text stringlengths 300 320k | source stringlengths 52 154 |
|---|---|
# Super-Exponential versus Exponential Growth in Compute Price-Performance
In recent months and years I have [seen](https://ourworldindata.org/grapher/gpu-price-performance) [sober](https://epochai.org/blog/trends-in-gpu-price-performance#appendix-a---dropping-data-before-2006) [analyses](https://browse.arxiv.org/pdf/... | https://www.lesswrong.com/posts/gLJP2sBqXDsQWLAgy/super-exponential-versus-exponential-growth-in-compute-price |
# Linking Alt Accounts
Sometimes people have something they want to say without it being traceable back to their main identity, and on internet forums it's common for people to use multiple accounts ("alts") for this. As machine learning software gets better, however, it becomes increasingly practical to link a person... | https://www.lesswrong.com/posts/xjuffm5FFphFDvLWF/linking-alt-accounts |
# Don't Dismiss Simple Alignment Approaches
I remember that when I first started in alignment there was this belief that in order to make alignment progress, you needed to be a genius. With the rise of interpretability, this expectation has moderated somewhat, but I still think it has influence and often leads people ... | https://www.lesswrong.com/posts/ziNCZEm7FE9LHxLai/don-t-dismiss-simple-alignment-approaches |
# Announcing Dialogues
As of today, everyone is able to create a new type of content on LessWrong: *Dialogues*.
In contrast with posts, which are for monologues, and comment sections, which are spaces for everyone to talk to everyone, a dialogue is a space for a few invited people to *speak with each other*.
I'm pe... | https://www.lesswrong.com/posts/kQuSZG8ibfW6fJYmo/announcing-dialogues-1 |
# Related Discussion from Thomas Kwa's MIRI Research Experience
One of the subthreads in [Thomas Kwa's MIRI research experience](https://www.lesswrong.com/posts/qbcuk8WwFnTZcXTd6/thomas-kwa-s-miri-research-experience) was about his experience working with Nate. In the comments, some other people brought up their own n... | https://www.lesswrong.com/posts/BGLu3iCGjjcSaeeBG/related-discussion-from-thomas-kwa-s-miri-research |
# Contra Nora Belrose on Orthogonality Thesis Being Trivial
[Disagreeing with Eliezer Y., Nora Belrose calls the Orthogonality Thesis "trivial, false or meaningless/confused"](https://twitter.com/norabelrose/status/1709598947470377139).
I think this is wrong. For instance, without the orthogonality thesis, one might ... | https://www.lesswrong.com/posts/6ZKKmiF9cnwNxiYa2/contra-nora-belrose-on-orthogonality-thesis-being-trivial |
# Fixing Insider Threats in the AI Supply Chain
Summary
=======
Insider threats are security risks caused by an organisation's staff. **Careless and intentional insider threats cause 25+% of cyber breaches**.[^2om770gv8d2]
In a [survey of existential risks involving cybersecurity and AI](https://forum.effectivealtru... | https://www.lesswrong.com/posts/htchppMWKQLdg3bz2/fixing-insider-threats-in-the-ai-supply-chain |
# Every Mention of EA in "Going Infinite"
Pinocchio is narrated by Jiminy Cricket; the Great Gatsby is narrated by Nick; Sam Bankman-Fried’s story is narrated by Michael Lewis. His book, *Going Infinite,* isn’t written from Sam’s perspective, exactly. Michael Lewis is a sympathetic fly on the wall documenting a tragic... | https://www.lesswrong.com/posts/GeKnGEX6GHzdcs3RM/every-mention-of-ea-in-going-infinite |
# Griffin Island
The park near our house has an interesting sign:
[](https://www.jefftk.com/griffin-island-sign-big.jpg)
A parent put it up, labeling the area you're able to explore from this point without crossing any streets. It's the large red area on my [child w... | https://www.lesswrong.com/posts/2PiwBEMQrWRvg5djs/griffin-island |
# Sam Altman's sister claims Sam sexually abused her -- Part 1: Introduction, outline, author's notes
Introduction
============
[Sam Altman](https://en.wikipedia.org/wiki/Sam_Altman)'s sister, Annie Altman, has claimed that Sam sexually abused her when she was a child for approximately 9 years, beginning when she was... | https://www.lesswrong.com/posts/QDczBduZorG4dxZiW/sam-altman-s-sister-claims-sam-sexually-abused-her-part-1-1 |
# A thought about the constraints of debtlessness in online communities
I notice that there's a lot of work done in online communities that's hard to price. We know so little about its price that it's a bit ambiguous as to whether it should be paid for or charged for:
* Is writing posts a duty or a privilege? You g... | https://www.lesswrong.com/posts/CoWM7hqYaeJJsixB9/a-thought-about-the-constraints-of-debtlessness-in-online |
# Comparing Anthropic's Dictionary Learning to Ours
Readers may have noticed many similarities between Anthropic's recent publication [Towards Monosemanticity: Decomposing Language Models With Dictionary Learning](https://transformer-circuits.pub/2023/monosemantic-features/) ([LW post](https://www.lesswrong.com/posts/... | https://www.lesswrong.com/posts/F4iogK5xdNd7jDNyw/comparing-anthropic-s-dictionary-learning-to-ours |
# The Gradient – The Artificiality of Alignment
[The Gradient](https://thegradient.pub/) is a “digital publication about artificial intelligence and the future,” founded by researchers at the Stanford Artificial Intelligence Laboratory. I found the latest essay, “The Artificiality of Intelligence,” by a PhD student at... | https://www.lesswrong.com/posts/BzbdYkBRMwzFAqLcg/the-gradient-the-artificiality-of-alignment |
# Bird-eye view visualization of LLM activations
I’m starting to learn about [mechanistic interpretability](https://www.neelnanda.io/mechanistic-interpretability/quickstart), and I’m seeing lots of great visualizations of transformer internals, but somehow I’ve never seen the whole large model’s internal state shown a... | https://www.lesswrong.com/posts/pjiiZJHbsYL7nqdpx/bird-eye-view-visualization-of-llm-activations |
# Linkpost: Are Emergent Abilities in Large Language Models just In-Context Learning?
A new preprint, [Lu et al.](https://arxiv.org/abs/2309.01809), published last month, has important implications for AI risk if true. It essentially suggests that emergent capabilities of current LLMs are (with some exceptions) mediat... | https://www.lesswrong.com/posts/qfawLPCp2MZgK44Yc/linkpost-are-emergent-abilities-in-large-language-models |
# Time is homogeneous sequentially-composable determination
*[Metadata: crossposted from [https://tsvibt.blogspot.com/2023/06/time-is-homogeneous-sequentially.html](https://tsvibt.blogspot.com/2023/06/time-is-homogeneous-sequentially.html). First completed June 30, 2023.]*
Time is the character of courses of events i... | https://www.lesswrong.com/posts/6qrLfAG7mDTyrHmh7/time-is-homogeneous-sequentially-composable-determination |
# FAQ: What the heck is goal agnosticism?
*Thanks to George Wang, Liron Shapira, Eliezer*[^xozj39d8axs]*, and probably dozens of other people at Manifest and earlier conferences that I can't immediately recall the names of for listening to me attempt to explain this in different ways and for related chit chat.*
I've ... | https://www.lesswrong.com/posts/G3aG5eLGYE3cnhEZq/faq-what-the-heck-is-goal-agnosticism |
# "The Heart of Gaming is the Power Fantasy", and Cohabitive Games
I quite liked [@mako yass](https://www.lesswrong.com/users/mako-yass?mention=user)' [Peacewagers/Cohabitive Games so Far](https://www.lesswrong.com/posts/bF353RHmuzFQcsokF/peacewagers-cohabitive-games-so-far). But I found myself wanting to link to this... | https://www.lesswrong.com/posts/vZ4ZrHSAtj9o78vfz/the-heart-of-gaming-is-the-power-fantasy-and-cohabitive |
# AI Alignment Breakthroughs this week (10/08/23)
\[Cross-posted from [https://midwitalignment.substack.com/p/ai-alignment-breakthroughs-this-week-a6e](https://midwitalignment.substack.com/p/ai-alignment-breakthroughs-this-week-a6e) \]
Note: I change to a new (much longer) format based off of feedback. Please le... | https://www.lesswrong.com/posts/N9afB9b6vzCT2YxZX/ai-alignment-breakthroughs-this-week-10-08-23 |
# We don't understand what happened with culture enough
This is a quick response to[ Evolution Provides No Evidence For the Sharp Left Turn](https://www.lesswrong.com/posts/hvz9qjWyv8cLX9JJR/evolution-provides-no-evidence-for-the-sharp-left-turn), due to it winning first prize in[ The Open Philanthropy Worldviews cont... | https://www.lesswrong.com/posts/wCtegGaWxttfKZsfx/we-don-t-understand-what-happened-with-culture-enough |
# New contractor role: Web security task force contractor for AI safety announcements
**Web security task force contractors for AI safety announcements**
Survival and Flourishing .Com is looking for white-hat hackers and security professionals who can join our task force of “on call” professionals for a week or so, o... | https://www.lesswrong.com/posts/TJyjLR7nhpxBztrtn/new-contractor-role-web-security-task-force-contractor-for |
# The case for aftermarket blind spot mirrors
Car side and rear-view mirrors are designed to give you a view around your car, but they typically don't show you the two spots directly to the side of your car (the "[blind spots](https://en.wikipedia.org/wiki/Vehicle_blind_spot)"). Luckily, you can increase the field of ... | https://www.lesswrong.com/posts/MsJpxMK8At98NGZvK/the-case-for-aftermarket-blind-spot-mirrors |
# NYT on the Manifest forecasting conference
Kevin Roose from New York Times just published an article about our [Manifest](https://manifestconference.net) conference, forecasting and prediction markets. And despite what you might assume from "NYT covering weird rationalist-y event", the article is very fair, perhaps ... | https://www.lesswrong.com/posts/yW9PitywvMuN6sxKQ/nyt-on-the-manifest-forecasting-conference |
# Truthseeking when your disagreements lie in moral philosophy
\[Status: latest entry in a longrunning series\]
My [last post](https://acesounderglass.com/2023/09/28/ea-vegan-advocacy-is-not-truthseeking-and-its-everyones-problem/) on truthseeking in EA vegan advocacy got a lot of comments, but there’s [one](https://... | https://www.lesswrong.com/posts/XYgZY57duhzKCkEBq/truthseeking-when-your-disagreements-lie-in-moral-philosophy |
# One: a story
The earliest living creatures were single cells, battered around by their environments. Over time they clumped together, to form multi-celled organisms capable of complex self-regulation. Those organisms formed mating pairs, to better spread their genes; then they formed social groups, to protect and sh... | https://www.lesswrong.com/posts/FqvR7hhS6wQxKvAyA/one-a-story |
# The Witching Hour
> *“This above all: to thine own self be true.”*
**Thursday**
Your successor is late again today. You already wrote your shift report, but you still need to onboard them personally, in case they have any questions that the report doesn’t answer. The servers you’re watching over are humming away s... | https://www.lesswrong.com/posts/oGJT5CoyGQcy5nqmC/the-witching-hour-1 |
# Non-superintelligent paperclip maximizers are normal
The [paperclip maximizer](https://en.wikipedia.org/wiki/Instrumental_convergence) is a thought experiment about a hypothetical superintelligent AGI that is obsessed with maximizing paperclips. It can be modeled as a utility-theoretic agent whose utility function i... | https://www.lesswrong.com/posts/Z8C29oMAmYjhk2CNN/non-superintelligent-paperclip-maximizers-are-normal |
# The Handbook of Rationality (2021, MIT press) is now open access
Of particular interest here is section 4.1 on subjective accounts of probability and section 8.4 which explicitly addresses Knightian uncertainty and its challenges for Bayesian accounts.
https://direct.mit.edu/books/oa-edited-volume/5525/The-Handbook... | https://www.lesswrong.com/posts/zcigMWSr5uKs2cPCv/the-handbook-of-rationality-2021-mit-press-is-now-open |
# Simple Terminal Colors
I work in the terminal a lot and I often want to write programs that color their output a bit. For example, perhaps I'm looking at a [sequencing read](https://www.jefftk.com/p/sequencing-intro) and I want to know what part of it was the reason it was flagged. I could use [colorama](https://pyp... | https://www.lesswrong.com/posts/oLK5J7xazS8L3p6fy/simple-terminal-colors |
# Epistemic Motif of Abstract-Concrete Cycles & Domain Expansion
*(crossposted from* [*here*](https://dalcybremin.github.io/posts/2023-10-09-epistemic-motif.html)*)*
I've noticed there are certain epistemic motifs (i.e. legible & consistent patterns of knowledge production) that come up in fairly wide circumstances a... | https://www.lesswrong.com/posts/x6xfX7nwf53uM8hBW/epistemic-motif-of-abstract-concrete-cycles-and-domain |
# Become a PIBBSS Research Affiliate
*TL;DR:* [***PIBBSS is hiring research affiliates!***](https://www.pibbss.ai/affiliate)
*PIBBSS is a research initiative facilitating work that draws on the parallels between intelligent behavior in natural and artificial systems, and leveraging these insights towards making AI sy... | https://www.lesswrong.com/posts/JvL3SC6tPjFyCiHad/become-a-pibbss-research-affiliate-1 |
# I'm a Former Israeli Officer. AMA
Hey everyone! Long time lurker, first time poster.
My name is Yovel Rom. I'm an Israeli data scientist. I served in the IDF in an elite program called Talpiot, which consists of a physics udergrad and broad military education (basically West Point for technology people, but more e... | https://www.lesswrong.com/posts/zJCKn4TSXcCXzc6fi/i-m-a-former-israeli-officer-ama |
# Childhood Roundup #3
It’s time once again for a roundup of all the childhood-related stories I’ve noticed since the last such post that don’t fit in elsewhere.
In addition to the standard post, I have a personal note, which is that unfortunately our current nanny’s health does not allow her to continue, and thus we... | https://www.lesswrong.com/posts/M3x2bGyJayxi52z3j/childhood-roundup-3 |
# Documenting Journey Into AI Safety
**Summary**
-----------
Recently, I have noticed that many individuals have become discouraged due to a seeming lack of progress in their attempts to break into a career in AI safety. It seems to me that a key bottleneck in the *beginning a career in AI safety* pipeline is access ... | https://www.lesswrong.com/posts/ozDWnEChJwuB5L5wg/documenting-journey-into-ai-safety |
# Announcing MIRI’s new CEO and leadership team
In 2023, MIRI has shifted focus in the direction of broad public communication—see, for example, our [recent TED talk](https://www.ted.com/talks/eliezer_yudkowsky_will_superintelligent_ai_end_the_world?language=en), our piece in TIME magazine “[Pausing AI Developments Is... | https://www.lesswrong.com/posts/NjtHt55nFbw3gehzY/announcing-miri-s-new-ceo-and-leadership-team |
# Examples of Low Status Fun
What is fun and what is high status are probably negatively correlated, *and* people probably mostly try to have high status fun. So I'd expect there to be several ways to have fun that are under-explored.
So, what are some example of things that are low status but potentially fun?
This ... | https://www.lesswrong.com/posts/igANvYQCrNeBycv7n/examples-of-low-status-fun |
# An explanation for every token: using an LLM to sample another LLM
Introduction
============
Much has been written about the [implications](https://www.lesswrong.com/posts/dcoxvEhAfYcov2LA6/agentized-llms-will-change-the-alignment-landscape) and [potential](https://www.lesswrong.com/posts/FRRb6Gqem8k69ocbi/external... | https://www.lesswrong.com/posts/5ZFgZbqp6Mi2xpYjK/an-explanation-for-every-token-using-an-llm-to-sample |
# Mini-Workshop on Applied Rationality
**Edit: we have reached 15 sign-ups. If you would like to attend then please fill out this the** [**sign-up form**](https://forms.gle/4FebNXQZopapzfsW7) **before Friday, October 20. People who are accepted in to the workshop will receive an email to that effect.**
John Steidley... | https://www.lesswrong.com/events/8NKocJj9y5xWefghM/mini-workshop-on-applied-rationality |
# Update on the UK AI Taskforce & upcoming AI Safety Summit
*While preparing for an upcoming Convergence Analysis post on the UK AI Taskforce and our recommendations, I looked into their reports, their plans for the upcoming AI safety summit, the recommendations from other orgs, and some miscellaneous UK AI events. I ... | https://www.lesswrong.com/posts/5fdcsWwtvG9jAtzGK/update-on-the-uk-ai-taskforce-and-upcoming-ai-safety-summit |
# You’re Measuring Model Complexity Wrong
**TLDR**: We explain why you should care about *model complexity*, why the *local learning coefficient* is arguably the correct measure of model complexity, and how to estimate its value.
In particular, we review [a new set of estimation techniques introduced by Lau et al. (2... | https://www.lesswrong.com/posts/6g8cAftfQufLmFDYT/you-re-measuring-model-complexity-wrong |
# Attributing to interactions with GCPD and GWPD
*This post provides background, motivation, and a nontechnical summary of the purely mathematical* [*https://arxiv.org/abs/2310.06686*](https://arxiv.org/abs/2310.06686)*.*
*Coauthors (alphabetical): Chris MacLeod, Jenny Nitishinskaya, Buck Shlegeris. Work done mostly ... | https://www.lesswrong.com/posts/hFaXe4Mi64xkE6Kqp/attributing-to-interactions-with-gcpd-and-gwpd |
# What I've been reading, October 2023: The stirrup in Europe, 19th-century art deco, and more
A ~monthly feature. Last month was busy for me with a lot of travel and a lot of focus on [The Roots of Progress as a nonprofit organization](https://rootsofprogress.org/), so I haven’t had as much time as I prefer for resea... | https://www.lesswrong.com/posts/9a7fixEi5pEfKbMAc/what-i-ve-been-reading-october-2023-the-stirrup-in-europe |
# Evolution Solved Alignment (what sharp left turn?)
Some people like to use the evolution of homo sapiens as an [argument by analogy](https://www.lesswrong.com/posts/GNhMPAWcfBCASy8e6/a-central-ai-alignment-problem-capabilities-generalization) concerning the apparent difficulty of aligning powerful optimization proce... | https://www.lesswrong.com/posts/xCCNdxNPpbJKD3x4W/evolution-solved-alignment-what-sharp-left-turn |
# Relevance of 'Harmful Intelligence' Data in Training Datasets (WebText vs. Pile)
*I believe the ideas in this post, though preliminary, will benefit many, especially those tackling the challenging aspects of the alignment problem. I think I've found an analogy that the alignment community should explore further.*
... | https://www.lesswrong.com/posts/Bm5QhjiWs95YL4Kgt/relevance-of-harmful-intelligence-data-in-training-datasets |
# Noticing confusion in physics
I recently had a Physics test, and the only thing I got wrong was the answer to this question: "What happens to the speed of sound in a gas when the gas is heated up?"
My answer was "The speed of sound decreases because heating up a gas decreases its density. Lower density means lower ... | https://www.lesswrong.com/posts/TCDZJ6deyKFpenoDf/noticing-confusion-in-physics |
# AI #33: Cool New Interpretability Paper
This has been a rough week for pretty much everyone. While I have had to deal with many things, and oh how I wish I could stop checking any new sources for a while, others have had it far worse. I am doing my best to count my blessings and to preserve my mental health, and her... | https://www.lesswrong.com/posts/pD5rkAvtwp25tyfRN/ai-33-cool-new-interpretability-paper |
# The International PauseAI Protest: Activism under uncertainty
This post is an attempt [to](https://forum.effectivealtruism.org/s/vw6tX5SyvTwMeSxJk) [summarize](https://www.astralcodexten.com/p/pause-for-thought-the-ai-pause-debate) the [crucial considerations](https://forum.effectivealtruism.org/topics/crucial-consi... | https://www.lesswrong.com/posts/ZSCBSJhwu4SLBQb4J/the-international-pauseai-protest-activism-under-uncertainty-1 |
# LoRA Fine-tuning Efficiently Undoes Safety Training from Llama 2-Chat 70B
*Produced as part of the* [*SERI ML Alignment Theory Scholars Program*](https://serimats.org/) *\- Summer 2023 Cohort, under the mentorship of Jeffrey Ladish.*
TL;DR LoRA fine-tuning undoes the safety training of Llama 2-Chat 70B with one GPU... | https://www.lesswrong.com/posts/qmQFHCgCyEEjuy5a7/lora-fine-tuning-efficiently-undoes-safety-training-from |
# unRLHF - Efficiently undoing LLM safeguards
*Produced as part of the* [*SERI ML Alignment Theory Scholars Program*](https://serimats.org/) *\- Summer 2023 Cohort, under the mentorship of Jeffrey Ladish. I'm grateful to Palisade Research for their support throughout this project.*
tl;dr: demonstrating that we can *c... | https://www.lesswrong.com/posts/3eqHYxfWb5x4Qfz8C/unrlhf-efficiently-undoing-llm-safeguards |
# What do Marginal Grants at EAIF Look Like? Funding Priorities and Grantmaking Thresholds at the EA Infrastructure Fund
The EA Infrastructure Fund (EAIF) offers targeted grants to amplify projects based on effective altruism principles, enhancing their access to talent, capital, and knowledge. They are currently fund... | https://www.lesswrong.com/posts/WC3E4Xf8DaGkiDJQh/what-do-marginal-grants-at-eaif-look-like-funding-priorities |
# Progress links digest, 2023-10-12: Dyson sphere thermodynamics and a cure for cavities
I’ve been traveling for a while, so this is a long one, covering the last ~month. I tried to cut it down, but there have been so many amazing announcements, opportunities, etc.! Feel free to skim and jump around:
**From the Roots... | https://www.lesswrong.com/posts/TPGDyE272dkcJABzR/progress-links-digest-2023-10-12-dyson-sphere-thermodynamics |
# Paper: Understanding and Controlling a Maze-Solving Policy Network
Mrinank, Austin, and Alex wrote a paper on the results from [Understanding and controlling a maze-solving policy network](https://www.lesswrong.com/posts/cAC4AXiNC5ig6jQnc/understanding-and-controlling-a-maze-solving-policy-network), [Maze-solving ag... | https://www.lesswrong.com/posts/DKtWikjcdApRj3rWr/paper-understanding-and-controlling-a-maze-solving-policy |
# Revisiting Guide Dogs and Blindness Prevention
In early effective altruism it [was common](https://www.jefftk.com/p/charity-variance-vision) [for people](https://80000hours.org/2011/11/it-is-effectiveness-not-overhead-that-matters/) [to give](https://80000hours.org/2012/09/is-there-such-thing-as-a-bad-charity/) [the... | https://www.lesswrong.com/posts/vRom5PmF2onedwEA8/revisiting-guide-dogs-and-blindness-prevention |
# LW UI features you might not have tried
Every once in a while I see a comment describing a problem with lesswrong that is fixable with existing features, if only the user knew about them. This post is an attempt to make these highly useful features more widely known.
I am technically a mod on lesswrong, which is w... | https://www.lesswrong.com/posts/tASjAvMsDEjXxGEhP/lw-ui-features-you-might-not-have-tried |
# Circle of Support (Oct 14th @ 10am PST)
Having spent some extended time in this community back in 2010s, I am well familiar with a lot of the stressors we can experience:
* Feeling of doom from AI (and other x-risks)
* Pressure to be more "effective" and more "altruistic"
* Fear of (or anger at) aging and dyi... | https://www.lesswrong.com/posts/fJLqDyu5T6mmy5jxN/circle-of-support-oct-14th-10am-pst |
# Combination Lock Boxes
Traditionally we control access to our houses with "something you have": a key. This works pretty well, until you have a visitor arrive when you're not home to give them a key. The traditional way to handle this is to hide a key (under the mat, in a flower pot, etc), turning this into "somethi... | https://www.lesswrong.com/posts/8Eypi8Ao8dPAE98BH/combination-lock-boxes |
# To open-source or to not open-source, that is (an oversimplification of) the question.
*\*\*\* This an edited and expanded version of *[*a post I made on X*](https://x.com/justinbullock14/status/1708678966800728120?s=46) *in response to*[GovAI’s new report“Open-Sourcing Highly Capable Foundation Models”](https://www... | https://www.lesswrong.com/posts/9HeZjGpkQQJfkcbqh/to-open-source-or-to-not-open-source-that-is-an |
# [Paper] All's Fair In Love And Love: Copy Suppression in GPT-2 Small
*This is a accompanying blog post to work done by Callum McDougall, Arthur Conmy and Cody Rushing as part of SERI MATS 2023. The work was mentored by Neel Nanda and Tom McGrath. You can find our full paper at* [*https://arxiv.org/abs/2310.04625*](h... | https://www.lesswrong.com/posts/ebezsHW6qJwxTFasX/paper-all-s-fair-in-love-and-love-copy-suppression-in-gpt-2 |
# Prediction markets covered in the NYT podcast “Hard Fork”
The latest episode of Hard Fork has a 20 min section dedicated to prediction markets! Kevin Roose and Casey Newton go into much more depth in this podcast than in [the earlier NYT article](https://www.nytimes.com/2023/10/08/technology/prediction-markets-manif... | https://www.lesswrong.com/posts/ADkyxynJaaykteLRt/prediction-markets-covered-in-the-nyt-podcast-hard-fork |
# Hiring: Lighthaven Events & Venue Lead
[**Apply for the position here**](https://airtable.com/appb8P36D2CSXg9AX/shrdqS6JXok99f6EX)
Lightcone is looking to hire someone to manage [*Lighthaven*](https://www.lesswrong.com/posts/memqyjNCpeDrveayx/the-lighthaven-campus-is-open-for-bookings), a campus located in Berkeley... | https://www.lesswrong.com/posts/kA8szQS3GDFF8BFju/hiring-lighthaven-events-and-venue-lead |
# Meta-Regulations
The United States has [a lot](https://www.forbes.com/sites/waynecrews/2017/08/15/how-many-rules-and-regulations-do-federal-agencies-issue/?sh=184eb64b1e64) of regulations:
![](https://substackcdn.com/image/fetch/f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.... | https://www.lesswrong.com/posts/F9kPmGd6zJDdQGxYs/meta-regulations |
# How to partition teams to move fast? Debating "low-dimensional cuts"
One of my favorite pastimes is to study historical examples of teams who accomplished ambitious projects effectively. Recently on this topic I wrote the post ["A Golden Age of Building? Excerpts and lessons from Empire State, Pentagon, Skunk Works ... | https://www.lesswrong.com/posts/B7NWqc88un2TQpGsY/how-to-partition-teams-to-move-fast-debating-low-dimensional |
# Disentangling Our Terminal and Instrumental Values
Disagreements related to what we value seem to explain maybe 10% of the
disagreements over AI safety. This post will try to explain how I think
about which values I care about perpetuating to the distant future.
Robin Hanson helped to clarify the choices in [Wh... | https://www.lesswrong.com/posts/HzxNfj3LJDANdoJaw/disentangling-our-terminal-and-instrumental-values |
# Dishonorable Gossip and Going Crazy
This is a fairly high-context conversation between me (Ben) and my friend (Ren).
Ren used to work at [CFAR](https://rationality.org/), and then ~4.5 years ago moved away to join the [Monastic Academy](https://www.monasticacademy.com/) in Vermont. Ren had recently visited me and ... | https://www.lesswrong.com/posts/DQjS59yetM64njSBS/dishonorable-gossip-and-going-crazy |
# RSPs are pauses done right
*COI: I am a research scientist at Anthropic, where I work on *[*model organisms of misalignment*](https://www.lesswrong.com/posts/ChDH335ckdvpxXaXX/model-organisms-of-misalignment-the-case-for-a-new-pillar-of-1)*; I was also involved in the drafting process for *[*Anthropic’s RSP*](https:... | https://www.lesswrong.com/posts/mcnWZBnbeDz7KKtjJ/rsps-are-pauses-done-right |
# Eight Magic Lamps
> *“Give me a lever long enough, and a fulcrum on which to place it, and I shall move the world.” - Archimedes*
Aladdin started with nothing; but after a sorcerer tasked him to retrieve a magic lamp, the lamp’s genie granted him wealth and fame. His fortune lasted until the sorcerer stole the lamp... | https://www.lesswrong.com/posts/o5PWN57svETZtQCir/eight-magic-lamps |
# The Gods of Straight Lines
> *“He intends only his own gain, and he is in this, as in many other cases, led by an invisible hand to promote an end which was no part of his intention.” - Adam Smith*
Cassandra was a priestess who was granted the gift of prophecy by Apollo. But after she rejected his advances, he chan... | https://www.lesswrong.com/posts/xkRtegmqL2iyhtDB3/the-gods-of-straight-lines |
# Is the Wave non-disparagement thingy okay?
A few weeks ago, Linch, Auckland (pseudonym), and I (Ruby) decided to chat about something and settled on this topic. What follows is a mix of distillation and edited chat (Slack) transcript.
Auckland and I (Ruby) start from the position that the Wave non-disparagement agr... | https://www.lesswrong.com/posts/7e8g4FYxJQafctKRs/is-the-wave-non-disparagement-thingy-okay-1 |
# Which Anaesthetic To Choose?
In causal decision theory, the perspective/indexical aspect could lead to reflective inconsistency. This is generally regarded as a problem. Here I present a thought experiment to show why this view may require further review.
Two Anaesthetics
----------------
Suppose you are about to... | https://www.lesswrong.com/posts/TkgZWZKXgcCLc3G55/which-anaesthetic-to-choose |
# Will no one rid me of this turbulent pest?
Last year, [I wrote about the promise of gene drives to wipe out mosquito species and end malaria.](https://denovo.substack.com/p/gene-drives-why-the-wait)
In the time since my previous writing, gene drives have still not been used in the wild, and [***over 600,000 peo... | https://www.lesswrong.com/posts/gjs3q83hA4giubaAw/will-no-one-rid-me-of-this-turbulent-pest |
# Natural Abstraction: Convergent Preferences Over Information Structures
The natural abstraction hypothesis claims (in part) that a wide variety of agents will learn to use the same abstractions and concepts to reason about the world.
What's the simplest possible setting where we can state something like this formal... | https://www.lesswrong.com/posts/gts62zc6roEWzZEsg/natural-abstraction-convergent-preferences-over-information-4 |
# The Puritans would one-box: evidential decision theory in the 17th century
Evidential decision theory was used way earlier than I would have expected, and its development was motivated theologically.
[Unconditional election](https://en.wikipedia.org/wiki/Unconditional_election) is the Protestant (mainly Calvinist a... | https://www.lesswrong.com/posts/2eJPcPoptvP5chjd3/the-puritans-would-one-box-evidential-decision-theory-in-the |
# In memory of Louise Glück
(Cross-posted from [my website](https://joecarlsmith.com/2023/10/15/in-memory-of-louise-gluck). Podcast version [here](https://joecarlsmithaudio.buzzsprout.com/2034731/13778055-in-memory-of-louise-gluck), or search "Joe Carlsmith Audio" on your podcast app.)
, here is my list of ... | https://www.lesswrong.com/posts/aCjvg5whn8DYqneg5/rationalist-horror-movies |
# Arguments for optimism on AI Alignment (I don't endorse this version, will reupload a new version soon.)
Or, why we probably don't need to worry about AI.
So this post is partially a response to Amalthea's comment on how I simply claimed that my side is right, and I responded by stating that I was going for a sho... | https://www.lesswrong.com/posts/99tD8L8Hk5wkKNY8Q/arguments-for-optimism-on-ai-alignment-i-don-t-endorse-this |
# More or Fewer Fights over Principles and Values?
I wrote:
> I wish that people would generally get into more conflicts over the values and principles that they hold
Vaniver wrote:
> I think we are unusually fractious and don't value peace treaties and fences as much as we should
I wrote:
> Fight me :)
Start of... | https://www.lesswrong.com/posts/GrzbTCtrTZDweEPGm/more-or-fewer-fights-over-principles-and-values |
# On Frequentism and Bayesian Dogma
I've heard that you believe that frequentism is correct. But that's obviously wrong, so what gives?
I guess first of all I should ask, what do you mean by "frequentism"?
I mean classical statistical frequentism. Though somewhat tongue-in-cheek, as I don't think it's fully correct,... | https://www.lesswrong.com/posts/34RGz5Be7swJJEKho/on-frequentism-and-bayesian-dogma |
# Taxonomy of AI-risk counterarguments
Partly inspired by [The Crux List](https://thezvi.substack.com/p/the-crux-list), the following is a non-comprehensive taxonomy of positions which imply that we should not be worried about existential risk from artificial superintelligence.
Each position individually is supposed ... | https://www.lesswrong.com/posts/BvFJnyqsJzBCybDSD/taxonomy-of-ai-risk-counterarguments |
# My AI Predictions 2023 - 2026
*Epistemic status: My, mostly intuitive, guesses - with only a few days dwelling on it, and no serious research beyond what I already knew.*
I work in the startup sphere, in field robotics, and I am about to have an opportunity to majorly shift what I am working on. To work out wha... | https://www.lesswrong.com/posts/btZPxeJLvuRGDqyke/my-ai-predictions-2023-2026 |
# Goodhart's Law in Reinforcement Learning
*Produced As Part Of The* [*OxAI Safety Labs program*](https://www.aisafetyhub.org/research-programme/)*, mentored by Joar Skalse.*
TL;DR
=====
This is a blog post introducing our new paper, "[*Goodhart's Law in Reinforcement Learning*](https://arxiv.org/abs/2310.09144)" (t... | https://www.lesswrong.com/posts/Eu6CvP7c7ivcGM3PJ/goodhart-s-law-in-reinforcement-learning |
# AI Alignment [Incremental Progress Units] this week (10/08/23)
\[edit: took out naming controversy stuff, as it was distracting from the point of the blog\]
I am introducing a new rating system for each alignment breakthrough. The rating system will go from 1 star⭐ to 5 stars ⭐⭐⭐⭐⭐.
A 1 star ⭐ “breakthroug... | https://www.lesswrong.com/posts/zkDCCSR3o2aBkTcbq/ai-alignment-incremental-progress-units-this-week-10-08-23 |
# The price is right
\[Saying an old thing in a new way\]
The United States Department of Transportation will pay $11.8 million to save a life. You know what that means? It means that if you come to the United States Department of Transportation with a plan (barriers around the Grand Canyon, wider lanes on the expre... | https://www.lesswrong.com/posts/Kx3BsHdFWp9EovMws/the-price-is-right |
# How should TurnTrout handle his DeepMind equity situation?
Ok, so the basic situation as I understand it is that as part of your Deepmind offer you would get a bunch of Google stock that would become available to you over the course of around 4 years or so. This means in-expectation, for 4 years, your net worth woul... | https://www.lesswrong.com/posts/Be3ertyJfwDdQucdd/how-should-turntrout-handle-his-deepmind-equity-situation |
# An EPUB of Arbital's AI Alignment section
2024-01-11 update: Sourcehut, the host I used for this repository, is down. I think I'll move to another setup to host these books.
I've created an EPUB of the [AI Alignment section](ht... | https://www.lesswrong.com/posts/Jrvm8YCAe6HYfJNsy/an-epub-of-arbital-s-ai-alignment-section |
# The Good Life in the face of the apocalypse
This is a dialogue Elizabeth and I wrote about our journeys trying to live well while caring about the world not ending. We used a chess timer to govern who was talking at a given time and to keep the pace snappy and time-boxed.
We talk about separating whether we were wo... | https://www.lesswrong.com/posts/jnac43sBH85L4kxsm/the-good-life-in-the-face-of-the-apocalypse |
# Late-talking kid part 3: gestalt language learning
*Previously in the series: *[*Late-talking kids and “Einstein syndrome” (2021)*](https://www.lesswrong.com/posts/WqjXPeALsLH6Dxfys/late-talking-kids-and-einstein-syndrome)*and* [*More notes from raising a late-talking kid (2022)*](https://www.lesswrong.com/posts/txu... | https://www.lesswrong.com/posts/DRscwvQhdBGvE4zny/late-talking-kid-part-3-gestalt-language-learning |
# Buffalo buffalo Buffalo buffalo buffalo buffalo Buffalo buffalo.
This is likely the least useful post I've published here. But maybe it's not *entirely* useless, plus I think it's somewhat entertaining, so I'll go ahead with it anyway.
The Sentence
------------
As you may or may not have heard before, "Buffalo buf... | https://www.lesswrong.com/posts/dYsbRc66opxNmNpAG/buffalo-buffalo-buffalo-buffalo-buffalo-buffalo-buffalo |
# Fertility Roundup #2
The world is slowly waking up to the fertility crisis. There is more acknowledgement of the problem, and there is more talk of potential practical solutions. I do not believe the topic is a realistic target for Balsa Policy Institute, but I will continue to keep an eye on the ball and issue peri... | https://www.lesswrong.com/posts/YjT88Avga8xTvtZzw/fertility-roundup-2 |
# Being patient with problems
In the writings of mathematician Alexander Grothendieck, there is a recurring discussion of two opposing strategies when confronting problems. One strategy, which he saw as the default among mathematicians and lamented leaning on early in his career, was to “attack problems,” that is, dir... | https://www.lesswrong.com/posts/2QJpcyKad5dDnoKMR/being-patient-with-problems |
# Trying to deconfuse some core AI x-risk problems
Max H commented on the [dialogues announcement](https://www.lesswrong.com/posts/kQuSZG8ibfW6fJYmo/announcing-dialogues-1), and with both of us being interested in having a conversation where we try to do something like "explore the basic case for AI X-risk, without mu... | https://www.lesswrong.com/posts/K8jJpdfKE3racyziC/trying-to-deconfuse-some-core-ai-x-risk-problems |
# Eliezer's example on Bayesian statistics is wr... oops!
This post was *going* to be an explanation of how an example Eliezer Yudkowsky frequently uses in discussion of statistics doesn't actually imply what he thinks it does. In the process of proving his mistake, I found out I was actually wrong. I'm still writing ... | https://www.lesswrong.com/posts/BeAHn5CisRgivuspA/eliezer-s-example-on-bayesian-statistics-is-wr-oops |
# When building an organization, there are lots of ways to prevent financial corruption of personnel. But what are the ways to prevent corruption via social status, political power, etc.?
Nowadays, even fairly sophisticated accounting and auditing concepts are well known by those with on-the-job work experience in any... | https://www.lesswrong.com/posts/JBjPreynFn5aDmJsj/when-building-an-organization-there-are-lots-of-ways-to |
# Worldwork for Ethics
**Abstract**: An alternative to the now-predominating models of alignment, corrigibility and "CEV", following a critique of these. The critique to show, in substance: CEV and corrigibility have the exact same problems - in effect, they're isomorphs of one another, and each equally unobtainable. ... | https://www.lesswrong.com/posts/Gd6Kbm2PaYwAYLEnC/worldwork-for-ethics |
# Investigating the learning coefficient of modular addition: hackathon project
As our project at the [Melbourne hackathon](https://devinterp.com/events/2023-q3-melbourne-hackathon) on [Singular Learning Theory](https://www.lesswrong.com/s/czrXjvCLsqGepybHC) and alignment (Oct. 7-8), we did some experiments to estimat... | https://www.lesswrong.com/posts/4v3hMuKfsGatLXPgt/investigating-the-learning-coefficient-of-modular-addition |
# Eleuther releases Llemma: An Open Language Model For Mathematics
> Today we release *Llemma*: 7 billion and 34 billion parameter language models for mathematics. The Llemma models were initialized with Code Llama weights, then trained on the Proof-Pile II, a 55 billion token dataset of mathematical and scientific do... | https://www.lesswrong.com/posts/HvgfNjihcDjCtEctE/eleuther-releases-llemma-an-open-language-model-for |
# Labs should be explicit about why they are building AGI
Three of the big AI labs say that they care about alignment and that they think misaligned AI poses a potentially existential threat to humanity. These labs continue to try to build AGI. I think this is a very bad idea.
The leaders of the big labs are clear th... | https://www.lesswrong.com/posts/6HEYbsqk35butCYTe/labs-should-be-explicit-about-why-they-are-building-agi |
# Hints about where values come from
Intro written a day later:
Spiracular and I discuss the nature and origin of values. The dialogue doesn't get into much of a single clear thread, but I had fun writing it and hope it has some interesting trailheads for others.
(You may wish to skip the methodological exchange at... | https://www.lesswrong.com/posts/fijSRFL6Z5pXBbCgi/hints-about-where-values-come-from |
# magnetic cryo-FTIR
## FTIR
[FTIR](https://en.wikipedia.org/wiki/Fourier-transform_infrared_spectroscopy) is a common way to determine what chemicals are present in a sample. For a molecule to absorb a photon, it must have some electric charges that can vibrate at about the same frequency as that photon. Many common... | https://www.lesswrong.com/posts/PdwQYLLp8mmfsAkEC/magnetic-cryo-ftir |
# At 87, Pearl is still able to change his mind
Judea Pearl is a famous researcher, known for Bayesian networks (the standard way of representing Bayesian models), and his statistical formalization of causality. Although [he has always been recommended reading here](https://www.lesswrong.com/posts/RiQYixgCdvd8eWsjg/re... | https://www.lesswrong.com/posts/uFqnB6BG4bkMW23LR/at-87-pearl-is-still-able-to-change-his-mind |
# On Interpretability's Robustness
*Léo Dana - Produced as part of the SERI ML Alignment Theory Scholars Program - Summer 2023 Cohort, as well as an internship at FAR AI, both under the mentorship of Claudia Shi.*
**Would you trust the** [**IOI**](https://arxiv.org/abs/2211.00593) **circuit?**
-----------------------... | https://www.lesswrong.com/posts/tNdSqrk6hpxfxmZqS/on-interpretability-s-robustness |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.