text stringlengths 300 320k | source stringlengths 52 154 |
|---|---|
# 11 heuristics for choosing (alignment) research projects
I recently spoke with Dane Sherburn about some of the most valuable things he learned as a SERI-MATS scholar.
Here are 11 heuristics he uses to prioritize between research projects:
1. **Impact:** Can I actually tell myself a believable story in which this ... | https://www.lesswrong.com/posts/WzbzvD7BxFQe4GqRo/11-heuristics-for-choosing-alignment-research-projects |
# Basics of Rationalist Discourse
**Introduction**
----------------
This post is meant to be a linkable resource. Its core is a [short list of guidelines](https://www.lesswrong.com/posts/XPv4sYrKnPzeJASuk/basics-of-rationalist-discourse-1#Guidelines__in_brief_) (you can link directly to the list) that are intended to... | https://www.lesswrong.com/posts/XPv4sYrKnPzeJASuk/basics-of-rationalist-discourse-1 |
# Jar of Chocolate
Truffles are pretty great: gooey fat-softened flavored chocolate inside, protected by a shell of hard chocolate so you don't get your fingers messy. Sometimes I'll make them with the kids:
[](https://www.jefftk.com/anna-dipping-truffles-big.jpg)
... | https://www.lesswrong.com/posts/Xs9kPafyNTZoukKSf/jar-of-chocolate |
# ChatGPT understands language
*Cross-posting this from my blog, since it seems relevant.*
**The case for GPT understanding language, by way of understanding the world**
There's a debate going on about whether or not language models similar to ChatGPT have the potential to be scaled up to something truly transformat... | https://www.lesswrong.com/posts/FuNuuoh2AzWqk6Bp8/chatgpt-understands-language |
# Assigning Praise and Blame: Decoupling Epistemology and Decision Theory
Your group/company/organization performs well, doing great work and dealing with new problems efficiently. As one of its leaders, you want to understand why, so that you can make it even more successful, and maybe emulate this success in other s... | https://www.lesswrong.com/posts/RKzNpJhamYgcLeAEv/assigning-praise-and-blame-decoupling-epistemology-and |
# The role of Bayesian ML in AI safety - an overview
*I want to thank Sebastian Farquhar, Laurence Midgley and Johan van den Heuvel, for feedback and discussion on this post. *
Some time ago I asked the question [“What is the role of Bayesian ML in AI safety/alignment?”](https://forum.effectivealtruism.org/posts/cpN8... | https://www.lesswrong.com/posts/RJZ7bwoDB6BWgt6St/the-role-of-bayesian-ml-in-ai-safety-an-overview |
# Scaling Laws Literature Review
Common shape of a scaling law, taken from [Hestness et ... | https://www.lesswrong.com/posts/6L9EhCa8Zo2GoThGB/scaling-laws-literature-review |
# Literature review of TAI timelines
We summarize and compare several models and forecasts predicting when transformative AI will be developed.
**Highlights**
--------------
* The review includes quantitative models, including both outside and inside view, and judgment-based forecasts by (teams of) experts.
* Wh... | https://www.lesswrong.com/posts/4eAnBaLxvnkydiavw/literature-review-of-tai-timelines |
# WaPo: "Big Tech was moving cautiously on AI. Then came ChatGPT."
Some good reporting here, got quotes from a lot of people. Here are the key paragraphs, in my opinion:
> In the past year or so, top AI researchers from Google have left to launch start-ups around large language models, including Character.AI, Cohere,... | https://www.lesswrong.com/posts/ab8wAd6FJvWAexYug/wapo-big-tech-was-moving-cautiously-on-ai-then-came-chatgpt |
# Spooky action at a distance in the loss landscape
> *Produced under the mentorship of Evan Hubinger as part of the* [*SERI ML Alignment Theory Scholars Program*](https://serimats.org/) *Winter 2022 Cohort*.
Not all global minima of the (training) loss landscape are created equal.
Even if they achieve equal perform... | https://www.lesswrong.com/posts/2N7eEKDuL5sHQou3N/spooky-action-at-a-distance-in-the-loss-landscape |
# A Simple Alignment Typology
I set out to [review the OpenAI alignment plan](https://www.lesswrong.com/posts/tvjLwataADYxCc9EN/openai-alignment-review-rationalists-versus-empiricists), and my brain at some point diverged to modeling the humans behind the arguments instead of the actual arguments.
So behold! A simpl... | https://www.lesswrong.com/posts/T2FGbZw8TYsjcpQBn/a-simple-alignment-typology |
# Reflections on Deception & Generality in Scalable Oversight (Another OpenAI Alignment Review)
Just like you can test your skill in experimental design by reviewing existing experiments, you can test[ your skill in alignment by reviewing existing alignment strategies](https://www.lesswrong.com/posts/8kB4rB8eaSFDWBbKi... | https://www.lesswrong.com/posts/tvjLwataADYxCc9EN/reflections-on-deception-and-generality-in-scalable |
# Advice I found helpful in 2022
Here are 10 of helpful pieces of advice I received in 2022:
1. **Understand things for yourself. **Figure out what you actually think about this AI stuff. What do you think will happen? What concrete risks and scenarios are you worried about? (*Credit to Oliver Habryka. Last year at ... | https://www.lesswrong.com/posts/aiSERWHsnDCgTLhGM/advice-i-found-helpful-in-2022 |
# On not getting contaminated by the wrong obesity ideas
[*A Chemical Hunger*](http://achemicalhunger.com/)([a](https://web.archive.org/web/20220430074603/http://achemicalhunger.com/)), a series by the authors of the blog [Slime Mold Time Mold](https://slimemoldtimemold.com/) (SMTM), argues that the obesity epidemic i... | https://www.lesswrong.com/posts/NRrbJJWnaSorrqvtZ/on-not-getting-contaminated-by-the-wrong-obesity-ideas |
# Optimality is the tiger, and annoying the user is its teeth
Just for fun, I wanted to see how far I could get using ChatGPT to implement [Optimality is the tiger, and agents are its teeth](https://www.lesswrong.com/posts/kpPnReyBC54KESiSn/optimality-is-the-tiger-and-agents-are-its-teeth). Its obviously not a good en... | https://www.lesswrong.com/posts/N7qE5o3jmoKoe4dHQ/optimality-is-the-tiger-and-annoying-the-user-is-its-teeth |
# LW Filter Tags (Rationality/World Modeling now promoted in Latest Posts)
In case you missed it, you can customize your Latest Posts feed on the frontpage based on tag. That is, you can promote or reduce the prominence of posts based on their tags such AI, Rationality, Practical, etc.
LessWrong's Core Tags[^ob9fjiyk... | https://www.lesswrong.com/posts/SnS392FnhkTb5xfJG/lw-filter-tags-rationality-world-modeling-now-promoted-in |
# Eli Dourado AMA on the Progress Forum
Eli Dourado—whom you might know from his writing on [geothermal](https://www.elidourado.com/p/geothermal), [supersonic](https://www.mercatus.org/students/research/research-papers/make-america-boom-again), [energy superabundance](https://www.thecgo.org/research/energy-superabunda... | https://www.lesswrong.com/posts/FQZxhX2nNrFGMPBuW/eli-dourado-ama-on-the-progress-forum |
# Stop-gradients lead to fixed point predictions
*Johannes Treutlein and Rubi Hudson worked on this post as part of* [*SERI MATS*](https://www.serimats.org/)*, under the mentorship of Evan Hubinger. Rubi has also received mentorship from Leo Gao. We thank Erik Jenner for helpful discussions and Alexander Pan for bring... | https://www.lesswrong.com/posts/i3v7WeCXyWiYfhihF/stop-gradients-lead-to-fixed-point-predictions |
# Why and How to Graduate Early [U.S.]
Students that pursue a bachelors degree in the U.S. traditionally spend four years in high school (graduating at ~18) and four years in university (~22). They often hold part-time jobs and internships, but only start accumulating “real” experience after finishing their education.... | https://www.lesswrong.com/posts/dkchEFnM9EYpW3bjK/why-and-how-to-graduate-early-u-s |
# Reverse RSS Stats
[In 2013](https://www.jefftk.com/p/reverse-rss-feed) I built a weird feature where clicking on " [◂◂ RSS](__REVERSE_RSS__)" at the top of a post gets you an RSS feed that runs in reverse from that post, with a 'new' old post with the publication of each new new post. It's a bit hard to think about,... | https://www.lesswrong.com/posts/5xBzWHurD2iNER2Yx/reverse-rss-stats |
# EA novel published on Amazon
Just so everyone knows, the novel I wrote to promote EA has been published on Amazon. This is not a request to purchase it, or try to directly support it in any particular way.
However if the blurb *does *sound interesting, go for it :)
https://www.amazon.com/dp/B0BSXHJRBQ | https://www.lesswrong.com/posts/F4biDYZwPmpjsaTSt/ea-novel-published-on-amazon |
# Compendium of problems with RLHF
*Epistemic status: This post is a distillation of many comments/posts. I believe that my list of problems is not the best organization of sub-problems. I would like to make it shorter, and simpler, because cool theories are generally simple unified theories, by identifying only 2 or ... | https://www.lesswrong.com/posts/d6DvuCKH5bSoT62DB/compendium-of-problems-with-rlhf |
# Structure, creativity, and novelty
*[Metadata: crossposted from [https://tsvibt.blogspot.com/2022/08/structure-creativity-and-novelty.html](https://tsvibt.blogspot.com/2022/08/structure-creativity-and-novelty.html). First completed 26 June 2022. I'm likely to not respond to comments promptly.]*
A high-level confu... | https://www.lesswrong.com/posts/5pHqQwCDWrvZGp8pX/structure-creativity-and-novelty |
# Takeaways from calibration training
**Summary.** Thoughts arising from doing calibration training on ~500 questions, consisting of both trivia questions and mundane real life events.
*Epistemic status*: Confident on own experiences, uncertain on generalizability and usefulness to others.
Intro
-----
I've practic... | https://www.lesswrong.com/posts/oRTzrjJDss6twSnBD/takeaways-from-calibration-training |
# Model-driven feedback could amplify alignment failures
Anthropic, DeepMind and Google Brain are all working on strategies to train language models on their own outputs. For a brief summary of the work so far:
1. [Red Teaming Language Models with Language Models](https://arxiv.org/pdf/2202.03286.pdf) (Perez et al.,... | https://www.lesswrong.com/posts/HhBcoRwnyJhQRJnxr/model-driven-feedback-could-amplify-alignment-failures |
# How Likely is Losing a Google Account?
Let's say you use a Google account as the root of your online identity: Gmail, Fi, "Sign in with Google", Google's password manager, etc. How much should you be worried that you'll suddenly find yourself unable to get into everything? Should you do something else instead?
To g... | https://www.lesswrong.com/posts/7baw7arBwzFX3TDrN/how-likely-is-losing-a-google-account |
# Against Boltzmann mesaoptimizers
1. I'm effectively certain that a weakly defined version of "mesaoptimizers"- like a component of a network which learns least squares optimization in some context- is practically reachable by SGD.
2. I'm pretty sure there exists some configuration of an extremely large neural netw... | https://www.lesswrong.com/posts/AncrLc5iSc4tmaYBJ/against-boltzmann-mesaoptimizers |
# Simulacra Levels Summary
For More Detail, Previously: [Simulacra Levels and Their Interactions](https://thezvi.wordpress.com/2020/06/15/simulacra-and-covid-19/), [Unifying the Simulacra Definitions](https://thezvi.wordpress.com/2020/08/03/unifying-the-simulacra-definitions/), [The Four Children of the Seder as the S... | https://www.lesswrong.com/posts/hLzwNuPyEvR4mfAce/simulacra-levels-summary |
# The Energy Requirements and Feasibility of Off-World Mining
Space Prospecting
-----------------
It’s been an enticing subject for more than half a century now and continues to capture the imaginations of extremely ambitious engineers and powerful capital allocators alike. It’s also somewhat viable: although most as... | https://www.lesswrong.com/posts/3899bP8tpK4ehL8h6/the-energy-requirements-and-feasibility-of-off-world-mining |
# What I mean by "alignment is in large part about making cognition aimable at all"
*(Epistemic status: attempting to clear up a misunderstanding about points I have attempted to make in the past. This post is not intended as an argument for those points.)*
I have long said that the lion's share of the AI alignment p... | https://www.lesswrong.com/posts/NJYmovr9ZZAyyTBwM/what-i-mean-by-alignment-is-in-large-part-about-making |
# 2022 Unofficial LessWrong General Census
**The Less Wrong General Census is unofficially back!**
In days of yore, there was an annual census of the site users. That census has come again in an unofficial form! [~~Click here~~](https://forms.gle/uXqi3rmKiBVbsL3m8) ~~to take the survey!~~ \[EDIT: SURVEY CLOSED, DO N... | https://www.lesswrong.com/posts/M2jGrtZcnWTrobirz/2022-unofficial-lesswrong-general-census |
# Why I hate the "accident vs. misuse" AI x-risk dichotomy (quick thoughts on "structural risk")
* I think the large majority of AI x-risk is "structural". Like climate change.
* Here's a good primer on structural risk (note that structural risk is *not* a synonym for "not caused by out-of-control AI"): [https://... | https://www.lesswrong.com/posts/6bpW2kyeKaBtuJuEk/why-i-hate-the-accident-vs-misuse-ai-x-risk-dichotomy-quick |
# Humans Can Be Manually Strategic
**Summary:** Humans don't always take the time to think and plan very well. That does not mean we can’t think about our goals and how we want to go about achieving them. This meetup is an attempt to prompt strategic thinking.
**Tags:** Repeatable, Investment, Highly Experimental. (... | https://www.lesswrong.com/posts/L6F8CfjbBaynf2BGu/humans-can-be-manually-strategic |
# Medical Image Registration: The obscure field where Deep Mesaoptimizers are already at the top of the benchmarks. (post + colab notebook)
Background: Introduction to Image Registration
----------------------------------------------
Image registration, or finding the corresponding points in a pair of, eg, MRI scans... | https://www.lesswrong.com/posts/FgXjuS4R9sRxbzE5w/medical-image-registration-the-obscure-field-where-deep |
# Saying things because they sound good
There is a memory that has always stuck with me. In sixth grade I remember sitting on the floor at a friend's house. We were supposed to write an essay together. I distinctly remember him proposing that I should write the first version and then he'll go over it afterwards and ma... | https://www.lesswrong.com/posts/nmQacPDn8h8shRrNr/saying-things-because-they-sound-good |
# EA & LW Forum Weekly Summary (23rd - 29th Jan '23)
*Supported by Rethink Priorities*
This is part of a weekly series summarizing the top posts on the EA and LW forums - you can see the full collection [here.](https://forum.effectivealtruism.org/s/W4fhpuN26naxGCBbN) The first post includes some details on purpos... | https://www.lesswrong.com/posts/3ckGT9dxeMmzqsiyq/ea-and-lw-forum-weekly-summary-23rd-29th-jan-23 |
# Apply to HAIST/MAIA’s AI Governance Workshop in DC (Feb 17-20)
The [Harvard AI Safety Team](https://haist.ai/) and [MIT AI Alignment](https://www.mitalignment.org/), supported by the Cambridge Boston Alignment Initiative ([CBAI](https://www.cbai.ai/)), are excited to announce a 3-day workshop on AI Governance, Polic... | https://www.lesswrong.com/posts/LXgEYLEFbcJnyzSEZ/apply-to-haist-maia-s-ai-governance-workshop-in-dc-feb-17-20 |
# Inner Misalignment in "Simulator" LLMs
*Alternate title: "Somewhat Contra Scott On Simulators".*
Scott Alexander has a [recent post](https://astralcodexten.substack.com/p/janus-simulators) up on large language models as simulators.
I generally agree with Part I of the post, which advocates thinking about LLMs as s... | https://www.lesswrong.com/posts/FLMyTjuTiGytE6sP2/inner-misalignment-in-simulator-llms |
# Beware of Fake Alternatives
Summary: many of us might occasionally find ourselves in situations where we’re considering taking action A, but decide against it because some option B seems superior. This is a good way to go only if B is an *actual* alternative (we’re actually going to do it), and not a *fake* alternat... | https://www.lesswrong.com/posts/emsHFEnyxXR8nEj6p/beware-of-fake-alternatives |
# New Hackathon: Robustness to distribution changes and ambiguity
EffiSciences is proud to announce a new hackathon organized with challenge-data-ens in France.
The objective of this event is to address a sub-problem of alignment that is particularly actionable: namely value extrapolation in the case of distribution ... | https://www.lesswrong.com/posts/WF5JpmpK8EM4xKyve/new-hackathon-robustness-to-distribution-changes-and |
# Mechanistic Interpretability Quickstart Guide
*This was written as a guide for Apart Research's Mechanistic Interpretability Hackathon as a compressed version of* [*my getting started post*](https://neelnanda.io/getting-started)*. The spirit is "how to speedrun your way to maybe doing something useful in mechanistic... | https://www.lesswrong.com/posts/jLAvJt8wuSFySN975/mechanistic-interpretability-quickstart-guide |
# Talk to me about your summer/career plans
Last May, I wrote a post called [talk to me about your summer plans](https://forum.effectivealtruism.org/posts/eREwP2tom9QKkvCqR/talk-to-me-about-your-summer-plans). I was impressed by the number & quality of people who reached out. Some people found it helpful, so I'm doing... | https://www.lesswrong.com/posts/R4otYvzrgvTcz4YkY/talk-to-me-about-your-summer-career-plans |
# No Really, Attention is ALL You Need - Attention can do feedforward networks
*\[Epistemic status: Mathematically proven, and I have* [*running code*](https://github.com/RobertHuben/ffn_via_attention) *that implements it.\]*
**Overview:** A [transformer](https://arxiv.org/abs/1706.03762) consists of two alternating ... | https://www.lesswrong.com/posts/K7AyY8LMrcKhwfbyj/no-really-attention-is-all-you-need-attention-can-do |
# [Linkpost] Human-narrated audio version of "Is Power-Seeking AI an Existential Risk?"
There's now a human-narrated audio version of my report on existential risk from power-seeking AI (text version [here](https://arxiv.org/pdf/2206.13353.pdf)). You can find the audio [here](https://joecarlsmithaudio.buzzsprout.com/2... | https://www.lesswrong.com/posts/PX8vBNaaM7dfnsENR/linkpost-human-narrated-audio-version-of-is-power-seeking-ai |
# Nice Clothes are Good, Actually
*This is just some casual thoughts about how to dress nice. I think dressing nice is a useful skill, but nobody told me in a way that convinced me. So this is written in the spirit of what I wish someone would have told 15 year old me.*
For most of my life I thought dressing nicely w... | https://www.lesswrong.com/posts/kJoyRgDDPzMg4Fo3Z/nice-clothes-are-good-actually |
# Criticism of the main framework in AI alignment
*Originally posted on the* [*EA Forum*](https://forum.effectivealtruism.org/s/MQf6pWJCnbkmKG3By/p/Cs8qhNakLuLXY4GvE) *for the Criticism and Red Teaming Contest.*
0\. Summary
===========
AI alignment research centred around the control problem works well for futures s... | https://www.lesswrong.com/posts/dTxGyKtshbmWsMWhn/criticism-of-the-main-framework-in-ai-alignment |
# Aiming for Convergence Is Like Discouraging Betting
## Summary
* In [a list of guidelines for rational discourse](https://www.lesswrong.com/posts/XPv4sYrKnPzeJASuk/basics-of-rationalist-discourse-1), Duncan Sabien proposes that one should "[a]im for convergence on truth, and behave as if your interlocutors are als... | https://www.lesswrong.com/posts/iThwqe3yPog56ytyq/aiming-for-convergence-is-like-discouraging-betting |
# The effect of horizon length on scaling laws
The scaling of optimal model size with compute is a key input into the [biological anchors framework](https://www.alignmentforum.org/posts/KrJfoZzpSDpnrv9va/draft-report-on-ai-timelines) for forecasting transformative AI. In particular, the "effective horizon length" intr... | https://www.lesswrong.com/posts/ouEQLAca3bbiimAFQ/the-effect-of-horizon-length-on-scaling-laws |
# Abstraction As Symmetry and Other Thoughts
*Epistemic status: slightly pruned brainstorming.*
I've been reading John Wentworth's posts on the [natural abstraction hypothesis](https://www.lesswrong.com/tag/natural-abstraction) recently, and wanted to add some of my own thoughts. These are very much "directions to lo... | https://www.lesswrong.com/posts/J7pFLC5xYacyLsZ4E/abstraction-as-symmetry-and-other-thoughts |
# Voting Results for the 2021 Review
Well, that's a wrap for the 2021 Review. We had 238 people cast votes. 452 posts were originally nominated, of which 149 posts received at least one review. The LessWrong moderation team will be awarding prizes and assembling posts into the Best of 2021 Books / Sequences soon. But ... | https://www.lesswrong.com/posts/zajNa9fdr8JYJpxrG/voting-results-for-the-2021-review |
# Subitisation of Self
Subitising is looking at a collection of objects and, as gestalt, assessing the number present without manual counting. It's about as basic of a System 1/2 split as you can get: children need to count manually, with each tallied increment taking effort; as we get used to seeing small collections... | https://www.lesswrong.com/posts/hTKu6BsjjBhcSA84h/subitisation-of-self |
# Trends in the dollar training cost of machine learning systems
Summary
=======
1. Using a [dataset](https://docs.google.com/spreadsheets/d/1beiholXkLpiYoeyDL64bOHLI-mXZ-_8ZnY1Ed6Sfjh0/edit?usp%3Dsharing) of 124 machine learning (ML) systems published between 2009 and 2022,[^obr3yxci06] I estimate that the cost of ... | https://www.lesswrong.com/posts/REBFQF43nwcJgp8Ge/trends-in-the-dollar-training-cost-of-machine-learning-1 |
# Language Models can be Utility-Maximising Agents
1 minute takeaways
------------------
* It's actually pretty easy to train and run a language model to function as an agent for a specific task, rather than as a [non-agentic simulator](https://www.lesswrong.com/posts/vJFdjigzmcXMhNTsx/simulators)
* The [resultin... | https://www.lesswrong.com/posts/523mueiug9RapHtWb/language-models-can-be-utility-maximising-agents |
# More findings on Memorization and double descent
Produced as part of the [SERI ML Alignment Theory Scholars Program](https://www.serimats.org/) \- Winter 2022 Cohort.
I’d like to thank Wes Gurnee, Aryan Bhatt, Eric Purdy and Stefan Heimersheim for discussions and Evan Hubinger, Neel Nanda, Adam Jermyn and Chris Ola... | https://www.lesswrong.com/posts/KzwB4ovzrZ8DYWgpw/more-findings-on-memorization-and-double-descent |
# AI Safety Arguments: An Interactive Guide
Crossposted to the [EA Forum](https://forum.effectivealtruism.org/posts/zSLjLvAHmEXJWaAZi/ai-safety-arguments-an-interactive-guide).
**TL/DR:** We developed an interactive guide to AI safety arguments, based on researcher interviews. [Go check it out!](https://ai-risk-discu... | https://www.lesswrong.com/posts/9YQby2miskbcKN9FB/ai-safety-arguments-an-interactive-guide |
# Schizophrenia as a deficiency in long-range cortex-to-cortex communication
*(Written in a hurry. I was almost going to title this “My poorly-researched pet theory of schizophrenia”. Hoping for feedback and pointers to relevant prior literature. I am **very** far from a schizophrenia expert. Really. I cannot emphasiz... | https://www.lesswrong.com/posts/H2epKysvFgPcTwC2f/schizophrenia-as-a-deficiency-in-long-range-cortex-to-cortex |
# Can we “cure” cancer?
In an excellent recent essay on “[big visions for biology](https://www.sam-rodriques.com/post/a-catalog-of-big-visions-for-biology),” Sam Rodriques writes:
> Ask most biologists about the cure for cancer, and they will tell you it doesn’t exist: cancer is many diseases that are mostly unrelate... | https://www.lesswrong.com/posts/xzf84rwi5C4JnKkWP/can-we-cure-cancer |
# Product safety is a poor model for AI governance
*Note: This post is intended to be accessible to readers with relatively little background in AI safety. People with a firm understanding of AI safety may find it too basic, though they may be interested in knowing which kinds of policies I have been encountering or t... | https://www.lesswrong.com/posts/XTd4xbFSc7NQhAFqh/product-safety-is-a-poor-model-for-ai-governance |
# Exercise is Good, Actually
*Yesterday I wrote about how* [*dressing nice is a good idea*](https://www.lesswrong.com/posts/kJoyRgDDPzMg4Fo3Z/nice-clothes-are-good-actually)*. This post is written similarly, but the topic is exercise. This is again somewhat casual writing that's in the spirit of what I wish someone ha... | https://www.lesswrong.com/posts/gibjCymJvNfENrwqk/exercise-is-good-actually |
# Focus on the places where you feel shocked everyone's dropping the ball
Writing down something I’ve found myself repeating in different conversations:
If you're looking for ways to help with the whole “[the world looks pretty doomed](https://www.lesswrong.com/posts/uMQ3cqWDPHhjtiesc/agi-ruin-a-list-of-lethalities)”... | https://www.lesswrong.com/posts/Zp6wG5eQFLGWwcG6j/focus-on-the-places-where-you-feel-shocked-everyone-s |
# “AI Risk Discussions” website: Exploring interviews from 97 AI Researchers
*Tldr:* Check out the new [AI Risk Discussions website](https://ai-risk-discussions.org)! It contains 83 transcripts of AI researcher [interviews](https://ai-risk-discussions.org/interviews), and a [quantitative analysis](https://forum.effect... | https://www.lesswrong.com/posts/g4nEtPFECTQW9tcff/ai-risk-discussions-website-exploring-interviews-from-97-ai |
# Interviews with 97 AI Researchers: Quantitative Analysis
TLDR: Last year, Vael Gates interviewed 97 AI researchers about their perceptions on the future of AI, focusing on risks from advanced AI. Among other questions, researchers were asked about the alignment problem, the problem of instrumental incentives, and th... | https://www.lesswrong.com/posts/Bok5RAPPjuHKvPyv2/interviews-with-97-ai-researchers-quantitative-analysis |
# A Brief Overview of AI Safety/Alignment Orgs, Fields, Researchers, and Resources for ML Researchers
Crossposted to EA Forum \[[Link](https://forum.effectivealtruism.org/posts/xMzXbnpPeKWpTi3Gt/a-brief-overview-of-ai-safety-alignment-orgs-fields)\]
TLDR: I’ve written an overview of the AI safety space, tagged by key... | https://www.lesswrong.com/posts/2eaLH7zp6pxdQwYSH/a-brief-overview-of-ai-safety-alignment-orgs-fields |
# A simple technique to reduce negative rumination
I have faced challenges with negative rumination and anxiety, especially when I am trying to relax or my mind is idle. However, I have recently discovered a simple and effective technique to overcome this.
I practice enumerating as many positive future events as I ca... | https://www.lesswrong.com/posts/PXwHzRRh2wRgLDfEa/a-simple-technique-to-reduce-negative-rumination |
# Hacked Account Spam
Whenever I post something publicly on Facebook mentioning hacking I get a ton of spammy comments:
[](https://www.jefftk.com/fb-hack-comments-big.jpg)
[](https://www.jefftk.com/fb-hack-comments-2-bi... | https://www.lesswrong.com/posts/roJHr9iioQsniysm7/hacked-account-spam |
# Progress links and tweets, 2023-02-01
**The Progress Forum**
----------------------
* [Eli Dourado AMA](https://progressforum.org/posts/acgrRfWETEdm96xJk/eli-dourado-ama) (lots of great questions and detailed answers!)
* [When did England start seeing itself as a commercial nation?](https://progressforum.org/po... | https://www.lesswrong.com/posts/ExqtXufBzahHf4uwu/progress-links-and-tweets-2023-02-01 |
# Research agenda: Formalizing abstractions of computations
*Big thanks to Leon Lang, Jérémy Scheurer, Adam Gleave, and Shoshannah Tekofsky for their feedback on a draft of this post, to Euan McLean (via FAR AI) for his feedback and a lot of help with editing, and to everyone else who discussed this agenda with me, in... | https://www.lesswrong.com/posts/L8LHBTMvhLDpxDaqv/research-agenda-formalizing-abstractions-of-computations-1 |
# You Don't Exist, Duncan
This is an experimental essay, not in the typical LessWrong or Duncan Sabien style.
Depending on how this goes, I might try writing a companion piece in the typical style, laying out the model clearly and explicitly and deriving concrete and specific recommendations from it.
But it seemed w... | https://www.lesswrong.com/posts/yepKvM5rsvbpix75G/you-don-t-exist-duncan |
# Epoch Impact Report 2022
[Epoch](https://epochai.org/blog/announcing-epoch) is a research group forecasting the development of transformative Artificial Intelligence. We try to understand how progr... | https://www.lesswrong.com/posts/WJJ5L5kMmsLpSnYWF/epoch-impact-report-2022 |
# Don't Judge a Tool by its Average Output
Epistemic status: this post’s main point itself is probably somewhat obvious. The concrete examples may leave some room for disagreement. No guarantees about my views on ChatGPT vs GPT3 in particular.
*Summary: Many tools and platforms out there provide users with certain in... | https://www.lesswrong.com/posts/gfeevXjXYDu3p9NHF/don-t-judge-a-tool-by-its-average-output |
# You are probably not a good alignment researcher, and other blatant lies
When people talk about research ability, a common meme I keep hearing goes something like this:
* Someone who would become a great alignment researcher will probably not be stopped by Confusions about a Thorny Technical Problem X that's Only O... | https://www.lesswrong.com/posts/ZRYqXHdiFrdxLAmue/you-are-probably-not-a-good-alignment-researcher-and-other |
# Covid 2/2/23: The Emergency Ends on 5/11
We finally have a date. On May 11, the official Covid Emergency ends. The actual emergency is, of course, long over, but for technical reasons the official change will have to wait a bit.
A related question is when it stops making sense to continue writing weekly Covid posts... | https://www.lesswrong.com/posts/srAcoXE9LbeEXLrXp/covid-2-2-23-the-emergency-ends-on-5-11 |
# Heritability, Behaviorism, and Within-Lifetime RL
I’m a firm subscriber to both:
* **(A)** The theory that people’s personalities are significantly predictable from their genes, and mostly independent of how their parents raised them (at least within the typical distribution, i.e. leaving aside cases of flagrant ... | https://www.lesswrong.com/posts/aodPs8H9dQxpXAcwk/heritability-behaviorism-and-within-lifetime-rl |
# More findings on maximal data dimension
Produced as part of the [SERI ML Alignment Theory Scholars Program](https://www.serimats.org/) \- Winter 2022 Cohort.
I’d like to thank Wes Gurnee, Aryan Bhatt, Eric Purdy and Stefan Heimersheim for discussions and Evan Hubinger, Neel Nanda, Adam Jermyn and Chris Olah for men... | https://www.lesswrong.com/posts/WfdxXhszxFc3BxZ8r/more-findings-on-maximal-data-dimension |
# How can I help inflammation-based nerve damage be temporary?
**Bounty \[closed\]: ** $50 for any suggestion that I consider promising enough to change my behavior for, conditional on my fully recovering by May 1.[^8321cw08gzx][^7kaxr1vjqk6][^b86v49r9of6][^57wrri2xi2]
**Question:** I recently got shingles; I'm now c... | https://www.lesswrong.com/posts/xd7c9CdyfBHBjGfiv/how-can-i-help-inflammation-based-nerve-damage-be-temporary |
# Andrew Huberman on How to Optimize Sleep
**Edit**: I am not as confident anymore as I have been in Andrew Huberman's general epistemics, so take this post with a grain of salt.
Introduction
============
In this post, I list all the advice from Andrew Huberman's podcast [Sleep Toolkit: Tools for Optimizing Sleep &... | https://www.lesswrong.com/posts/byRksvtDhK7sSc5Zm/andrew-huberman-on-how-to-optimize-sleep |
# Normative vs Descriptive Models of Agency
*Produced as part of the* [*SERI ML Alignment Theory Scholars Program*](https://serimats.org/) *\- Winter 2022 Cohort.*
I think there's an important distinction to be made between work in agent foundations which is concerned with normative models, and work which is concerne... | https://www.lesswrong.com/posts/nzRh8yQHi3bx9bLsD/normative-vs-descriptive-models-of-agency-2 |
# Conditioning Predictive Models: Large language models as predictors
_This is the first of seven posts in the [Conditioning Predictive Models Sequence](https://www.alignmentforum.org/s/n3utvGrgC2SGi9xQX) based on the paper “[Conditioning Predictive Models: Risks and Strategies](https://arxiv.org/abs/2302.00805)” by E... | https://www.lesswrong.com/posts/XwXmedJAo5m4r29eu/conditioning-predictive-models-large-language-models-as |
# Conditioning Predictive Models: Outer alignment via careful conditioning
_This is the second of seven posts in the [Conditioning Predictive Models Sequence](https://www.alignmentforum.org/s/n3utvGrgC2SGi9xQX) based on the paper “[Conditioning Predictive Models: Risks and Strategies](https://arxiv.org/abs/2302.00805)... | https://www.lesswrong.com/posts/3kkmXfvCv9DmT3kwx/conditioning-predictive-models-outer-alignment-via-careful |
# Halifax Monthly Meetup: Introduction to Effective Altruism
In the interest of growing the LW/EA/ACX community in Halifax, we are hosting monthly meetups. Newcomers are especially welcome!
This month we'll be discussing the elementary ideas of effective altruism. What is effective altruism? What do effective altruis... | https://www.lesswrong.com/events/uwq6CZ53gvfiELtrq/halifax-monthly-meetup-introduction-to-effective-altruism |
# What fact that you know is true but most people aren't ready to accept it?
Understanding and updating beliefs on deeply engrained topics can take enormous efforts, but sometimes it can be so hard that the listener cannot even in principle accept the new reality. The listener is simply not ready, he lacks a vast back... | https://www.lesswrong.com/posts/twdjDGDZtyHqZvLiH/what-fact-that-you-know-is-true-but-most-people-aren-t-ready |
# I don't think MIRI "gave up"
*Disclaimer: I haven't run this by Nate or Eliezer, if they think it mischaracterizes them, whoops.*
I have seen many people assume MIRI (or Eliezer) ((or Nate?)) has "given up", and been telling other people to give up.
On one hand: if you're only half paying attention, and skimmed a ... | https://www.lesswrong.com/posts/DTLPdtMdgMgvetokW/i-don-t-think-miri-gave-up |
# Taboo P(doom)
I think it's actively unhelpful to talk about P(doom) because it fails to distinguish between literally everyone dying and humanity failing to capture 99.99% of the value of the future under a total utilitarian view but in practice, everyone who's born lives a very good life. These are very different o... | https://www.lesswrong.com/posts/omDu7vNy3YyKXsvCd/taboo-p-doom |
# Housing and Transit Roundup #3
Previously: [#1](https://thezvi.substack.com/p/housing-and-transit-thoughts-and), [#2](https://thezvi.substack.com/p/housing-and-transit-thoughts-and-c07)
In honor of much of northern California coming face to face with the Builder’s Remedy, I’m doing this first, ahead of the monthly ... | https://www.lesswrong.com/posts/zjjMxeKmArpwz4zDY/housing-and-transit-roundup-3 |
# Reply to Duncan Sabien on Strawmanning
> DAVID: [...] Because that's all we can ever have of each other: an imitation, a Copy. All we can ever know about are the portraits of each other inside our own skulls.
> LORAINE: Is that all you think I am? An idea in your head?
> DAVID: No! But if it's all I have, then i... | https://www.lesswrong.com/posts/5zjucvhSFvp92eYbE/reply-to-duncan-sabien-on-strawmanning |
# Many AI governance proposals have a tradeoff between usefulness and feasibility
*Epistemic status: We think this is a simple and common idea in discussions about AI governance proposals. We don’t expect this to be controversial, but we think it might be useful to put a *[*label on it *](https://www.lesswrong.com/pos... | https://www.lesswrong.com/posts/JHE3ZKCrtvCuxFEMe/many-ai-governance-proposals-have-a-tradeoff-between |
# [Linkpost] Google invested $300M in Anthropic in late 2022
A few quotes from the article:
Google has invested about $300mn in artificial intelligence start-up Anthropic, making it the latest tech giant to throw its money and computing power behind a new generation of companies trying to claim a place in the booming... | https://www.lesswrong.com/posts/aF8Hnnmn5au5DLvYk/linkpost-google-invested-usd300m-in-anthropic-in-late-2022 |
# 2+2=π√2+n
This [post](https://www.lesswrong.com/posts/yepKvM5rsvbpix75G/you-don-t-exist-duncan) mentions an equation from Ender's shadow
$$
2+2=\pi\sqrt{2+n}
$$
Which kind of annoys me because either:
1. It's trivial to solve using algebra
2. The joke is that the answer is irrational so it takes "forever" to so... | https://www.lesswrong.com/posts/fT8vsyFqcumtJ62rE/2-2-p-2-n |
# Update on Book Review Dominant Assurance Contract
This is a follow-up to [Consider paying for literature or book reviews using bounties and dominant assurance contracts](https://forum.effectivealtruism.org/posts/dMhscQ9sTXCaegYju/consider-paying-for-literature-or-book-reviews-using) (LessWrong cross-post [here](http... | https://www.lesswrong.com/posts/yYvgmKGMtzKxbJZew/update-on-book-review-dominant-assurance-contract |
# Small Talk is Good, Actually
*You all liked the* [*last*](https://www.lesswrong.com/posts/kJoyRgDDPzMg4Fo3Z/nice-clothes-are-good-actually) [*two*](https://www.lesswrong.com/posts/gibjCymJvNfENrwqk/exercise-is-good-actually)*, so I'm going to keep it going so long as I have ideas and time. This is another casual po... | https://www.lesswrong.com/posts/bigpXkTkuci6cTfZm/small-talk-is-good-actually |
# Fucking Goddamn Basics of Rationalist Discourse
1. [Don't say false shit](https://www.lesswrong.com/posts/XPv4sYrKnPzeJASuk/basics-of-rationalist-discourse-1#1__Don_t_say_straightforwardly_false_things_) omg this one's so basic what are you even doing. And to be perfectly fucking clear "false shit" includes exagger... | https://www.lesswrong.com/posts/gPPdYTwPkdvAEYyEK/fucking-goddamn-basics-of-rationalist-discourse |
# Path-Dependence in ChatGPT's Political Outputs
There's a [meme](https://twitter.com/stillgray/status/1620880361407545345) [going](https://twitter.com/KatrinaPanova/status/1621587583313772549) [around](https://twitter.com/Trinhducanhkho3/status/1621213930155831296) where ChatGPT will happily say good things about an... | https://www.lesswrong.com/posts/qm3hLejG92DmRArWB/path-dependence-in-chatgpt-s-political-outputs |
# AXRP Episode 19 - Mechanistic Interpretability with Neel Nanda
[YouTube link](https://www.youtube.com/watch?v=3YbE7zybc5k&list=PLmjaTS1-AiDeqUuaJjasfrM6fjSszm9pK&index=19)
How good are we at understanding the internal computation of advanced machine learning models, and do we have a hope at getting better? In this ... | https://www.lesswrong.com/posts/r2yTwkGt3kbQG2mXi/axrp-episode-19-mechanistic-interpretability-with-neel-nanda |
# Empathy as a natural consequence of learnt reward models
**Epistemic Status:** *Pretty speculative but built on scientific literature. This post builds off my previous post on* [*learnt reward models*](https://www.lesswrong.com/posts/RorXWkriXwErvJtvn/agi-will-have-learnt-utility-functions)*. Crossposted from my* [*... | https://www.lesswrong.com/posts/zaER5ziEprE7aNm6u/empathy-as-a-natural-consequence-of-learnt-reward-models |
# The Future of Structured Self Improvement
Two years ago, we founded the Guild of the ROSE to provide structure and community for people interested in self improvement. Since then, we've grown from a three-month Alpha project to a year-round workshop host.
If you've been in our [Discord server](https://discord.gg/rn... | https://www.lesswrong.com/posts/utxfCAife3gSR6Bmn/the-future-of-structured-self-improvement |
# Who invented knitting? The plot thickens
Last time on Eukaryote Writes Blog: [You learned about knitting history](https://eukaryotewritesblog.com/2022/08/04/fiber-arts-mysterious-dodecahedrons-and-waiting-on-eureka/).
You thought you were done learning about knitting history? You fool. You buffoon. I wanted to do... | https://www.lesswrong.com/posts/iZYMDpxnw2EkTTmdq/who-invented-knitting-the-plot-thickens |
# Modal Fixpoint Cooperation without Löb's Theorem
*Followed By:* [*Payor's Lemma in Natural Language*](https://www.lesswrong.com/posts/iCDBQtby4L2fZ7yns/payor-s-lemma-in-natural-language)
*TL;DR: This post introduces a novel logical approach to achieving group-scale cooperation, based on modal fixpoint theory. ... | https://www.lesswrong.com/posts/2WpPRrqrFQa6n2x3W/modal-fixpoint-cooperation-without-loeb-s-theorem |
# I hired 5 people to sit behind me and make me productive for a month
***Warning: this is not in typical LessWrong "style", but nevertheless I think it is of interest to people here.***
Most people approach productivity from the bottom up. They notice something about a process that feels inefficient, so they set out... | https://www.lesswrong.com/posts/gp9pmgSX3BXnhv8pJ/i-hired-5-people-to-sit-behind-me-and-make-me-productive-for |
# Podcast with Oli Habryka on LessWrong / Lightcone Infrastructure
OK sorry to over-advertise but it seemed like this one would be of interest to the LessWrong and EA communities. Episode description below, audio is [here](https://youtu.be/4lbeutseEJQ), or search for "The Filan Cabinet Habryka" wherever you listen to ... | https://www.lesswrong.com/posts/jBPNK4iREjZr36sDx/podcast-with-oli-habryka-on-lesswrong-lightcone |
# Pandemic Prediction Checklist: H5N1 (6/14)
H5N1 likely has spread among minks, the first major spread from mammal to mammal
*Changelog available at bottom of post.*
Here's what we can do about ... | https://www.lesswrong.com/posts/G5cDmYLY9GzpzDeFe/pandemic-prediction-checklist-h5n1-6-14 |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.