https://new-savanna.blogspot.com/2023/08/why-i-hang-out-at-lesswrong-and-why-you.html), is not directed at a LessWrong audience. However,, some ... | https://www.lesswrong.com/posts/nd9qnKxmYPr76Yi5s/why-i-hang-out-at-lesswrong-and-why-you-should-check-in |
# Biosecurity Culture, Computer Security Culture
While I've only worked in biosecurity for [about a year](https://www.jefftk.com/p/leaving-google-joining-the-nucleic-acid-observatory) and my computer security background consists of things I picked up while working on other aspects of software engineering, the cultures... | https://www.lesswrong.com/posts/bTteXdzcpAsGQE5Qc/biosecurity-culture-computer-security-culture |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.