text stringlengths 300 320k | source stringlengths 52 154 |
|---|---|
# Speculative Technologies launch and Ben Reinhardt AMA
Last week a new R&D organization launched: [Speculative Technologies](https://www.spec.tech/), a private DARPA-like approach to creating fundamental new technologies. I’ve been following the work of the founder, Ben Reinhardt, for a few years, and I’m very excite... | https://www.lesswrong.com/posts/Ems557utFmFvF4NGb/speculative-technologies-launch-and-ben-reinhardt-ama |
# Bankless Podcast: 159 - We’re All Gonna Die with Eliezer Yudkowsky
YouTube description:
> We wanted to do an episode on AI… and we went deep down the rabbit hole. As we went down, we discussed ChatGPT and the new generation of AI, digital superintelligence, the end of humanity, and if there’s anything we can do to ... | https://www.lesswrong.com/posts/FKNtgZrGYwgsz3nHT/bankless-podcast-159-we-re-all-gonna-die-with-eliezer |
# What AI companies can do today to help with the most important century
I’ve been writing about tangible things we can do today to help the [most important century](https://www.cold-takes.com/most-important-century/) go well. Previously, I wrote about [helpful messages to spread](https://www.cold-takes.com/spreading-... | https://www.lesswrong.com/posts/eidpSCMQhWmucijBi/what-ai-companies-can-do-today-to-help-with-the-most |
# EIS IX: Interpretability and Adversaries
Part 9 of 12 in the [Engineer’s Interpretability Sequence](https://www.alignmentforum.org/s/a6ne2ve5uturEEQK7).
Thanks to Nikolaos Tsilivis for helpful discussions.
The studies of interpretability and adversaries are inseparable.
===========================================... | https://www.lesswrong.com/posts/kYNMXjg8Tmcq3vjM6/eis-ix-interpretability-and-adversaries |
# A circuit for Python docstrings in a 4-layer attention-only transformer
*Produced as part of the* [*SERI ML Alignment Theory Scholars Program*](https://serimats.org/) *under the supervision of Neel Nanda - Winter 2022 Cohort.*
TL;DR: We found a circuit in a pre-trained 4-layer attention-only transformer language mo... | https://www.lesswrong.com/posts/u6KXXmKFbXfWzoAXn/a-circuit-for-python-docstrings-in-a-4-layer-attention-only |
# There are no coherence theorems
\[Written by EJT as part of the CAIS Philosophy Fellowship. Thanks to Dan for help posting to the Alignment Forum\]
**Introduction**
================
For about fifteen years, the AI safety community has been discussing [coherence arguments](https://www.lesswrong.com/tag/coherence-ar... | https://www.lesswrong.com/posts/yCuzmCsE86BTu9PfA/there-are-no-coherence-theorems |
# Validator models: A simple approach to detecting goodharting
*Crossposted from my personal* [*blog*](https://www.beren.io/2023-02-19-validator-models-simple-approach-detecting-goodhearting/)*.*
A naive approach to aligning an AGI, and an abstract version of what is currently used in SOTA approaches such as RLHF, is... | https://www.lesswrong.com/posts/r6f9DPBZYpWFw8Qrb/validator-models-a-simple-approach-to-detecting-goodharting |
# Behavioral and mechanistic definitions (often confuse AI alignment discussions)
**TL;DR: ***It’s important to distinguish between *behavioral *definitions – which categorize objects based on outside observable properties – and*mechanistic *definitions – which categorize objects based on their internal mechanisms.* *... | https://www.lesswrong.com/posts/Si52fuEGSJJTXW9zs/behavioral-and-mechanistic-definitions-often-confuse-ai |
# On second thought, prompt injections are probably examples of misalignment
Changed my mind.
All of my contentions about whether or not OpenAI actually cares about this problem seem valid to me. However, while prompt injections *are* exploits developed by humans to get ChatGPT to do something off-brand, they're pro... | https://www.lesswrong.com/posts/guGNszinGLfm58cuJ/on-second-thought-prompt-injections-are-probably-examples-of |
# AI alignment researchers don't (seem to) stack
_(Status: another point I find myself repeating frequently.)_
One of the reasons I suspect we need a lot of [serial time](https://www.lesswrong.com/s/v55BhXbpJuaExkpcD/p/vQNJrJqebXEWjJfnz) to solve the alignment problem is that alignment researchers don't seem to me t... | https://www.lesswrong.com/posts/4ujM6KBN4CyABCdJt/ai-alignment-researchers-don-t-seem-to-stack |
# Feelings are Good, Actually
*This is another casually written post in a* [*series*](https://www.lesswrong.com/s/mifkJ4GTdnpHspv7E) *of posts written in the spirit of advice to my past self.*
It's hard to remember exactly when I started to try to control my emotions. Probably I could not have been very old, only 4 o... | https://www.lesswrong.com/posts/CqgCtfjLHQSiudiLk/feelings-are-good-actually |
# Instrumentality makes agents agenty
You could describe the behavior of untuned GPT-like model[^xcl3sjm7lr] using a (peculiar) utility function. The fact that the loss function and training didn't explicitly involve a reward function doesn't mean a utility function can't represent what's learned, after all.
Coming f... | https://www.lesswrong.com/posts/EBKJq2gkhvdMg5nTQ/instrumentality-makes-agents-agenty |
# Medlife Crisis: "Why Do People Keep Falling For Things That Don't Work?"
In this video, Dr. Rohin Francis introduces the term "mechanistic bias" to describe the phenomenon where people tend to believe that a medical treatment would work if it has a plausible sounding mechanism of action, even when a randomized contr... | https://www.lesswrong.com/posts/z3zge49C6FGCC5eYR/medlife-crisis-why-do-people-keep-falling-for-things-that |
# Breaking the Optimizer’s Curse, and Consequences for Existential Risks and Value Learning
Multi-Armed Bandits Considered Harmful
--------------------------------------
People frequently analyze the process of artificial agents gathering knowledge in the framework of explore/exploit strategies for multi-armed bandit... | https://www.lesswrong.com/posts/ZqTQtEvBQhiGy6y7p/breaking-the-optimizer-s-curse-and-consequences-for-1 |
# Basic facts about language models during training
*We thank Eric Winsor, Lee Sharkey, Dan Braun, Carlos Ramon Guevara, and Misha Wagner for helpful suggestions and comments on this post.*
*This post builds upon *[*our last post*](https://www.lesswrong.com/posts/PDLfpRwSynu73mxGw/basic-facts-about-language-model-int... | https://www.lesswrong.com/posts/2JJtxitp6nqu6ffak/basic-facts-about-language-models-during-training-1 |
# You're not a simulation, 'cause you're hallucinating
I've found that the "[Simulators](https://www.lesswrong.com/posts/vJFdjigzmcXMhNTsx/simulators)" post is excellent for breaking prior assumptions about large language models - these algorithms are not agents, nor genies, nor Oracles. They are currently something v... | https://www.lesswrong.com/posts/dMBmZNwdjQ6yHvWZ5/you-re-not-a-simulation-cause-you-re-hallucinating |
# AI #1: Sydney and Bing
Previous AI-related recent posts: [Jailbreaking ChatGPT on Release Day](https://thezvi.substack.com/p/jailbreaking-the-chatgpt-on-release), [Next Level Seinfeld](https://thezvi.substack.com/p/next-level-seinfeld), [Escape Velocity From Bullshit Jobs](https://thezvi.substack.com/p/escape-veloci... | https://www.lesswrong.com/posts/WkchhorbLsSMbLacZ/ai-1-sydney-and-bing |
# Deceptive Alignment is <1% Likely by Default
*Thanks to Wil Perkins, Grant Fleming, Thomas Larsen, Declan Nishiyama, and Frank McBride for feedback on this post. Thanks also to Paul Christiano, Daniel Kokotajlo, and Aaron Scher for comments on the original post that helped clarify the argument. Any mistakes are my o... | https://www.lesswrong.com/posts/RTkatYxJWvXR4Qbyd/deceptive-alignment-is-less-than-1-likely-by-default |
# EIS X: Continual Learning, Modularity, Compression, and Biological Brains
Part 10 of 12 in the [Engineer’s Interpretability Sequence](https://www.alignmentforum.org/s/a6ne2ve5uturEEQK7).
The science of interpretability is part of a larger picture.
=============================================================
The ... | https://www.lesswrong.com/posts/fHnwCDDbDHWqbJ8Nd/eis-x-continual-learning-modularity-compression-and |
# A Stranger Priority? Topics at the Outer Reaches of Effective Altruism (my dissertation)
(Cross-posted from [my website](https://joecarlsmith.com/2023/02/21/a-stranger-priority-topics-at-the-outer-reaches-of-effective-altruism).)
After many years of focusing on other stuff, I recently completed my doctorate in phil... | https://www.lesswrong.com/posts/pHx8nu4rfecDdo6vM/a-stranger-priority-topics-at-the-outer-reaches-of-effective |
# Pretraining Language Models with Human Preferences
*This post summarizes the main results from our recently released paper *[*Pretraining Language Models with Human Preferences*](https://arxiv.org/abs/2302.08582)*, and puts them in the broader context of AI safety. For a quick summary of the paper, take a look at *[... | https://www.lesswrong.com/posts/8F4dXYriqbsom46x5/pretraining-language-models-with-human-preferences |
# What is it like doing AI safety work?
How do you know if you’ll like AI safety work? What’s the day-to-day work like? What are the best parts of the job? What are the worst?
To better answer these questions, we talked to ten AI safety researchers in a variety of organizations, roles, and subfields. If you’re intere... | https://www.lesswrong.com/posts/TrfCQiutKZz5CJfeG/what-is-it-like-doing-ai-safety-work |
# The Emotional Type of a Decision
Our minds have built-in tools that we use intuitively to evaluate decisions as we go about our day-to-day lives. Sometimes, making a decision can feel particularly challenging on an emotional level. This can lead to procrastination in making the decision or just going with the defaul... | https://www.lesswrong.com/posts/ZwjQRZtsPR9favCvc/the-emotional-type-of-a-decision |
# Fighting For Our Lives - What Ordinary People Can Do
tl;dr: Get organized and start brainstorming
\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_\_
*If you're reading this, you are the resistance.*
I've always wanted to say that. But seriously,... | https://www.lesswrong.com/posts/CqmDWHLMwybSDTNFe/fighting-for-our-lives-what-ordinary-people-can-do |
# A proof of inner Löb's theorem
This is a short post that offers a slightly different take on the standard proof of Löb's theorem. It offers nothing else of any value :)
We seek to prove the "inner" version, which we write as:
$$\Box P \leftrightarrow \Box (\Box P \rightarrow P)$$
The proof uses quining to build a ... | https://www.lesswrong.com/posts/QgZAbFHtgSGjx4aTS/a-proof-of-inner-loeb-s-theorem |
# Candyland is Terrible
Candyland seriously suffers from not being aimed at a coherent age: either you're old enough that you could instead play a game where you make choices, or you're young enough that the disappointment of getting a card that—out of nowhere—sends you nearly back to the beginning is crushing. Growin... | https://www.lesswrong.com/posts/PF2KwbiGS6679tHKg/candyland-is-terrible |
# The shallow reality of 'deep learning theory'
> *Produced under the mentorship of Evan Hubinger as part of the* [*SERI ML Alignment Theory Scholars Program*](https://serimats.org/) *\- Winter 2022 Cohort*
Most results under the umbrella of "deep learning theory" are not actually deep, about learning, or even theori... | https://www.lesswrong.com/posts/BDTfddkttFXHqGnEi/the-shallow-reality-of-deep-learning-theory |
# Power-Seeking = Minimising free energy
Epistemic Status: *Pretty certain there are better ways to describe this but I believe that the underlying intuition holds and that it might be an exciting formalisation of power. *
*Thank you to Viktor Rehnberg and Roman Levantov for some great discussions leading up to... | https://www.lesswrong.com/posts/KYxpkoh8ppnPfmuF3/power-seeking-minimising-free-energy |
# Intervening in the Residual Stream
*This post describes work done at SERI-MATS with mentorship from Neel Nanda. This work has benefitted from discussions with Neel Nanda, Stefan Heimersheim, Kajetan Janiak, Wes Gurnee, Marius Hobbhahn, Aryan Bhatt, Mykyta Baliesnyi, and Laura Vaughn.*
, a private DARPA-like research organization. See also coverage in [Forbes](https://www.forbes.com/sites/johncumbers/2023/02/15/ben-... | https://www.lesswrong.com/posts/wTHmxxNNNE9trz9Jv/progress-links-and-tweets-2023-02-22 |
# Can submarines swim?
*\[Note: This was written for a general audience; most of it is probably too basic for LessWrong. Thanks to the commenters who critiqued the original draft, this is the revised and published version.\]*
Did any science fiction predict that when AI arrived, it would be unreliable, often illogica... | https://www.lesswrong.com/posts/pdfKJGyhfAxag2Kes/can-submarines-swim-1 |
# Building and Entertaining Couples
Cross-posted, as always, [from Putanumonit](https://putanumonit.com/2023/02/11/building-and-entertaining-couples/).
* * *
A friend told me recently that she’s looking for a husband to settle have kids, but she’s facing a problem: most of the men she goes on dates with fail to exci... | https://www.lesswrong.com/posts/gcse5Cqr3rgHtLAiY/building-and-entertaining-couples |
# EIS XI: Moving Forward
Part 11 of 12 in the [Engineer’s Interpretability Sequence](https://www.alignmentforum.org/s/a6ne2ve5uturEEQK7).
So far, this sequence has discussed a number of topics in interpretability research, all building toward this post. Its goal is to explain some approaches that may be valuable movi... | https://www.lesswrong.com/posts/L5Rua9aTndviy8dvc/eis-xi-moving-forward |
# Video/animation: Neel Nanda explains what mechanistic interpretability is
Nice little video - audio is Neel Nanda explaining what mechanistic interpretability is and why he does it, and it's illustrated by the illustrious [Hamish Doodles](https://hamishdoodles.com/). Excerpted from the [AXRP episode](https://www.les... | https://www.lesswrong.com/posts/JtrqxAae5JAFthkkb/video-animation-neel-nanda-explains-what-mechanistic |
# Short Notes on Research Process
I've transitioned from [reading about alignment and theorizing](https://www.lesswrong.com/s/aK4hyeJyM2sYNS4Yc) about the problem landscape, to doing things. And by things I mean MVPs of bigger ideas, where each bigger idea roughly falls under the umbrella of *Collective Human Intellig... | https://www.lesswrong.com/posts/oFHo6Ei4kcom5ug65/short-notes-on-research-process |
# Big Mac Subsidy?
I was reading a [Faunalytics report](https://faunalytics.org/veg-obstacle-analysis/) on people who stopped being vegan or vegetarian, [via the EA Forum](https://forum.effectivealtruism.org/posts/otfCm4TMFGrs8vLng/faunalytics-analysis-on-reasons-for-abandoning-veg-n-diets), and was surprised to see:
... | https://www.lesswrong.com/posts/PuCqfK9DcaZDHZbAx/big-mac-subsidy |
# Interpersonal alignment intuitions
Let's try this again...
The problem of aligning superhuman AGI is very difficult. We don't have access to superhuman general intelligences. We have access to superhuman narrow intelligences, and human-level general intelligences.
There's an idea described [here](https://squirr... | https://www.lesswrong.com/posts/ypvs4asdFq7riDWmd/interpersonal-alignment-intuitions |
# Full Transcript: Eliezer Yudkowsky on the Bankless podcast
[*This podcast*](https://www.youtube.com/watch?v=gA1sNLL6yg4) *has gotten a lot of traction, so we're posting a full transcript of it, lightly edited with ads removed, for those who prefer reading over audio. *
[Intro](https://www.youtube.com/watch?v=gA... | https://www.lesswrong.com/posts/Aq82XqYhgqdPdPrBA/full-transcript-eliezer-yudkowsky-on-the-bankless-podcast |
# Covid 2/23/23: Your Best Possible Situation
No news is (often) good news.
At old Magic: The Gathering tournaments, judge Tony Parodi would often tell us, ‘if your opponent did not show up, that is your best possible situation.’
Every week, when I set out to write the Covid update, [I held out the hope that at some... | https://www.lesswrong.com/posts/mehFn6sWqZhndeZP9/covid-2-23-23-your-best-possible-situation |
# EIS XII: Summary
Part 12 of 12 in the [Engineer’s Interpretability Sequence](https://www.alignmentforum.org/s/a6ne2ve5uturEEQK7).
**TAISIC = “the AI safety interpretability community”**
**MI = “mechanistic interpretability” **
There might be some addenda later, but for now, this is the final post in The Engineer... | https://www.lesswrong.com/posts/aymbce8ge9ve2C4Po/eis-xii-summary |
# Somerville Roads Getting More Dangerous?
In a local discussion of whether our attempts to reduce traffic fatalities were working someone posted:
> I've lived in Somerville for over 40 years and have visited even longer. The roads are more dangerous than ever.
With all the [Vision Zero](https://www.somervillema.gov... | https://www.lesswrong.com/posts/tkQczpaD3xidWXziS/somerville-roads-getting-more-dangerous |
# Why I'm Skeptical of De-Extinction
*This piece was co-authored with* [*@Metacelsus*](https://www.lesswrong.com/users/metacelsus?mention=user)
*“For a successful technology, reality must take precedence over public relations, for nature cannot be fooled.”*
— [Richard Feynman](https://www.refsmmat.com/files/reflect... | https://www.lesswrong.com/posts/ihq24ri5g5svwwmjx/why-i-m-skeptical-of-de-extinction |
# Searching for a model's concepts by their shape – a theoretical framework
**Produced as part of the** [**SERI ML Alignment Theory Scholars Program**](https://serimats.org/) **\- Winter 2022 Cohort**
Introduction
============
I think that [Discovering Latent Knowledge in Language Models Without Supervision](https:/... | https://www.lesswrong.com/posts/Go5ELsHAyw7QrArQ6/searching-for-a-model-s-concepts-by-their-shape-a |
# AI that shouldn't work, yet kind of does
There are some things that work surprisingly well in AI. For example, AI that transfers [the style of one image to the content of another.](https://www.tensorflow.org/tutorials/generative/style_transfer) Why is the approach described here a hack. It starts with a neural net t... | https://www.lesswrong.com/posts/NK2CeDNKMEY9gRZp2/ai-that-shouldn-t-work-yet-kind-of-does |
# Teleosemantics!
I wanted to write a long, detailed, analytic post about this, somewhat like my [Radical Probabilism](https://www.lesswrong.com/posts/xJyY5QkQvNJpZLJRo/radical-probabilism-1) post (to me, this is a similarly large update). However, I haven't gotten around to it for a long while. And perhaps it is bett... | https://www.lesswrong.com/posts/WGEPBmErv8ufrq8Fc/teleosemantics |
# [Link] A community alert about Ziz
The start of the post is copy-pasted below. **Note that the post is anonymous, and I am not claiming to have written it.**
* * *
Some people in the rationalist community are concerned about risks of physical violence from Ziz and some of her associates. Following discussions with... | https://www.lesswrong.com/posts/T5RzkFcNpRdckGauu/link-a-community-alert-about-ziz |
# AGI systems & humans will both need to solve the alignment problem
*Epistemic status: brainstorm-y musings about goal preservation under self-improvement and a really really bad plan for trading with human-level AGI systems to solve alignment.*
### When will AGI systems want to solve the alignment problem?
At some... | https://www.lesswrong.com/posts/wZAa9fHZfR6zxtdNx/agi-systems-and-humans-will-both-need-to-solve-the-alignment |
# Are there rationality techniques similar to staring at the wall for 4 hours?
I'm wondering if one exists, and what is the name of that technique or that family of techniques. It seems like something that CFAR has researched extensively and layered a lot of things on top of.
4 hours is necessary for anything substa... | https://www.lesswrong.com/posts/hribPfHkMsapg3oYM/are-there-rationality-techniques-similar-to-staring-at-the |
# Consent Isn't Always Enough
After the [Time article](https://time.com/6252617/effective-altruism-sexual-harassment/) on sexual harassment and abuse within Effective Altruism and [Owen Cotton-Barratt's resignation](https://forum.effectivealtruism.org/posts/9JCkkjKMNL4Hmg4qP/ev-uk-board-statement-on-owen-s-resignation... | https://www.lesswrong.com/posts/7DBzgWJHFWkFv27hd/consent-isn-t-always-enough |
# How major governments can help with the most important century
I’ve been writing about tangible things we can do today to help the [most important century](https://www.cold-takes.com/most-important-century/) go well. Previously, I wrote about [helpful messages to spread](https://www.cold-takes.com/spreading-messages... | https://www.lesswrong.com/posts/xtJQcmv7THCE8Topa/how-major-governments-can-help-with-the-most-important |
# Relationship Orientations
Jacob Falkovich has written a good post on [Entertaining vs Building](https://www.lesswrong.com/posts/gcse5Cqr3rgHtLAiY/building-and-entertaining-couples) orientations, which made me realize I should probably cross-post my own articles on these sorts of things.
*(Note: these articles r... | https://www.lesswrong.com/posts/4dN4ySTzXdNd2q9Dq/relationship-orientations |
# Meta "open sources" LMs competitive with Chinchilla, PaLM, and code-davinci-002 (Paper)
As the title says, Meta trained 4 foundational models with 7B, 13B, 33B, and 65B parameters respectively, and is open sourcing them for research.[^ubd95r91ii] You can get their code on their Github repo: [https://github.com/faceb... | https://www.lesswrong.com/posts/ChbRgvuGaG2dAtr6i/meta-open-sources-lms-competitive-with-chinchilla-palm-and |
# Sam Altman: "Planning for AGI and beyond"
*(OpenAI releases a blog post detailing their AGI roadmap. I'm copying the text below, though see the linked blog post for better formatted version)*
* * *
Our mission is to ensure that artificial general intelligence—AI systems that are generally smarter than humans—[bene... | https://www.lesswrong.com/posts/zRn6aQyD8uhAN7qCc/sam-altman-planning-for-agi-and-beyond |
# Are you stably aligned?
Epistimic status: posing questions I think are probably useful, without a strong guess at the answers.
If we made an AGI just like you, would it be aligned?
Would it stay aligned over a long time span?
The purpose of the first question is to explore the possibility raised by [shard theor... | https://www.lesswrong.com/posts/Sf99QEqGD76Z7NBiq/are-you-stably-aligned |
# How popular is ChatGPT? Part 1: more popular than Taylor Swift
_Harlan Stewart, 23 February 2023_
Introduction
------------
Public attention toward AI seems much higher after the release of ChatGPT at the end of November. But how much higher is it? To better understand this, I looked at search data from Google Tre... | https://www.lesswrong.com/posts/uqnimhFSjXyWbpTwd/how-popular-is-chatgpt-part-1-more-popular-than-taylor-swift |
# Retrospective on the 2022 Conjecture AI Discussions
At the end of 2022, following the success of the [2021 MIRI Conversations](https://www.lesswrong.com/s/n945eovrA3oDueqtq), [Conjecture](https://conjecture.dev/) started a project to host discussions about AGI and alignment with key people in the field. The goal was... | https://www.lesswrong.com/posts/BEyAWbCdtWpSGxmun/retrospective-on-the-2022-conjecture-ai-discussions |
# Christiano (ARC) and GA (Conjecture) Discuss Alignment Cruxes
The following are the summary and transcript of a discussion between Paul Christiano (ARC) and Gabriel Alfour, hereafter GA (Conjecture), which took place on December 11, 2022 on Slack. It was held as part of a series of discussions between [Conjecture](h... | https://www.lesswrong.com/posts/pgpFHLJnv7AdSi3qS/christiano-arc-and-ga-conjecture-discuss-alignment-cruxes |
# Agents vs. Predictors: Concrete differentiating factors
_Thanks to Paul Christiano and Kate Woolverton for useful conversations and feedback._
In "[Conditioning Predictive Models](https://www.alignmentforum.org/s/n3utvGrgC2SGi9xQX)," we devote a lot of effort into trying to understand how likely predictive models a... | https://www.lesswrong.com/posts/eQ4eLQAmPvp9anJcB/agents-vs-predictors-concrete-differentiating-factors |
# What kind of place is this?
I started getting LessWrong posts in my email about a year ago—I don't remember signing up, but I must have done it intentionally. I like most of what I've been reading so far: it's a civil forum in which people think about the process of thought and its application, though some of the sp... | https://www.lesswrong.com/posts/GjuiqXAx9r4w6Fmxv/what-kind-of-place-is-this |
# Buddhist Psychotechnology for Withstanding Apocalypse Stress
I could try writing the well cited and argued version of this post, but that would result in procrastination and get stuck in drafts, so you get the stream of consciousness version instead. Happy to answer clarifying questions in the comments. This post is... | https://www.lesswrong.com/posts/nnCjjdXEXdRhzTAPE/buddhist-psychotechnology-for-withstanding-apocalypse-stress |
# Pink Shoggoths: What does alignment look like in practice?
I believe it is commonly accepted around these parts that we are doomed, due to our utter lack of attempts to resolve the Control Problem. Perhaps efforts will be made that are more substantial, but as for right now, chances for doom are high, if not 100%, a... | https://www.lesswrong.com/posts/9y5RpyyFJX4GaqPLC/pink-shoggoths-what-does-alignment-look-like-in-practice |
# The Practitioner's Path 2.0: the Empiricist Archetype
Earlier this month, we [introduced](https://guildoftherose.org/articles/the-future-of-structured-self-improvement) the Practitioner's Path 2.0, our new framework for structured self improvement. The heart of the Path is the Skilltree, which is divided into three ... | https://www.lesswrong.com/posts/HsCtB9G4HXNysyFid/the-practitioner-s-path-2-0-the-empiricist-archetype |
# [Prediction] Humanity will survive the next hundred years
Definition: Living *Homo sapiens* will live on the surface of the planet Earth exactly 100 years from the publication of this post. This is similar to the Caplan-Yudkowsky bet except I extend the duration from 13 to 100 years and remove the reference to the c... | https://www.lesswrong.com/posts/nP8DdBbxju93t76Sh/prediction-humanity-will-survive-the-next-hundred-years |
# Cognitive Emulation: A Naive AI Safety Proposal
*This is part of the work done at *[*Conjecture*](https://www.conjecture.dev)*.*
*This post has been reviewed before publication as per our *[*infohazard policy*](https://www.lesswrong.com/posts/Gs29k3beHiqWFZqnn/conjecture-internal-infohazard-policy)*. We thank our e... | https://www.lesswrong.com/posts/ngEvKav9w57XrGQnb/cognitive-emulation-a-naive-ai-safety-proposal |
# Making Implied Standards Explicit
Standards are everywhere:
"That movie was kind of bad"
"You're so pretty!"
"I'm so disappointed in you"
By exploring standards, I've been able to notice them in my life. By noticing them, I've less of a need to "measure up".
By investigating the many areas standards appear, I h... | https://www.lesswrong.com/posts/nNdoP5qnASFT9FYx6/making-implied-standards-explicit |
# Nod posts
Earlier today I was reading [How to Read Papers Efficiently: Fast-then-Slow Three pass method](https://www.lesswrong.com/posts/sAyJsvkWxFTkovqZF/how-to-read-papers-efficiently-fast-then-slow-three-pass). I liked it and upvoted it.
I really like what it's saying. If you sit down and try to read a paper fro... | https://www.lesswrong.com/posts/KFM8diDgprhCWtEgz/nod-posts |
# Would more model evals teams be good?
If multiple model evaluations organizations similar to the ARC evals team emerge and compete for big tech clients, does this drive a "race to the bottom," where more lenient evals teams get larger market share? I'm imaging a scenario in which there are government restrictions on... | https://www.lesswrong.com/posts/fiqDjKCGzaopWoz7s/would-more-model-evals-teams-be-good |
# Politics is the Fun-Killer
In 2007, Eliezer wrote [Politics is the Mind-Killer](https://www.lesswrong.com/posts/9weLK2AJ9JEt2Tt8f/politics-is-the-mind-killer). I agree that it is a mind-killer, but even moreso than being a mind-killer, I think that it is a fun-killer. Let me explain.
I'll start by summarizing the i... | https://www.lesswrong.com/posts/vzkGa2CcyGBdB7pkz/politics-is-the-fun-killer |
# A mechanistic explanation for SolidGoldMagikarp-like tokens in GPT2
This post seems to have an important bug in its code that changes its conclusions, as pointed out to me by user [@ckkissane](https://www.lesswrong.com/users/ckkissane?mention=user) . At a high level, there are probably more hard to speak tokens and ... | https://www.lesswrong.com/posts/dFbfCLZA4pejckeKc/a-mechanistic-explanation-for-solidgoldmagikarp-like-tokens |
# Incentives and Selection: A Missing Frame From AI Threat Discussions?
## Epistemic Status
Written quickly, originally as a [Twitter thread](https://twitter.com/CineraVerinia/status/1629639514229489671).
___
# Thesis
I think a missing frame from AI threat discussion is incentives (especially economic) and sel... | https://www.lesswrong.com/posts/5HvCSt3vDSDuuXkfz/incentives-and-selection-a-missing-frame-from-ai-threat |
# "Rationalist Discourse" Is Like "Physicist Motors"
Imagine being a student of physics, and coming across a blog post proposing a list of guidelines for "physicist motors"—motor designs informed by the knowledge of physicists, unlike ordinary motors.
Even if most of the things on the list seemed like sensible advice... | https://www.lesswrong.com/posts/SX6wQEdGfzz7GKYvp/rationalist-discourse-is-like-physicist-motors |
# The Preference Fulfillment Hypothesis
Short version
=============
Humans have an innate motivation ("preference fulfillment", PF) to fulfill the preferences of those they care about. It corresponds to at least some of the senses of the word "love", as well as related words such as "kindness" and "compassion".
I hy... | https://www.lesswrong.com/posts/Kf6sKZudduhJmykTg/the-preference-fulfillment-hypothesis |
# A library for safety research in conditioning on RLHF tasks
There is a couple of discussions regarding conditioning / decision transformer training. "Conditioning" as in, placing your reward model's reward in the prompt. We then train our language model to create a completion that follows the specified reward.
See ... | https://www.lesswrong.com/posts/aHPmGPWtmK259J8ou/a-library-for-safety-research-in-conditioning-on-rlhf-tasks |
# [Link] Petition on brain preservation: Allow global access to high-quality brain preservation as an option rapidly after death
I've put a few extracts I found notable below, but check out the URL for the full text and to sign the petition. Note that signing the petition doesn't mean you necessarily want to preserve ... | https://www.lesswrong.com/posts/sjYsihT7spr4FMMsg/link-petition-on-brain-preservation-allow-global-access-to |
# Taboo "human-level intelligence"
Maybe calls for tabooing terms can become a bit inflationary. However, talking about if and when AI will reach or surpass “human-level intelligence” can misleadingly frame the discussion, so I recommend avoiding that term.
What is "human-level intelligence" and when will AI sur... | https://www.lesswrong.com/posts/izsSEG9MRQjtgv5hh/taboo-human-level-intelligence |
# Reflections on Zen and the Art of Motorcycle Maintenance
I read *Zen and the Art of Motorcycle Maintenance* by Robert Pirsig twice in 2019. I wrote reflections on many of the chapters as I went through the second time. I posted them to a Goodreads book group at the time, but I've decided to collect them here as well... | https://www.lesswrong.com/posts/BapmiGY8axSFmb3aH/reflections-on-zen-and-the-art-of-motorcycle-maintenance |
# [Link Post] Cyber Digital Authoritarianism (National Intelligence Council Report)
**Disclaimer**: The U.S. government selectively declassified this report for unknown reasons, and anti-Chinese-Communist-Party sentiment is high in Washington D.C. While I doubt this means the contents are factually inaccurate, it is l... | https://www.lesswrong.com/posts/DfcGHqgxAWAL6BCst/link-post-cyber-digital-authoritarianism-national |
# Bayesian Scenario: Snipers & Soldiers
*You are a soldier, planning to attack an enemy position. You have a foolproof strategy for taking positions held by ordinary enemies, which will get you killed if you use it on positions held by snipers. You can call in an airstrike that will reliably devastate the enemy positi... | https://www.lesswrong.com/posts/prQjoBnTY2LEPHZgG/bayesian-scenario-snipers-and-soldiers |
# Learning How to Learn (And 20+ Studies)
I have been interested in cognitive science and effective learning methods for years. I've read multiple books and articles and put many ideas to test. In this post, I aim to synthesize my notes and provide references to scientific studies.
Summary (TL;DR)
---------------
##... | https://www.lesswrong.com/posts/BzxbZqWoP9DRrznFi/learning-how-to-learn-and-20-studies |
# Respect Chesterton-Schelling Fences
*This post is inspired by the recent Ziz-revelations posted here and elsewhere.*
[Chesterton Fence](https://wiki.lesswrong.com/wiki/Chesterton%27s_Fence) injunction: Do not remove a protective barrier unless you know why it was put there in the first place.
[Schelling fence](htt... | https://www.lesswrong.com/posts/p2Qq4WWQnEokgjimy/respect-chesterton-schelling-fences |
# [Simulators seminar sequence] #2 Semiotic physics - revamped
***Update February 21st**:* *After the initial publication of this article (January 3rd) we received a lot of feedback and several people pointed out that propositions 1 and 2 were incorrect as stated. That was unfortunate as it distracted from the broader... | https://www.lesswrong.com/posts/TTn6vTcZ3szBctvgb/simulators-seminar-sequence-2-semiotic-physics-revamped |
# Something Unfathomable: Unaligned Humanity and how we're racing against death with death
I fear I may be becoming a mini-Yudkowsky.
I write this in response to multiple exclamatory remarks I've seen in recent weeks, excited over the prospect of all jobs being automated, of ultra-high unemployment, basic income, an... | https://www.lesswrong.com/posts/aEhjHksoKZpvLt6ND/something-unfathomable-unaligned-humanity-and-how-we-re |
# Fertility Rate Roundup #1
Previously: [On Car Seats as Contraception](https://thezvi.substack.com/p/on-car-seats-as-contraception)
\[Editor’s Note: This post assumes the perspective that more people having more children is good, actually. I will not be engaging with any of the arguments against this, of any quality... | https://www.lesswrong.com/posts/qXyc6vA6hWcpuHscr/fertility-rate-roundup-1 |
# Normie response to Normie AI Safety Skepticism
The following post is an adaptation of a
[comment I posted](https://www.reddit.com/r/MachineLearning/comments/11ada91/d_to_the_ml_researchers_and_practitioners_here_do/j9szujb/)
on the [Machine Learning subreddit](https://www.reddit.com/r/MachineLearning/)
in response t... | https://www.lesswrong.com/posts/XzgmJoB8CFo5N3z9n/normie-response-to-normie-ai-safety-skepticism-1 |
# Milk EA, Casu Marzu EA
Pretty much everyone starts off drinking milk, and while adult consumption varies culturally, genetically, and ethically, if I put milk on my morning bran flakes that's a neutral choice around here. If my breakfast came up in talking with a friend they might think it was dull, but they wouldn'... | https://www.lesswrong.com/posts/29xpPQ6SnB945dtuH/milk-ea-casu-marzu-ea |
# Counting-down vs. counting-up coherence
*[Metadata: crossposted from [https://tsvibt.blogspot.com/2022/10/counting-down-vs-counting-up-coherence.html](https://tsvibt.blogspot.com/2022/10/counting-down-vs-counting-up-coherence.html). First completed 25 October 2022.]*
Counting-down coherence is the coherence of a mi... | https://www.lesswrong.com/posts/svpnmmeJresYs23rY/counting-down-vs-counting-up-coherence |
# Prediction Thread: Make Predictions About How Different Factors Affect AGI X-Risk.
In this post you can make several predictions for how different factors affect the probability that the creation of AGI leads to an extinction level catastrophe. This might be useful for planning.
Please let me know if you have othe... | https://www.lesswrong.com/posts/xFoJNijbTFFnxBhmD/prediction-thread-make-predictions-about-how-different |
# Beginning to feel like a conspiracy theorist
I see myself, and is often described by others, as a logical person and high in the openness trait. I am in university and have the last 2 years done some various amateur litterature review reasearch on different topics, especially on health and nutrition.
On many topic... | https://www.lesswrong.com/posts/GBZAYhLTjwsnvAyZf/beginning-to-feel-like-a-conspiracy-theorist |
# A case for capabilities work on AI as net positive
Edit: Changed the title.
Or, why I no longer agree with the standard LW position on AI anymore.
In a sense, this is sort of a weird post compared to what LW usually posts on AI.
A lot of this is going to depend on some posts that changed my worldview on AI risk,... | https://www.lesswrong.com/posts/ix4Tx3BMR4EwHdnS3/a-case-for-capabilities-work-on-ai-as-net-positive |
# The Birth and Death of Sydney — The Bayesian Conspiracy Podcast
We discuss Microsoft’s new Bing search chatbot – Sydney. It received a wave of media coverage due to erratic behavior, before being neutered by Microsoft. A dive into its behavior, human responses, and implications.
**Links**
[The Birth and Death of S... | https://www.lesswrong.com/posts/wj2WrFHqxYTow5hXt/the-birth-and-death-of-sydney-the-bayesian-conspiracy |
# [S] D&D.Sci: All the D8a. Allllllll of it. Evaluation and Ruleset
This is a follow-up to [](https://www.lesswrong.com/posts/cesr2yGoZ2y327qKQ/d-and-d-sci-divination-nine-black-doves) [last week's D&D.Sci scenario](https://www.lesswrong.com/posts/SjsuGzrf7YZRYEdcP/s-d-and-d-sci-all-the-d8a-allllllll-of-it): if you i... | https://www.lesswrong.com/posts/KsbmHM5cYoJoN7TBp/s-d-and-d-sci-all-the-d8a-allllllll-of-it-evaluation-and |
# Conversationism
conversationism, the mathematics behind communication and understanding, used to ra... | https://www.lesswrong.com/posts/HpHyERTmsmhDiRHtY/conversationism |
# Transcript: Testing ChatGPT's Performance in Engineering
*This is a supplement to* [*an article I wrote elsewhere*](https://ez.analog.com/ez-blogs/b/engineering-mind/posts/the-limits-of-ai-testing-chatgpt-s-performance-in-engineering)*, to provide the complete transcript of an interaction with ChatGPT. *
I evaluate... | https://www.lesswrong.com/posts/RsvKae3KSXEaivpgJ/transcript-testing-chatgpt-s-performance-in-engineering |
# Transcript: Yudkowsky on Bankless follow-up Q&A
**Head over to Rob Bensinger's** [**improved**](https://www.lesswrong.com/posts/e4pYaNt89mottpkWZ/yudkowsky-on-agi-risk-on-the-bankless-podcast) **transcript, combined with the original podcast.** ***(This one has been updated correspondingly)*****.**
**\-\-\-**
*Thi... | https://www.lesswrong.com/posts/jfYnq8pKLpKLwaRGN/transcript-yudkowsky-on-bankless-follow-up-q-and-a |
# $20 Million in NSF Grants for Safety Research
After a year of negotiation, the NSF has [announced](https://beta.nsf.gov/funding/opportunities/safe-learning-enabled-systems) a **$20 million** request for proposals for empirical AI safety research.
[Here is the detailed program description](https://www.nsf.gov/pubs/2... | https://www.lesswrong.com/posts/jwe6jpubuMiuSRqff/usd20-million-in-nsf-grants-for-safety-research |
# What is the fertility rate?
I recently read [Zvi's post on fertility rate](https://www.lesswrong.com/posts/qXyc6vA6hWcpuHscr/fertility-rate-roundup-1). In it, they discuss the falling fertility rate around the world, and how to increase it. Before discussing whether or not we should increase the fertility rate (whic... | https://www.lesswrong.com/posts/9wJk9uuXF2MsuxHLK/what-is-the-fertility-rate |
# Digital Molecular Assemblers: What synthetic media/generative AI actually represents, and where I think it's going
[I had an epiphany about synthetic media back in December 2017](https://www.reddit.com/r/artificial/comments/7lwrep/media_synthesis_and_personalized_content_my/). In that epiphany, I came to the conclus... | https://www.lesswrong.com/posts/C5mmznTc8G3B9RCtY/digital-molecular-assemblers-what-synthetic-media-generative |
# What does Bing Chat tell us about AI risk?
_Image from [here](https://astralcodexten.substack.com/p/janus-simulators?ref=cold-takes) via [this tweet](https://twitter.com/repligate/status/16144161900... | https://www.lesswrong.com/posts/QHnW6AwbrLwbYs9yo/what-does-bing-chat-tell-us-about-ai-risk |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.