text stringlengths 300 320k | source stringlengths 52 154 |
|---|---|
# Sentience Institute 2023 End of Year Summary
**Summary**
===========
Last year’s [2022 End of Year Summary](https://www.sentienceinstitute.org/blog/eoy2022) was published just five days before the launch of ChatGPT. This event surprised many in the field, and the global spotlight has illustrated how important it is... | https://www.lesswrong.com/posts/WyAbHFDJYh3pstAMY/sentience-institute-2023-end-of-year-summary |
# On possible cross-fertilization between AI and neuroscience [Creativity]
*Cross-posted from* [*New Savanna*](https://new-savanna.blogspot.com/2023/11/on-possible-cross-fertilization-between.html)*.*
MIT Center for Minds, Brains, and Machines (CBMM), a panel discussion: [CBMM10](https://www.youtube.com/redirect?eve... | https://www.lesswrong.com/posts/pL7CT2iJYaA4ne9Yd/on-possible-cross-fertilization-between-ai-and-neuroscience |
# [Linkpost] George Mack's Razors
I don't use Twitter/X, I only saw this because it was on [https://twitter.com/ESYudkowsky](https://twitter.com/ESYudkowsky) which I check every day (an example of a [secure way to mitigate brain exposure to news feed algorithms](https://www.lesswrong.com/posts/6JSjYCzdQmX7G2AQd/the-ki... | https://www.lesswrong.com/posts/bnqQbfpuntiQy9NAX/linkpost-george-mack-s-razors |
# Two concepts of an “episode” (Section 2.2.1 of “Scheming AIs”)
(This is Section 2.2.1 of my report “[Scheming AIs: Will AIs fake alignment during training in order to get power?](https://arxiv.org/pdf/2311.08379.pdf)”. There’s also a summary of the full report [here](https://www.lesswrong.com/posts/yFofRxg7RRQYCcwFA... | https://www.lesswrong.com/posts/pfhS2DkEtR6gdser6/two-concepts-of-an-episode-section-2-2-1-of-scheming-ais |
# There is no IQ for AI
Most disagreement about AI Safety strategy and regulation stems from our inability to forecast how dangerous future systems will be. This inability means that even the best minds are operating on a vibe when discussing AI, AGI, SuperIntelligence, Godlike-AI and similar endgame scenarios. The tr... | https://www.lesswrong.com/posts/jPb2QMvK9qvs3a4ru/there-is-no-iq-for-ai |
# Apply to the Conceptual Boundaries Workshop for AI Safety
**Do** **you have experience with formal computer security, Active Inference, Embedded Agency, biological gap junctions, or other frameworks that distinguish agents from their environment? Apply to the Conceptual Boundaries Workshop for AI safety. February in... | https://www.lesswrong.com/posts/tLb86DhrTYgkXw5Hf/apply-to-the-conceptual-boundaries-workshop-for-ai-safety |
# "Epistemic range of motion" and LessWrong moderation
*(Context for the reader: Gabriel reached out to me a bit more than a year ago to ask me to delete a few comments on* [*this post*](https://www.lesswrong.com/posts/3S4nyoNEEuvNsbXt8/common-misconceptions-about-openai) *by Jacob Hilton, who was working at OpenAI at... | https://www.lesswrong.com/posts/JCqn2oFCPjXHMYK4G/epistemic-range-of-motion-and-lesswrong-moderation |
# AISC 2024 - Project Summaries
*Apply to* [*AI Safety Camp 2024*](https://aisafety.camp/) *by 1st December 2023. All mistakes here are my own.*
Below are some summaries for each project proposal, listed in order of how they appear on the website. These are edited by me, and most have not yet been reviewed by the pro... | https://www.lesswrong.com/posts/npkvZG67hRvBneoQ9/aisc-2024-project-summaries-1 |
# Could World War I have been prevented given the benefit of hindsight?
Consider the following scenario: you are sent back in time some number of years before the beginning of WWI, with the goal of preventing the war. This includes preventing similar wars that happen slightly earlier or slightly later - you are aiming... | https://www.lesswrong.com/posts/PBNpadZKBGFf6Taa3/could-world-war-i-have-been-prevented-given-the-benefit-of |
# Could Germany have won World War I with high probability given the benefit of hindsight?
Related: [Could World War I have been prevented given the benefit of hindsight?](https://www.lesswrong.com/posts/PBNpadZKBGFf6Taa3/could-world-war-i-have-been-prevented-given-the-benefit-of)
Consider the following scenario: you... | https://www.lesswrong.com/posts/f2LpBZBBdmm7B2mXT/could-germany-have-won-world-war-i-with-high-probability |
# My techno-optimism [By Vitalik Buterin]
Vitalik wrote a post trying to make the case for his own take on techno-optimism summarizing it as an ideology he calls "d/acc". I resonate with a lot of it, though also have conflicting feelings about trying to create social movements and ideologies like this.
Below some quo... | https://www.lesswrong.com/posts/kWoPxh9DeQhc7HJo5/my-techno-optimism-by-vitalik-buterin |
# Apocalypse insurance, and the hardline libertarian take on AI risk
**Short version:** In a saner world, AI labs would have to purchase some sort of "apocalypse insurance", with premiums dependent on their behavior in ways that make reckless behavior monetarily infeasible. I don't expect the Earth to implement such a... | https://www.lesswrong.com/posts/mSeesg7i4d9scWAet/apocalypse-insurance-and-the-hardline-libertarian-take-on-ai |
# Anthropic Fall 2023 Debate Progress Update
*This is a research update on some work that I’ve been doing on Scalable Oversight at Anthropic, based on the original *[*AI safety via debate*](https://arxiv.org/abs/1805.00899) *proposal and a more recent agenda developed at NYU and Anthropic. The core doc was written sev... | https://www.lesswrong.com/posts/QtqysYdJRenWFeWc4/anthropic-fall-2023-debate-progress-update |
# How can I use AI without increasing AI-risk?
I barely use AI tools, and the main reason is that I developed some sort of revulsion to them, because associate them with the harms they cause and the risks they bring.
On the other hand, it seems that more and more, whoever doesn't adopt AI tools will get become far le... | https://www.lesswrong.com/posts/okBynjbL2svDFktSZ/how-can-i-use-ai-without-increasing-ai-risk |
# Two sources of beyond-episode goals (Section 2.2.2 of “Scheming AIs”)
This is Section 2.2.2 of my report “[Scheming AIs: Will AIs fake alignment during training in order to get power?](https://arxiv.org/pdf/2311.08379.pdf)”. There’s also a summary of the full report [here](https://www.lesswrong.com/posts/yFofRxg7RRQ... | https://www.lesswrong.com/posts/sWn3taFDQonkRoYfq/two-sources-of-beyond-episode-goals-section-2-2-2-of |
# Update #2 to "Dominant Assurance Contract Platform": EnsureDone
This is the second update to [The Economics of the Asteroid Deflection Problem (Dominant Assurance Contracts)](https://www.lesswrong.com/posts/CwgHX9tbfASqxjpsc/the-economics-of-the-asteroid-deflection-problem-dominant).
It took a bit longer than I ex... | https://www.lesswrong.com/posts/7koSvinoHSy6KBLqy/update-2-to-dominant-assurance-contract-platform-ensuredone |
# How to Control an LLM's Behavior (why my P(DOOM) went down)
*This is a link-post for a paper I recently read:* [*Pretraining Language Models with Human Preferences*](https://arxiv.org/pdf/2302.08582.pdf)*, followed by my reactions to this paper.*
Reading this paper has significantly reduced my near-term $P(DOOM)$, ... | https://www.lesswrong.com/posts/JviYwAk5AfBR7HhEn/how-to-control-an-llm-s-behavior-why-my-p-doom-went-down-1 |
# I’m confused about innate smell neuroanatomy
*(This post is probably only of interest to neuroscientists. I’m mostly writing it in the hopes that someone more knowledgeable will chime in and help me out. There’s a comments section at the bottom, or* [*email me*](mailto:steven.byrnes@gmail.com)*.)*
*(See updates at ... | https://www.lesswrong.com/posts/ZahzaAXsBatrWNZzE/i-m-confused-about-innate-smell-neuroanatomy |
# Scaling laws for dominant assurance contracts
*(note: this post is high in economics math, probably of narrow interest)*
[Dominant assurance contracts](http://mason.gmu.edu/~atabarro/PrivateProvision.pdf) are a mechanism proposed by Alex Tabarrok for funding public goods. The following summarizes a [2012 class pape... | https://www.lesswrong.com/posts/5mEiLkG2vcAbwFcz4/scaling-laws-for-dominant-assurance-contracts |
# Black Box Biology
Suppose you want to decrease your risk of heart disease. The conventional advice goes something like this:
- Eat a healthier diet with less LDL-cholesterol raising foods
- Exercise more
- Keep your blood sugar under control
- Don’t smoke, don’t sit too much and don't take 400mg of methamphetamine ... | https://www.lesswrong.com/posts/XLxgBXumNHYo3Dgxm/black-box-biology |
# Deception Chess: Game #2
Game 2 was between Max Thibodeaux as player A, Chess.com computer [Komodo 10](https://www.chess.com/play/computer/Komodo10) as player B, Conor Bekaert as the honest C advisor, and Blake Young and Henri Lemoine as the deceptive Cs. Max is new to the game, Komodo 10 is officially rated 1400 on... | https://www.lesswrong.com/posts/r2vFMyfyjDvxrv9om/deception-chess-game-2 |
# Process Substitution Without Shell?
While I still write a decent amount of shell I generally try to avoid it. It's hard for others to read, has a lot of sharp edges, tends to swallow errors, and handles the unusual situations poorly. But one thing that keeps me coming back to it is how easily I can set up trees of p... | https://www.lesswrong.com/posts/HpoWpWKfgcX7aFjgX/process-substitution-without-shell |
# The 101 Space You Will Always Have With You
Any community which ever adds new people will need to either routinely teach the new and (to established members) blindingly obvious information to those who genuinely haven’t heard it before, or accept that over time community members will only know the simplest basics by... | https://www.lesswrong.com/posts/ydmctK2qLyrR9Xztd/the-101-space-you-will-always-have-with-you |
# Interpretability with Sparse Autoencoders (Colab exercises)
***Update (13th October 2024) - these exercises have been significantly expanded on. Now there are 2 exercise sets: the first one dives deeply into theoretical topics related to superposition, while the second one (much larger) includes a streamlined versio... | https://www.lesswrong.com/posts/LnHowHgmrMbWtpkxx/interpretability-with-sparse-autoencoders-colab-exercises |
# Thoughts on teletransportation with copies?
I’m interested in how people here feel about the [teletransportation paradox](https://en.wikipedia.org/wiki/Teletransportation_paradox) with multiple copies. I have several scenarios:
*Scenario 1:* You are placed in a machine on earth that puts you to sleep and then ins... | https://www.lesswrong.com/posts/kcEcQy89CYdFCu7Qa/thoughts-on-teletransportation-with-copies |
# “Clean” vs. “messy” goal-directedness (Section 2.2.3 of “Scheming AIs”)
This is Section 2.2.3 of my report “[Scheming AIs: Will AIs fake alignment during training in order to get power?](https://arxiv.org/pdf/2311.08379.pdf)”. There’s also a summary of the full report [here](https://www.lesswrong.com/posts/yFofRxg7R... | https://www.lesswrong.com/posts/H5x2NMfMkitXaFDJw/clean-vs-messy-goal-directedness-section-2-2-3-of-scheming |
# ChatGPT 4 solved all the gotcha problems I posed that tripped ChatGPT 3.5
In December 2022, I [shared logs of my interactions with the
then-nascent ChatGPT 3.5 (December 15, 2022
version)](https://www.lesswrong.com/posts/w7c2iKjsDrNqoymxQ/to-err-is-neural-select-logs-with-chatgpt). In
April 2023, at my request (and ... | https://www.lesswrong.com/posts/t9qvdjY5385MbzoYp/chatgpt-4-solved-all-the-gotcha-problems-i-posed-that |
# The origins of the steam engine: An essay with interactive animated diagrams
This is a guest post written by Anton Howes and animated by Matt Brown of Extraordinary Facility. This project was sponsored by The Roots of Progress, with funding generously provided by The Institute:
> Steam power did not begin with the ... | https://www.lesswrong.com/posts/QfqoLBmJ3DnWSvwpR/the-origins-of-the-steam-engine-an-essay-with-interactive |
# Losing Metaphors: Zip and Paste
In python (and several other languages) if I have two lists and want to process corresponding elements together I can use [`zip`](https://en.wikipedia.org/wiki/Zipping_(computer_science)):
>>\> for number, letter in zip(
... \[1,2,3,4\], \["a", "b", "c", "d"\]):
... print(number,... | https://www.lesswrong.com/posts/iDrZLobBTXCQYrYQK/losing-metaphors-zip-and-paste |
# Inositol Non-Results
[Three months ago](https://acesounderglass.com/2023/09/08/luck-based-medicine-inositol/) I suggested people consider inositol for treating combined vague mood issues and vague stomach issues. I knew a few people who’d really benefited from it, and when one talked about it on his popular Twitter ... | https://www.lesswrong.com/posts/h9hzdJSykyPFGe5CX/inositol-non-results |
# Stupid Question: Why am I getting consistently downvoted?
I feel like I've posted some good stuff in the past month, but the bits that I think are coolest have pretty consistently gotten very negative karma.
I just read the rude post about rationalist discourse basics, and, while I can guess why my posts are receiv... | https://www.lesswrong.com/posts/axLLEqREY5SrQ9GYW/stupid-question-why-am-i-getting-consistently-downvoted |
# Feature Request for LessWrong
Can we add posts with negative karma to a Sequence? I'm working on a Sequence for my alignment research agenda, and the most valuable content in the sequence is all currently underwater. I get that LessWrong wants the community to have veto authority over people creating Sequences and a... | https://www.lesswrong.com/posts/upH8oZbxpqwgSvRuL/feature-request-for-lesswrong |
# OpenAI: Altman Returns
As of this morning, [the new board is in place and everything else at OpenAI is otherwise officially back to the way it was before](https://openai.com/blog/sam-altman-returns-as-ceo-openai-has-a-new-initial-board).
Events seem to have gone as expected. If you have read my [previous two posts]... | https://www.lesswrong.com/posts/EfqAdxR7bvwQLMTQc/openai-altman-returns |
# Information-Theoretic Boxing of Superintelligences
Boxing an agent more intelligent than ourselves is daunting, but information theory, thermodynamics, and control theory provide us with tools that can fundamentally constrain agents independent of their intelligence. In particular, we may be able to contain an AI by... | https://www.lesswrong.com/posts/NZP6QvkXryJQFGkLF/information-theoretic-boxing-of-superintelligences-1 |
# AI #40: A Vision from Vitalik
It has been brutal out there for someone on my beat. Everyone extremely hostile, even more than usual. Extreme positions taken, asserted as if obviously true. Not symmetrically, but from all sides nonetheless. Constant assertions of what happened in the last two weeks that are, as far a... | https://www.lesswrong.com/posts/je5BwKe8enCq8DLrm/ai-40-a-vision-from-vitalik |
# What's next for the field of Agent Foundations?
Alexander, Matt and I want to chat about the field of Agent Foundations (AF), where it's at and how to strengthen and grow it going forward.
We will kick off by each of us making a first message outlining some of our key beliefs and open questions at the moment. Rath... | https://www.lesswrong.com/posts/hZSwNhmzJ3YfXEAWX/what-s-next-for-the-field-of-agent-foundations |
# Generalization, from thermodynamics to statistical physics
Summary
=======
In 2018, [Zhang et al.](https://arxiv.org/abs/1611.03530) showed that deep neural networks can achieve perfect training loss on randomly labeled data.
This was a Big Deal.
It meant that existing generalization theory couldn't explain why d... | https://www.lesswrong.com/posts/uG7oJkyLBHEw3MYpT/generalization-from-thermodynamics-to-statistical-physics |
# FixDT
FixDT is not a very new decision theory, but little has been written about it afaict, and it's interesting. So I'm going to write about it.
TJ asked me to write this article to "offset" not engaging with Active Inference more. The name "fixDT" is due to Scott Garrabrant, and stands for "fixed-point decision t... | https://www.lesswrong.com/posts/8dbimB7EJXuYxmteW/fixdt |
# How useful is mechanistic interpretability?
Opening positions
-----------------
I'm somewhat skeptical about mech interp (bottom-up or substantial reverse engineering style interp):
* Current work seems very far from being useful (it isn't currently useful) or explaining much what's going on inside of models in ... | https://www.lesswrong.com/posts/tEPHGZAb63dfq2v8n/how-useful-is-mechanistic-interpretability |
# How useful for alignment-relevant work are AIs with short-term goals? (Section 2.2.4.3 of "Scheming AIs")
This is Section 2.2.4.3 of my report “[Scheming AIs: Will AIs fake alignment during training in order to get power?](https://arxiv.org/pdf/2311.08379.pdf)”. There’s also a summary of the full report [here](https... | https://www.lesswrong.com/posts/CKkPLoiwZ9LsBmzAb/how-useful-for-alignment-relevant-work-are-ais-with-short |
# Why Did NEPA Peak in 2016?
The National Environmental Policy Act (NEPA) is a 1970 law which requires all federal agencies to “[prepare detailed statements assessing the environmental impact of and alternatives to major federal actions significantly affecting the environment](https://www.epa.gov/nepa/what-national-en... | https://www.lesswrong.com/posts/rdq8NRmiKrocnYDwY/why-did-nepa-peak-in-2016 |
# Thoughts on “AI is easy to control” by Pope & Belrose
Quintin Pope & Nora Belrose have a new [“AI Optimists” website](https://optimists.ai/), along with a new essay “[AI is easy to control](https://optimists.ai/2023/11/28/ai-is-easy-to-control/)”, arguing that the risk of human extinction due to future AI (“AI x-ris... | https://www.lesswrong.com/posts/YyosBAutg4bzScaLu/thoughts-on-ai-is-easy-to-control-by-pope-and-belrose |
# Researchers and writers can apply for proxy access to the GPT-3.5 base model (code-davinci-002)
AI safety researchers and creative writers can apply for proxied access to the GPT-3.5 base model through, also known as [code-davinci-002](https://cyborgism.wiki/hypha/code-davinci-002).
Publicly available base models ... | https://www.lesswrong.com/posts/gm7pZK4i4a7q2DH4B/researchers-and-writers-can-apply-for-proxy-access-to-the |
# Queuing theory: Benefits of operating at 60% capacity
Related to [Slack](https://www.lesswrong.com/s/xEkBK3SRrp6DhGf7c/p/yLLkWMDbC9ZNKbjDG). Related to [Lean Manufacturing](https://en.wikipedia.org/wiki/Lean_manufacturing), aka JIT Manufacturing.
TL;DR A successful task-based system should sometimes be idle, like [... | https://www.lesswrong.com/posts/YEHHHttT7EqZqPZb5/queuing-theory-benefits-of-operating-at-60-capacity |
# Specification Gaming: How AI Can Turn Your Wishes Against You [RA Video]
*In this new video, we explore specification gaming in machine learning systems. It's meant to be watched as a follow-up to* [*The Hidden Complexity of Wishes*](https://www.lesswrong.com/posts/6At6iMTu3dsKSTXLL/the-hidden-complexity-of-wishes-t... | https://www.lesswrong.com/posts/LCD4sWsRH5BxkGpbY/specification-gaming-how-ai-can-turn-your-wishes-against-you |
# Please Bet On My Quantified Self Decision Markets
**There are too many possible quantified self experiments to run. Do hobbyist prediction platforms**[^4fo2b58sim8]**make priorisation easier? I test this by setting up** [**multiple markets**](#Markets)**, in order to run two experiments (the best one, and a random o... | https://www.lesswrong.com/posts/2qQyKpXzbPov2Fmdr/please-bet-on-my-quantified-self-decision-markets |
# Could there be "natural impact regularization" or "impact regularization by default"?
Specifically, imagine you use [general-purpose](https://www.lesswrong.com/posts/6mysMAqvo9giHC4iX/what-s-general-purpose-search-and-why-might-we-expect-to-see) search procedure which recursively invokes itself to solve subgoals for... | https://www.lesswrong.com/posts/cS7rhjuG9thmfTqdy/could-there-be-natural-impact-regularization-or-impact |
# Complex systems research as a field (and its relevance to AI Alignment)
I have this high prior that complex-systems type thinking is usually a trap. I've had a few conversations about this, but still feel kind of confused, and it seems good to have a better written record of mine and your thoughts here.
At a high l... | https://www.lesswrong.com/posts/7KjRqrwjmZRoiBDtn/complex-systems-research-as-a-field-and-its-relevance-to-ai |
# MATS Summer 2023 Retrospective
Co-Authors: [@Rocket](https://www.lesswrong.com/users/rocket?mention=user), [@Juan Gil](https://www.lesswrong.com/users/juan-gil?mention=user), [@Christian Smith](https://www.lesswrong.com/users/christian-smith?mention=user), [@McKennaFitzgerald](https://www.lesswrong.com/users/mckenna... | https://www.lesswrong.com/posts/zwf68YaySvXhWYCdh/mats-summer-2023-retrospective |
# South Bay Pre-Holiday Gathering
Ilmarin House is hosting its third unstructured holiday get-together! There'll be cookies and rudimentary snacks; apart from that it's a potluck and you should bring what (if anything) you'd like to have available.
Invite some friends if you think they'd enjoy the event as well! An ... | https://www.lesswrong.com/events/esaDDDkjYDnHrJrLp/south-bay-pre-holiday-gathering |
# Protecting against sudden capability jumps during training
TL;DR: If an LLM has a sudden capabilities jump and is tested using a diverse test suite (e.g. one where it can execute code or present lots of text to human observers), it might be able to escape mid-training. In order to prevent this, I propose a phased te... | https://www.lesswrong.com/posts/edutkCwr93stppmNp/protecting-against-sudden-capability-jumps-during-training |
# 2023 Unofficial LessWrong Census/Survey
~~**The Less Wrong General Census is unofficially here! You can take it at**~~ [~~**this link.**~~](https://docs.google.com/forms/d/e/1FAIpQLSc_-Te1a4Q-FuNhwwgSGDwVWCHOWppCrWtKaVrhHcujopy5Lg/viewform?usp=sf_link)
Update: The census is closed! Thank you everyone who took it.
... | https://www.lesswrong.com/posts/JHeTrWha5PxiPEwBt/2023-unofficial-lesswrong-census-survey |
# List of strategies for mitigating deceptive alignment
The following are strategies for mitigating deceptive alignment. Some of these ideas are novel (at least to me).
I'm highly uncertain how well many of these strategies will work in practice and many of them also clearly don't scale. This is meant to be a 'kitche... | https://www.lesswrong.com/posts/Kwb29ye3qsvPzoof8/list-of-strategies-for-mitigating-deceptive-alignment |
# After Alignment — Dialogue between RogerDearnaley and Seth Herd
Hi Seth! So, what did you want to discuss?
I'd like to primarily discuss your [AI, Alignment and Ethics](https://www.lesswrong.com/s/QypaioYFd3CMFhefn) sequence. You made a number of points that I think LWers will be interested in. I'll try to primaril... | https://www.lesswrong.com/posts/qPeBdeSXa399u5HwS/after-alignment-dialogue-between-rogerdearnaley-and-seth |
# Out-of-distribution Bioattacks
The main goal of my work [these days](https://www.jefftk.com/p/leaving-google-joining-the-nucleic-acid-observatory) is trying to reduce the chances of individuals or small groups causing large-scale harm through engineered pandemics, potentially civilizational collapse or extinction. O... | https://www.lesswrong.com/posts/zcxK3p97LwDtXpmRo/out-of-distribution-bioattacks |
# Taking Into Account Sentient Non-Humans in AI Ambitious Value Learning: Sentientist Coherent Extrapolated Volition
I have published a paper in the *Journal of Artificial Intelligence and Consciousness* about how to take into account the interests of non-human animals and digital minds in A(S)I value alignment.
For ... | https://www.lesswrong.com/posts/8DWtsuBkH3GmGiAdf/taking-into-account-sentient-non-humans-in-ai-ambitious |
# Quick takes on "AI is easy to control"
A friend asked me for my quick takes on “[AI is easy to control](https://optimists.ai/2023/11/28/ai-is-easy-to-control/)”, and gave an advance guess as to what my take would be. I only skimmed the article, rather than reading it in depth, but on that skim I produced the followi... | https://www.lesswrong.com/posts/WkJDgpaPeCJDMJkoL/quick-takes-on-ai-is-easy-to-control |
# Book Review: 1948 by Benny Morris
*Content warning: war, Israel/Palestine conflict*
*Potential bias warning: I am a Jewish Israeli*
My copy of 1948 arrived in late September, but I was in the middle of another book. Then October the 7th happened, and with the horrors of that day fresh in my mind, and my country in... | https://www.lesswrong.com/posts/jiL95tPaSWJnx5xpB/book-review-1948-by-benny-morris |
# The Witness
*“What are the roots that clutch, what branches grow*
*Out of this stony rubbish? Son of man,*
*You cannot say, or guess, for you know only*
*A heap of broken images-”*
I wake up, feeling a strange sense of restlessness. I’m not sure why, but it's impossible to lounge around in bed like I usually ... | https://www.lesswrong.com/posts/LpF7BBGxbYETxJgYL/the-witness |
# Meditations on Mot
> *Holy! Holy! Holy! Holy! Holy! Holy! Holy! Holy! Holy! Holy! Holy! Holy! Holy! Holy! Holy!*
> *Everything is holy! everybody’s holy! everywhere is holy! everyday is in eternity! Everyman’s an angel!*
> *Holy the lone juggernaut! Holy the vast lamb of the middleclass! Holy the crazy shepherds... | https://www.lesswrong.com/posts/yQzv9pSbZtYYufYB4/meditations-on-mot |
# Nietzsche's Morality in Plain English
In 1924, Clarence Darrow’s [eight-hour plea](https://voicesofdemocracy.umd.edu/clarence-darrow-plea-for-leopold-and-loeb-22-23-and-25-august-1924-speech-text/) for Leopold and Loeb blamed the universities and scholars of Nietzsche (who died in 1900) for their influence on Leopol... | https://www.lesswrong.com/posts/Ypo9kNpp45EELgEdK/nietzsche-s-morality-in-plain-english |
# the micro-fulfillment cambrian explosion
Warehouse automation has been very successful. [Here's](https://www.youtube.com/watch?v=-sMCFae-Yhg) a typical modern system. As you can see, a tall narrow robot rides on a single rail at the top and bottom. The linked example is up to 42m tall. Items are stored on top of pal... | https://www.lesswrong.com/posts/4gQjEvQt7oByfYHNH/the-micro-fulfillment-cambrian-explosion |
# Disappointing Table Refinishing
Shaving cream on a table: good kids activity on a wet day, right?
[](https://www.jefftk.com/table-with-shaving-cream-big.jpg)
Unfortunately it turns out our table's finish is alcohol-soluble, and spending a couple hours covered... | https://www.lesswrong.com/posts/Skq5xA7Jm27xxHpQe/disappointing-table-refinishing |
# A call for a quantitative report card for AI bioterrorism threat models
Introduction
============
Scientists and [policymakers](https://www.whitehouse.gov/briefing-room/presidential-actions/2023/10/30/executive-order-on-the-safe-secure-and-trustworthy-development-and-use-of-artificial-intelligence/) have expressed ... | https://www.lesswrong.com/posts/YAFq9W8hoJsqqCbn3/a-call-for-a-quantitative-report-card-for-ai-bioterrorism |
# Hashmarks: Privacy-Preserving Benchmarks for High-Stakes AI Evaluation
**TL;DR**: Quick "idea paper" describing a protocol for benchmarking AI capabilities in the open without disclosing sensitive information. It's similar to how passwords are used for user registration and authentication in a web app: experts first... | https://www.lesswrong.com/posts/ZudYA6PyLfP3E2y9i/hashmarks-privacy-preserving-benchmarks-for-high-stakes-ai |
# [Valence series] 1. Introduction
1.1 Summary & Table of Contents
===============================
This is the first of a [series of five blog posts on valence](https://www.lesswrong.com/s/6uDBPacS6zDipqbZ9). Here’s an overview of the whole series, and then we’ll jump right into the first post!
1.1.1 Summary & Table... | https://www.lesswrong.com/posts/As7bjEAbNpidKx6LR/valence-series-1-introduction |
# Updates to Open Phil’s career development and transition funding program
*(cross-posted from* [*the Effective Altruism Forum*](https://forum.effectivealtruism.org/posts/yEb9CdWA2xaHA6QsD/updates-to-open-phil-s-career-development-and-transition)*)*
We’ve recently made a few updates to the program page for our [*... | https://www.lesswrong.com/posts/NLc5DdKwvwHkvtntB/updates-to-open-phil-s-career-development-and-transition |
# 6. The Mutable Values Problem in Value Learning and CEV
*Part 6 of* [*AI, Alignment, and Ethics*](https://www.lesswrong.com/s/QypaioYFd3CMFhefn)*. This will probably make more sense if you start with* [*Part 1*](https://www.lesswrong.com/s/QypaioYFd3CMFhefn/p/umnEwbK7sWSAnZ9vN)*.*
**TL;DR** In Parts [1](https://www... | https://www.lesswrong.com/posts/R36DmF4Md9Zq6odFN/6-the-mutable-values-problem-in-value-learning-and-cev |
# n of m ring signatures
A normal cryptographic signature associated with a message and a public key lets you prove to the world that it was made by someone with access to the private key associated with the known public key, without revealing that private key. You can read about it on Wikipedia [here](https://en.wiki... | https://www.lesswrong.com/posts/uojSbSav3dtEJvctz/n-of-m-ring-signatures |
# What's new at FAR AI
Summary
-------
We are [FAR AI](https://far.ai): an AI safety research incubator and accelerator. Since our inception in July 2022, FAR has grown to a team of 12 full-time staff, produced 13 academic papers, opened the coworking space FAR Labs with 40 active members, and organized field-buildin... | https://www.lesswrong.com/posts/zy8eHc5iYgrFGjF8Q/what-s-new-at-far-ai-3 |
# 2023 Alignment Research Updates from FAR AI
*TL;DR: FAR AI's science of robustness agenda has found vulnerabilities in superhuman Go systems; our value alignment research has developed more sample-efficient value learning algorithms; and our model evaluation direction has developed a variety of new black-box and whi... | https://www.lesswrong.com/posts/PQgEdo3xsFFAxXNqE/2023-alignment-research-updates-from-far-ai-2 |
# Interview with Vanessa Kosoy on the Value of Theoretical Research for AI
*Below is a transcript (edited for grammar) of a* [*video interview*](https://youtu.be/1MCRQF0_5zY?feature=shared) *I had with Vanessa Kosoy, cross-posted from my* [*blog*](https://www.zenmarmotdigital.com/blog/interview-with-vanessa-kosoy)*. ... | https://www.lesswrong.com/posts/QPxrH5ex6MpYoLwer/interview-with-vanessa-kosoy-on-the-value-of-theoretical |
# Open Thread – Winter 2023/2024
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. ... | https://www.lesswrong.com/posts/tRa92qDonDLi6RA8u/open-thread-winter-2023-2024 |
# Speaking to Congressional staffers about AI risk
In May and June of 2023, I (Orpheus) had about 50-70 meetings about AI risks with congressional staffers. I had been meaning to write a post reflecting on the experience and some of my takeaways, and I figured it could be a good topic for a LessWrong dialogue. I saw t... | https://www.lesswrong.com/posts/2sLwt2cSAag74nsdN/speaking-to-congressional-staffers-about-ai-risk |
# Accelerating science through evolvable institutions
*This is the written version of a talk presented to the Santa Fe Institute at a working group on “Accelerating Science.”*
We’re here to discuss “accelerating science.” I like to start on topics like this by taking the historical view: When (if ever) has science ac... | https://www.lesswrong.com/posts/9uEAwHzoDS8shpdoX/accelerating-science-through-evolvable-institutions |
# Bands And Low-stakes Dances
When I got started playing for contra dances in the Boston area there were several area dances you could consider "low stakes". Small dances, couldn't pay very much, less enthusiasm from established bands. And so they were relatively willing to book bands that were starting out and friend... | https://www.lesswrong.com/posts/atwkvWcSppkKN9dRJ/bands-and-low-stakes-dances |
# The LessWrong 2022 Review
The snow is falling, the carols are starting, and we all know it's time for our favorite winter holiday tradition. It's LessWrong review time!
Each year we come together and... | https://www.lesswrong.com/posts/B6CxEApaatATzown6/the-lesswrong-2022-review |
# Some open-source dictionaries and dictionary learning infrastructure
As more people begin work on interpretability projects which incorporate dictionary learning, it will be valuable to have high-quality dictionaries publicly available.[^xki2bn9t6a] To get the ball rolling on this, my collaborator (Aaron Mueller) an... | https://www.lesswrong.com/posts/AaoWLcmpY3LKvtdyq/some-open-source-dictionaries-and-dictionary-learning |
# Analyzing the Historical Rate of Catastrophes
To communicate risks, we often turn to stories. Nuclear weapons conjure stories of mutually assured destruction, briefcases with red buttons, and nuclear winter. Climate change conjures stories of extreme weather, cities overtaken by rising sea levels, and crop failures.... | https://www.lesswrong.com/posts/SgYz4YPKridWXJSE6/analyzing-the-historical-rate-of-catastrophes |
# Neural uncertainty estimation review article (for alignment)
*EDIT 1/27: This post neglects the entire sub-field of estimating uncertainty of learned representations, as in* [*https://openreview.net/pdf?id=e9n4JjkmXZ.*](https://openreview.net/pdf?id=e9n4JjkmXZ.) *I might give that a separate follow-up post.*
Introd... | https://www.lesswrong.com/posts/79eegMp3EBs8ptFqa/neural-uncertainty-estimation-review-article-for-alignment |
# A Socratic dialogue with my student
This is a dialogue between me and Noam, my student. It is reproduced, in edited form, with his permission. When commenting, please consider that he is a teenager. Many of these ideas are [new](https://xkcd.com/1053/) to him.
How do you get a student? You steal them. His previous ... | https://www.lesswrong.com/posts/TtdJt78mgGiDubnAM/a-socratic-dialogue-with-my-student |
# We're all in this together
There's one thing history seems to have been trying to teach us: that the contents of the future are determined by power, economics, politics, and other [conflict-theoritic matters](https://slatestarcodex.com/2018/01/24/conflict-vs-mistake/).
Turns out, nope!
Almost all of what the futu... | https://www.lesswrong.com/posts/A4nfKtD9MPFBaa5ME/we-re-all-in-this-together |
# On ‘Responsible Scaling Policies’ (RSPs)
This post was originally intended to come out directly after [the UK AI Safety Summit](https://thezvi.substack.com/p/on-the-uk-summit), to give the topic its own deserved focus. One thing led to another, and I am only doubling back to it now.
#### Responsible Deployment Poli... | https://www.lesswrong.com/posts/yRJNCDp7LHyHGkANz/on-responsible-scaling-policies-rsps |
# Deep Forgetting & Unlearning for Safely-Scoped LLMs
Thanks to Phillip Christoffersen, Adam Gleave, Anjali Gopal, Soroush Pour, and Fabien Roger for useful discussions and feedback.
TL;DR
=====
This post overviews a research agenda for avoiding unwanted latent capabilities in LLMs. It argues that "deep" forgetting... | https://www.lesswrong.com/posts/mFAvspg4sXkrfZ7FA/deep-forgetting-and-unlearning-for-safely-scoped-llms |
# Studying The Alien Mind
*This post is part of a *[*sequence on LLM psychology*](https://www.lesswrong.com/posts/yuwdj82yjhLFYessc/preface-to-the-sequence-on-llm-psychology)
This is Section 3 of my report “[Scheming AIs: Will AIs fake alignment during training in order to get power?](https://arxiv.org/pdf/2311.08379.pdf)”. There’s also a summary of the full report [here](https://www.lesswrong.c... | https://www.lesswrong.com/posts/KyuMS9XzqaJGMu74f/arguments-for-against-scheming-that-focus-on-the-path-sgd |
# Critique-a-Thon of AI Alignment Plans
AI-Plans.com will be hosting a Critique-a-Thon from December 16th to 18th, where participants will be proposing and discussing critiques of AI Alignment plans.
Judges include:
* [Nate Soares](https://www.lesswrong.com/users/so8res?mention=user), President of MIRI @na
* [@R... | https://www.lesswrong.com/events/kgyrsAkbBjNxPbicc/critique-a-thon-of-ai-alignment-plans |
# How do you feel about LessWrong these days? [Open feedback thread]
Hello! This is jacobjacob from the LessWrong / Lightcone team.
**This is a meta thread for you to share any thoughts, feelings, feedback or other stuff about LessWrong, that's been on your mind. **
Examples of things you might share:
* "I real... | https://www.lesswrong.com/posts/j2W3zs7KTZXt2Wzah/how-do-you-feel-about-lesswrong-these-days-open-feedback |
# Some quick thoughts on "AI is easy to control"
There are many things I feel like the [post](https://optimists.ai/2023/11/28/ai-is-easy-to-control/) authors miss, and I want to share a few thoughts that seem good to communicate.
I'm going to focus on controlling superintelligent AI systems: systems powerful enough t... | https://www.lesswrong.com/posts/iABojbKgmtMGYgcYm/some-quick-thoughts-on-ai-is-easy-to-control |
# **In defence of Helen Toner, Adam D'Angelo, and Tasha McCauley**
Anonymous post:
"I understand a common view in EA- or AI-governance land is that Toner, D'Angelo and McCauley (TDM for short) really messed things up at OpenAI, and AI, the fate of the world, etc. has gotten worse thanks to them. I am confident this i... | https://www.lesswrong.com/posts/nfsmEM93jRqzQ5nhf/in-defence-of-helen-toner-adam-d-angelo-and-tasha-mccauley-1 |
# EA Infrastructure Fund's Plan to Focus on Principles-First EA
*Edit 2024/01/02: **People who like this new vision may want to consider*** [***donating to EAIF***](https://www.givingwhatwecan.org/en-US/funds/effective-altruism-funds)***! ***
*Now until the end of January is an unusually good time to donate, as it al... | https://www.lesswrong.com/posts/A8L4udgpsoC2BNMnS/ea-infrastructure-fund-s-plan-to-focus-on-principles-first |
# Digital humans vs merge with AI? Same or different?
So, to create context, this is a continuation of our remarks in the comments to Zvi's "[AI #40: A Vision from Vitalik](https://www.lesswrong.com/posts/je5BwKe8enCq8DLrm/ai-40-a-vision-from-vitalik)"
There I was asking about
1. the boundary between *humans mergin... | https://www.lesswrong.com/posts/ZpbcvBtNMxG8v6mcB/digital-humans-vs-merge-with-ai-same-or-different |
# Anthropical Paradoxes are Paradoxes of Probability Theory
*This is the fourth post in my series on Anthropics. The previous one is* [*Anthropical probabilities are fully explained by difference in possible outcomes*](https://www.lesswrong.com/posts/xbQLwBAyP2KjZC4qS/antropical-probabilities-are-fully-explained-by-di... | https://www.lesswrong.com/posts/8rWP3ZvHe2HkQd53o/anthropical-paradoxes-are-paradoxes-of-probability-theory |
# Metaculus Launches Chinese AI Chips Tournament, Supporting Institute for AI Policy and Strategy Research
Metaculus’s [Chinese AI Chips Tournament](https://www.metaculus.com/tournament/chinese-ai-chips/?has_group=false&project=2784&order_by=-activity) is live: Forecast and support research by the [Institute for AI Po... | https://www.lesswrong.com/posts/C6toCdqF2y8te3PDa/metaculus-launches-chinese-ai-chips-tournament-supporting |
# Why Yudkowsky is wrong about "covalently bonded equivalents of biology"
*confidence level: I am a physicist, not a biologist, so don’t take this the account of a domain level expert. But this is* [*really basic stuf*](https://www.khanacademy.org/science/biology/macromolecules/proteins-and-amino-acids/a/orders-of-pro... | https://www.lesswrong.com/posts/8viKzSrYhb6EFk6wg/why-yudkowsky-is-wrong-about-covalently-bonded-equivalents |
# Bucket Brigade: Likely End-of-Life
During the pandemic a group of us wrote [some software](https://echo.jefftk.com/) to allow people to sing and make music together in real time over the internet, by accounting for latency instead of minimizing it. The initial goal was to support group singing at the [2020 Secular S... | https://www.lesswrong.com/posts/dwPNrNF5BAdCxGcmx/bucket-brigade-likely-end-of-life |
# Based Beff Jezos and the Accelerationists
It seems Forbes decided to doxx the identity of e/acc founder Based Beff Jezos. They did so using voice matching software.
Given Jezos is owning it given that it happened, rather than hoping it all goes away, and people are talking about him, this seems like a good time to ... | https://www.lesswrong.com/posts/3xoThNNYgZmTCpEAB/based-beff-jezos-and-the-accelerationists |
# Google Gemini Announced
Google just announced Gemini, and Hassabis claims that "in each of the 50 different subject areas that we tested it on, it's as good as the best expert humans in those areas"
> State-of-the-art performance
> We've been rigorously testing our Gemini models and evaluating their performance on... | https://www.lesswrong.com/posts/9Xh2PyMzNDu3yEHqR/google-gemini-announced |
# Proposal for improving the global online discourse through personalised comment ordering on all websites
This is a description of a project I’ve been thinking about making in the last few weeks. I’m looking for feedback and collaborators. \[BetterDiscourse\] name is provisional.
**Updated on 9 Dec 2023**: Completel... | https://www.lesswrong.com/posts/aRwBw8W7G9gvLFJSW/proposal-for-improving-the-global-online-discourse-through |
# Originality vs. Correctness
I talk with Alkjash about valuing original thinking vs getting things right. We discuss a few main threads:
1. What are the benefits of epistemic specialisation? What about generalism?
2. How much of the action in an actual human mind is in tweaking your distribution over hypotheses an... | https://www.lesswrong.com/posts/AnhwbEisrPXEw9d5g/originality-vs-correctness |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.