text stringlengths 300 320k | source stringlengths 52 154 |
|---|---|
# "Pivotal Acts" means something specific
The term [Pivotal Act](https://arbital.com/p/pivotal/) was written up on Arbital in 2015. I only started hearing it discussed in 2020, and then it rapidly started seeing more traction when the [MIRI 2021 Conversations](https://www.lesswrong.com/sequences/n945eovrA3oDueqtq) wer... | https://www.lesswrong.com/posts/PKBXczqhry7iK3Ruw/pivotal-acts-means-something-specific |
# Puddle Temperature Alarm
I cook pancakes for my kids most mornings, and the griddle takes a little while to heat up. If you look around for articles on how to know when your pan is hot enough you'll see people suggest dropping in a little water and seeing how it behaves. The problem is, this requires you to be payin... | https://www.lesswrong.com/posts/vb2snk7pdACJBMuM4/puddle-temperature-alarm |
# Staying Split: Sabatini and Social Justice
*Author's note: This is a contemporary post about an ongoing current event. It's not a timeless essay in the way that most of my essays are, though it does contain insight and thoughts on timeless topics.*
* * *
**Background**
--------------
David Sabatini is a molecula... | https://www.lesswrong.com/posts/YTJp5WBcktBimdxBG/staying-split-sabatini-and-social-justice |
# Eliciting Latent Knowledge (ELK) - Distillation/Summary
This post was inspired by the [AI safety distillation contest](https://forum.effectivealtruism.org/posts/ei4pYFJKcbGAdGnNb/calling-for-student-submissions-ai-safety-distillation). It turned out to be more of a summary than a distillation for two reasons. Firstl... | https://www.lesswrong.com/posts/rxoBY9CMkqDsHt25t/eliciting-latent-knowledge-elk-distillation-summary |
# Entitlement as a major amplifier of unhappiness
A crude simplification of an important part of Buddhist thought is
that [desire is the root cause of all suffering](https://www.google.com/search?q=desire+is+the+root+of+all+suffering). There are many
nuances to this, with the central theme being that [*attachment* to
... | https://www.lesswrong.com/posts/psMczuJ9KbujgrgLq/entitlement-as-a-major-amplifier-of-unhappiness |
# Has anyone actually tried to convince Terry Tao or other top mathematicians to work on alignment?
This has been discussed several times in the past, see:
* [Have You Tried Hiring People?](https://www.lesswrong.com/posts/ido3qfidfDJbigTEQ/have-you-tried-hiring-people), a LW post
* It talks about [this ACX co... | https://www.lesswrong.com/posts/SG6fcAFhhJjys2WtH/has-anyone-actually-tried-to-convince-terry-tao-or-other-top |
# Comment reply: my low-quality thoughts on why CFAR didn't get farther with a "real/efficacious art of rationality"
Hi! I was writing this originally as a comment-reply to [this thread](https://www.lesswrong.com/posts/McsqZQBbwJnDXimNp/what-is-going-on-with-cfar?commentId=EC8L8t75aP9bkQBX3), but my reply is long, so... | https://www.lesswrong.com/posts/B9kP6x5rpmuCzpfWb/comment-reply-my-low-quality-thoughts-on-why-cfar-didn-t-get |
# [Linkpost & Discussion] AI Trained on 4Chan Becomes ‘Hate Speech Machine’ [and outperforms GPT-3 on TruthfulQA Benchmark?!]
I just came across [this story](https://www.vice.com/en/article/7k8zwx/ai-trained-on-4chan-becomes-hate-speech-machine), which seems potentially relevant to the community, both from the perspec... | https://www.lesswrong.com/posts/jwrciTJLSJyinBgbR/linkpost-and-discussion-ai-trained-on-4chan-becomes-hate |
# The “mind-body vicious cycle” model of RSI & back pain
*(Written in a hurry, but hopefully better than nothing.)*
Here are three alternative models of what might be happening in people diagnosed with chronic back pain, and repetitive strain injury (RSI), and carpal tunnel syndrome, and maybe other things too:
_Click lower right to download or find on Apple Podcasts, Spotify, Stitcher, etc._
I've been working on a new series of posts about the [most important century](https://www.c... | https://www.lesswrong.com/posts/oBBzqkZwkxDvsKBGB/ai-could-defeat-all-of-us-combined |
# Website For Yoda Timers
[Yoda Timers](https://www.lesswrong.com/tag/five-minute-timers) are pretty awesome. From the Wiki:
> Sometimes, all you need to solve an impossible problem is to actually think for 5 minutes. (Sometimes it is actually 15 minutes, or two hours. But, the first five minutes is helpful for trans... | https://www.lesswrong.com/posts/9ZkhTkCok7vvJ6iGF/website-for-yoda-timers |
# Covid 6/9/22: Nice
Things seem to have peaked in the Northeast once again. Nice.
Things are getting worse in the South as the summer hits. Not as nice, not that unexpected. At this point we can pattern match.
You know what else is nice? Not having to care about the pandemic. I do my best to care about it so you do... | https://www.lesswrong.com/posts/5kwstcMX9aTcu2Afb/covid-6-9-22-nice |
# Book Review: How the World Became Rich
In [*How the World Became Rich*](https://amzn.to/3avF169), Mark Koyama and Jared Rubin have written what will almost certainly become a beloved economic history classic, as [*Power and Plenty*](https://amzn.to/3tn0gOa) or [*The British Industrial Revolution in Global Perspecti... | https://www.lesswrong.com/posts/wgFX779mgmAFWkZeg/book-review-how-the-world-became-rich |
# If there was a millennium equivalent prize for AI alignment, what would the problems be?
If someone were to create one or a number of huge (e.g 1 to 10 million dollar) prizes for solving important problems related to AI alignment, it might have a decent chance of focusing a lot of researchers on the topic, which mig... | https://www.lesswrong.com/posts/nEy2JDjvqE6o5E6kx/if-there-was-a-millennium-equivalent-prize-for-ai-alignment |
# Why it's bad to kill Grandma
1.
==
In college, I had a friend who was into debate competitions. One weekend, the debate club funded him to go to a nearby city for a tournament. When I asked him how it went, he said:
> Oh, I didn’t have enough time to prepare, so I just skipped the tournament and partied with some ... | https://www.lesswrong.com/posts/HBbXgoQxL7cstWCAG/why-it-s-bad-to-kill-grandma |
# Operationalizing two tasks in Gary Marcus’s AGI challenge
I’ve been thinking about the Gary Marcus/Elon Musk AGI challenge. As you may know, Elon Musk casually tweeted that he expected to see AGI in 2029. Gary Musk decided to take him seriously and bet him $100,000 that it wouldn’t happen. [He specified five criteri... | https://www.lesswrong.com/posts/rB2oHSQdBuevfenbD/operationalizing-two-tasks-in-gary-marcus-s-agi-challenge |
# Progress links and tweets, 2022-06-08
Best of my Twitter, May 31–June 6 (a few days behind, still ironing out the process… For the latest, [follow me on Twitter: @jasoncrawford](https://twitter.com/intent/follow?screen_name=jasoncrawford).)
After [last time](https://rootsofprogress.org/links-and-tweets-2022-05-30) ... | https://www.lesswrong.com/posts/h2h4QA7fhGjr5QKNq/progress-links-and-tweets-2022-06-08 |
# How Do Selection Theorems Relate To Interpretability?
One pretty major problem with today’s interpretability methods (e.g. work by [Chris Olah & co](https://transformer-circuits.pub/2021/framework/index.html)) is that we have to redo a bunch of work whenever a new net comes out, and even more work when a new archite... | https://www.lesswrong.com/posts/A7QgKwWvAkuXonAy5/how-do-selection-theorems-relate-to-interpretability |
# Forestalling Atmospheric Ignition
It's April, 1945 and you're a high ranking scientist in the Manhattan Project. You also happen to have the ear of President Truman, so your words and actions have a major impact on world outcomes. Everyone has been working hard on the bomb, determined to defeat the Nazis, except tha... | https://www.lesswrong.com/posts/cawxKpee3ATwpQg9o/forestalling-atmospheric-ignition |
# You Only Get One Shot: an Intuition Pump for Embedded Agency
_This is a short attempt to articulate a framing which I sometimes find useful for thinking about [embedded agency](https://www.lesswrong.com/tag/embedded-agency). I noticed that I wanted to refer to it a few times in conversations and other writings._
A ... | https://www.lesswrong.com/posts/HiufALieNbWHqR9en/you-only-get-one-shot-an-intuition-pump-for-embedded-agency |
# Bureaucracy of AIs
In my post about [AI Alignment Strategies](https://www.lesswrong.com/posts/8ibDJeoiDuxJkPwfa/various-alignment-strategies-and-how-likely-they-are-to-work), I strongly endorsed an approach I call Bureaucracy of AIs.
Specifically, I made this claim:
> In the first place, \[Bureaucracy of AIs\] is ... | https://www.lesswrong.com/posts/p7XnbyP5ehh33fEY7/bureaucracy-of-ais |
# why assume AGIs will optimize for fixed goals?
When I read posts about AI alignment on LW / AF/ Arbital, I almost always find a particular bundle of assumptions taken for granted:
* An AGI has a single terminal **goal**[^5a3i0rmzhi].
* The goal is a **fixed** part of the AI's structure. The internal dynamics ... | https://www.lesswrong.com/posts/dKTh9Td3KaJ8QW6gw/why-assume-agis-will-optimize-for-fixed-goals |
# Open Problems in AI X-Risk [PAIS #5]
*This is the fifth post in* [*a sequence of posts*](https://www.alignmentforum.org/posts/bffA9WC9nEJhtagQi/introduction-to-pragmatic-ai-safety-pragmatic-ai-safety-1) *that describe our models for Pragmatic AI Safety.*
Dan Hendrycks (an author of this post), Nicholas Carlini, Joh... | https://www.lesswrong.com/posts/5HtDzRAk7ePWsiL2L/open-problems-in-ai-x-risk-pais-5 |
# A plausible story about AI risk.
This was originally supposed to be a response to the new AGI Safety FAQ-in-progress, but it got a bit too long. Anonymous writes:
> A lot of the AI risk arguments seem to come... with a very particular transhumanist aesthetic about the future (nanotech, ... etc.). I find these thi... | https://www.lesswrong.com/posts/LtdbPZxLuYktYhveL/a-plausible-story-about-ai-risk |
# Tao, Kontsevich & others on HLAI in Math
I found this 2015 panel with Terence Tao and some other eminent mathematicians to be interesting. The panel covered various topics but got into the question of when computers will be able to do research-level mathematics. Most interestingly, Maxim Kontsevich was alone in pre... | https://www.lesswrong.com/posts/ouvFSmzDwbBojJMCm/tao-kontsevich-and-others-on-hlai-in-math |
# [linkpost] The final AI benchmark: BIG-bench
There is a Google project called [BIG-bench](https://github.com/google/BIG-bench), with contributions from OpenAI and other big orgs. They've crowdsourced >200 of highly diverse text tasks (from answering scientific questions to predicting protein interacting sites to mea... | https://www.lesswrong.com/posts/qjproXBGPQSAF9Hbd/linkpost-the-final-ai-benchmark-big-bench |
# I No Longer Believe Intelligence to be "Magical"
### Epistemic Status
[This started out as a shortform](https://www.lesswrong.com/posts/zhhYwM7gk8LZsDzxj/dragongod-s-shortform?commentId=9WoCvnBZe6KsgG4q5) and is held to the same epistemic standards that I do for [my Twitter account](https://twitter.com/CineraVerini... | https://www.lesswrong.com/posts/rqAvheoRHwSDsXrLw/i-no-longer-believe-intelligence-to-be-magical |
# Leaving Google, Joining the Nucleic Acid Observatory
In 2017 I [rejoined Google](https://www.jefftk.com/p/rejoining-google) to earn money to donate. At the time I thought earning to give was probably not where I could have the most impact, but I [wasn't able to find](https://www.jefftk.com/p/thinking-about-going-bac... | https://www.lesswrong.com/posts/HxH2MGeWEZD83riNo/leaving-google-joining-the-nucleic-acid-observatory |
# Another plausible scenario of AI risk: AI builds military infrastructure while collaborating with humans, defects later.
As was said [before](https://www.lesswrong.com/posts/LtdbPZxLuYktYhveL/a-plausible-story-about-ai-risk): "A lot of the AI risk arguments seem to come... with a very particular transhumanist aesthe... | https://www.lesswrong.com/posts/8z7WE8RGkoYAevzzv/another-plausible-scenario-of-ai-risk-ai-builds-military |
# How dangerous is human-level AI?
*I don't expect anything here to be original; this is just me thinking things through, and perhaps this post will be useful as a relatively stand-alone argument.*
Typically, you can envision many many more future states than there are future states you could bring about. You can ima... | https://www.lesswrong.com/posts/z6H46mAuKwvyykCKp/how-dangerous-is-human-level-ai |
# Summary of "AGI Ruin: A List of Lethalities"
Overview
========
Introduction
------------
This post is a summary of Eliezer Yudkowsky’s post “AGI Ruin: A List of Lethalities”. I wrote it because I think the original post is longer and less organized than I would like it to be. The purpose of this post is to summari... | https://www.lesswrong.com/posts/Wic2P2bGejbFH3Sxb/summary-of-agi-ruin-a-list-of-lethalities |
# How fast can we perform a forward pass?
_Thanks to Hao Zhang, Kayvon Fatahalian, and Jean-Stanislas Denain for helpful discussions and comments._
Over the last month, I’ve spent a lot of time trying to answer the following question:
> How quickly can we perform one forward pass in a transformer model?
By a transf... | https://www.lesswrong.com/posts/gPmGTND8Kroxgpgsn/how-fast-can-we-perform-a-forward-pass |
# Show LW: YodaTimer.com
I just launched https://yoda-timer.netlify.app/. It's a simple little website that makes it a bit more fun to use [Yoda Timers](https://www.lesswrong.com/tag/five-minute-timers). See [Website For Yoda Timers](https://www.lesswrong.com/posts/9ZkhTkCok7vvJ6iGF/website-for-yoda-timers) for the in... | https://www.lesswrong.com/posts/YueZyWQThn9yfxn4y/show-lw-yodatimer-com |
# The Mountain Troll
It was a sane world. A Rational world. A world where every developmentally normal teenager was taught Bayesian probability.
Saundra's math class was dressed in their finest robes. Her teacher, Mr Waze, had invited the monk Ryokan to come speak. It was supposed to be a formality. Monks rarely came... | https://www.lesswrong.com/posts/exvuvFZniGh2R5nvd/the-mountain-troll |
# Steganography and the CycleGAN - alignment failure case study
*1\. This is a (lightly edited) transcript of a lightning talk I gave during the* [*Virtual AI Safety Camp 2022*](https://aisafety.camp/)*. The original talk in video format can be found* [*here*](https://drive.google.com/file/d/1v-nfm6aAyGn3zGQbXc8sd2VJ6... | https://www.lesswrong.com/posts/uutXLm2DRcCtFBZ2D/steganography-and-the-cyclegan-alignment-failure-case-study |
# Godzilla Strategies
> Clutching a bottle of whiskey in one hand and a shotgun in the other, John scoured the research literature for ideas... He discovered several papers that described software-assisted hardware recovery. The basic idea was simple: if hardware suffers more transient failures as it gets smaller, why... | https://www.lesswrong.com/posts/DwqgLXn5qYC7GqExF/godzilla-strategies |
# Why don't you introduce really impressive people you personally know to AI alignment (more often)?
I have a few questions to the subset of readers who:
1. Believe technical AI alignment research is both important and hard to make significant progress in
2. Have a personal connection with a person who doesn't know... | https://www.lesswrong.com/posts/YmrFWphM3Gne29iLJ/why-don-t-you-introduce-really-impressive-people-you |
# Poorly-Aimed Death Rays
**Alternate framing:** [Optimality is the tiger, and agents are its teeth](https://www.lesswrong.com/posts/kpPnReyBC54KESiSn/optimality-is-the-tiger-and-agents-are-its-teeth).
**Tonally relevant:** [Godzilla Strategies](https://www.lesswrong.com/posts/DwqgLXn5qYC7GqExF/godzilla-strategies).
... | https://www.lesswrong.com/posts/EHSJD8qTnuFHG73fd/poorly-aimed-death-rays |
# Beauty and the Beast
`**I remember when I first realized I was hideous. Generally, my autobiographical memory is poor** ` \- but pain is the best a*nti*-*amnesic* and it was among the more painful days of my life.
Before that first day of Kindergarten, my parents kept me isolated. They say they did this to protect ... | https://www.lesswrong.com/posts/Ke7DiT2DHMyGiv3s2/beauty-and-the-beast |
# Why has no person / group ever taken over the world?
I'm actually kind of surprised at this, now that I think of it. Clearly there have been numerous people who have *wanted* to take over the world, and there have been several individuals or empires that have taken over *significant fractions* of the world. Are the... | https://www.lesswrong.com/posts/AARyrAYuyjGwibGpA/why-has-no-person-group-ever-taken-over-the-world |
# How much stupider than humans can AI be and still kill us all through sheer numbers and resource access?
It is reasonably clear that not just a super-human but "just" a human-level AGI that can "clone itself" can defeat humanity through concerted effort and determination, should they choose to. I wonder how much les... | https://www.lesswrong.com/posts/9q9WzG6hwNhifTK2m/how-much-stupider-than-humans-can-ai-be-and-still-kill-us |
# A claim that Google's LaMDA is sentient
https://cajundiscordian.medium.com/is-lamda-sentient-an-interview-ea64d916d917 apparently posted by a Google engineer.
It could be an elaborate hoax, and has remnants of gwern's idea (https://www.gwern.net/fiction/Clippy) of a transformer waking up and having internal experie... | https://www.lesswrong.com/posts/vqgpDoY4eKyNnWoFd/a-claim-that-google-s-lamda-is-sentient |
# How To: A Workshop (or anything)
*Author's note: while the whole piece is a 55min read, there's a ~7min standalone essay at the top, and then a break and a shift into a different kind of content. You really actually for real can just read the 7 minutes and then stop, if that satisfies the need you had going in. Or ... | https://www.lesswrong.com/posts/WotMLgX9hHXzBxKCB/how-to-a-workshop-or-anything |
# Why all the fuss about recursive self-improvement?
*This article was outlined by Nate Soares, inflated by Rob Bensinger, and then edited by Nate. Content warning: the tone of this post feels defensive to me. I don't generally enjoy writing in "defensive" mode, but I've had this argument thrice recently in surprising... | https://www.lesswrong.com/posts/8NKu9WES7KeKRWEKK/why-all-the-fuss-about-recursive-self-improvement |
# Intuitive Explanation of AIXI
[AIXI](https://www.lesswrong.com/tag/aixi) is a mathematical notion of an ideal RL agent, developed by Marcus Hutter. The famous formula for AIXI is as follows:
$$
a_k = \text{argmax}_{a_k} \sum_{o_kr_k} \text{max}_{a_{k+1}} \sum_{o_{k+1} r_{k+1}} \dots \sum_{o_mr_m} [r_k + \dots + r_... | https://www.lesswrong.com/posts/DFdSD3iKYwFS29iQs/intuitive-explanation-of-aixi |
# How are compute assets distributed in the world?
I own probably about 10^13 FLOP/s, mostly in my pc (currently in Amsterdam). Google owns more than that, in various datacenters (e.g. one in The Dalles, Oregon).
I'd like to get a better picture of the distribution of FLOP/s in the world. Any high level broadly relev... | https://www.lesswrong.com/posts/ty38fuTWv6rdoFTaF/how-are-compute-assets-distributed-in-the-world |
# How much does cybersecurity reduce AI risk?
AI is here, and AGI is coming. It's quite possible that any work being done now will be futile in comparison to reducing AI risk.
This is one of those things that's unsettling for me as someone who did a Ph. D. in a non-AI area of computer science.
But one of the mai... | https://www.lesswrong.com/posts/aC8pwXNzckJ7LRAPt/how-much-does-cybersecurity-reduce-ai-risk |
# Grokking “Semi-informative priors over AI timelines”
*Notes: *
* *I give visual explanations for Tom Davidson’s report,* [*Semi-informative priors over AI timelines*](https://www.openphilanthropy.org/semi-informative-priors)*, and summarise the key assumptions and intuitions*
* *The diagrams can be found* [*her... | https://www.lesswrong.com/posts/Mj4CWRauhF3DzpLvu/grokking-semi-informative-priors-over-ai-timelines |
# Code Quality and Rule Consequentialism
(See also [Taking the outside view on code quality](https://adamzerner.bearblog.dev/taking-the-outside-view-on-code-quality/))
Code quality. Such a divisive topic. To overgeneralize[^1], managers always want things released soon, which means a quick and dirty version of the co... | https://www.lesswrong.com/posts/ckWnPziauk5hcrros/code-quality-and-rule-consequentialism |
# Why so little AI risk on rationalist-adjacent blogs?
I read a lot of rationalist-adjacents. Outside of LessWrong and ACX, I hardly ever see posts on AI risk. Tyler Cowen of Marginal Revolution [writes](https://marginalrevolution.com/marginalrevolution/2022/02/are-nuclear-weapons-or-rogue-ai-the-more-dangerous-existe... | https://www.lesswrong.com/posts/RsDwRmHGvf6GqaQkE/why-so-little-ai-risk-on-rationalist-adjacent-blogs |
# The beautiful magical enchanted golden Dall-e Mini is underrated
\[Showcasing some of the pictures that my daughter and I created with the open source generative model [Dalle-Mini](https://huggingface.co/spaces/dalle-mini/dalle-mini). No deeper insights, just an appreciation how well it works in some cases.\]
I hav... | https://www.lesswrong.com/posts/NciL5tukFCAihrKNN/the-beautiful-magical-enchanted-golden-dall-e-mini-is |
# What's the "This AI is of moral concern." fire alarm?
Given the [recent noise](https://www.lesswrong.com/posts/vqgpDoY4eKyNnWoFd/a-claim-that-google-s-lamda-is-sentient) on this issue around LaMDA, I thought it might be a good idea to have some discussion around this point. I'm curious about what possible evidence w... | https://www.lesswrong.com/posts/ZGgneqEJXJLJxrBZD/what-s-the-this-ai-is-of-moral-concern-fire-alarm |
# D&D.Sci June 2022 Evaluation and Ruleset
This is a followup to [the D&D.Sci post](https://www.lesswrong.com/posts/FjgS5mZCQedarYdCb/d-and-d-sci-june-2022-a-goddess-tried-to-reincarnate-me-into) I made ten days ago; if you haven’t already read it, you should do so now before spoiling yourself.
[Here](https://h-b-p.g... | https://www.lesswrong.com/posts/WvvX2KX2zavNtTBDA/d-and-d-sci-june-2022-evaluation-and-ruleset |
# On A List of Lethalities
Response to (Eliezer Yudkowsky): [A List of Lethalities](https://www.lesswrong.com/posts/uMQ3cqWDPHhjtiesc/agi-ruin-a-list-of-lethalities).
_Author’s Note: I do not work in AI Safety, lack technical domain knowledge and in many ways am going to be wrong. I wasn’t going to write this to avoi... | https://www.lesswrong.com/posts/LLRtjkvh9AackwuNB/on-a-list-of-lethalities |
# Training Trace Priors
I’m worried about scenarios involving deceptive models. We’ve failed at inner alignment so the model has goals that are not aligned with ours. It can somehow detect when it’s in training, and during training it pretends to share our goals. During deployment, surprise! The model paperclips the u... | https://www.lesswrong.com/posts/hjrqXjEpaw9ogScPh/training-trace-priors |
# Contra EY: Can AGI destroy us without trial & error?
> NB: I've never published on LW before and apologize if my writing skills are not in line with LW's usual style. This post is an edited copy of the [same article](https://nsokolsky.substack.com/p/can-agi-destroy-us-without-trial?s=w) in my blog.
EY published an... | https://www.lesswrong.com/posts/qfDgEreMoSEtmLTws/contra-ey-can-agi-destroy-us-without-trial-and-error |
# Crypto-fed Computation
## Overview
The idea of this post is to describe, discuss, and if warranted understand how to create, a model of *crypto-fed computation.*[^1] The basic idea is that high-powered GPU (or other ML-specialized) hardware could be equipped with in-chip hardware locks such that the computational c... | https://www.lesswrong.com/posts/RX4k3svXibctfQkJm/crypto-fed-computation |
# Continuity Assumptions
This post will try to explain what I mean by continuity assumptions (and discontinuity assumptions), and why differences in these are upstream of many disagreements about AI safety. None of this is particularly new, but seemed worth reiterating in the form of a short post.
What I mean by cont... | https://www.lesswrong.com/posts/cHJxSJ4jBmBRGtbaE/continuity-assumptions |
# OpenAI: GPT-based LLMs show ability to discriminate between its own wrong answers, but inability to explain how/why it makes that discrimination, even as model scales
This seems concerning. Not an expert so unable to tell how concerning it is. Wanted to start a discussion! Full text: [https://openai.com/blog/critiqu... | https://www.lesswrong.com/posts/yTp9s4LrJn6ppLvvM/openai-gpt-based-llms-show-ability-to-discriminate-between |
# Why multi-agent safety is important
Alternative Title: Mo’ Agents, Mo’ Problems
*This was a semi-*[*adversarial collaboration*](https://en.wikipedia.org/wiki/Adversarial_collaboration#:~:text=In%20science%2C%20adversarial%20collaboration%20is,with%20opposing%20views%20work%20together.) *with Robert Kirk; views are... | https://www.lesswrong.com/posts/pkfKRG9dQr6unrhQT/why-multi-agent-safety-is-important |
# Investigating causal understanding in LLMs
This is a project by [Marius Hobbhahn](https://www.mariushobbhahn.com/aboutme/) and [Tom Lieberum](https://tomfrederik.github.io/). David Seiler contributed ideas and guidance.
Executive summary
=================
We want to investigate the quality of LLMs causal world mo... | https://www.lesswrong.com/posts/yZb5eFvDoaqB337X5/investigating-causal-understanding-in-llms |
# Was the Industrial Revolution The Industrial Revolution?
There was an Industrial Revolution, and it was slow.
The debate’s been over for two decades. The gradualists, armed with new growth accounting techniques, have won, exorcising forever the Ashtonian vision of an eighteenth-century leap into exponential growth.... | https://www.lesswrong.com/posts/eMdHBDTZkvfcFGkAb/was-the-industrial-revolution-the-industrial-revolution |
# How Do You Quantify [Physics Interfacing] Real World Capabilities?
### Disclaimer
Sorry for the length of this "question". The post in which I cover the preliminary context hasn't been published yet (4,000+ words and I'm < 40% through my outline).
Introduction
============
I have done some thinking of how o... | https://www.lesswrong.com/posts/6NSangDDgqsHx2d3R/how-do-you-quantify-physics-interfacing-real-world |
# Blake Richards on Why he is Skeptical of Existential Risk from AI
I have recently interviewed Blake Richards, an Assistant Professor in the Montreal Neurological Institute and the School of Computer Science at McGill University and a Core Faculty Member at MiLA. Below you will find some quotes summarizing his takes ... | https://www.lesswrong.com/posts/dtpLLRbKjoDXs7b2a/blake-richards-on-why-he-is-skeptical-of-existential-risk |
# Slow motion videos as AI risk intuition pumps
***tl;dr:** When making the case for AI as a risk to humanity, trying showing people an evocative illustration of what differences in processing speeds can look like, such as* [*this video*](https://vimeo.com/83664407)*.*
Over the past ~12 years of making the case for A... | https://www.lesswrong.com/posts/Ccsx339LE9Jhoii9K/slow-motion-videos-as-ai-risk-intuition-pumps |
# Yes, AI research will be substantially curtailed if a lab causes a major disaster
There's a narrative that [Chapman](https://twitter.com/Meaningness/status/1536575132302331905?cxt=HHwWgsCiuaK8gdMqAAAA) and other smart people seem to endorse that goes:
> People say a public AI disaster would rally public opinion aga... | https://www.lesswrong.com/posts/LzKJMx9jqt3bDujij/yes-ai-research-will-be-substantially-curtailed-if-a-lab |
# Will vague "AI sentience" concerns do more for AI safety than anything else we might do?
*\[Cross-posting from something I* [*posted on Facebook*](https://www.facebook.com/permalink.php?story_fbid=pfbid032sJE4mmE14XTSvP5dBEgY7k85PS1jYCFn8i1Jfy6DRMvSHme3TTDr3DTbiCF3cP7l&id=100010608396052)*.\]*
Prediction 1 (high co... | https://www.lesswrong.com/posts/nWnCsNnoLwL3fjqkA/will-vague-ai-sentience-concerns-do-more-for-ai-safety-than |
# Favourite new AI productivity tools?
[Transformative AI](https://www.lesswrong.com/tag/transformative-ai) will revolutionize the ways we interact with the world. But maybe narrow [tool AIs](https://www.lesswrong.com/tag/tool-ai) are already starting to do so: Every week, there are new startups, websites, and tools l... | https://www.lesswrong.com/posts/Nribr7pqa3kP9d5Xf/favourite-new-ai-productivity-tools |
# A Butterfly's View of Probability
Cross-posted from [my blog](https://gabrieldwu.github.io/butterfly-probability).
*Thanks to Alex Cai and Haneul Shin for discussing these ideas with me. Thanks to Eric Neyman for feedback.*
![](https://39669.cdn.cke-cs.com/rQvD3VnunXZu34m86e5f/images/daea1a03459e6eae26beddf21c2cb6... | https://www.lesswrong.com/posts/ihMvymdJs4TrgLqnb/a-butterfly-s-view-of-probability |
# Alignment Risk Doesn't Require Superintelligence
Outsize destructive coordination is rare
----------------------------------------
I was 9 years old when 9/11 happened. My gut reaction was, basically, "wait, doesn't this sort of thing happen all the time?" I don't mean to suggest I was wise or world-weary or anythi... | https://www.lesswrong.com/posts/EPEvBANNYN9rxRQFB/alignment-risk-doesn-t-require-superintelligence |
# Our mental building blocks are more different than I thought
I recently had a conversation with one of my best friends about her [Aphantasia](https://en.wikipedia.org/wiki/Aphantasia). In short, she can’t create mental images. When she thinks about the world, she thinks in text, relationships, sounds, and much more ... | https://www.lesswrong.com/posts/LKxcdtZpsxBDrw7wz/our-mental-building-blocks-are-more-different-than-i-thought |
# A central AI alignment problem: capabilities generalization, and the sharp left turn
(*This post was factored out of a larger post that I (Nate Soares) wrote, with help from Rob Bensinger, who also rearranged some pieces and added some text to smooth things out. I'm not terribly happy with it, but am posting it anyw... | https://www.lesswrong.com/posts/GNhMPAWcfBCASy8e6/a-central-ai-alignment-problem-capabilities-generalization |
# Berlin AI Safety Open Meetup June 2022
Hello everyone,
we're a group of ~5 people trying to do independent AI alignment research in (some of) our free time. Once a month after one of our sessions, we have a meetup that's open for everyone who's interested in AI alignment (yes, that's also you!).
__If you're doubtf... | https://www.lesswrong.com/events/qPQ9kuyckNpsvgtCD/berlin-ai-safety-open-meetup-june-2022 |
# Georgism, in theory
*Last year, Astral Codex had [a series](https://astralcodexten.substack.com/p/does-georgism-work-part-2-can-landlords?s=r) of interesting posts on Georgism. The claims felt very counterintuitive to me, so I dove into it until I could at least sort out the theory. And I emerged with a better under... | https://www.lesswrong.com/posts/XoYDmCzeKiB87rs7a/georgism-in-theory |
# What are all the AI Alignment and AI Safety Communication Hubs?
For effective work on complex topics, multiple people need to work together\[citation needed\].
For AI Alignment, LessWrong and its sister site, the [Alignment Forum](https://www.alignmentforum.org/), are arguably the Shelling points. But there are oth... | https://www.lesswrong.com/posts/vmLfa5PEcuAyZX3cC/what-are-all-the-ai-alignment-and-ai-safety-communication |
# FYI: I’m working on a book about the threat of AGI/ASI for a general audience. I hope it will be of value to the cause and the community
The TLDR is the title of this post.
Hi all, long time EA/Rationalist, first time poster (apologies if formatting is off). I’m posting to mention that I’m 30,000 words into a draft... | https://www.lesswrong.com/posts/E5MwmuuryBLF3wWRZ/fyi-i-m-working-on-a-book-about-the-threat-of-agi-asi-for-a |
# I applied for a MIRI job in 2020. Here's what happened next.
I recently saw Rob Bensinger [asking](https://www.lesswrong.com/posts/keiYkaeoLHoKK4LYA/?commentId=ynHnQGG3Qx3cnpecp) what MIRI could do to improve hiring. I tried to get a job at MIRI in February 2020 and I have some feedback to give. I am sorry, because ... | https://www.lesswrong.com/posts/HPBdzWYp95zH3mQCa/i-applied-for-a-miri-job-in-2020-here-s-what-happened-next |
# Progress links and tweets, 2022-06-13
**Links**
---------
* [Davis Kedrosky reviews *How the World Became Rich,* by Koyama & Rubin](https://daviskedrosky.substack.com/p/book-review-how-the-world-became)
* [Study: keeping Diablo Canyon would avoid blackouts; reduce power costs, emissions and natural gas; and hel... | https://www.lesswrong.com/posts/Gmtsgp57BCXEKP9py/progress-links-and-tweets-2022-06-13 |
# Contra Hofstadter on GPT-3 Nonsense
[Douglas Hofstadter wrote in the Economist recently](https://www.economist.com/by-invitation/2022/06/09/artificial-neural-networks-today-are-not-conscious-according-to-douglas-hofstadter) on whether modern large language models are conscious. As part of this, he and his colleague ... | https://www.lesswrong.com/posts/ADwayvunaJqBLzawa/contra-hofstadter-on-gpt-3-nonsense |
# Lifeguards
**1**
=====
Many readers will be familiar with Peter Singer’s [Drowning Child experiment](https://www.thelifeyoucansave.org/child-in-the-pond/):
> On your way to work, you pass a small pond. On hot days, children sometimes play in the pond, which is only about knee-deep. The weather’s cool today, though... | https://www.lesswrong.com/posts/STPbaybu6KBbo4mZj/lifeguards |
# What if LaMDA is indeed sentient / self-aware / worth having rights?
Most pundits ridicule Blake Lemoine and his claims that LaMDA is sentient and deserves rights.
What if they're wrong?
The more thoughtful criticisms of his claims could be summarized as follows:
* The presented evidence (e.g. chatbot listing... | https://www.lesswrong.com/posts/9DteqyykFRe44po78/what-if-lamda-is-indeed-sentient-self-aware-worth-having |
# Ten experiments in modularity, which we'd like you to run!
This is the third post describing our team’s work on selection theorems for modularity, as part of a project mentored by John Wentworth ([see](https://www.lesswrong.com/posts/JzTfKrgC7Lfz3zcwM/theories-of-modularity-in-the-biological-literature) [here](https... | https://www.lesswrong.com/posts/99WtcMpsRqZcrocCd/ten-experiments-in-modularity-which-we-d-like-you-to-run |
# Against Active Shooter Drills
Epistemic Status: You really, _really_ should know this already. Reference post.
We have a $2.7 billion industry devoted to causing the traumatizing of children.
They refer to it as ‘active shooter drill training.’
I was initially motivated to write this when [I saw a claim that the ... | https://www.lesswrong.com/posts/hgpbwzvm4gYsioPsh/against-active-shooter-drills |
# Covid 6/16/22: Do Not Hand it to Them
You (finally!) no longer need to hand anyone a negative test in order to travel internationally.
You also most definitely do not have to hand it to the FDA for their work solving the bottlenecks they created.
Paxlovid remains something that you will probably not by default get... | https://www.lesswrong.com/posts/Y3RfXTYtyg9vKrniP/covid-6-16-22-do-not-hand-it-to-them |
# Don't Over-Optimize Things
or *Optimizing Optimization*
============================
The [definition](https://dictionary.cambridge.org/dictionary/english/optimize) of *optimize* is:
> to make something as good as possible
It's hard to argue with that. It's no coincidence that a lot of us have something of an opti... | https://www.lesswrong.com/posts/4zJoFQqf35SvFjrmZ/don-t-over-optimize-things |
# Perils of optimizing in social contexts
As a special case of [not over-optimizing things](https://www.lesswrong.com/posts/4zJoFQqf35SvFjrmZ/don-t-over-optimize-things), I want to point out that an optimizing approach can have a bunch of subtle negative effects in social contexts. The slogan is something like "**opti... | https://www.lesswrong.com/posts/zhBPhkca7oviZgnBq/perils-of-optimizing-in-social-contexts |
# Breaking Down Goal-Directed Behaviour
When we speak about entities 'wanting' things, or having '[goal-directed](https://www.alignmentforum.org/tag/goal-directedness) behaviour', what do we mean?
Because most of the actors that we (my human readers and I) attentively interact frequently with are (presumably) computa... | https://www.lesswrong.com/posts/WhETfFgkfNSTShc4y/breaking-down-goal-directed-behaviour |
# [Link] "The madness of reduced medical diagnostics" by Dynomight
Link:
- [The madness of reduced medical diagnostics](https://dynomight.net/diagnostics/)
This is (or seems to me now to be) an obvious-in-hindsight and I'm sad that I've never (or don't remember having) encountered it; at least not so succinctly.
I'... | https://www.lesswrong.com/posts/BYnyxpESz8Doxs7CE/link-the-madness-of-reduced-medical-diagnostics-by-dynomight |
# AI Risk, as Seen on Snapchat
The news tiles on Snapchat are utter drivel. Looking at my phone, today's headlines are: "Zendaya Pregnancy Rumors!", "Why Hollywood Won't Cast Lucas Cruikshank", and "The Most Dangerous Hood in Puerto Rico". Essentially, Snapchat news is the Gen-Z equivalent of the tabloid section at a ... | https://www.lesswrong.com/posts/8Ce9oBsiHTkzzk8tN/ai-risk-as-seen-on-snapchat |
# A possible AI-inoculation due to early "robot uprising"
I have talked about it before in the comments, but given the near-certain doomsday predictions by Eliezer, it might be worth posting it separately: there is a chance of a scenario that will sharply change the AI development trajectory before it is certain to wi... | https://www.lesswrong.com/posts/e4WHZcuEwFiAxrS2N/a-possible-ai-inoculation-due-to-early-robot-uprising |
# Towards Gears-Level Understanding of Agency
**Epistemic status:** Highly speculative, much of this is probably wrong.
**Summary:** [The "centerpiece"](https://www.lesswrong.com/posts/4pRPmFSfCLvKGEnFx/a-gears-level-model-of-agency#2__How_Is_Agency_Formed_) of this post is my attempt to describe how a full-fledged c... | https://www.lesswrong.com/posts/4pRPmFSfCLvKGEnFx/towards-gears-level-understanding-of-agency |
# Humans are very reliable agents
*This post has been recorded as part of the LessWrong Curated Podcast, and an be listened to on* [*Spotify*](https://open.spotify.com/episode/7qxVJWIR9YJuzFH9bLaTae?si=c29f23a272a74fd3)*,* [*Apple Podcasts*](https://podcasts.apple.com/us/podcast/humans-are-very-reliable-agents-by-alys... | https://www.lesswrong.com/posts/28zsuPaJpKAGSX4zq/humans-are-very-reliable-agents |
# BBC Future covers progress studies
BBC Future has [an article on progress studies](https://www.bbc.com/future/article/20220615-do-we-need-a-better-understanding-of-progress). I was interviewed, along with Tyler Cowen, Holden Karnofsky, and others. It’s well-researched and, although somewhat critical, it is pretty fa... | https://www.lesswrong.com/posts/uJWaydJZcaHmM2DGc/bbc-future-covers-progress-studies |
# A transparency and interpretability tech tree
_Thanks to Chris Olah, Neel Nanda, Kate Woolverton, Richard Ngo, Buck Shlegeris, Daniel Kokotajlo, Kyle McDonell, Laria Reynolds, Eliezer Yudkowksy, Mark Xu, and James Lucassen for useful comments, conversations, and feedback that informed this post._
The more I have th... | https://www.lesswrong.com/posts/nbq2bWLcYmSGup9aF/a-transparency-and-interpretability-tech-tree |
# A Litany Missing from the Canon
I wish to take whatever action will best bring about what I value.
I wish to become whatever person will most manifest what I value.
There is no virtue in clinging to yesterday’s expectations,
and no vice in hunting what matters.
I can make the best possible choice - and I can d... | https://www.lesswrong.com/posts/uaCxAM3pW3rJhuZo6/a-litany-missing-from-the-canon |
# wrapper-minds are the enemy
This post is a follow-up to "[why assume AGIs will optimize for fixed goals?](https://www.lesswrong.com/posts/dKTh9Td3KaJ8QW6gw/why-assume-agis-will-optimize-for-fixed-goals)". I'll assume you've read that one first.
I ended the earlier post by saying:
> \[A\]gents with the "wrapper st... | https://www.lesswrong.com/posts/Mrz2srZWc7EzbADSo/wrapper-minds-are-the-enemy |
# What's it like to have sex with Duncan?
Content warning: the title. While this essay is neither explicit nor vulgar, it talks frankly about the topic at hand in a PG-13 sort of way, and may be NSFW for many instances of W. I secured approval from LW leadership before posting.
, an EA-adjacent org created to promote impactful research, is organizing a bootcamp to develop machine learning skills to get started in AI Safety. The bootcamp will be held in France and is **intended for French ... | https://www.lesswrong.com/posts/kJDsi59FHs9jukSh9/apply-to-the-machine-learning-for-good-bootcamp-in-france |
# Announcing the Clearer Thinking Regrants program
Do you have an idea for a project, or run an existing project, startup, or organization that could one day have a big positive impact on the future of the world? [Apply now to our brand new Clearer Thinking Regrants program!](https://programs.clearerthinking.org/FTX_r... | https://www.lesswrong.com/posts/8895vQgRXjZFtHYkL/announcing-the-clearer-thinking-regrants-program |
# Value extrapolation vs Wireheading
Talk given by Rebecca Gorman and Stuart Armstrong at the CHAI 2022 Asilomar Conference. We present an example of AI wireheading (an AI taking over its own reward channel), and show how value extrapolation can be ... | https://www.lesswrong.com/posts/snmwFzoDMQMhyTirN/value-extrapolation-vs-wireheading |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.