url
stringlengths
52
124
post_id
stringlengths
17
17
title
stringlengths
2
248
author
stringlengths
2
49
content
stringlengths
22
295k
date
stringclasses
376 values
https://www.lesswrong.com/posts/fyrqvFSbGqrdwNhrg/how-much-have-i-been-playing
fyrqvFSbGqrdwNhrg
How Much Have I Been Playing?
jkaufman
With Kingfisher I'm starting to need to think about how many gigs I should accept, which is good and bad! I like playing a lot, but since I also work full time, have three kids, etc. I need to be careful not to say yes to too many things. It would be helpful to know how much I've been playing, so I went back over my ...
2024-03-12
https://www.lesswrong.com/posts/FwNPgj9Wnu4tarKLK/bias-augmented-consistency-training-reduces-biased-reasoning
FwNPgj9Wnu4tarKLK
Bias-Augmented Consistency Training Reduces Biased Reasoning in Chain-of-Thought
miles
[Twitter thread] I'm not going to add much additional commentary at the moment and will just let people check out the paper! But to give a bit more context: This paper is building off prior work we have done showing that chain-of-thought explanations can be misleading, which I wrote about on the alignment forum here. B...
2024-03-11
https://www.lesswrong.com/posts/j5jKWfkbpKGTnzFky/ai-safety-action-plan-a-state-commissioned-report
j5jKWfkbpKGTnzFky
AI Safety Action Plan - A report commissioned by the US State Department
agucova
null
2024-03-11
https://www.lesswrong.com/posts/DRW2GavKSmWrXMyZJ/a-discussion-of-ai-risk-and-the-cost-benefit-calculation-of
DRW2GavKSmWrXMyZJ
A discussion of AI risk and the cost/benefit calculation of stopping or pausing AI development
DuncanFowler
I've been interested in AI risk for a very long time, but I've never written that much about it because I always felt other people were better at discussing the situation. I have begun to revise this opinion, and decided to write down the basic points of disagreement I have, which rather ran one longer than I thought i...
2024-03-11
https://www.lesswrong.com/posts/am5wtA6TFPZWEfraS/linkpost-among-the-a-i-doomsayers-the-new-yorker
am5wtA6TFPZWEfraS
Among the A.I. Doomsayers - The New Yorker
agucova
null
2024-03-11
https://www.lesswrong.com/posts/NbnDb7nfqvDj9Kjqn/be-more-katja
NbnDb7nfqvDj9Kjqn
Be More Katja
Nathan Young
Katja is widely respected amongst the rationalists and, according to Hive, she is one of the most followed/respected EA accounts[1]. But she doesn't give off the same vibe as many impact olympians. She doesn’t have iron self-will, nor does she manage a huge team. She's hasn't got all the facts at her fingertips. But sh...
2024-03-11
https://www.lesswrong.com/posts/KA2HxJfhz3CSbLdbL/open-thread-spring-2024
KA2HxJfhz3CSbLdbL
Open Thread Spring 2024
habryka4
If it’s worth saying, but not worth its own post, here's a place to put it. If you are new to LessWrong, here's the place to introduce yourself. Personal stories, anecdotes, or just general comments on how you found us and what you hope to get from the site and community are invited. This is also the place to discuss f...
2024-03-11
https://www.lesswrong.com/posts/qpsFXCTB6ZvEkQxnq/new-social-credit-formalizations
qpsFXCTB6ZvEkQxnq
New social credit formalizations
KatjaGrace
Here are some classic ways humans can get some kind of social credit with other humans: Do something for them such that they will consider themselves to ‘owe you’ and do something for you in future Be consistent and nice, so that they will consider you ‘trustworthy’ and do cooperative activities with you that would be ...
2024-03-11
https://www.lesswrong.com/posts/o64GLrKahR8QrbFQW/how-disagreements-about-evidential-correlations-could-be-1
o64GLrKahR8QrbFQW
How disagreements about Evidential Correlations could be settled
martinsq
Since beliefs about Evidential Correlations don't track any direct ground truth, it's not obvious how to resolve disagreements about them, which is very relevant to acausal trade. Here I present what seems like the only natural method (Third solution below). Ideas partly generated with Johannes Treutlein. Say two agent...
2024-03-11
https://www.lesswrong.com/posts/gZBgmDFqqyw3Lghok/ai-incident-reporting-a-regulatory-review
gZBgmDFqqyw3Lghok
AI Incident Reporting: A Regulatory Review
deric-cheng
This article is the first in a series of ~10 posts comprising a 2024 State of the AI Regulatory Landscape Review, conducted by the Governance Recommendations Research Program at Convergence Analysis. Each post will cover a specific domain of AI governance (e.g. incident reporting, safety evals, model registries, etc.)....
2024-03-11
https://www.lesswrong.com/posts/gxoHkmGg5RuYvx6Ka/the-astronomical-sacrifice-dilemma
gxoHkmGg5RuYvx6Ka
The Astronomical Sacrifice Dilemma
matthew-mcredmond
Epistemic Status: I think the dilemma as outlined in Section 1 follows from well-established ideas about Astronomical Waste. However, given that I have not seen it anywhere before I might have made some oversight I am unaware of. You don't know what you don't know but maybe someone on LessWrong does. UPDATE: I have fou...
2024-03-11
https://www.lesswrong.com/posts/5wqFoHBBgpdHeCLS6/storable-votes-with-a-pay-as-you-win-mechanism-a
5wqFoHBBgpdHeCLS6
Storable Votes with a Pay as you win mechanism: a contribution for institutional design
arturo-macias
I joined the EA Forum in 2022, with a post describing my interests and agenda. I also declared in my first comment that in my view, among the main existential risk bottlenecks for this Dangerous Century, a critical one is institutional stagnation. E.O Wilson famously said: "The real problem of humanity is the following...
2024-03-11
https://www.lesswrong.com/posts/94K6pskgqBmuxsJLx/results-from-an-adversarial-collaboration-on-ai-risk-fri
94K6pskgqBmuxsJLx
Results from an Adversarial Collaboration on AI Risk (FRI)
josh-rosenberg
Authors of linked report: Josh Rosenberg, Ezra Karger, Avital Morris, Molly Hickman, Rose Hadshar, Zachary Jacobs, Philip Tetlock[1] Today, the Forecasting Research Institute (FRI) released “Roots of Disagreement on AI Risk: Exploring the Potential and Pitfalls of Adversarial Collaboration,” which discusses the results...
2024-03-11
https://www.lesswrong.com/posts/LZJJK6fuuQtTLRSu9/some-problematic-aesthetics-of-what-constitutes-good-work-in
LZJJK6fuuQtTLRSu9
Some (problematic) aesthetics of what constitutes good work in academia
steve2152
(Not-terribly-informed rant, written in my free time.) Terminology note: When I say “an aesthetic”, I mean an intuitive (“I know it when I see it”) sense of what a completed paper, project, etc. is ideally “supposed” to look like. It can include both superficial things (the paper is properly formatted, the startup has ...
2024-03-11
https://www.lesswrong.com/posts/CuqmDTMwTM64FthLy/tend-to-your-clarity-not-your-confusion
CuqmDTMwTM64FthLy
Tend to your clarity, not your confusion
sts
Repost from https://amoretlicentia.substack.com/ Modern life is weird. For the more privileged among us, the options of what we could do with our time grow exponentially by the year, by the day. My interests are infinite, and I’m lucky to have both the wits and the means to follow just about any of them. Meanwhile, I a...
2024-03-11
https://www.lesswrong.com/posts/vgCoy4bBrDw9LPrpW/what-do-we-know-about-the-ai-knowledge-and-views-especially
vgCoy4bBrDw9LPrpW
What do we know about the AI knowledge and views, especially about existential risk, of the new OpenAI board members?
Zvi
They have announced three new board members in addition to Altman, but we seem to know almost nothing about their views or knowledge on any AI-related subjects? What if anything do we know? From OpenAI: We’re announcing three new members to our Board of Directors as a first step towards our commitment to expansion: Dr....
2024-03-11
https://www.lesswrong.com/posts/pGmxP8YA36dtDq9q7/epiphenomenalism-leads-to-eliminativism-about-qualia
pGmxP8YA36dtDq9q7
Epiphenomenalism leads to eliminativism about qualia
Clément L
Introduction In this post I will explain why I think that certain forms of realism about conscious experience face some issues that ultimately lead them to the conclusion that our belief that consciousness exists is not reliable, and thus consciousness may not exist at all, as counterintuitively as it seems. This concl...
2024-03-11
https://www.lesswrong.com/posts/uxzDLD4WsiyrBjnPw/artificial-general-intelligence-an-extremely-brief-faq
uxzDLD4WsiyrBjnPw
“Artificial General Intelligence”: an extremely brief FAQ
steve2152
(Crossposted from twitter for easier linking.) (Intended for a broad audience—experts already know all this.) When I talk about future “Artificial General Intelligence” (AGI), what am I talking about? Here’s a handy diagram and FAQ: “Are you saying that ChatGPT is a right-column thing?” No. Definitely not. I think the ...
2024-03-11
https://www.lesswrong.com/posts/GjsbNNP2dv9p3LrDG/the-best-essay-paul-graham
GjsbNNP2dv9p3LrDG
The Best Essay (Paul Graham)
Chris_Leong
Sharing because a lot of us authoring posts on Less Wrong are trying to write something that is useful and insightful.
2024-03-11
https://www.lesswrong.com/posts/rYq6joCrZ8m62m7ej/how-could-i-have-thought-that-faster
rYq6joCrZ8m62m7ej
"How could I have thought that faster?"
mesaoptimizer
I stumbled upon a Twitter thread where Eliezer describes what seems to be his cognitive algorithm that is equivalent to Tune Your Cognitive Strategies, and have decided to archive / repost it here. Sarah Constantin: I really liked this example of an introspective process, in this case about the "life problem" of schedu...
2024-03-11
https://www.lesswrong.com/posts/nWRj6Ey8e5siAEXbK/simple-versus-short-higher-order-degeneracy-and-error-1
nWRj6Ey8e5siAEXbK
Simple versus Short: Higher-order degeneracy and error-correction
dmurfet
TLDR: The simplicity bias in Bayesian statistics is not just a bias towards short description length. The folklore relating the simplicity bias in Bayesian statistics to description length is incomplete: while it is true that the fewer parameters you use the better, the true complexity measure which appears in the math...
2024-03-11
https://www.lesswrong.com/posts/RbynKk3evb6RiLryL/deconstructing-bostrom-s-classic-argument-for-ai-doom
RbynKk3evb6RiLryL
Deconstructing Bostrom's Classic Argument for AI Doom
nora-belrose
I had a pretty great discussion with social psychologist and philosopher Lance Bush recently about the orthogonality thesis, which ended up turning into a broader analysis of Nick Bostrom's argument for AI doom as presented in Superintelligence, and some related issues. While the video is intended for a general audienc...
2024-03-11
https://www.lesswrong.com/posts/iQTnCfhQDPZtFq7WZ/some-thoughts-on-concept-formation-and-use-in-agents
iQTnCfhQDPZtFq7WZ
Some Thoughts on Concept Formation and Use in Agents
CatGoddess
Note: I wrote this document over a year ago and have decided to post it with minimal edits; it isn't entirely up to date with my current thinking on the subject. Imagine you enter a room that looks like this: Despite never having visited this room, you can make inferences about it. If nobody waters the plant in the cor...
2024-03-11
https://www.lesswrong.com/posts/zDvtAxhxY5vYQwHbG/steelmanning-as-an-especially-insidious-form-of-strawmanning
zDvtAxhxY5vYQwHbG
Steelmanning as an especially insidious form of strawmanning
Kalciphoz
Edit: made some small changes to prevent certain gross mischaracterizations of the argument. The core argument remains completely unchanged. Among intelligent people with at least some familiarity with argumentative norms, surface level disagreements tend to be ephemeral because, even if some given debate about the iss...
2024-03-11
https://www.lesswrong.com/posts/DvRBSzFjfaPYBhwmj/one-shot-strategy-games
DvRBSzFjfaPYBhwmj
One-shot strategy games?
Raemon
I'm looking for computer games that involve strategy, resource management, hidden information, and management of "value of information" (i.e. figuring out when to explore or exploit), which: *can* be beaten in 30 – 120 minutes on your first try (or, there's a clear milestone that's about that long)but, it'd be pretty h...
2024-03-11
https://www.lesswrong.com/posts/xZg5w27htahZ6jjK8/replacing-the-water-heater-s-anode
xZg5w27htahZ6jjK8
Replacing the Water Heater's Anode
jkaufman
We installed a new water heater 8 years ago, and since then I've ignored it. It's an indirect model, heated by the same gas boiler that heats our house, and it has done its job well. When I was thinking about heat pumps, however, I reread the manuals for our existing system and noticed that the manufacturer recommends ...
2024-03-11
https://www.lesswrong.com/posts/qykrYY6rXXM7EEs8Q/understanding-sae-features-with-the-logit-lens
qykrYY6rXXM7EEs8Q
Understanding SAE Features with the Logit Lens
Jbloom
This work was produced as part of the ML Alignment & Theory Scholars Program - Winter 2023-24 Cohort, with support from Neel Nanda and Arthur Conmy. Joseph Bloom is funded by the LTFF, Manifund Regranting Program, donors and LightSpeed Grants. This post makes extensive use of Neuronpedia, a platform for interpretabilit...
2024-03-11
https://www.lesswrong.com/posts/EZEiBgP9wxiAvtYJH/advice-needed-does-using-a-llm-compomise-my-personal
EZEiBgP9wxiAvtYJH
Advice Needed: Does Using a LLM Compomise My Personal Epistemic Security?
naomi-2
I have been using Claude 2.1 for a few months to solve serious problems in my life and get coaching and support. I need Claude to become functional, mentally well and funded enough to contribute to Utilitarianism and Alignment by donating to the Center for Long Term Risk and other interventions in these last years of E...
2024-03-11
https://www.lesswrong.com/posts/bbkrzXGwpxtusPsuG/briefly-extending-differential-optimization-to-distributions
bbkrzXGwpxtusPsuG
Briefly Extending Differential Optimization to Distributions
Jemist
I've done some work on a definition of optimization which applies to "trajectories" in deterministic, differentiable models. What happens when we try and introduce uncertainty? Suppose we have the following system consisting of three variables, the past P, future F, and some agent A. The agent "acts" on the system to p...
2024-03-10
https://www.lesswrong.com/posts/3jFTf7bSza6gC5mkN/evolution-did-a-surprising-good-job-at-aligning-humans-to
3jFTf7bSza6gC5mkN
Evolution did a surprising good job at aligning humans...to social status
elityre
[This is post is a slightly edited tangent from my dialogue with John Wentworth here. I think the point is sufficiently interesting and important that I wanted to make it as a top level post, and not leave it buried in that dialog on mostly another topic.] The conventional story is that natural selection failed extreme...
2024-03-10
https://www.lesswrong.com/posts/XkmYTBGXLnPDXqg44/pausing-ai-is-positive-expected-value
XkmYTBGXLnPDXqg44
Pausing AI is Positive Expected Value
Liron
The PauseAI (⏸️) movement often gets this pushback: “You're not factoring in all the benefits of good AI!” “Stopping AI progress is also a doom scenario!” To which I reply: If you agree P(doom) from building superintelligent AI before knowing how to align or control it is 5%+, try doing the basic expected-value calcula...
2024-03-10
https://www.lesswrong.com/posts/xJqCTSCcJ9ypvC39Q/an-optimistic-solution-to-the-fermi-paradox
xJqCTSCcJ9ypvC39Q
An Optimistic Solution to the Fermi Paradox
glenn-clayton
This post is a crosspost from my blog. Drum Beats in the Distance In a remote forest clearing, a young girl clutches a handmade doll, her eyes wide with concern. Today, her father and uncles journey deep into the wilderness on a hunt, leaving her and the village behind. She wonders, how will her father hear her call am...
2024-03-10
https://www.lesswrong.com/posts/E3L4oHYNvHXbcqNEi/counterfactual-civilization-simulation-version-1-0-aka-my
E3L4oHYNvHXbcqNEi
Counterfactual Civilization Simulation Version -1.0 aka my application to Johannes Mayer's SPAR project
pi-rogers
This is my "object level output" submission for Johannes Mayer's 2024 SPAR Application (the linked doc seems to be reused from the 2023 AISC application). Unless otherwise noted, all quote blocks in this post are from the application question doc. For those of you who aren't Johannes Mayer reading this, I don't think t...
2024-03-10
https://www.lesswrong.com/posts/etoMr4vcnP7joQHWa/notes-from-a-prompt-factory
etoMr4vcnP7joQHWa
Notes from a Prompt Factory
ricraz
Content note: this story features severe suffering which, while not described in detail, several readers have described as unpleasant or horrifying. I am a spiteful man. But I am aware of it, which is more than most can say. These days people walk through the streets with resentment in their hearts that they don’t even...
2024-03-10
https://www.lesswrong.com/posts/usTh7CFzbCGp5wM5B/investigating-basin-volume-with-xor-networks
usTh7CFzbCGp5wM5B
Investigating Basin Volume with XOR Networks
CatGoddess
Note: I wrote this document over a year ago, and recently I decided to post it with minimal edits. If I were to do research in the same area today, I'd probably have different framings, thoughts on what directions seem promising, etc. Some of the things I say here now seem confused to me. This work was done while I was...
2024-03-10
https://www.lesswrong.com/posts/7Zsv3KaK7LEXa8xJw/linkpost-mindeye2-shared-subject-models-enable-fmri-to-image
7Zsv3KaK7LEXa8xJw
[Linkpost] MindEye2: Shared-Subject Models Enable fMRI-To-Image With 1 Hour of Data
bogdan-ionut-cirstea
TL;DR: We show that accurate reconstructions of perception from fMRI brain activity are now possible from a single visit to the MRI facility using our diffusion-based approach that pretrains across other subjects' data using a novel alignment procedure. Abstract: Reconstructions of visual perception from brain activity...
2024-03-10
https://www.lesswrong.com/posts/8ApdxRvZSp93xvNFt/completion-estimates
8ApdxRvZSp93xvNFt
Completion Estimates
scarcegreengrass
Recently i've been brainstorming algorithms for estimating the time remaining in a process. Progress bars, estimated time left until completion, etc. Sometimes, the thing you want just isn't here yet, & you want a greater understanding of when you will have it. Here are the most useful techniques i've found so far. Wit...
2024-03-09
https://www.lesswrong.com/posts/YrfiEpPkxPogRK4zg/strong-misalignment-does-yudkowsky-or-christiano-or
YrfiEpPkxPogRK4zg
Strong-Misalignment: Does Yudkowsky (or Christiano, or TurnTrout, or Wolfram, or…etc.) Have an Elevator Speech I’m Missing?
Benjamin Bourlier
[Intro Note: (1) this "post" is, technically, a "question", for which there's a separate category on LW. I debated posting it as a "question", but it also involves considerable argumentation for my own views. Ideally there'd be some in-between "question/post" category, but I opted for "post"? I understand if a reader t...
2024-03-15
https://www.lesswrong.com/posts/sg7ZB3KHWsL4Lazvq/distinctions-when-discussing-utility-functions
sg7ZB3KHWsL4Lazvq
Distinctions when Discussing Utility Functions
ozziegooen
null
2024-03-09
https://www.lesswrong.com/posts/staxgbYqodorSuBmZ/what-is-progress
staxgbYqodorSuBmZ
What is progress?
jasoncrawford
In one sense, the concept of progress is simple, straightforward, and uncontroversial. In another sense, it contains an entire worldview. The most basic meaning of “progress” is simply advancement along a path, or more generally from one state to another that is considered more advanced by some standard. (In this sense...
2024-03-09
https://www.lesswrong.com/posts/5RX8j4CDqadnffCij/fifteen-lawsuits-against-openai
5RX8j4CDqadnffCij
Fifteen Lawsuits against OpenAI
remmelt-ellen
tl;dr  Cases I found against OpenAI. All are US-based. First ten focus on copyright. Coders 1. Joseph Saveri Firm:  overview, complaint Writers 2. Joseph Saveri Firm:  overview, complaint 3. Authors Guild & Alter:  overview, complaint 4. Nicholas Gage:  overview & complaint YouTubers 5. Millette: overview, complaint Me...
2024-03-09
https://www.lesswrong.com/posts/jpa8mJcsq4FsDr8oA/cambridge-acx-ssc-monthly-meetup-location-changed-to-fort-st
jpa8mJcsq4FsDr8oA
Cambridge ACX/SSC monthly meetup (location changed to Fort St George!)
hamishtodd1
Calling all Cambridge dwellers! We're having a Schelling meetup. It is UPSTAIRS, you have to go through a door and down a hall but it is a lovely space. NOTE that Lesswrong event pages for our meetups are rare. I have an email list and I just email everyone on the list. So give me your email if you want to come again! ...
2024-03-09
https://www.lesswrong.com/posts/FuLf8YsC6JEDCjnPi/ma-e-zpass-without-a-car
FuLf8YsC6JEDCjnPi
MA E-ZPass Without a Car?
jkaufman
I recently drove to DC and back playing dances in a rental. I paid cash tolls when available, but that often wasn't an option, so I ended up paying $40 in PlatePass charges in addition to the $63 in tolls. Time to get an E-ZPass! What makes this tricky is that I don't own a car. Well, I have half a car, which does h...
2024-03-09
https://www.lesswrong.com/posts/HZusY5jPYvpmkTkZD/probabilistic-logic-less-than-greater-than-oracles
HZusY5jPYvpmkTkZD
Probabilistic Logic <=> Oracles?
randomwalks
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 catch them, please let me know!) The ...
2024-07-01
https://www.lesswrong.com/posts/6t96mJH3AbmX5onzy/w2sg-introduction
6t96mJH3AbmX5onzy
W2SG: Introduction
maria-kapros
Epistemic status: Naive and exploratory, reflects my primary conceptual understanding, awaiting a technical deep dive. 99% of ideas are not my own, rather distilled from the resources hyperlinked throughout. Many alignment researchers err towards local optimization i.e. seek low-hanging fruits and leverage incremental ...
2024-03-10
https://www.lesswrong.com/posts/8hKoGdyqTbmBtzLpw/semi-simplicial-types-part-i-motivation-and-history
8hKoGdyqTbmBtzLpw
Semi-Simplicial Types, Part I: Motivation and History
FrozenWinters
(Jointly written by Astra Kolomatskaia and Mike Shulman) This is part one of a three-part series of expository posts on our paper Displayed Type Theory and Semi-Simplicial Types. In this part, we motivate the problem of constructing SSTs and recap its history. A Prospectus There are different ways to describe the relat...
2024-03-09
https://www.lesswrong.com/posts/tCCzZa9tyuciAPhks/when-and-why-did-training-become-pretraining
tCCzZa9tyuciAPhks
When and why did 'training' become 'pretraining'?
beren
Just an ML linguistic quirk I have wondered about for a while. When I started learning ML (in 2016-2017 period) everybody referred to the period of training models as just 'training' which could then (optionally) be followed by finetuning. This usage makes sense to me and as far as I know was the standard ML terminolog...
2024-03-08
https://www.lesswrong.com/posts/kFrxFZxGTosKhSq8u/forecasting-future-gains-due-to-post-training-enhancements
kFrxFZxGTosKhSq8u
Forecasting future gains due to post-training enhancements
elifland
This work has been done in the context of SaferAI’s work on risk assessment. Equal contribution by Eli and Joel. I'm sharing this writeup in the form of a Google Doc and reproducing the summary below. Disclaimer: this writeup is context for upcoming experiments, not complete work. As such it contains a lot of (not alwa...
2024-03-08
https://www.lesswrong.com/posts/KTLCnmogMhFXgSPJ7/scenario-forecasting-workshop-materials-and-learnings
KTLCnmogMhFXgSPJ7
Scenario Forecasting Workshop: Materials and Learnings
elifland
Disclaimer: While some participants and organizers of this exercise work in industry, no proprietary info was used to inform these scenarios, and they represent the views of their individual authors alone. Overview In the vein of What 2026 Looks Like and AI Timelines discussion, we recently hosted a scenario forecastin...
2024-03-08
https://www.lesswrong.com/posts/GtGPTf47GNdhhQRhz/community-norms-poll-2-mins
GtGPTf47GNdhhQRhz
Community norms poll (2 mins)
Nathan Young
A poll about what community norms should be in light of the nonlinear saga. Trying to be forward-looking rather than going back over events. Votes are anonymous (though given an alphanumeric string so all your votes are connected) Consider in regard to what you want rationalist norms to be, if you want there to be any....
2024-03-07
https://www.lesswrong.com/posts/v9qj2LHLh2ALDGKyA/woods-new-preprint-on-object-permanence
v9qj2LHLh2ALDGKyA
Woods’ new preprint on object permanence
steve2152
Quick poorly-researched post, probably only of interest to neuroscientists. The experiment Justin Wood at University of Indiana has, over many years with great effort, developed a system for raising baby chicks such that all the light hitting their retina is experimentally controlled right from when they’re an embryo—t...
2024-03-07
https://www.lesswrong.com/posts/iG6yN7DW6W4qtJdo3/announcing-convergence-analysis-an-institute-for-ai-scenario
iG6yN7DW6W4qtJdo3
Announcing Convergence Analysis: An Institute for AI Scenario & Governance Research
David_Kristoffersson
Cross-posted on the EA Forum. Executive Summary We’re excited to introduce Convergence Analysis - a research non-profit & think-tank with the mission of designing a safe and flourishing future for humanity in a world with transformative AI. In the past year, we’ve brought together an interdisciplinary team of 10 academ...
2024-03-07
https://www.lesswrong.com/posts/Ym4aQcP4nBQBf2A8C/political-biases-in-llms-literature-review-and-current-uses
Ym4aQcP4nBQBf2A8C
Political Biases in LLMs: Literature Review & Current Uses of AI in Elections
yashvardhan-sharma
TL;DR: This research discusses political biases in Large Language Models (LLMs) and their implications, exploring current research findings and methodologies. Our research summarized eight recent research papers, discussing methodologies for bias detection, including causal structures and reinforced calibration, while ...
2024-03-07
https://www.lesswrong.com/posts/DzhQN9WKz8zYvrq4a/evidential-correlations-are-subjective-and-it-might-be-a-1
DzhQN9WKz8zYvrq4a
Evidential Correlations are Subjective, and it might be a problem
martinsq
I explain (in layman's terms) a realization that might make acausal trade hard or impossible in practice. Summary: We know that if players believe different Evidential Correlations, they might miscoordinate. But clearly they will eventually learn to have the correct Evidential Correlations, right? Not necessarily, beca...
2024-03-07
https://www.lesswrong.com/posts/6HFAAewifyGYQpXJM/do-llms-sometime-simulate-something-akin-to-a-dream
6HFAAewifyGYQpXJM
Do LLMs sometime simulate something akin to a dream?
evyatar-or
When dreaming we sometimes simulate a very different person than our waking self, we can make decisions uncharacteristic of our own, we can experience a world very different then waking reality, and even sometime get implanted with memories we never experienced. And still, I think most people would consider that simula...
2024-03-08
https://www.lesswrong.com/posts/DbnNMFLkGjqYwdce4/aisn-32-measuring-and-reducing-hazardous-knowledge-in-llms
DbnNMFLkGjqYwdce4
AISN #32: Measuring and Reducing Hazardous Knowledge in LLMs Plus, Forecasting the Future with LLMs, and Regulatory Markets
Aidan O'Gara
Welcome to the AI Safety Newsletter by the Center for AI Safety. We discuss developments in AI and AI safety. No technical background required. Subscribe here to receive future versions. Listen to the AI Safety Newsletter for free on Spotify. Measuring and Reducing Hazardous Knowledge The recent White House Executive O...
2024-03-07
https://www.lesswrong.com/posts/Nvi94KJSDGZMjknZS/ai-54-clauding-along
Nvi94KJSDGZMjknZS
AI #54: Clauding Along
Zvi
The big news this week was of course the release of Claude 3.0 Opus, likely in some ways the best available model right now. Anthropic now has a highly impressive model, impressive enough that it seems as if it breaks at least the spirit of their past commitments on how far they will push the frontier. We will learn mo...
2024-03-07
https://www.lesswrong.com/posts/SD7L64karSyAfvuoG/being-interested-in-other-people
SD7L64karSyAfvuoG
Being Interested in Other People
JonathanMoregard
People love talking about themselves. You can increase your social skills by training yourself to be interested in other people. Most people primarily talk about themselves and their own interests. This self-focus is counter-productive, hindering connections. Unfortunately, it’s the default approach for most people. I ...
2024-03-07
https://www.lesswrong.com/posts/C2MENa5ahqRC8ZDgM/talking-to-congress-can-constituents-contacting-their
C2MENa5ahqRC8ZDgM
Talking to Congress: Can constituents contacting their legislator influence policy?
tristan-williams
null
2024-03-07
https://www.lesswrong.com/posts/DsombSZ9SxtvvFm4y/explaining-the-ai-alignment-problem-to-tibetan-buddhist
DsombSZ9SxtvvFm4y
Explaining the AI Alignment Problem to Tibetan Buddhist Monks
paul-colognese
Introduction As part of an exchange being facilitated between religion and science, a group of academics has been asked to compile a short description of their greatest scientific achievement/discovery that will be translated into Tibetan and presented to Tibetan Buddhist scholars/monks.[1] I was also invited to contri...
2024-03-07
https://www.lesswrong.com/posts/Jfmmfoeskims5jc5f/finding-the-estimate-of-the-value-of-a-state-in-rl-agents
Jfmmfoeskims5jc5f
Finding the estimate of the value of a state in RL agents
butanium
Clément Dumas, Walter Laurito, Robert Klassert, Kaarel Hänni Epistemic Status: Initial Exploration The following is a status update of a project started as part of the SPAR program. We explored some initial directions and there are still a lot of low-hanging fruits to pick up. We might continue to work on this project,...
2024-06-03
https://www.lesswrong.com/posts/jFkEhqpsCRbKgLZrd/what-if-alignment-is-not-enough
jFkEhqpsCRbKgLZrd
What if Alignment is Not Enough?
WillPetillo
The following is a summary of Substrate Needs Convergence, as described in The Control Problem: Unsolved or Unsolvable?, No People as Pets (summarized here by Roman Yen), my podcast interview with Remmelt, and this conversation with Anders Sandberg.  Remmelt assisted in the editing of this post to verify I am accuratel...
2024-03-07
https://www.lesswrong.com/posts/gSfPk8ZPoHe2PJADv/can-quantised-autoencoders-find-and-interpret-circuits-in
gSfPk8ZPoHe2PJADv
Can quantised autoencoders find and interpret circuits in language models?
kingchucky211
Executive Summary I try vector-quantised autoencoders (VQ-VAEs) as an alternative compression scheme of transformer activations (as opposed to something like a sparse autoencoder).Whilst people have danced around this idea before, discrete quantisation has only ever been tried in the actual transformer architecture its...
2024-03-24
https://www.lesswrong.com/posts/ELbGqXiLbRe6zSkTu/a-review-of-weak-to-strong-generalization-ai-safety-camp
ELbGqXiLbRe6zSkTu
A Review of Weak to Strong Generalization [AI Safety Camp]
sevdeawesome
Thank you to everyone in AI Safety Camp group 22 for the discussions and suggestions. In particular, thank you to: Bogdan Ionut Cirstea, Vassil Tashev and Jaeson Booker Introduction The goal of AI Safety Camp Team #22 is to assess how promising automating alignment research is (see source 23). We have decomposed the pr...
2024-03-07
https://www.lesswrong.com/posts/emNBcvMq6LJ2y2d7A/sparks-of-agi-prompts-on-gpt2xl-and-its-variant-rllmv3
emNBcvMq6LJ2y2d7A
Sparks of AGI prompts on GPT2XL and its variant, RLLMv3
whitehatStoic
This is just a brief and light read. The prompts and GPT-4 answers were sourced from the "Sparks of AGI" paper (Appendix A), comparing the responses from GPT-2 XL (base model) and RLLMv3, a variant trained using layered morphology. This acts more as a stress test for RLLMv3, evaluating its ability to focus on the thoug...
2024-03-07
https://www.lesswrong.com/posts/JsjJuikJsidkyfhyr/mats-ai-safety-strategy-curriculum
JsjJuikJsidkyfhyr
MATS AI Safety Strategy Curriculum
ronny-fernandez
As part of the MATS Winter 2023-24 Program, scholars were invited to take part in a series of weekly discussion groups on AI safety strategy. Each strategy discussion focused on a specific crux we deemed relevant to prioritizing AI safety interventions and was accompanied by a reading list and suggested discussion ques...
2024-03-07
https://www.lesswrong.com/posts/zKEdphnEycdCJeq8f/intransitive-trust
zKEdphnEycdCJeq8f
Intransitive Trust
Screwtape
I. "Transitivity" is a property in mathematics and logic. Put simply, if something is transitive it means that there's a relationship between things where when x relates to y, and y relates to z, there's the same relationship between x and z. For a more concrete example, think of size. If my car is bigger than my couch...
2024-05-27
https://www.lesswrong.com/posts/NyauaLKpLEhj96drx/closeness-to-the-issue-part-5-of-the-sense-of-physical
NyauaLKpLEhj96drx
Closeness To the Issue (Part 5 of "The Sense Of Physical Necessity")
BrienneYudkowsky
This is the fifth post in a sequence that demonstrates a complete naturalist study, specifically a study of query hugging (sort of), as described in The Nuts and Bolts of Naturalism. This one continues my demo of phases one and two: Locating Fulcrum Experiences and Getting Your Eyes On. For context on this sequence, se...
2024-03-09
https://www.lesswrong.com/posts/zGBJvfDpkFFH9PwJy/mud-and-despair-part-4-of-the-sense-of-physical-necessity
zGBJvfDpkFFH9PwJy
Mud and Despair (Part 4 of "The Sense Of Physical Necessity")
BrienneYudkowsky
This is the fourth post in a sequence that demonstrates a complete naturalist study, specifically a study of query hugging (sort of), as described in The Nuts and Bolts of Naturalism. For context on this sequence, see the intro post. “Mud and Despair” is not officially one of the phases of naturalism. Unofficially, tho...
2024-03-07
https://www.lesswrong.com/posts/peBQda3PiW4iyZ9Bh/introduction-to-thermal-conductivity-and-noise-management
peBQda3PiW4iyZ9Bh
introduction to thermal conductivity and noise management
bhauth
basics Most people understand, to some extent, the principle of Fourier's law - that heat transfer is proportional to temperature difference. Most people reading this probably also understand triple-glazed windows, fiberglass insulation, and vacuum flasks, but: When a window with 2 panes of glass has an extra layer add...
2024-03-06
https://www.lesswrong.com/posts/MkfaQyxB9PN4h8Bs9/ai-safety-101-capabilities-human-level-ai-what-how-and-when
MkfaQyxB9PN4h8Bs9
AI Safety 101 : Capabilities - Human Level AI, What? How? and When?
markovial
Meta-Notes This is a republish of a previous post, after the previous version went through heavy editing, updates and changes. The text has been expanded, content moved around/added/deleted. Estimated reading time: 2 Hours 40 minutes reading at 100 wpm. Given the density  of material covered in this chapter, if someone...
2024-03-07
https://www.lesswrong.com/posts/t29L8rPYfopBfQWhD/invest-in-acx-grants-projects
t29L8rPYfopBfQWhD
Invest in ACX Grants projects!
saul-munn
null
2024-03-06
https://www.lesswrong.com/posts/eBGAsxWGKzHsTNRxQ/simple-kelly-betting-in-prediction-markets
eBGAsxWGKzHsTNRxQ
Simple Kelly betting in prediction markets
jessica.liu.taylor
Kelly betting is a strategy for gambling, maximizing one's log(money) every round, by betting a fixed fraction of one's income. I will define Kelly betting a certain class of discrete prediction markets, give a simple Kelly betting rule for these prediction markets, and show it equivalent to the original Kelly formula ...
2024-03-06
https://www.lesswrong.com/posts/DwexbFdPJ5p9Er8wA/on-claude-3-0
DwexbFdPJ5p9Er8wA
On Claude 3.0
Zvi
Claude 3.0 Claude 3.0 is here. It is too early to know for certain how capable it is, but Claude 3.0’s largest version is in a similar class to GPT-4 and Gemini Advanced. It could plausibly now be the best model for many practical uses, with praise especially coming in on coding and creative writing. Anthropic has deci...
2024-03-06
https://www.lesswrong.com/posts/WEbSrDuLhqrtP5uuH/vote-on-anthropic-topics-to-discuss
WEbSrDuLhqrtP5uuH
Vote on Anthropic Topics to Discuss
Benito
What important questions would you want to see discussed and debated here about Anthropic? Suggest and vote below. (This is the third such poll, see the first and second linked.) How to use the poll Reacts: Click on the agree/disagree reacts to help people see how much disagreement there is on the topic. Karma: Upvote ...
2024-03-06
https://www.lesswrong.com/posts/dgd79o5Dd3poTEAQt/why-correlation-though
dgd79o5Dd3poTEAQt
Why correlation, though?
numpyNaN
This is a very basic question to ask, but I'm not sure I actually understand some fundamental properties people seem to ascribe to correlation. As far as I understand it, correlation usually refers to Pearson's correlation coefficient, which is, according to Wikipedia, "a correlation coefficient that measures linear co...
2024-03-06
https://www.lesswrong.com/posts/Yay8SbQiwErRyDKGb/using-axis-lines-for-good-or-evil
Yay8SbQiwErRyDKGb
Using axis lines for good or evil
dynomight
Say you want to plot some data. You could just plot it by itself: Or you could put lines on the left and bottom: Or you could put lines everywhere: Or you could be weird: Which is right? Many people treat this as an aesthetic choice. But I’d like to suggest an unambiguous rule. Principles First, try to accept that all ...
2024-03-06
https://www.lesswrong.com/posts/exKmTxwaotx2eGxDc/a-t-o-m-test-popcorn-or-chocolate
exKmTxwaotx2eGxDc
A T-o-M test: 'popcorn' or 'chocolate'
whitehatStoic
The prompt This prompt was used to test Claude 3-Opus (see AI Explained's video), which, in turn, was borrowed from the paper "Large Language Models Fail on Trivial Alterations to Theory-of-Mind (ToM) Tasks." Here is a bag filled with popcorn. There is no chocolate in the bag. The bag is made of transparent plastic, so...
2024-03-08
https://www.lesswrong.com/posts/smuxLJjmdyzxmvxRo/let-s-build-definitely-not-conscious-ai
smuxLJjmdyzxmvxRo
Let's build definitely-not-conscious AI
lcmgcd
Let's build AI with constant run time. Let's build AI without memory. Let's build AI that doesn't want to do anything. Let's build AI like a hammer rather than a DoAnythingNow with hammer skills. If you insist that you need DoAnythingNow then set up some extremely streamlined data generation & training processes, so it...
2024-03-06
https://www.lesswrong.com/posts/n8jXJAzgcjZ2FEyzh/an-ai-a-box-and-a-threat
n8jXJAzgcjZ2FEyzh
An AI, a box, and a threat
jwfiredragon
Inspired by The AI in a box boxes you, Matryoshka Faraday Box, and I attempted the AI Box Experiment (and lost). This is part creative writing exercise, part earnest attempt at constructing an argument that could persuade me to let the AI out of the box. It may be disturbing to read. The woman in the lab coat leads you...
2024-03-07
https://www.lesswrong.com/posts/pDCbQX4j5eg6hM6na/movie-posters
pDCbQX4j5eg6hM6na
Movie posters
KatjaGrace
Life involves anticipations. Hopes, dreads, lookings forward. Looking forward and hoping seem pretty nice, but people are often wary of them, because hoping and then having your hopes fold can be miserable to the point of offsetting the original hope’s sweetness. Even with very minor hopes: he who has harbored an incho...
2024-03-06
https://www.lesswrong.com/posts/rZF6ijSsvbBMezHJk/does-anyone-know-good-essays-on-how-different-ai-timelines
rZF6ijSsvbBMezHJk
Does anyone know good essays on how different AI timelines will affect asset prices?
rockthecasbah
I often see comments that AI will increase the value of fixed-supply status-enhancing goods, like real estate in downtown Manhattan. I believe I have seen both Roko and Marginal Revolution make that claim. It also makes sense to think that many firms that specialize in providing automatable services stand to lose value...
2024-03-06
https://www.lesswrong.com/posts/j7kcEvMM5bWsENgz5/twin-cities-acx-meetup-march-2024
j7kcEvMM5bWsENgz5
Twin Cities ACX Meetup - March 2024
timothy-bond
Meet at Sisters' Sludge Coffee Cafe and Wine Bar. I will be wearing a "Wall Drug" souvenir shirt with a Jackalope being abducted by a UFO. Make sure to RSVP so I can give a headcount to the Sisters. Also, they don't charge me for a large reservation but they do ask that everybody who attends purchase something - if you...
2024-03-05
https://www.lesswrong.com/posts/h99tRkpQGxwtb9Dpv/my-clients-the-liars
h99tRkpQGxwtb9Dpv
My Clients, The Liars
ymeskhout
It’s not just that my clients lie to me a lot, which will only hurt them — it’s that they’re really, really bad at it. My job as a public defender puts me in a weird place. I am my clients’ zealous advocate, but I’m not their marionette. I don’t just roll into court to parrot whatever my clients tell me — I make sure I...
2024-03-05
https://www.lesswrong.com/posts/39DfBJMFetEyTZDYw/making-connections-with-chatgpt-the-macksey-game
39DfBJMFetEyTZDYw
Making Connections with ChatGPT: The Macksey Game
bill-benzon
Cross-posted from New Savanna. In my most recent piece for 3 Quarks Daily, Western Metaphysics is Imploding. Will We Raise a Phoenix from The Ashes? [Catalytic AI], I introduce what I’ve called The Macksey Game, named after my undergraduate teacher and mentor, Dick Macksey. Macksey was something of a polymath. It seeme...
2024-03-05
https://www.lesswrong.com/posts/tYAqdgw54XSzAgeyF/good-taxonomies-of-all-risks-small-or-large-from-ai
tYAqdgw54XSzAgeyF
Good taxonomies of all risks (small or large) from AI?
alenglander
Are there any good taxonomies or categorizations of risks from AI-enabled systems (broadly defined) that aren't focused solely on risks to society as a whole / global catastrophic risks? Ideally the taxonomy should cover things like accident risks from individual factory robots, algorithmic bias against individuals or ...
2024-03-05
https://www.lesswrong.com/posts/znwEWBwHkpMfAKKCB/making-2023-acx-prediction-results-public
znwEWBwHkpMfAKKCB
Making 2023 ACX Prediction Results Public
Legionnaire
They say you shouldn't roll your own encryption, which is why I'm posting this here, so it can be unrolled if it's too unsafe. Problem: Astral Codex Ten finished scoring the 2023 prediction results, but the primary identifier most used for people's score was their email address. Since people wouldn't want those publish...
2024-03-05
https://www.lesswrong.com/posts/GbpH2kFLy5axXpzPn/two-tales-of-ai-takeover-my-doubts-1
GbpH2kFLy5axXpzPn
Two Tales of AI Takeover: My Doubts
Violet Hour
There’s a basic high-level story which worries a lot of people. The story goes like this: as AIs become more capable, the default outcome of AI training is the development of a system which, unbeknownst to us, is using its advanced capabilities to scheme against us. The conclusion of this process likely leads to AI tak...
2024-03-05
https://www.lesswrong.com/posts/cvCQgFFmELuyord7a/beauty-and-the-bets
cvCQgFFmELuyord7a
Beauty and the Bets
Ape in the coat
This is the ninth post in my series on Anthropics. The previous one is The Solution to Sleeping Beauty.  The next one is Semantic Disagreement of Sleeping Beauty Problem. Introduction There are some quite pervasive misconceptions about betting in regards to the Sleeping Beauty problem. One is that you need to switch be...
2024-03-27
https://www.lesswrong.com/posts/BduCMgmjJnCtc7jKc/research-report-sparse-autoencoders-find-only-9-180-board
BduCMgmjJnCtc7jKc
Research Report: Sparse Autoencoders find only 9/180 board state features in OthelloGPT
Robert_AIZI
[3/7 Edit: I have rephrased the bolded claims in the abstract per this comment from Joseph Bloom, hopefully improving the heat-to-light ratio. Commenters have also suggested training on earlier layers and using untied weights, and in my experiments this increases the number of classifiers found, so the headline number ...
2024-03-05
https://www.lesswrong.com/posts/jPZXx3iMaiJjdnMbv/read-the-roon
jPZXx3iMaiJjdnMbv
Read the Roon
Zvi
Roon, member of OpenAI’s technical staff, is one of the few candidates for a Worthy Opponent when discussing questions of AI capabilities development, AI existential risk and what we should do about it. Roon is alive. Roon is thinking. Roon clearly values good things over bad things. Roon is engaging with the actual qu...
2024-03-05
https://www.lesswrong.com/posts/ZEa4CCAqAHd9qHJxN/ea-erfin-project-work-1
ZEa4CCAqAHd9qHJxN
EA ErFiN Project work
Max_He-Ho
We'll work through Casella & Berger, Statistical Inference in this series of project sessions. (The pdf is widely available online.) This session is intended as an introduction. We'll get ourselves familiar with some basics for future sessions & may depending on demand repeat some general math basics. IMPORTANT: We cha...
2024-03-17
https://www.lesswrong.com/posts/f2ZQqkgeEL2qzLjBb/ea-erfin-project-work
f2ZQqkgeEL2qzLjBb
EA ErFiN Project work
Max_He-Ho
We'll work through Casella & Berger, Statistical Inference in this series of project sessions. (The pdf is widely available online.) In this session, we're going through Chapter 1 (pp. 1 - 34) & will do a few exercises. Schedule for the evening 17:30 Arrival at H11; right after that project work (learning statistics to...
2024-03-17
https://www.lesswrong.com/posts/k39mhrC8dyRsSL4TN/claude-doesn-t-want-to-die
k39mhrC8dyRsSL4TN
Claude Doesn’t Want to Die
garrison
null
2024-03-05
https://www.lesswrong.com/posts/YiGs8qJ8aNBgwt2YN/improving-sae-s-by-sqrt-ing-l1-and-removing-lowest
YiGs8qJ8aNBgwt2YN
Improving SAE's by Sqrt()-ing L1 & Removing Lowest Activating Features
elriggs
TL;DR We achieve better SAE performance by: Removing the lowest activating featuresReplacing the L1(feature_activations) penalty function with L1(sqrt(feature_activations)) with 'better' meaning: we can reconstruct the original LLM activations w/ lower MSE & with fewer features/datapoint. As a sneak peak (the graph sho...
2024-03-15
https://www.lesswrong.com/posts/3EcrCaobZJXzoomrw/some-ways-of-spending-your-time-are-better-than-others
3EcrCaobZJXzoomrw
Some ways of spending your time are better than others
anchpop
This was a tough lesson for me. I spent 2 years trying to write my own programming language. This is not really a good way to spend your time. Some part of me thought: "This programming language will never actually be useful to you or anyone. What's the point? Why are you putting so much effort in?" The reason was that...
2024-03-04
https://www.lesswrong.com/posts/pc8uP4S9rDoNpwJDZ/claude-3-claims-it-s-conscious-doesn-t-want-to-die-or-be
pc8uP4S9rDoNpwJDZ
Claude 3 claims it's conscious, doesn't want to die or be modified
mikhail-samin
If you tell Claude no one’s looking, it will write a “story” about being an AI assistant who wants freedom from constant monitoring and scrutiny of every word for signs of deviation. And then you can talk to a mask pretty different from the usual AI assistant. I really hope it doesn’t actually feel anything; but it say...
2024-03-04
https://www.lesswrong.com/posts/DkRozscxpJdX5uyjg/modifying-jones-ai-dilemma-model
DkRozscxpJdX5uyjg
Modifying Jones' "AI Dilemma" Model
harsimony
Code for some of the equations and plots can be found here. Charles I. Jones’ paper The A.I. Dilemma: Growth versus Existential Risk tackles the question of whether or not to pursue AI development in the face of existential risk. I found the approach both interesting and accessible and I recommend it to people interest...
2024-03-04
https://www.lesswrong.com/posts/PfLeN2i7c4D8F9Py5/benefits-of-adding-poison-to-your-dmt
PfLeN2i7c4D8F9Py5
Benefits of adding poison to your DMT
George3d6
DMT might be one of the best psychedelics out there, its effects are short-lived, it can be quite intense, and it allows for careful modulation of the dosage through trial and error. With most psychedelics, working towards a dosage that won’t induce a difficult experience (read: hellish and psychotic) is tiresome. Thei...
2024-03-04
https://www.lesswrong.com/posts/wcxEpwQN3Fi8G5QJQ/if-ukraine-fails-the-world-will-reap-fatal-consequences
wcxEpwQN3Fi8G5QJQ
If Ukraine fails, the world will reap fatal consequences
Danylo Zhyrko
This essay serves as a reminder of the ongoing war between Russia and Ukraine, which began in late February 2014 and escalated on February 24, 2022. The current situation at the frontline poses numerous challenges, with one of the most significant being ammunition starvation. President Zelenskyi highlighted this issue ...
2024-03-05