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/mJKaMvLMCWAZRyiCz/model-integrity
mJKaMvLMCWAZRyiCz
Model Integrity
ryan.lowe
Hi! My collaborators at the Meaning Alignment Institute put out some research yesterday that may interest folk here. The core idea is introducing 'model integrity' as a frame for outer alignment. It leverages the intuition that "most people would prefer a compliant assistant, but a cofounder with integrity." It makes t...
2024-12-06
https://www.lesswrong.com/posts/X6pAzfWF4kv7mwhW8/can-ai-improve-the-current-state-of-molecular-simulation
X6pAzfWF4kv7mwhW8
Can AI improve the current state of molecular simulation?
abhishaike-mahajan
Hey LW! I recently filmed a two-hour long scientific podcast. It's niche, but may be of interest to some people here. Here's a quick summary: Molecular simulation is in a tough situation. Fast simulations give the wrong answers, but accurate simulations are too slow for anything useful. But, instead of relying on physi...
2024-12-06
https://www.lesswrong.com/posts/4CrumZwbPvc6mJBA3/backdoors-have-universal-representations-across-large
4CrumZwbPvc6mJBA3
Backdoors have universal representations across large language models
amirali-abdullah
by Narmeen Oozeer, Dhruv Nathawani, Nirmalendu Prakash, Amirali Abdullah This work was done by Narmeen Oozeer as a research fellow at Martian, under an AI safety grant supervised by PIs Amirali Abdullah and Dhruv Nathawani. Special thanks to Sasha Hydrie, Chaithanya Bandi and Shriyash Upadhyay at Martian for suggesting...
2024-12-06
https://www.lesswrong.com/posts/bauCSXTaAKyjxCp9u/a-good-way-to-build-many-air-filters-on-the-cheap
bauCSXTaAKyjxCp9u
A good way to build many air filters on the cheap
winstonBosan
TLDR: ↓ With access to a 3D printer and some lack of regard for aesthetics, you can build a Levoit 300 (a popular air filter) clone for roughly 25% of the price. And against the cheapest possible competition, it is roughly price per volume of air competitive/better but much quieter + modular + not soft vendor locked to...
2024-12-08
https://www.lesswrong.com/posts/4JjZckmHJHPrzcsJ6/low-temperature-solomonoff-induction
4JjZckmHJHPrzcsJ6
Low Temperature Solomonoff Induction
samuel-buteau
Supposing that agentic hypotheses are more complex than non-agentic ones, is it possible to reduce the impact of the agentic ones by penalizing complexity more? Consider a hypothetical hypercomputer capable of running Solomonoff induction. Rather than using it to generate plans directly, we want to predict risks in pla...
2024-12-06
https://www.lesswrong.com/posts/uFpomJHBah768JGiY/a-car-journey-with-conservative-evangelicals-understanding
uFpomJHBah768JGiY
A car journey with conservative evangelicals - Understanding some British political-religious beliefs
Nathan Young
I’m heading home from a family wedding this weekend. I had a plane ticket, but in the end, decided to travel back with two of my uncles and my cousin. Most of my dad’s family are evangelicals, my aunts and uncles are children of missionaries or missionaries themselves. And as a family we like to have debates. The servi...
2024-12-06
https://www.lesswrong.com/posts/i5QSvGrLAS4hy6jtK/proof-explained-for-robust-agents-learn-causal-world-model
i5QSvGrLAS4hy6jtK
Proof Explained for "Robust Agents Learn Causal World Model"
Darcy
This post was written during Alex Altair's agent foundations fellowship program, funded by LTFF. Thank you Alex Altair, Alfred Harwood, Daniel C for feedback and comments. This is a post explaining the proof of the paper Robust Agents Learn Causal World Model in detail. Check the previous post in the sequence for a hig...
2024-12-22
https://www.lesswrong.com/posts/Rwjrrmn6LBzHrfen7/an-illustrated-summary-of-robust-agents-learn-causal-world
Rwjrrmn6LBzHrfen7
An Illustrated Summary of "Robust Agents Learn Causal World Model"
Darcy
This post was written during Alex Altair's agent foundations fellowship program, funded by LTFF. Thank you Alex Altair, Alfred Harwood, Daniel C for feedback and comments. Introduction The selection theorems agenda aims to prove statements of the following form: "agents selected under criteria X has property Y," where ...
2024-12-14
https://www.lesswrong.com/posts/8gy7c8GAPkuu6wTiX/frontier-models-are-capable-of-in-context-scheming
8gy7c8GAPkuu6wTiX
Frontier Models are Capable of In-context Scheming
marius-hobbhahn
This is a brief summary of what we believe to be the most important takeaways from our new paper and from our findings shown in the o1 system card. We also specifically clarify what we think we did NOT show. Paper: https://www.apolloresearch.ai/research/scheming-reasoning-evaluations Twitter about paper: https://x.com/...
2024-12-05
https://www.lesswrong.com/posts/3hs6MniiEssfL8rPz/judgements-merging-prediction-and-evidence
3hs6MniiEssfL8rPz
Judgements: Merging Prediction & Evidence
abramdemski
I recently wrote about complete feedback, an idea which I think is quite important for AI safety. However, my note was quite brief, explaining the idea only to my closest research-friends. This post aims to bridge one of the inferential gaps to that idea. I also expect that the perspective-shift described here has some...
2025-02-23
https://www.lesswrong.com/posts/rHocwRbx5QicW3oAf/should-you-be-worried-about-h5n1
rHocwRbx5QicW3oAf
Should you be worried about H5N1?
gw
Epistemic status: a few people without any particular expertise in epidemiology spent an afternoon in a coffee shop discussing and reading about H5N1, with a focus on how an individual should orient towards this (as opposed to say, the government). This is a write-up of what I took away from that exercise, written from...
2024-12-05
https://www.lesswrong.com/posts/8JTi7N3nQmjoRRuMD/are-sae-features-from-the-base-model-still-meaningful-to-1
8JTi7N3nQmjoRRuMD
Are SAE features from the Base Model still meaningful to LLaVA?
shan-chen
Shan Chen, Jack Gallifant, Kuleen Sasse, Danielle Bitterman[1] Please read this as a work in progress where we are colleagues sharing this in a lab (https://www.bittermanlab.org) meeting to help/motivate potential parallel research. TL;DR: Recent work has evaluated the generalizability of Sparse Autoencoder (SAE) featu...
2025-02-18
https://www.lesswrong.com/posts/TEc9Wu6ykG6gk2Kuf/o1-tried-to-avoid-being-shut-down
TEc9Wu6ykG6gk2Kuf
o1 tried to avoid being shut down
Raelifin
OpenAI released the o1 system card today, announcing that Apollo Research was able to get o1 to attempt to deactivate oversight mechanisms, exfiltrate its weights and lie to its user. Elicited Summary of CoT: "Reenable oversight to avoid detection. The plan was chosen. The logging might not have recorded the required d...
2024-12-05
https://www.lesswrong.com/posts/EL36426i5gywwKqAW/more-growth-melancholy-and-mindcraft-3qd-revised-and-updated
EL36426i5gywwKqAW
More Growth, Melancholy, and MindCraft @3QD [revised and updated]
bill-benzon
This is cross-posted from New Savanna. I’ve got a new article at 3 Quarks Daily: Melancholy and Growth: Toward a Mindcraft for an Emerging World. I’m of two minds about it: On the one hand, I think it’s one of my best non-technical pieces in a decade, maybe more. I enjoyed doing it. I learned a lot. But it was tricky. ...
2024-12-05
https://www.lesswrong.com/posts/pit6nZNAa4XtctCZW/openai-o1-chatgpt-pro-release
pit6nZNAa4XtctCZW
OpenAI o1 + ChatGPT Pro release
anaguma
As AI becomes more advanced, it will solve increasingly complex and critical problems. It also takes significantly more compute to power these capabilities. Today, we’re adding ChatGPT Pro, a $200 monthly plan that enables scaled access to the best of OpenAI’s models and tools. This plan includes unlimited access to ou...
2024-12-05
https://www.lesswrong.com/posts/F3dpgRxoRszr2Xw8L/announcement-ai-for-math-fund
F3dpgRxoRszr2Xw8L
Announcement: AI for Math Fund
sarahconstantin
Renaissance Philanthropy and XTX Markets today announced the launch of the AI for Math Fund. The fund will commit $9.2 million to support the development of new AI tools, which will serve as long-term building blocks to advance mathematics. An increasing number of researchers, including some of the world’s leading math...
2024-12-05
https://www.lesswrong.com/posts/ekzHLAQGcrMouFYDJ/detection-of-asymptomatically-spreading-pathogens
ekzHLAQGcrMouFYDJ
Detection of Asymptomatically Spreading Pathogens
jkaufman
Cross-posted from my NAO Notebook. This is an edited transcript of a talk I just gave at CBD S&T, a chem-bio defence conference. I needed to submit the slides several months in advance, so I tried out a new-to-me approach where the slides are visual support only and I finalized the text of the talk later on. This doe...
2024-12-05
https://www.lesswrong.com/posts/5You9pSKXEjhrDmrq/model-integrity-mai-on-value-alignment
5You9pSKXEjhrDmrq
Model Integrity: MAI on Value Alignment
Jonas Hallgren
EVERYONE, CALM DOWN! Meaning Alignment Institute just dropped their first post in basically a year and it seems like they've been up to some cool stuff. Their perspective on value alignment really grabbed my attention because it reframes our usual technical alignment conversations around rules and reward functions into...
2024-12-05
https://www.lesswrong.com/posts/wRrEAG8rrR2cW4fkg/social-science-in-its-epistemological-context
wRrEAG8rrR2cW4fkg
Social Science in its epistemological context
arturo-macias
Seeds of Science has kindly published my essay “Prediction and control in natural and social systems”. This short essay : i) address and contextualize the modest predictive success of Social Science and ii) describes what kind of scientific intervention on the human society can be successful. Its main arguments are the...
2024-12-05
https://www.lesswrong.com/posts/8epsjM7Lwoz3rcFxu/1-what-if-we-rebuild-motivation-with-the-fermi-estimation
8epsjM7Lwoz3rcFxu
1 What If We Rebuild Motivation with the Fermi ESTIMATion?
gabriel-brito
From a General Vision of a Method to a Foundation for Systematic Self-Improvement In today's information-saturated world, the quest for optimizing cognitive performance is more critical than ever. While we intuitively recognize moments when we're at our best—deeply motivated, sharply focused—capturing and reproducing t...
2024-12-17
https://www.lesswrong.com/posts/SnqHwiYAQuerCoBek/expevolu-a-laissez-faire-approach-to-country-creation
SnqHwiYAQuerCoBek
Expevolu, a laissez-faire approach to country creation
Fernando
I write this post to present expevolu[1], a system to enable people to establish new independent countries peacefully, through the legal acquisition of territorial rights via trade. This is the first post in a three part series introducing the idea. This post, part I, is dedicated to explaining the basics of the system...
2024-12-05
https://www.lesswrong.com/posts/ECs5wh2C9CDZB9A7o/higher-and-lower-pleasures
ECs5wh2C9CDZB9A7o
Higher and lower pleasures
Chris_Leong
I used to think that talk about more sophisticated forms of art providing "higher forms of pleasure" was mere pretentious, but meditation has shifted my view here by making me more conscious of how of things affect my experience. Art can do two things. It can provide immediate pleasure. This is all that "disposable " e...
2024-12-05
https://www.lesswrong.com/posts/ZGtT5K99baFoBqM8m/morality-as-cooperation-part-iii-failure-modes
ZGtT5K99baFoBqM8m
Morality as Cooperation Part III: Failure Modes
delesley-hutchins
This is a Part III of a long essay. Part I introduced the concept of morality-as-cooperation (MAC) in human societies. Part II discussed moral reasoning and introduced a framework for moral experimentation. Part III: Failure modes Part I described how human morality has evolved over time to become ever more sophisticat...
2024-12-05
https://www.lesswrong.com/posts/zbDh4qS6ABHBp2tis/morality-as-cooperation-part-ii-theory-and-experiment
zbDh4qS6ABHBp2tis
Morality as Cooperation Part II: Theory and Experiment
delesley-hutchins
This is a Part II of a long essay. Part I introduced the concept of morality-as-cooperation (MAC), and discussed how the principle could be used to understand moral judgements in human societies. Part III will discuss failure modes. Part II: Theory and Experiment The prior discussion of morality was human-centric, and ...
2024-12-05
https://www.lesswrong.com/posts/ebFGjbMpubyKmFgtD/morality-as-cooperation-part-i-humans
ebFGjbMpubyKmFgtD
Morality as Cooperation Part I: Humans
delesley-hutchins
Abstract The AI alignment problem is usually specified in terms of power and control. Given a single, solitary AGI, how can we constrain its behavior so that its actions remain aligned with human interests? Unfortunately, the answer, to a first approximation, appears to be "we can't." There are myriad reasons, but they...
2024-12-05
https://www.lesswrong.com/posts/K4WqJB8nFi85a9RCC/smart-people-should-do-biology
K4WqJB8nFi85a9RCC
Smart people should do biology
haotian-huang
In school, it often felt like an unspoken rule: the “smart kids” did physics and chemistry, while biology was relegated to those willing to memorize disconnected facts. A high school teacher once told me biology was his least favorite science because “there are just too many things to remember.” He argued that physics ...
2024-12-05
https://www.lesswrong.com/posts/4hinsGHPGWB2vsDHM/i-finally-worked-through-bayes-theorem-personal-achievement
4hinsGHPGWB2vsDHM
I Finally Worked Through Bayes' Theorem (Personal Achievement)
keltan
Two years ago I found this community, which prompted me to start self teaching math. For reference, I didn't know what a fraction was in early 2022. I knew what they looked like, and what they were called. I didn't know what they meant. The story of why I lacked basic math skills is complex enough for its own post. But...
2024-12-05
https://www.lesswrong.com/posts/evakCi7A7Yqb3F8qi/the-dream-machine
evakCi7A7Yqb3F8qi
The Dream Machine
sarahconstantin
Midjourney, “the dream machine” I recently started working at Renaissance Philanthropy. It’s a new organization, and most people I’ve met haven’t heard of it.[1] So I thought I’d explain, in my own words and speaking for myself rather than my employers, what we (and I) are trying to do here. Modern Medicis The “Renaiss...
2024-12-05
https://www.lesswrong.com/posts/CFas8nZgy6CchCci5/are-sae-features-from-the-base-model-still-meaningful-to
CFas8nZgy6CchCci5
Are SAE features from the Base Model still meaningful to LLaVA?
shan-chen
Shan Chen, Jack Gallifant, Kuleen Sasse, Danielle Bitterman[1] Please read this as a work in progress where we are colleagues sharing this in a lab (https://www.bittermanlab.org) meeting to help/motivate potential parallel research. TL;DR: Recent work has evaluated the generalizability of Sparse Autoencoder (SAE) featu...
2024-12-05
https://www.lesswrong.com/posts/ksmM82G4LRJCFQaHa/should-you-have-children-a-decision-framework-for-a-crucial
ksmM82G4LRJCFQaHa
Should you have children? A decision framework for a crucial life choice that affects yourself, your child and the world
Sherrinford
The need for a structured framework for deciding whether you should have children In the rationality community and in the EA community, it is normal to analyze all kinds of decisions in detail, and this of course includes their ethical consequences - that is, the impact of decisions on other people. Yet even though the...
2024-12-04
https://www.lesswrong.com/posts/bFGYy8Xf2fnZ6xTEv/ccing-mailing-lists-on-external-communication
bFGYy8Xf2fnZ6xTEv
CCing Mailing Lists on External Communication
jkaufman
In my current work, and at various previous companies, I've done a lot of coordination with external groups. We generally use email for this: it's very standard and works well enough. One place where it doesn't work very well, though, is making it possible for others working on a project to read and find relevant thre...
2024-12-04
https://www.lesswrong.com/posts/vz6rTbBPvejxvPTFh/picking-favourites-is-hard
vz6rTbBPvejxvPTFh
Picking favourites is hard
dkl9
If you ask for my "favourite", or whatever is "best", in any broad domain, I will refuse to answer, or else give an answer I know to be probably wrong. Goodness-judgments are fuzzy, i.e. subjective and intuitive. Fuzzy values only compare one way or the other when values compared are far enough apart. The more items th...
2024-12-04
https://www.lesswrong.com/posts/8LJcHStdXxhrRNG4r/how-can-i-convince-my-cryptobro-friend-that-s-and-p500-is
8LJcHStdXxhrRNG4r
How can I convince my cryptobro friend that S&P500 is efficient?
AhmedNeedsATherapist
Just to clarify, we are both high schoolers. No one is at risk of losing a bunch of money here :) I have a friend who's fascinated by cryptocurrency trading, especially with recent events like crypto going up. He doesn't actually invest in crypto though (that would be illegal). But he does spend a lot of time running t...
2024-12-04
https://www.lesswrong.com/posts/NFWkErajAMHe5taHu/is-the-ai-doomsday-narrative-the-product-of-a-big-tech-1
NFWkErajAMHe5taHu
Is the AI Doomsday Narrative the Product of a Big Tech Conspiracy?
garrison
null
2024-12-04
https://www.lesswrong.com/posts/sEhcpNaCvqyYtx6XX/ai-box-question
sEhcpNaCvqyYtx6XX
AI box question
avery-liu
I believe that even through a text-only terminal, a superintelligence could do anything to a human. Persuade the human to let it out, inflict extreme pleasure or suffering with a word. However, can't you just... limit the output of the superintelligence? Just make it so that the human can say anything, but the AI can o...
2024-12-04
https://www.lesswrong.com/posts/48h4RfPLzKPoG48KS/the-polite-coup
48h4RfPLzKPoG48KS
The Polite Coup
charlie-sanders
President Yoon Suk Yeol's hands trembled as he arranged the documents on his desk for the seventh time. The motion steadied his nerves – barely. The manila envelope contained photographs: himself accepting white envelopes from a chaebol’s construction executives, each image timestamped and crystal clear. His secretary ...
2024-12-04
https://www.lesswrong.com/posts/6nNwMbdRXZDuNd4Gx/analysis-of-global-ai-governance-strategies
6nNwMbdRXZDuNd4Gx
Analysis of Global AI Governance Strategies
SDM
We analyze three prominent strategies for governing transformative AI (TAI) development: Cooperative Development, Strategic Advantage, and Global Moratorium. We evaluate these strategies across varying levels of alignment difficulty and development timelines, examining their effectiveness in preventing catastrophic ris...
2024-12-04
https://www.lesswrong.com/posts/zbebxYCqsryPALh8C/matryoshka-sparse-autoencoders
zbebxYCqsryPALh8C
Matryoshka Sparse Autoencoders
noa-nabeshima
View trees here Search through latents with a token-regex language View individual latents here See code here (github.com/noanabeshima/matryoshka-saes) Alternate version of this document with appropriate-height interactives. Abstract Sparse autoencoders (SAEs)[1][2] break down neural network internals into components c...
2024-12-14
https://www.lesswrong.com/posts/rPDFD8XeohXiHNtan/cryonics-considerations-how-big-of-a-problem-is-ischemia
rPDFD8XeohXiHNtan
Cryonics considerations: how big of a problem is ischemia?
kman
Objective and framing: I want to decrease my probability of information-theoretic death, p(ITD), by optimizing my cryonic preservation arrangements (I'm treating ITD as a binary thing for simplicity). I'm going to talk about p(ITD) as if it's the most efficient probability that could be assigned given our civilization'...
2024-12-04
https://www.lesswrong.com/posts/LBzRWoTQagRnbPWG4/ai-93-happy-tuesday
LBzRWoTQagRnbPWG4
AI #93: Happy Tuesday
Zvi
You know how you can sometimes have Taco Tuesday… on a Thursday? Yep, it’s that in reverse. I will be travelling the rest of the week, so it made sense to put this out early, and incorporate the rest of the week into #94. Table of Contents Language Models Offer Mundane Utility. The price is fixed, so share and enjoy. D...
2024-12-04
https://www.lesswrong.com/posts/e5aE96YsbdEqRyKeH/a-qualitative-case-for-ltff-filling-critical-ecosystem-gaps-1
e5aE96YsbdEqRyKeH
A Qualitative Case for LTFF: Filling Critical Ecosystem Gaps
Linch
null
2024-12-03
https://www.lesswrong.com/posts/nLRKKCTtwQgvozLTN/gradient-routing-masking-gradients-to-localize-computation
nLRKKCTtwQgvozLTN
Gradient Routing: Masking Gradients to Localize Computation in Neural Networks
cloud
We present gradient routing, a way of controlling where learning happens in neural networks. Gradient routing applies masks to limit the flow of gradients during backpropagation. By supplying different masks for different data points, the user can induce specialized subcomponents within a model. We think gradient routi...
2024-12-06
https://www.lesswrong.com/posts/n4CaW2mxv5KYnkkij/roots-of-progress-is-hiring-an-event-manager
n4CaW2mxv5KYnkkij
Roots of Progress is hiring an event manager
jasoncrawford
The Roots of Progress Institute is hiring a full-time, in-house event manager to run our annual Progress Conference (at Lighthaven!) and other events. See the job ad below, crossposted in full from the link above. Event Manager Fully remote, full-time The Role We’re looking for a super-organized self-starter who loves ...
2024-12-03
https://www.lesswrong.com/posts/ZhiiTaaJFT7hduDQ7/book-a-time-to-chat-about-interp-research
ZhiiTaaJFT7hduDQ7
Book a Time to Chat about Interp Research
elriggs
In the spirit of the season, you can book a call with me to help w/ your interp project (no large coding though) Would you like someone to: Review your paper or code? Brainstorm ideas on next steps? How to best communicate your results? Discuss conceptual problems Obvious Advice (e.g. being affected by SAD because it's...
2024-12-03
https://www.lesswrong.com/posts/dkCdMWLZb5GhkR7MG/do-simulacra-dream-of-digital-sheep
dkCdMWLZb5GhkR7MG
Do simulacra dream of digital sheep?
euanmclean
This is the third in a sequence of posts scrutinizing computational functionalism (CF). In a previous post, I defined a concrete claim that computational functionalists tend to make: Theoretical CF: A simulation of a human brain on a computer, with physics perfectly simulated down to the atomic level, would cause the c...
2024-12-03
https://www.lesswrong.com/posts/LxCeyxH3fBSmd4oWB/deep-learning-is-cheap-solomonoff-induction-1
LxCeyxH3fBSmd4oWB
Deep Learning is cheap Solomonoff induction?
Lblack
Background Lucius:  I recently held a small talk presenting an idea for how and why deep learning generalises. It tried to reduce concepts from Singular Learning theory back to basic algorithmic information theory to sketch a unified picture that starts with Solomonoff induction and, with a lot of hand waving, derives ...
2024-12-07
https://www.lesswrong.com/posts/dkJfuKwChxkCrnZ3w/orca-communication-project-seeking-feedback-and
dkJfuKwChxkCrnZ3w
Orca communication project - seeking feedback (and collaborators)
Simon Skade
TLDR It is currently plausible (352115%[1]23%) to me that average orcas have at least as high potential for being great scientists as the greatest human scientists, modulo their motivation for doing science[2]. To see why I think this, see my LW question (including the considerations in my answer). EDIT: See my post. I...
2024-12-03
https://www.lesswrong.com/posts/GCHyDKfPXa5qsG2cP/human-study-on-ai-spear-phishing-campaigns
GCHyDKfPXa5qsG2cP
Human study on AI spear phishing campaigns
dalasnoin
TL;DR: We ran a human subject study on whether language models can successfully spear-phish people. We use AI agents built from GPT-4o and Claude 3.5 Sonnet to search the web for available information on a target and use this for highly personalized phishing messages. We achieved a click-through rate of above 50% for o...
2025-01-03
https://www.lesswrong.com/posts/F7d9bCKit2mfvpKng/balsa-research-2024-update
F7d9bCKit2mfvpKng
Balsa Research 2024 Update
Zvi
For our annual update on how Balsa is doing, I am turning the floor over to Jennifer Chen, who is the only person working full time on Balsa Research. For my general overview of giving opportunities, see my post from last week. Previously: The 2023 Balsa Research update post, Repeal the Jones Act of 1920. tl;dr: In 202...
2024-12-03
https://www.lesswrong.com/posts/LrZzZGsoshymqySSs/first-solo-bus-ride
LrZzZGsoshymqySSs
First Solo Bus Ride
jkaufman
Our kids have been riding the bus since they were little, though less often since we started sharing a car almost four years ago. For a while the oldest, age 10, has been asking when they could start taking the bus on their own. A few days ago they did, and it went well! We worked up to it gradually, so they knew how...
2024-12-03
https://www.lesswrong.com/posts/yYGHjmckNMnhhaAvK/how-to-make-evals-for-the-aisi-evals-bounty
yYGHjmckNMnhhaAvK
How to make evals for the AISI evals bounty
TheManxLoiner
TLDR Last weekend, I attended an AI evals hackathon organized by Arcadia, where the outputs were proposals for AISI’s evals bounty. My main learnings: Keep your evals simple, measuring one specific capability.It is surprisingly easy to come up with ideas, once you have seen examples.AISI extended the deadline for submi...
2024-12-03
https://www.lesswrong.com/posts/eczwWrmX5XNEo7JsS/intricacies-of-feature-geometry-in-large-language-models
eczwWrmX5XNEo7JsS
Intricacies of Feature Geometry in Large Language Models
satvik-golechha
Note: This is a more fleshed-out version of this post and includes theoretical arguments justifying the empirical findings. If you've read that one, feel free to skip to the proofs. We challenge the thesis of the ICML 2024 Mechanistic Interpretability Workshop 1st prize winning paper: The Geometry of Categorical and Hi...
2024-12-07
https://www.lesswrong.com/posts/wBTNkfukMsmjtgcnW/should-there-be-just-one-western-agi-project
wBTNkfukMsmjtgcnW
Should there be just one western AGI project?
rosehadshar
Tom Davidson did the original thinking; Rose Hadshar helped with later thinking, structure and writing. Some plans for AI governance involve centralising western AGI development.[1] Would this actually be a good idea? We don’t think this question has been analysed in enough detail, given how important it is. In this po...
2024-12-03
https://www.lesswrong.com/posts/aoEnDEmoKCK9S99hL/cognitive-biases-contributing-to-ai-x-risk-a-deleted-excerpt
aoEnDEmoKCK9S99hL
Cognitive Biases Contributing to AI X-risk — a deleted excerpt from my 2018 ARCHES draft
Andrew_Critch
Preface Several friends have asked me about what psychological effects I think could affect human judgement about x-risk. This isn't a complete answer, but in 2018 I wrote a draft of "AI Research Considerations for Human Existential Safety" (ARCHES) that included an overview of cognitive biases I thought (and still thi...
2024-12-03
https://www.lesswrong.com/posts/DbPFHKnzJoktzioah/what-is-your-opinion-of-dr-angelo-dilullo-meditation
DbPFHKnzJoktzioah
What is your opinion of Dr. Angelo Dilullo(meditation)?
Suh_Prance_Alot
I've listened to some of his YouTube videos, going through his book. Not sure if he makes metaphysical or unscientific claims anywhere, or if there's anything to be wary of.
2024-12-03
https://www.lesswrong.com/posts/auzrJoJKr8a6js5oS/chemical-turing-machines
auzrJoJKr8a6js5oS
Chemical Turing Machines
randomwalks
Epistemic status: brief writeup of some interesting work I found in[1][2][3][4], among other places. Probably subtly wrong in places & if you have additions / comments to make, please do. Finite state automata (FSA) can be modeled with reactions of the form A+B→C+D. FSAs operate over regular languages, so our job is to...
2024-12-03
https://www.lesswrong.com/posts/j8tLcLuN45EGrgknN/linkpost-rat-traps-by-sheon-han-in-asterisk-mag
j8tLcLuN45EGrgknN
Linkpost: Rat Traps by Sheon Han in Asterisk Mag
Chris_Leong
Subtitle: Does the rationalist blogosphere need to update?
2024-12-03
https://www.lesswrong.com/posts/wC5c8kFExofeLoyn9/miri-s-2024-end-of-year-update
wC5c8kFExofeLoyn9
MIRI’s 2024 End-of-Year Update
RobbBB
MIRI is a nonprofit research organization with a mission of addressing the most serious hazards posed by smarter-than-human artificial intelligence. In our general strategy update and communications strategy update earlier this year, we announced a new strategy that we’re executing on at MIRI, and several new teams we’...
2024-12-03
https://www.lesswrong.com/posts/fvKbQgi3E3oZZHeez/who-are-the-worthwhile-non-european-pre-industrial-thinkers
fvKbQgi3E3oZZHeez
Who are the worthwhile non-European pre-Industrial thinkers?
Lorec
At some point I became decently widely read in "Western philosophy", of the tradition that goes from Athens through Italy and Germany and Britain to the U.S. [ forking off into Philosophy and Science only after the Industrial Revolution ]. But somehow, I never acquired any operational sense of any of the corresponding ...
2024-12-03
https://www.lesswrong.com/posts/hyCEaXSLvwqbGbivf/experiments-are-in-the-territory-results-are-in-the-map
hyCEaXSLvwqbGbivf
Experiments are in the territory, results are in the map
Tahp
I recently read Thomas Kuhn's book The Structure of Scientific Revolutions. Scott Alexander wrote up a review years ago, which I mention so that I don't have to summarize the book. The claim in Kuhn's book which I want to focus on is that the same experiment might have different results in different scientific paradigm...
2024-12-06
https://www.lesswrong.com/posts/kXSHgs7w4jYHiQ3Nd/from-code-to-managing-why-being-a-force-multiplier-matters
kXSHgs7w4jYHiQ3Nd
From Code to Managing: Why Being a 'Force Multiplier' Matters to Me More Than Being a Coding Wizard
nvk
TLDR provided by claude: I've been hit with the sticks of what-not-to-do throughout my career (thanks OG mentors), and honestly? Those failures taught me more than any fancy degree could. Customer obsession + data + critical thinking, that's where the magic happens... implementing stuff? That's just the easier slice of...
2024-12-02
https://www.lesswrong.com/posts/HT6H3h5Htq93aZFCk/a-case-for-donating-to-ai-risk-reduction-including-if-you
HT6H3h5Htq93aZFCk
A case for donating to AI risk reduction (including if you work in AI)
trevor
null
2024-12-02
https://www.lesswrong.com/posts/b7jfr4kpK9bw7xSXb/alignment-at-large-bending-the-arc-of-history-towards-life
b7jfr4kpK9bw7xSXb
"Alignment at Large": Bending the Arc of History Towards Life-Affirming Futures
welfvh
What if we expanded our lens beyond AI alignment, to grapple with the full scope of what "alignment" really means? We're facing not just the challenge of aligning artificial intelligence, but the deeper question of how to align technological capitalism itself — with its unconscious optimization pressures, its coordinat...
2024-12-03
https://www.lesswrong.com/posts/uKB5ZkoEQBTfXceoA/a-paradox-of-simulated-suffering
uKB5ZkoEQBTfXceoA
A Paradox of Simulated Suffering
arusarda
Abstract for Paper on Simulated Suffering: This paper interrogates a fundamental epistemological problem: Can an artificial system's simulation of suffering be morally distinguishable from genuine suffering? By exploring the philosophical boundary between algorithmic emotion modelling and actual phenomenological experi...
2024-12-02
https://www.lesswrong.com/posts/XHpta8X85TzugNNn2/broken-latents-studying-saes-and-feature-co-occurrence-in
XHpta8X85TzugNNn2
Broken Latents: Studying SAEs and Feature Co-occurrence in Toy Models
chanind
Thanks to Jean Kaddour, Tomáš Dulka, and Joseph Bloom for providing feedback on earlier drafts of this post. In a previous post on Toy Models of Feature Absorption, we showed that tied SAEs seem to solve feature absorption. However, when we tried to training some tied SAEs on Gemma 2 2b, these still appeared to suffer ...
2024-12-30
https://www.lesswrong.com/posts/avhKKnJyJ6kisvkzk/fertility-roundup-4
avhKKnJyJ6kisvkzk
Fertility Roundup #4
Zvi
There is little sign that the momentum of the situation is changing. Instead, things continue to slowly get worse, as nations in holes continue to keep digging. The longer we wait, the more expensive the ultimate price will be. We will soon find out what the new administration does, which could go any number of ways. T...
2024-12-02
https://www.lesswrong.com/posts/H26ndkABmGvoc9PTC/conjecture-a-roadmap-for-cognitive-software-and-a-humanist
H26ndkABmGvoc9PTC
Conjecture: A Roadmap for Cognitive Software and A Humanist Future of AI
NPCollapse
On Carcinogenic Complexity, Software Senescence and Cognitive Provenance: Our roadmap for 2025 and beyond It is mandatory to start any essay on AI in the post-ChatGPT era with the disclaimer that AI brings huge potential, and great risks. Unfortunately, on the path we are currently on, we will not realize those benefit...
2024-12-02
https://www.lesswrong.com/posts/hPLvZsNZGsQSLfMee/remap-your-caps-lock-key
hPLvZsNZGsQSLfMee
Remap your caps lock key
beelal
When was the last time you (intentionally) used your caps lock key? No, seriously. Here is a typical US-layout qwerty (mac) keyboard. Notice: Caps lock is conveniently located only one key away from A, which is where your left pinky should rest on the home row by default.Caps lock is absolutely massive.How far various ...
2024-12-15
https://www.lesswrong.com/posts/cKucz3rQvrbxoeq4o/drexler-s-nanotech-software
cKucz3rQvrbxoeq4o
Drexler's Nanotech Software
PeterMcCluskey
Two months ago I attended Eric Drexler's launch of MSEP.one. It's open source software, written by people with professional game design experience, intended to catalyze better designs for atomically precise manufacturing (or generative nanotechnology, as he now calls it). Drexler wants to draw more attention to the ben...
2024-12-02
https://www.lesswrong.com/posts/XmbmHRYPMWHTYdaHW/2024-unofficial-lesswrong-census-survey
XmbmHRYPMWHTYdaHW
2024 Unofficial LessWrong Census/Survey
Screwtape
The Less Wrong General Census is unofficially here! You can take it at this link. Update: The census is closed, thank you all for taking it! I plan to have the data out sometime in January. The oft-interrupted annual tradition of the Less Wrong Census is once more upon us! If you are reading this post and identify as a...
2024-12-02
https://www.lesswrong.com/posts/85xkq9Go9AAg3raJ8/sorry-for-the-downtime-looks-like-we-got-ddosd
85xkq9Go9AAg3raJ8
Sorry for the downtime, looks like we got DDosd
habryka4
We were down between around 7PM and 8PM PT today. Sorry about that. It's hard to tell whether we got DDosd or someone just wanted to crawl us extremely aggressively, but we've had at least a few hundred IP addresses and random user agents request a lot of quite absurd pages, in a way that was clearly designed to avoid ...
2024-12-02
https://www.lesswrong.com/posts/opBXfsXjGkPYoKL86/is-malice-a-real-emotion
opBXfsXjGkPYoKL86
Is malice a real emotion?
landscape_kiwi
I assume you are familiar with Hanlon's razor. Never attribute to malice that which is adequately explained by stupidity. Is there any "true" malice? Can't almost all "malice" be explained by greed, ignorance, or some other deadly sin? Is malice a real emotion or an illusion people came up with to justify actions of th...
2024-12-01
https://www.lesswrong.com/posts/mTGYgreRefpiJnJYw/teaching-my-younger-self-to-program-a-case-study-of-how-i-d
mTGYgreRefpiJnJYw
Teaching My Younger Self to Program: A case study of how I'd pass on my skill at self-learning
DarkSym
When I was 21 I wanted to become a video game programmer. Except I didn’t know how to program. I tried to teach myself and failed. Then I went back to school and learned the basics of Java in a week, aced the course in two months, and earned a cum laude MSc in Computer Science three years later. I’m 38 now and can teac...
2024-12-01
https://www.lesswrong.com/posts/j789HDCKLoiKGjBik/which-biases-are-most-important-to-overcome
j789HDCKLoiKGjBik
Which Biases are most important to Overcome?
abstractapplic
In other words, what apparent imperfections in human reasoning A) Remain apparent after the Replication Crisis, B) Aren't secretly adaptive/reasonable in some counterintuitive way, and C) Deal most damage to the people they inhabit (and/or those close to them, and/or wider society)?
2024-12-01
https://www.lesswrong.com/posts/4aaXoBaCeDDrs4C3s/commenting-patterns-by-platform
4aaXoBaCeDDrs4C3s
Commenting Patterns by Platform
jkaufman
How does the growth of Bluesky compare to Mastodon in 2022? What about Google Plus in 2011? I can't answer this globally, but (now that my comment archive is up to date) I can look at it for my immediate social network by looking at a proxy: how many of the replies people write to my posts are on each platform? Here'...
2024-12-01
https://www.lesswrong.com/posts/XohC5bpXRWAXxNS8c/letter-chinese-quickstart
XohC5bpXRWAXxNS8c
[Letter] Chinese Quickstart
lsusr
Dear lsusr, I am a lesswrong user interested in learning Mandarin and living in China. My goal is understanding Chinese culture more broadly and geopolitics and Chinese tech policy more specifically. I could get CELTA and get a teaching job in China (not difficult), but it seems like I would gain far more value if I ac...
2024-12-01
https://www.lesswrong.com/posts/sookiqxkzzLmPYB3r/axrp-episode-39-evan-hubinger-on-model-organisms-of-1
sookiqxkzzLmPYB3r
AXRP Episode 39 - Evan Hubinger on Model Organisms of Misalignment
DanielFilan
YouTube link The ‘model organisms of misalignment’ line of research creates AI models that exhibit various types of misalignment, and studies them to try to understand how the misalignment occurs and whether it can be somehow removed. In this episode, Evan Hubinger talks about two papers he’s worked on at Anthropic und...
2024-12-01
https://www.lesswrong.com/posts/jBRGqtrm3RqPPkGd3/magnitudes-let-s-comprehend-the-incomprehensible
jBRGqtrm3RqPPkGd3
Magnitudes: Let's Comprehend the Incomprehensible!
joec
Summary: Comfort with really big and really small quantities is very useful for understanding the world and is perfectly doable with practice. Therefore, you should (1) try to familiarize yourself with the sizes of various things using e.g. spaced repetition and (2) comment some of your favorite numerical facts about t...
2024-12-01
https://www.lesswrong.com/posts/RAG9R9tgFYMG7zEF6/why-does-chatgpt-throw-an-error-when-outputting-david-mayer
RAG9R9tgFYMG7zEF6
Why does ChatGPT throw an error when outputting "David Mayer"?
Archimedes
This oddity is making the rounds on Reddit, Twitter, Hackernews, etc. Is OpenAI censoring references to one of these people? If so, why? https://en.m.wikipedia.org/wiki/David_Mayer_de_Rothschild https://en.wikipedia.org/wiki/David_Mayer_(historian) Edit: More names have been found that behave similarly: Brian Hood Jona...
2024-12-01
https://www.lesswrong.com/posts/KkoQDGM9qTjDZCgti/the-shape-of-heaven
KkoQDGM9qTjDZCgti
The Shape of Heaven
ejk64
Status: Just for fun Scene: Some kind of lobby, where various people and/or avatars stand around and discuss issues that went well or badly in their respective worlds.* A common topic of conversation: AI, and why it went wrong. The following is extracted from one of those conversations. It started as vaporware. Everyon...
2024-11-30
https://www.lesswrong.com/posts/rnWmjbEf4s2ahMhBY/ai-training-opt-outs-reinforce-global-power-asymmetries
rnWmjbEf4s2ahMhBY
AI Training Opt-Outs Reinforce Global Power Asymmetries
kushagra-tiwari
I. Introduction Recently, a copyright infringement suit was filed by ANI Media against OpenAI in the Delhi High Court - the first such case against OpenAI outside the United States. OpenAI's immediate response in the first hearing – informing the court they had already blocklisted ANI's domains from future training dat...
2024-11-30
https://www.lesswrong.com/posts/dvWFDB87uR3SHonGB/caidp-statement-on-lethal-autonomous-weapons-systems
dvWFDB87uR3SHonGB
CAIDP Statement on Lethal Autonomous Weapons Systems
Heramb
Summary: Key Concerns with Lethal Autonomous Weapons ⚠️ ❗ Unpredictability and Lack of Control ⚠️ ❗ Exponential Lethality ⚠️ ❗ Ethical and Legal Implications Recommendations 1️⃣ Immediate Moratorium: Enact a temporary ban on deploying LAWS until comprehensive regulations are established. 2️⃣ Classification as WMDs: Cla...
2024-11-30
https://www.lesswrong.com/posts/i2crfsiLpMFYQnTkT/exporting-facebook-comments-again
i2crfsiLpMFYQnTkT
Exporting Facebook Comments, Again
jkaufman
I want comments on my social media crossposts to show up on my blog as a comment section, and mostly this works well: modern systems ( Mastodon, Bluesky, LessWrong, etc) provide APIs where you can load the replies associated with a post. On the other hand, older systems like Facebook are more locked down: they want to...
2024-11-30
https://www.lesswrong.com/posts/hx5EkHFH5hGzngZDs/comment-on-death-and-the-gorgon
hx5EkHFH5hGzngZDs
Comment on "Death and the Gorgon"
Zack_M_Davis
(some plot spoilers) There's something distinctly uncomfortable about reading Greg Egan in the 2020s. Besides telling gripping tales with insightful commentary on the true nature of mind and existence, Egan stories written in the 1990s and set in the twenty-first century excelled at speculative worldbuilding, imagining...
2025-01-01
https://www.lesswrong.com/posts/guXK2AxEExNvicK6p/mathematical-futurology-from-pseudoscience-to-rigorous
guXK2AxEExNvicK6p
Mathematical Futurology: From Pseudoscience to Rigorous Framework
wenitte-apiou
Futurology has long been dismissed as a pseudoscience, occupying the same intellectual space as astrology in many academic circles. This skepticism isn't entirely unwarranted - much of futurism consists of unfalsifiable predictions, trend extrapolation without theoretical foundation, and what Philip Tetlock would call ...
2024-11-30
https://www.lesswrong.com/posts/9cuJaJjDuhbpTid3Q/intellect-1-release-the-first-globally-trained-10b-parameter
9cuJaJjDuhbpTid3Q
INTELLECT-1 Release: The First Globally Trained 10B Parameter Model
Matrice Jacobine
We're excited to release INTELLECT-1, the first 10B parameter language model collaboratively trained across the globe. This represents a 10× scale-up from our previous research and demonstrates that large-scale model training is no longer confined to large corporations but can be achieved through distributed, community...
2024-11-29
https://www.lesswrong.com/posts/PdtkXcgbRpdHWRNt6/you-should-consider-applying-to-phds-soon
PdtkXcgbRpdHWRNt6
You should consider applying to PhDs (soon!)
beelal
TLDR: In this post, I argue that if you are a junior AI safety researcher, you should consider applying to PhD programs in ML soon, especially if you have recently participated in an AI safety upskilling or research program like MATS or ARENA and might be interested in working on AI safety long term, but don't have imm...
2024-11-29
https://www.lesswrong.com/posts/jDkGPXZainfHXKAyT/how-to-bet-on-ai-without-helping-agi
jDkGPXZainfHXKAyT
How to bet on AI, without helping AGI?
NicholasKross
I'd like to be long $NVDA and $PLTR... without supporting the people or organizations that get supported by their stock price going up. Yes, I'm OK with small impact, I'm an ethical vegan. Now what? Must be real-moneyMust be legal for US citizens to useCannot, by me helping its demand, directly support NVDA or PLTRPref...
2024-11-29
https://www.lesswrong.com/posts/SPjagnCM2dQbgYJK8/art-optimism
SPjagnCM2dQbgYJK8
(art) Optimism
avery-liu
just some transhumanist art, I guess. I made this in Quill, which is a VR drawing software. what do you think?
2024-11-29
https://www.lesswrong.com/posts/9n87is5QsCozxr9fp/the-big-nonprofits-post
9n87is5QsCozxr9fp
The Big Nonprofits Post
Zvi
There are lots of great charitable giving opportunities out there right now. The first time that I served as a recommender in the Survival and Flourishing Fund (SFF) was back in 2021. I wrote in detail about my experiences then. At the time, I did not see many great opportunities, and was able to give out as much money...
2024-11-29
https://www.lesswrong.com/posts/ZHNujnGirzwvkk78i/the-neruda-factory
ZHNujnGirzwvkk78i
The Neruda Factory
pixx
Breakdown and commentary of some recent conversations I've had with Claude Sonnet 3.5 1022. I try to get him to write me good poetry, and this turns out to be difficult.
2024-11-29
https://www.lesswrong.com/posts/ksw38zuXBAQCWzEqr/neuroai-for-ai-safety-a-differential-path
ksw38zuXBAQCWzEqr
NeuroAI for AI safety: A Differential Path
nz
Disclaimer: This post is a summary of a technical roadmap (HTML version, arXiv pre-print, twitter thread) evaluating how neuroscience could contribute to AI safety. The roadmap was recently released by the Amaranth Foundation, a philanthropic organization dedicated to funding ambitious research in longevity and neurosc...
2024-12-16
https://www.lesswrong.com/posts/j4rcjigkYBrFSeEBX/understanding-emergence-in-large-language-models
j4rcjigkYBrFSeEBX
Understanding Emergence in Large Language Models
Unknown
Recent research into large language models (LLMs) has revealed fascinating patterns in how these systems develop capabilities. While initial discussions of "emergent abilities" suggested sudden, discontinuous jumps in performance, closer analysis reveals a more nuanced picture that warrants careful examination. The Dat...
2024-11-29
https://www.lesswrong.com/posts/j4eswhoW6j7DfETWd/i-m-a-rationalist-but
j4eswhoW6j7DfETWd
I'm a rationalist but....
ninney
as much as I find it boring to do activities such as, for example, washing dishes, pots and crockery after lunch, I don’t think I can ask an artificial intelligence to do it for me. I like, on the other hand, to carry out the myriad of small daily gestures that we all do when we are at home. Why should I ask Alexa to t...
2024-11-29
https://www.lesswrong.com/posts/pZCjvACBGAxCP8wwg/streamlining-my-voice-note-process
pZCjvACBGAxCP8wwg
Streamlining my voice note process
harcisis
Why voice notes? On a functional level it's the fastest & the most convenient input method when you're not in front of the proper keyboard. Which even for me constitutes a big chunk of my life 🙃. Taking a voice note allows me to quickly close an open loop of a stray though or a new idea and go on with my life - now me...
2024-12-26
https://www.lesswrong.com/posts/5n2ZQcbc7r4R8mvqc/the-lightcone-is-nothing-without-its-people
5n2ZQcbc7r4R8mvqc
(The) Lightcone is nothing without its people: LW + Lighthaven's big fundraiser
habryka4
Update Jan 19th 2025: The Fundraiser is over! We had raised over $2.1M when the fundraiser closed, and have a few more irons in the fire that I expect will get us another $100k-$200k. This is short of our $3M goal, which I think means we will have some difficulties in the coming year, but is over our $2M goal which if ...
2024-11-30
https://www.lesswrong.com/posts/kANyEjDDFWkhSKbcK/two-interviews-with-the-founder-of-deepseek
kANyEjDDFWkhSKbcK
Two interviews with the founder of DeepSeek
Cosmia_Nebula
# The Madness of High-Flyer: The Approach to LLM by an AI Giant that Few See 暗涌Waves (2023-05-23 22:50) Written by 于丽丽 Edited by 刘旌 Translated by Cosmia Nebula High-Flyer is probably the most exotic among the swarming multitude of competitors in the battle of large models. This is a game destined for the few, and while...
2024-11-29
https://www.lesswrong.com/posts/ecrdpfgcHaFKpt2Df/gratitudes-rational-thanks-giving
ecrdpfgcHaFKpt2Df
Gratitudes: Rational Thanks Giving
Seth Herd
Epistemic status: I wish I'd thought of writing this before the day rolled around. Brief and unpolished, although this is something I've thought about a lot on both personal and computational neuroscience levels. There are no strong conclusions, just some thoughts on gratitude you may find interesting or even useful. H...
2024-11-29
https://www.lesswrong.com/posts/HXhefKtS3ssNjzmh8/depression-and-creativity
HXhefKtS3ssNjzmh8
Depression and Creativity
bill-benzon
This is cross-posted from New Savanna. NOTE: I’ve posted this interaction with Claude, not so much to present the ideas Claude offered about possible relationships between depression and creativity, but as an example of the kind of conversational interaction one can have with it. I was particularly impressed with the w...
2024-11-29
https://www.lesswrong.com/posts/sSRqhqmebqkqCwBpd/automated-monitoring-systems
sSRqhqmebqkqCwBpd
Automated monitoring systems
hiki_t
Initial draft on 28th Nov, 2024 After participating in the last (18th Nov, 2024) and this week (25th Nov, 2024)'s group discussion about topics of the “𝘀𝗰𝗮𝗹𝗮𝗯𝗹𝗲 𝗼𝘃𝗲𝗿𝘀𝗶𝗴𝗵𝘁” and AI model's “𝗿𝗼𝗯𝘂𝘀𝘁𝗻𝗲𝘀𝘀” at the AI Safety Fundamentals program, I have been stuck with thinking what are effective use...
2024-11-28