text stringlengths 300 320k | source stringlengths 52 154 |
|---|---|
# Analysis of key AI analogies
The following is an analysis of seven prominent AI analogies: aliens, the brain, climate change, electricity, the Industrial Revolution, the neocortex, & nuclear fission. You can find longer versions of these as separate blogposts on [my substack](https://machinocene.substack.com/).
**0... | https://www.lesswrong.com/posts/srG2RguMfp6ubGnJa/analysis-of-key-ai-analogies |
# A “Scaling Monosemanticity” Explainer
*Coauthored by Fedor Ryzhenkov and Dmitrii Volkov (Palisade Research)*
*At Palisade, we often discuss latest safety results with policymakers and think tanks who seek to understand the state of current technology. This document condenses and streamlines the various internal not... | https://www.lesswrong.com/posts/wg6E3oJJrNnmJezNz/a-scaling-monosemanticity-explainer |
# Datasets that change the odds you exist
1.
--
It’s October 1962. The Cuban missile crisis just happened, thankfully without apocalyptic nuclear war. But still:
* Apocalyptic nuclear war easily *could* have happened.
* Crises as serious as the Cuban missile crisis clearly aren’t *that* rare, since one just happ... | https://www.lesswrong.com/posts/RfC4mkYuLksukyzns/datasets-that-change-the-odds-you-exist |
# My 5-step program for losing weight
I can’t claim to be an expert on the subject of weight loss but wanted to share the steps I’ve followed to get from ~81 kg (178 pounds) to ~70 kg (154 pounds) over a period of 8 months. Several people have asked me for tips, so I’ve figured I might as well write it all down for fu... | https://www.lesswrong.com/posts/Ejt6yHChar5TmxhfP/my-5-step-program-for-losing-weight |
# LLMs Universally Learn a Feature Representing Token Frequency / Rarity
Summary
-------
* LLMs appear to *universally* learn a feature in their embeddings representing the frequency / rarity of the tokens they were trained on
* This feature is observed across model sizes, in base models, instruction tuned models... | https://www.lesswrong.com/posts/rofQseMb5ojESyASw/llms-universally-learn-a-feature-representing-token |
# Review of METR’s public evaluation protocol
**TLDR:** We reviewed METR’s “Example evaluation protocol” and found a couple of points where there is room for improvement or unclear information. We also make a couple of suggestions around scoring, outsourcing, etc.
* This review was done by 2 people without previous... | https://www.lesswrong.com/posts/orKjt6TneueAezyio/review-of-metr-s-public-evaluation-protocol |
# In Defense of Lawyers Playing Their Part
*This is a linkpost for In* [*Defense of Lawyers Playing Their Part*](https://outsidetheasylum.blog/in-defense-of-lawyers-playing-their-part/).
[Michael Huemer](https://en.wikipedia.org/wiki/Michael_Huemer) writes about [why he believes it’s wrong for lawyers to pursue unjus... | https://www.lesswrong.com/posts/FMKnFxgbtCLxPPS4J/in-defense-of-lawyers-playing-their-part |
# Important open problems in voting
**Strategy-resistance**
Identify, or prove impossibility, of a voting system which incentivizes—
1. A *strictly* sincere ranking of all candidates in the zero-information setting, where it implements a "good" social choice rule such as the relative (normalized) utilitarian rule, ... | https://www.lesswrong.com/posts/HJp3C3z8XefwBeQcR/important-open-problems-in-voting |
# Probabilistic Logic <=> Oracles?
*Epistemic status: this is a draft I wrote at the end of MATS that I decided to make public in case that people with more experience with this machinery wanted to give constructive feedback. Is very unpolished!!! And likely quite very wrong in some cases / makes false claims (if you... | https://www.lesswrong.com/posts/HZusY5jPYvpmkTkZD/probabilistic-logic-less-than-greater-than-oracles |
# Whirlwind Tour of Chain of Thought Literature Relevant to Automating Alignment Research.
**This post is inspired by a series of comments by Bogdan: **[initial comment](https://www.lesswrong.com/posts/HmQGHGCnvmpCNDBjc/current-ais-provide-nearly-no-data-relevant-to-agi-alignment#mcA57W6YK6a2TGaE2)[follow-up 1](https:... | https://www.lesswrong.com/posts/WtsmwfzedJfSccynt/whirlwind-tour-of-chain-of-thought-literature-relevant-to |
# How good are LLMs at doing ML on an unknown dataset?
I just ran two evaluation tests on each of the three leading LLM chatbots, GPT4o, Claude Sonnet 3.5 and Gemini advanced. In the challenge the models were presented with a novel dataset, and were asked to develop a ML model to do supervised classification of the d... | https://www.lesswrong.com/posts/Fr6eJkjYWG9Mw6XQc/how-good-are-llms-at-doing-ml-on-an-unknown-dataset |
# Rationalists As People Who Build Piles Of Rocks
I recently attended the LessOnline event at the LightHaven campus.
It was fantastic, and I’ve got plenty of posts coming down the pipeline sparked by conversations or sessions held during the event. I got to meet a lot of my own personal heroes.
It was great.
Someth... | https://www.lesswrong.com/posts/yXkw4nrrTKwQFjzDK/rationalists-as-people-who-build-piles-of-rocks |
# Self-censoring on AI x-risk discussions?
I catch myself sometimes thinking of ideas / scenarios that support higher p(doom), typically as counter-examples to points folks make for lower p(doom), and I wonder how much self-censorship I should apply, given that AI can read these conversations.
My CoT:
1. I sure don... | https://www.lesswrong.com/posts/7aJ9vdqjq6HZjomRi/self-censoring-on-ai-x-risk-discussions |
# New Executive Team & Board — PIBBSS
TLDR:[*PIBBSS*](http://pibbss.ai) *is changing its core team. Nora is stepping down as director due to joining ARIA, and Lucas Teixeira and Dusan Nesic are taking over her leadership role. Nora joins the board, alongside Tan Zhi Xuan, Alexander Gietelink Oldenziel, Ben Goldhaber ... | https://www.lesswrong.com/posts/MqDoZtMZYckCpZGSS/new-executive-team-and-board-pibbss |
# Interpreting Preference Models w/ Sparse Autoencoders
This is the real reward output for an OS preference model. The bottom "jailbreak" completion was manually created by looking at reward-relevant SAE... | https://www.lesswrong.com/posts/5XmxmszdjzBQzqpmz/interpreting-preference-models-w-sparse-autoencoders |
# You Have Now Subscribed To Bug Facts
*Programming note: This meetup is taking place at a non-standard time (6:45) and location (Columbia Lake); plan accordingly!*
This Thursday, we'll be meeting at the Columbia Lake disc golf course. Bring water and bug spray and dress appropriately for getting real close to natur... | https://www.lesswrong.com/events/9n5LHtZtGujRHtLxg/you-have-now-subscribed-to-bug-facts |
# Covert Malicious Finetuning
This post discusses our recent paper [Covert Malicious Finetuning: Challenges in Safeguarding LLM Adaptation](https://arxiv.org/abs/2406.20053) and comments on its implications for AI safety.
What is Covert Malicious Finetuning?
====================================
Covert Malicious Fine... | https://www.lesswrong.com/posts/33emJkmw5bMAXZHHt/covert-malicious-finetuning-1 |
# OthelloGPT learned a bag of heuristics
*Work performed as a part of Neel Nanda's MATS 6.0 (Summer 2024) training program.*
TLDR
----
This is an interim report on reverse-engineering [Othello-GPT](https://arxiv.org/pdf/2210.13382.pdf), an 8-layer transformer trained to take sequences of Othello moves and predict le... | https://www.lesswrong.com/posts/gcpNuEZnxAPayaKBY/othellogpt-learned-a-bag-of-heuristics-1 |
# How Congressional Offices Process Constituent Communication
For anyone interested in facilitating constituent communication (emailing, calling, meeting with the legislators that represent you) to influence policy, it will be helpful to understand how offices process these various communication channels. This is what... | https://www.lesswrong.com/posts/gFN7hknQwPfKQuEdF/how-congressional-offices-process-constituent-communication |
# Economics Roundup #2
Previously: [Economics Roundup #1](https://thezvi.substack.com/p/economics-roundup-1?utm_source=publication-search)
Let’s take advantage of the normality while we have it. In all senses.
#### Insane Tax Proposals
There is Trump’s proposal to replace income taxes with tariffs, but he is not al... | https://www.lesswrong.com/posts/MMtWB8wAu5Buc6sve/economics-roundup-2 |
# Decomposing the QK circuit with Bilinear Sparse Dictionary Learning
*This work was produced as part of Lee Sharkey's stream in the *[*ML Alignment & Theory Scholars Program*](https://www.matsprogram.org/) *\- Winter 2023-24 Cohort*
Intro and Motivation
====================
Sparse dictionary learning (SDL) has attr... | https://www.lesswrong.com/posts/2ep6FGjTQoGDRnhrq/decomposing-the-qk-circuit-with-bilinear-sparse-dictionary |
# List of Collective Intelligence Projects
During the last [Foresight Intelligent Cooperation Workshop](https://foresight.org/2024-intelligent-cooperation-workshop/) I got very curious about what collective intelligence tools currently exist. A list:
* [Pol.is](https://pol.is/): "Input Crowd, Output Meaning"
* ... | https://www.lesswrong.com/posts/vcuBJgfSCvyPmqG7a/list-of-collective-intelligence-projects |
# An AI Race With China Can Be Better Than Not Racing
__Frustrated by all your bad takes, I write a Monte-Carlo analysis
of whether a transformative-AI-race between the PRC and the USA
would be good. To my surprise, I find that it is *better* than not
racing. Advocating for an international project to build TAI instea... | https://www.lesswrong.com/posts/WT3u2tK2AJpYKvaZd/an-ai-race-with-china-can-be-better-than-not-racing |
# How ARENA course material gets made
**TL;DR**
=========
In this post, I describe my methodology for building new material for ARENA. I'll mostly be referring to the exercises on IOI, Superposition and Function Vectors as case studies. I expect this to be useful for people who are interested in designing material fo... | https://www.lesswrong.com/posts/ZBZtWxsf9iXN7R2D5/how-arena-course-material-gets-made |
# Creating an Exobrain
**Workshop description:**
-------------------------
We will learn how to choose and manage a note-taking system, a task-management system, and a calendar system. Collectively, these tools are affectionately referred to as an "Exobrain".
That said, this isn't a typical productivity course. "Pro... | https://www.lesswrong.com/events/46tDCWWzTA46EpDum/creating-an-exobrain |
# Open Sourcing Metaculus
*This is a linkpost for an announcement by Metaculus CEO,* [*Deger Turan*](https://x.com/degerturann)*, originally published June 5th, 2024. In addition to spreading the word here, we are interested to gather feedback and ideas from the EA community.*
We make reasoning and coordination tools... | https://www.lesswrong.com/posts/myBQH6Abp5wa6dEJp/open-sourcing-metaculus |
# Announcing the AI Forecasting Benchmark Series | July 8, $120k in Prizes
**On July 8th we’re launching the first in a series of quarterly tournaments benchmarking the state of the art in AI forecasting and how it compares to the best human forecasting on real-world questions. You can find the** [**landing page for t... | https://www.lesswrong.com/posts/hmnB9zpY5wB7dKMQ3/announcing-the-ai-forecasting-benchmark-series-or-july-8 |
# 3C's: A Recipe For Mathing Concepts
Opening Example: Teleology
--------------------------
When people say “the heart’s purpose is to pump blood” or “a pencil’s function is to write”, what does that mean physically? What are “purpose” or “function”, not merely in intuitive terms, but in terms of math and physics? Th... | https://www.lesswrong.com/posts/fEvCxNte6FKSRNFvN/3c-s-a-recipe-for-mathing-concepts |
# What percent of the sun would a Dyson Sphere cover?
I disagreed with a bunch of the implications of [this comment](https://www.lesswrong.com/posts/87EzRDAHkQJptLthE/but-why-would-the-ai-kill-us?commentId=LgjwepWmnJKjXvoQK), but I was curious about the specific question "Would a dyson sphere made out of the solar sys... | https://www.lesswrong.com/posts/EKhNDspRxLGcZAuJz/what-percent-of-the-sun-would-a-dyson-sphere-cover |
# Static Analysis As A Lifestyle
I’ve been watching French Top Chef (the best Top Chef, fight me) wi... | https://www.lesswrong.com/posts/XEuArCYEALQ6XecW7/static-analysis-as-a-lifestyle |
# Musings on LLM Scale (Jul 2024)
In a [recent interview](https://www.youtube.com/watch?v=xm6jNMSFT7g), Dario Amodei [claimed](https://www.youtube.com/watch?v=xm6jNMSFT7g&t=802s) that cost of training is (starting with models already available)
> Right now, $100 million. There are models in training today that are mor... | https://www.lesswrong.com/posts/T3tDQfkAjFsScHL3C/musings-on-llm-scale-jul-2024 |
# When Are Results from Computational Complexity Not Too Coarse?
Tl;dr, While an algorithm's computational complexity may be exponential in general (worst-case), it is often possible to stratify its input via some dimension $k$ that makes it polynomial for a fixed $k$, and only exponential in $k$. Conceptually, this q... | https://www.lesswrong.com/posts/Ja9NP3NJpEd7BXMnW/when-are-results-from-computational-complexity-not-too |
# Notes on Tuning Metacognition
*Summary*: Reflections and practice notes on a metacognitive technique aimed at refining the process of thinking, rather than the thoughts themselves.
*Epistemic Status*: Experimental and observational, based on personal practice and reflections over a brief period.
Introduction
=====... | https://www.lesswrong.com/posts/jgjGJBFLSzBYJjxAh/notes-on-tuning-metacognition |
# 80,000 hours should remove OpenAI from the Job Board (and similar EA orgs should do similarly)
*I haven't shared this post with other relevant parties – my experience has been that private discussion of this sort of thing is more paralyzing than helpful. I might change my mind in the resulting discussion, but, I pre... | https://www.lesswrong.com/posts/8qCwuE8GjrYPSqbri/80-000-hours-should-remove-openai-from-the-job-board-and |
# How predictive processing solved my wrist pain
This is a post about a period of major suffering I went through, and how I acquired the knowledge and skill to eventually lift me out of it.
Disclaimer: this is not medical advice — consider it a personal narrative, perhaps a starting point for exploration.
A fire b... | https://www.lesswrong.com/posts/8jGb5hRGGEcs7Be78/how-predictive-processing-solved-my-wrist-pain |
# Introduction to French AI Policy
*This post was written as part of the AI Governance Fundamentals course by BlueDot. I thank Charles Beasley and the students from my cohort for their feedback and encouragements.*
**Disclaimer**: The French policy landscape is in rapid flux, after president Macron called for a snap ... | https://www.lesswrong.com/posts/GFeyXGib7DD3ooTEN/introduction-to-french-ai-policy |
# The Dumbification of our smart screens
Here we are at the peak of an eternal technology era.
AI agents are on their way. Driverless taxis are making a [debut](https://www.sfchronicle.com/sf/article/waymo-s-f-19532311.php). 4G is an old fashion. We’re heading to a seemingly great utopia.
Perhaps the greatest invent... | https://www.lesswrong.com/posts/T7fd4paHJRiaACoTe/the-dumbification-of-our-smart-screens |
# AI #71: Farewell to Chevron
Chevron deference is no more. How will this impact AI regulation?
The obvious answer is it is now much harder for us to ‘muddle through via existing laws and regulations until we learn more,’ because the court narrowed our affordances to do that. And similarly, if and when Congress does ... | https://www.lesswrong.com/posts/AYJcL6GD3FLkL4yNC/ai-71-farewell-to-chevron |
# Consider the humble rock (or: why the dumb thing kills you)
When people think about street-fights and what they should do when they find themselves in the unfortunate position of being in one, they tend to stumble across a pretty concerning thought relatively early on: "What if my attacker has a knife?" . Then they ... | https://www.lesswrong.com/posts/fWEnZqgxA2BcxZXF3/consider-the-humble-rock-or-why-the-dumb-thing-kills-you |
# Libs vs Frameworks, Middle-Level Regularities vs Theories
Recently, a friend of mine brought out t... | https://www.lesswrong.com/posts/hqpDh6jtxHQTsPA7H/libs-vs-frameworks-middle-level-regularities-vs-theories |
# Finding the Wisdom to Build Safe AI
We may soon build [superintelligent](https://www.lesswrong.com/tag/superintelligence) AI. Such AI poses an [existential threat](https://nickbostrom.com/existential/risks.pdf) to humanity, and all life on Earth, if it is not aligned with our flourishing. Aligning superintelligent A... | https://www.lesswrong.com/posts/pchAuTJhfKJS2s8it/finding-the-wisdom-to-build-safe-ai |
# Doomsday Argument and the False Dilemma of Anthropic Reasoning
Doomsday Inference
------------------
Can we use probability theory to estimate how many people there will be throughout the whole human history? Sure. We can build a probability model, that takes into account birth rates, possible existential hazards, ... | https://www.lesswrong.com/posts/YgSKfAG2iY5Sxw7Xd/doomsday-argument-and-the-false-dilemma-of-anthropic |
# ML4Good Summer Bootcamps - Applications Open [deadline extended]
ML4Good bootcamps are 10-day bootcamps focusing on upskilling in technical AI safety, exploring governance, and delving into conceptual topics. ML4Good is a non-profit project and freeof charge for the participants. We're seeking motivated individuals ... | https://www.lesswrong.com/posts/i5xLe5brs2Pf9heA9/ml4good-summer-bootcamps-applications-open-deadline-extended |
# Minimalist And Maximalist Type Systems
I have two completely opposed aesthetic preferences inside ... | https://www.lesswrong.com/posts/HuWaETuQiRHd87jRv/minimalist-and-maximalist-type-systems |
# [Interim research report] Activation plateaus & sensitive directions in GPT2
*This part-report / part-proposal describes ongoing research, but I'd like to share early results for feedback. I am especially interested in any comment finding mistakes or trivial explanations for these results. I will work on this propos... | https://www.lesswrong.com/posts/LajDyGyiyX8DNNsuF/interim-research-report-activation-plateaus-and-sensitive-1 |
# D&D.Sci: Whom Shall You Call?
*This is a D&D.Sci scenario: a puzzle where players are given a dataset to analyze and an objective to pursue using information from that dataset.*
*Intended Difficulty: Average*
Good news! After an interminable period of waiting for lawyers to finish arguing over the will, you’ve inh... | https://www.lesswrong.com/posts/vrB8xYK8pFPmP5ts8/d-and-d-sci-whom-shall-you-call |
# A "Bitter Lesson" Approach to Aligning AGI and ASI
**TL;DR**: I discuss the challenge of aligning AGI/ASI, and outline an extremely simple approach to aligning an LLM: train entirely on a synthetic dataset that always shows the AI acting aligned (even when the humans behave badly), and use a conditional training/in... | https://www.lesswrong.com/posts/oRQMonLfdLfoGcDEh/a-bitter-lesson-approach-to-aligning-agi-and-asi-1 |
# What progress have we made on automated auditing?
One use case for model internals work is to perform automated auditing of models:
[https://www.alignmentforum.org/posts/cQwT8asti3kyA62zc/automating-auditing-an-ambitious-concrete-technical-research](https://www.alignmentforum.org/posts/cQwT8asti3kyA62zc/automating-... | https://www.lesswrong.com/posts/v5jBXHH9rDwBttWNp/what-progress-have-we-made-on-automated-auditing |
# Travel Buffer
I recently finished a [nine-day road trip](https://www.jefftk.com/p/kingfisher-summer-tour-2024), playing a series of dances. This combined driving long distances with arrival deadlines: you don't want to arrive late. Not only is it unprofessional, but it's stressful rushing to set up and you'll likely... | https://www.lesswrong.com/posts/N7owFAHWgALNWs25d/travel-buffer |
# Indecision and internalized authority figures
A [trauma book I was reading](https://kauppa.viisaselama.fi/sivu/tuote/lapsuuden-kehityksellinen-trauma/4428065) had an interesting claim that indecision is often because the person looks for the approval of an internalized authority figure (the writer is a Jungian thera... | https://www.lesswrong.com/posts/7dW9dHdcj5XHFxui5/indecision-and-internalized-authority-figures |
# Links and brief musings for June
Links in English
================
Schrödinger’s Ursula
--------------------
Apparently the concept of Schrödinger’s cat [got popularized thanks to Ursula Le Guin](https://physicsworld.com/a/ursula-le-guin-the-pioneering-author-we-should-thank-for-popularizing-schrodingers-cat/).
>... | https://www.lesswrong.com/posts/qHgyGguajSjP4wGEf/links-and-brief-musings-for-june |
# AI Alignment Research Engineer Accelerator (ARENA): Call for applicants v4.0
TL;DR
=====
**We are excited to announce the fourth iteration of** [**ARENA**](https://www.arena.education/)**(Alignment Research Engineer Accelerator), a 4-5 week ML bootcamp with a focus on AI safety! ARENA’s mission is to provide talent... | https://www.lesswrong.com/posts/6SE2b247B45hKxziF/ai-alignment-research-engineer-accelerator-arena-call-for-2 |
# Can agents coordinate on randomness without outside sources?
There are situations where two agents that can read each other’s source code want to have a bit of random information (e.g., they want to cooperate and split an indivisible thing by deciding randomly who’ll have it and having half of it in expectation).
I... | https://www.lesswrong.com/posts/uQciccRgxjxjYKXAL/can-agents-coordinate-on-randomness-without-outside-sources |
# [Linkpost] A Case for AI Consciousness
Just wanted to share a new paper on AI consciousness with Simon Goldstein that members of this community might be interested in. Here's the abstract:
*It is generally assumed that existing artificial systems are not phenomenally conscious, and that the construction of phenomen... | https://www.lesswrong.com/posts/hXsbrQSbgD7JnbW9p/linkpost-a-case-for-ai-consciousness |
# An AI Manhattan Project is Not Inevitable
Early last month, Leopold Aschenbrenner released a [long essay](https://situational-awareness.ai/) and [podcast](https://www.dwarkeshpatel.com/p/leopold-aschenbrenner) outlining his projections for the future of AI. Both of these sources are full of interesting arguments and... | https://www.lesswrong.com/posts/QffwXEExuGZadwqBK/an-ai-manhattan-project-is-not-inevitable |
# Scalable oversight as a quantitative rather than qualitative problem
\[Many of these ideas were developed in conversation with Ryan Greenblatt and Ansh Radhakrishnan; a lot of this isn’t original but I haven’t seen it written up\]
A lot of the time when people talk about scalable oversight, which I'll define for th... | https://www.lesswrong.com/posts/6AT4vhYzww56CR6cm/scalable-oversight-as-a-quantitative-rather-than-qualitative |
# Thresholding
*(This is a linkpost for Duncan Sabien's article "Thresholding" which was published July 6th, 2024. I (Screwtape) am crossposting a linkpost version because I want to nominate it for the Best of LW 2024 review - I'm not the original author.)*
If I were in some group or subculture and I wanted to do as ... | https://www.lesswrong.com/posts/tbNp6eu745bkProw8/thresholding |
# LK-99 in retrospect
About a year ago, there was a lot of public interest in a supposed room-temperature superconductor called [LK-99](https://en.wikipedia.org/wiki/LK-99). What I publicly said at the time was, basically:
1. We should remember the possibility that apparent levitation is from ferromagnetism or parama... | https://www.lesswrong.com/posts/3in2vMnfeEeM8aCvA/lk-99-in-retrospect |
# Reflections on Less Online
**Meta:** *This post turned out longer, slower, and less well-written than I hoped. I don’t see any similar posts in a quick search, though, so
I'm posting it anyway. I’ve tried to front-load
feedback that might be useful to the organizers, and put more personal
stuff towards the end. For ... | https://www.lesswrong.com/posts/NdfXRzWdNoEJgwxsN/reflections-on-less-online |
# Beyond Biomarkers: Understanding Multiscale Causality
Photo by [Jigar Panchal](https://unsplash.com/@brave4_heart?utm_content=creditCopyText&utm_medium=referral&utm_source=unsplash) on [Unsplash](https://unsplash.com/photos/a-large-group-of-... | https://www.lesswrong.com/posts/8XtYr3av6rpEbjGTh/beyond-biomarkers-understanding-multiscale-causality |
# An Extremely Opinionated Annotated List of My Favourite Mechanistic Interpretability Papers v2
*This post represents my personal hot takes, not the opinions of my team or employer. This is a massively updated version of* [*a similar list*](https://www.alignmentforum.org/posts/SfPrNY45kQaBozwmu/an-extremely-opinionat... | https://www.lesswrong.com/posts/NfFST5Mio7BCAQHPA/an-extremely-opinionated-annotated-list-of-my-favourite |
# On saying "Thank you" instead of "I'm Sorry"
Back in 2016 or so, I ran into an idea going around the self-help / trauma-informed-therapy / cognitive-behavioral internet: *Learn to say “thank you” instead of “I’m sorry”*. It’s turned out to be one of the most transformative pieces of advice I’ve ever taken. I’d like ... | https://www.lesswrong.com/posts/ascu5u6qF4Thu3YP7/on-saying-thank-you-instead-of-i-m-sorry |
# Controlled Creative Destruction
This is a crosspost from [https://250bpm.substack.com/p/controlled-creative-destruction](https://250bpm.substack.com/p/controlled-creative-destruction).
Creative destruction is a process which lets the inefficient fail and the efficient survive. It's the economists' way of saying of ... | https://www.lesswrong.com/posts/quvunrHrJ3ByTegYy/controlled-creative-destruction |
# Poker is a bad game for teaching epistemics. Figgie is a better one.
_Editor's note: Somewhat after I posted this on my own blog, [Max Chiswick](https://aipokertutorial.com/about) cornered me at LessOnline / Manifest and gave me a whole new perspective on this topic. I now believe that there is a way to use poker to... | https://www.lesswrong.com/posts/PypgeCxFHLzmBENK4/poker-is-a-bad-game-for-teaching-epistemics-figgie-is-a |
# On scalable oversight with weak LLMs judging strong LLMs
**Abstract**
============
Scalable oversight protocols aim to enable humans to accurately supervise superhuman AI. In this paper we study *debate*, where two AI's compete to convince a human judge; *consultancy*, where a single AI tries to convince a human ju... | https://www.lesswrong.com/posts/Qn3ZDf9WAqGuAjWQe/on-scalable-oversight-with-weak-llms-judging-strong-llms |
# Towards shutdownable agents via stochastic choice
We[^usnm40pukij] have a [new paper](https://arxiv.org/abs/2407.00805) testing the [POST-Agents Proposal (PAP)](https://www.lesswrong.com/posts/YbEbwYWkf8mv9jnmi/the-shutdown-problem-incomplete-preferences-as-a-solution). The abstract and main-text is below. Appendice... | https://www.lesswrong.com/posts/dzvnAGDPsisMY8h7b/towards-shutdownable-agents-via-stochastic-choice |
# Games of My Childhood: The Troops
The Troops (Finnish “joukot”, could also be translated as “the armies” or “the forces”) was a game of pretend that I played the most with my friend Eero; I believe Aleksi also joined in.
The central premise was that each time that you played a video game and killed, recruited, resc... | https://www.lesswrong.com/posts/NxBPFfKCCdrK7gKGi/games-of-my-childhood-the-troops |
# Response to Dileep George: AGI safety warrants planning ahead
*(Target audience: Dileep George himself, and anyone coming from a similar place.)*
Dileep George is a researcher working at the intersection of AI and neuroscience. He started his career by co-founding Numenta in 2005 with Jeff Hawkins (while a Stanford... | https://www.lesswrong.com/posts/LJD4C7KAr64onL8fq/response-to-dileep-george-agi-safety-warrants-planning-ahead |
# Announcing The Techno-Humanist Manifesto: A new philosophy of progress for the 21st century
Humans are a curious species: We have a need not only to *do*, but to *explain* what we are doing—to each other and above all to ... | https://www.lesswrong.com/posts/TDmkDKzD5zdZruua7/announcing-the-techno-humanist-manifesto-a-new-philosophy-of |
# Dialogue introduction to Singular Learning Theory
**Alice:** A lot of people are talking about Singular Learning Theory. Do you know what it is?
**Bob:** I do. (*pause)* Kind of.
**Alice:** Well, I don't. Explanation time?
**Bob:** Uh, I'm not really an expert on it. You know, there's a [lot](https://www.lesswron... | https://www.lesswrong.com/posts/CmcarN6fGgTGwGuFp/dialogue-introduction-to-singular-learning-theory |
# The Golden Mean of Scientific Virtues
I recently discovered [this nice post on the scientific vir... | https://www.lesswrong.com/posts/GwKzezDsTdQCbTrBK/the-golden-mean-of-scientific-virtues |
# Pantheon Interface
Pantheon is an **experimental LLM interface** exploring a different type of human-AI interaction. We created this as a part of the [cyborgism](https://www.lesswrong.com/posts/bxt7uCiHam4QXrQAA/cyborgism) project, with the abstract motivation of augmenting the human ability to think by integrating ... | https://www.lesswrong.com/posts/JHsfMWtwxBGGTmb8A/pantheon-interface |
# Advice to junior AI governance researchers
This summer, I’m supervising some research fellows through [Cambridge’s ERA AI Fellowship](https://erafellowship.org/). The program started last week, and I’ve had conversations with about 6 fellows about their research projects & summer goals.
In this post, I’ll highligh... | https://www.lesswrong.com/posts/ABSEaJvtap4v4nzcM/advice-to-junior-ai-governance-researchers |
# Sample Prevalence vs Global Prevalence
*Cross-posted from my [NAO Notebook](https://data.securebio.org/jefftk-notebook/sample-vs-global-prevalence). Thanks to Evan Fields and Mike McLaren for editorial feedback on this post.*
In [Detecting Genetically Engineered Viruses With Metagenomic Sequencing](https://naobserv... | https://www.lesswrong.com/posts/D3o3ed8WRspbgnPGr/sample-prevalence-vs-global-prevalence |
# "The Singularity Is Nearer" by Ray Kurzweil - Review
Over the past week I read Ray Kurzweil's new book "The Singularity Is Nearer". This is my review.
First off, if you've read The Singularity Is Near, which was published 19 years ago in 2005, you should be aware that the sequel book is a lot less technical. The Si... | https://www.lesswrong.com/posts/QXHKLbHzC5nPvJnXd/the-singularity-is-nearer-by-ray-kurzweil-review |
# Robin Hanson & Liron Shapira Debate AI X-Risk
Robin and I just had an interesting 2-hour AI doom debate. We picked up where the [Hanson-Yudkowsky Foom Debate](https://www.lesswrong.com/tag/the-hanson-yudkowsky-ai-foom-debate) left off in 2008, revisiting key arguments in the light of recent AI advances.
, discus... | https://www.lesswrong.com/posts/gXesaqCJzNzcextmm/wtf-is-with-the-infancy-gospel-of-thomas-a-deep-dive-into |
# Consent across power differentials
I'd like to put forward another description of a basic issue that's been around for a while. I don't know if there's been significant progress on a solution, and would be happy to pointed to any such progress. I've opted to go for a relatively rough and quick post that doesn't dive... | https://www.lesswrong.com/posts/CZpZFYPqtk6wMTQwR/consent-across-power-differentials |
# Medical Roundup #3
This time around, we cover the Hanson/Alexander debates on the value of medicine, and otherwise we mostly have good news.
#### Technology Advances
Regeneron administers a single shot in a genetically deaf child’s ear, [and they can hear after a few months](https://twitter.com/cremieuxrecueil/sta... | https://www.lesswrong.com/posts/6GhemtgJxF9sSDNrq/medical-roundup-3 |
# What and Why: Developmental Interpretability of Reinforcement Learning
**Introduction**
================
I happen to be in that happy stage in the research cycle where I ask for money so I can continue to work on things I think are important. Part of that means justifying what I want to work on to the satisfaction ... | https://www.lesswrong.com/posts/Bczmi8vjiugDRec7C/what-and-why-developmental-interpretability-of-reinforcement |
# UC Berkeley course on LLMs and ML Safety
The UC Berkeley course I co-taught now has lecture videos available: [https://www.youtube.com/playlist?list=PLJ66BAXN6D8H_gRQJGjmbnS5qCWoxJNfe](https://www.youtube.com/playlist?list=PLJ66BAXN6D8H_gRQJGjmbnS5qCWoxJNfe)
Course site: [Understanding LLMs: Foundations and Safety]... | https://www.lesswrong.com/posts/FENtGM5waoxtYazzM/uc-berkeley-course-on-llms-and-ml-safety |
# Paper Summary: The Effects of Communicating Uncertainty on Public Trust in Facts and Numbers
by Anne Marthe van der Bles, Sander van der Linden, Alexandra L. J. Freeman, and David J. Spiegelhalter. (2020)
_[https://www.pnas.org/doi/pdf/10.1073/pnas.1913678117](https://www.pnas.org/doi/pdf/10.1073/pnas.1913678117)._... | https://www.lesswrong.com/posts/oNxPqB3kyCd2kYnXH/paper-summary-the-effects-of-communicating-uncertainty-on |
# Summer Tour Stops
Last week, Lily, Anna, Cecilia, Harris, and I drove to St Louis and back playing dances. I wrote about one aspect [last week](https://www.jefftk.com/p/travel-buffer), and will probably continue splitting thoughts across posts. Today I'm thinking some about our approach to stopping along the way.
I... | https://www.lesswrong.com/posts/TvCQJHnHcMCuLL3dD/summer-tour-stops |
# Causal Graphs of GPT-2-Small's Residual Stream
*Thanks to the many people I've chatted with this about over the past many months. And special thanks to* [*Cunningham et al.*](https://arxiv.org/abs/2309.08600)*,* [*Marks et al.*](https://arxiv.org/abs/2403.19647)*,* [*Joseph Bloom*](https://www.lesswrong.com/users/jo... | https://www.lesswrong.com/posts/uNGAjA8wCNDZHJxu8/causal-graphs-of-gpt-2-small-s-residual-stream |
# Pondering how good or bad things will be in the AGI future
Yesterday I heard a podcast where someone said he hoped AGI would be developed in his lifetime. This confused me, and I realized that it might be useful - at least for me - to write down this confusion.
Consider that for some reasons - different history, di... | https://www.lesswrong.com/posts/cnBLj6KQSeZ4CA9NS/pondering-how-good-or-bad-things-will-be-in-the-agi-future |
# On passing Complete and Honest Ideological Turing Tests (CHITTs)
*\[Cross-posted from Facebook\]*
"In the moment when I truly understand my enemy, understand him well enough to defeat him, then in that very moment I also love him. I think it’s impossible to really understand somebody, what they want, what they beli... | https://www.lesswrong.com/posts/erE6jmJviBAarWgQL/on-passing-complete-and-honest-ideological-turing-tests |
# DIY RLHF: A simple implementation for hands on experience
*Many thanks to Diogo de Lucena, Cameron Berg, Judd Rosenblatt, and Philip Gubbins for support and feedback on this post.*
**TL;DR**
Reinforcement Learning from Human Feedback (RLHF) is one of the leading methods for fine-tuning foundational models to be he... | https://www.lesswrong.com/posts/BmJZKtuoroqgBpWTe/diy-rlhf-a-simple-implementation-for-hands-on-experience |
# New page: Integrity
There's a [new page](https://ailabwatch.org/resources/integrity/) collecting integrity incidents at the frontier AI labs.
Also a month ago I made a [page](https://ailabwatch.org/resources/company-advocacy/) on labs' policy advocacy.
If you have suggestions to improve these pages, or have ideas ... | https://www.lesswrong.com/posts/KxkMmwJnhmCFBjhTi/new-page-integrity |
# Antitrust as Controlled Creative Destruction
[
](https://substackcdn.com/image/fetch/f_auto,q_au... | https://www.lesswrong.com/posts/usyCLLT3PxGqNuqap/antitrust-as-controlled-creative-destruction |
# Fluent, Cruxy Predictions
*The latest in the* [*Feedback Loop Rationality*](https://www.lesswrong.com/sequences/skTdjtx87XKurdvWf) *series.*
Periodically, people (including me) try to operationalize predictions, or bets, and... it doesn't seem to help much.
I think I recently "got good" at making "actually useful ... | https://www.lesswrong.com/posts/wDpXshpakpYDcTtug/fluent-cruxy-predictions-1 |
# [EAForum xpost] A breakdown of OpenAI's revenue
We estimate that, as of June 12, 2024, OpenAI has an annualized revenue (ARR) of:
**$1.9B for ChatGPT Plus** (7.7M global subscribers),
**$714M from ChatGPT Enterprise** (1.2M seats),
**$510M from the API**, and
** $290M from ChatGPT Team** (from 980k seats)
... | https://www.lesswrong.com/posts/BmqKxL3BedDTCjer3/eaforum-xpost-a-breakdown-of-openai-s-revenue |
# Managing Emotional Potential Energy
I really need a holiday.
I’m feeling stressed and trapped by ... | https://www.lesswrong.com/posts/7LWKzJd3L3YJ8Ys6D/managing-emotional-potential-energy |
# Reliable Sources: The Story of David Gerard
[ career advice needed (asking for a friend).
Let's assume, for the sake of discussion, that [Leopold Aschenbrenner](https://situational-awareness.ai/wp-content/uploads/2024/0... | https://www.lesswrong.com/posts/tgrdfvN8f3WzvZGcr/what-other-lines-of-work-are-safe-from-ai-automation |
# The Best Bits From Build, Baby, Build
[Build, Baby, Build](https://www.amazon.com/Build-Baby-Science-Housing-Regulation/dp/1952223415/ref=sr_1_1?sr=8-1) is George Mason Economist Bryan Caplan’s latest book and [second](https://www.amazon.com/Open-Borders-Science-Ethics-Immigration/dp/1250316960/ref=sr_1_2?sr=8-2) gr... | https://www.lesswrong.com/posts/Xtv9FcFpJQfqLqubQ/the-best-bits-from-build-baby-build |
# AI #72: Denying the Future
The Future. It is coming.
A surprising number of economists deny this when it comes to AI. Not only do they deny the future that lies in the future. They also deny the future that is here, but which is unevenly distributed. Their predictions and projections do not factor in even what the ... | https://www.lesswrong.com/posts/xAoXxjtDGGCP7tBDY/ai-72-denying-the-future |
# Video Intro to Guaranteed Safe AI
*Many thanks to Evan Miyazono, Nora Amman, Philip Gubbins, and Judd Rosenblatt for valuable feedback towards making this video.*
We created a video introduction to the paper [Towards Guaranteed Safe AI](https://arxiv.org/abs/2405.06624) to highlight its concepts[^xvm8n3q7e5g] and ... | https://www.lesswrong.com/posts/dQHg2sKtnuXeNydFS/video-intro-to-guaranteed-safe-ai |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.