text stringlengths 300 320k | source stringlengths 52 154 |
|---|---|
# The Dumbest Possible Gets There First
*This was written as part of the first *[*Refine*](https://www.lesswrong.com/posts/5uiQkyKdejX3aEHLM/how-to-diversify-conceptual-alignment-the-model-behind) *blog post day.*
A sneaking suspicion that I’ve found difficult to shake off while following AI risk discussion is that c... | https://www.lesswrong.com/posts/BPJLzkEpx8Btz9ywq/the-dumbest-possible-gets-there-first |
# Refine's First Blog Post Day
Yesterday was the first blog post day at Refine. It came after a week of readings, discussions and exercises about epistemology (more concretely [shapes of mind](https://www.alignmentforum.org/posts/RrirwtP7cNmHtJRxE/shapes-of-mind-and-pluralism-in-alignment), [productive mistakes](https... | https://www.lesswrong.com/posts/8jewbj9am99H2oYio/refine-s-first-blog-post-day |
# An extended rocket alignment analogy
*This work was done while at* [*Conjecture*](https://conjecture.dev)*.*
*This post has been written for the first *[*Refine *](https://www.alignmentforum.org/posts/D7epkkJb3CqDTYgX9/refine-an-incubator-for-conceptual-alignment-research-bets)*blog post day, at the end of the week... | https://www.lesswrong.com/posts/xuYdCDgoBno5haJB6/an-extended-rocket-alignment-analogy |
# Cultivating Valiance
When I discovered existential risk from AGI I was surprised to learn most people find the topic distressing or depressing. I realized everyone dying with a high probability *should* feel bad ... but it just didn't. Instead I felt the desire to punch the problem in the face (don't ask me how that... | https://www.lesswrong.com/posts/qGtmNeMqAw2oQJQcs/cultivating-valiance |
# Evolution is a bad analogy for AGI: inner alignment
**TL;DR**: The dynamics of human learning processes and reward circuitry are more relevant than evolution for understanding how inner values arise from outer optimization criteria.
This post is related to Steve Byrnes’ [Against evolution as an analogy for how hum... | https://www.lesswrong.com/posts/FyChg3kYG54tEN3u6/evolution-is-a-bad-analogy-for-agi-inner-alignment |
# A brief note on Simplicity Bias
The idea of 'simplicity bias' is a popular one in informal discussions about the functions that neural networks implement. I recently tried to think a little bit more about what this meant. This brief note was written mostly for my own benefit, but could be of benefit to others. Havin... | https://www.lesswrong.com/posts/Gyggp2DJRMRLSnhid/a-brief-note-on-simplicity-bias-1 |
# Brain-like AGI project "aintelope"
Steven Byrnes called for a [brain-like AGI research agenda](https://www.lesswrong.com/s/HzcM2dkCq7fwXBej8/p/tj8AC3vhTnBywdZoA), and three guys from the Hamburg EA community listened.
We are excited about Steven's five-star program [15.2.1.2 "Reverse-engineer human social instinct... | https://www.lesswrong.com/posts/c2tEfqEMi6jcJ4kdg/brain-like-agi-project-aintelope |
# All the posts I will never write
*This post has been written for the first* [*Refine*](https://www.alignmentforum.org/posts/5uiQkyKdejX3aEHLM/how-to-diversify-conceptual-alignment-the-model-behind) *blog post day, at the end of the week of readings, dicussions, and exercises about epistomology for doing good concept... | https://www.lesswrong.com/posts/EeTq9vbzMT4Zb4oWo/all-the-posts-i-will-never-write |
# A Mechanistic Interpretability Analysis of Grokking
**A significantly updated version of this work is** [**now on Arxiv**](https://arxiv.org/abs/2301.05217) **and was published as a spotlight paper at ICLR 2023**
*aka, how the best way to do modular addition is with Discrete Fourier Transforms and trig identities*
... | https://www.lesswrong.com/posts/N6WM6hs7RQMKDhYjB/a-mechanistic-interpretability-analysis-of-grokking |
# Seeking Interns/RAs for Mechanistic Interpretability Projects
**UPDATE:** The deadline for applying for this is 11:59pm PT Sat 27 Aug. I'll be selecting candidates following a 2 week (paid) work trial, 10hr/week starting Sept 12 where participants pick a concrete research idea and try to make progress on it.
* * *
... | https://www.lesswrong.com/posts/Kx9K7tLFf8rxcnNLT/seeking-interns-ras-for-mechanistic-interpretability |
# Extreme Security
What makes "password" a bad password?
You might say that it's because everybody else is already using it, and maybe you'd be correct that everybody in the world deciding to no longer use the password "password" could eventually make it acceptable again.
But consider this: if everyone who was usi... | https://www.lesswrong.com/posts/QXKzSjje3pauhepPN/extreme-security |
# And the Revenues Are So Small
The ‘Inflation Reduction Act’ includes a corporate minimum tax provision. The top 150 American corporations have to pay a minimum percentage of book profits as tax.
The whole thing has strange implications, with Marginal Revolution offering several [claims](https://marginalrevolution.c... | https://www.lesswrong.com/posts/LYctjqrExscgwYXea/and-the-revenues-are-so-small |
# The Parable of the Boy Who Cried 5% Chance of Wolf
*Epistemic status: a parable making a moderately strong claim about statistics*
Once upon a time, there was a boy who cried "there's a 5% chance there's a wolf!"
The villagers came running, saw no wolf, and said "He *said* there was a wolf and there was not. Thus ... | https://www.lesswrong.com/posts/tE7y8FZe7wSSzoRaS/the-parable-of-the-boy-who-cried-5-chance-of-wolf |
# Are there practical exercises for developing the Scout mindset?
I'm thinking about doing a LessWrong meetup around the Scout mindset. I prefer to have meetups where there are a lot of two-person exercises as those are really good to get nerds who have trouble with normal small talk to connect to each other in additi... | https://www.lesswrong.com/posts/xrJjuFfBFD33eHsiP/are-there-practical-exercises-for-developing-the-scout |
# "What Mistakes Are You Making Right Now?"
(That's^ all.)
(Crazy, though, that [*you successfully*](https://arbital.com/p/Vinge_law/#:~:text=Vinge%27s%20Law%20(as,Deep%20Blue%20would.)) [noticed your mistake](https://www.lesswrong.com/posts/SA79JMXKWke32A3hG/original-seeing) from just a prompt...) | https://www.lesswrong.com/posts/oKzNLWnrYrT3pWsxj/what-mistakes-are-you-making-right-now |
# What's General-Purpose Search, And Why Might We Expect To See It In Trained ML Systems?
[Benito](https://www.lesswrong.com/users/benito) has an interesting job. Here’s some of the stuff he’s had to do over the past couple years:
* build a prototype of an office
* resolve neighbor complaints at a party
* find ... | https://www.lesswrong.com/posts/6mysMAqvo9giHC4iX/what-s-general-purpose-search-and-why-might-we-expect-to-see |
# I’m mildly skeptical that blindness prevents schizophrenia
*(Low confidence, written in a hurry.)*
I was (randomly) trying to make sense of schizophrenia the other day, and there’s a piece of the puzzle that just seems not to fit with everything else: namely, [the claim](https://www.psychologytoday.com/us/blog/the-... | https://www.lesswrong.com/posts/z9Syf3pGffpvHwfr4/i-m-mildly-skeptical-that-blindness-prevents-schizophrenia |
# Dwarves & D.Sci: Data Fortress Evaluation & 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/wmG6kdcacNoB3HWJs/dwarves-and-d-sci-data-fortress): if you intend to play that, and h... | https://www.lesswrong.com/posts/ukvT6pLaG7yocAWdb/dwarves-and-d-sci-data-fortress-evaluation-and-ruleset |
# What Makes an Idea Understandable? On Architecturally and Culturally Natural Ideas.
Midjourney generating a HD image of "a m... | https://www.lesswrong.com/posts/bQkp4pi5Ra4SgSSxn/what-makes-an-idea-understandable-on-architecturally-and |
# Deception as the optimal: mesa-optimizers and inner alignment
*This is a brief distillation of* [*Risks from Learned Optimization in Advanced Machine Learning Systems*](https://arxiv.org/abs/1906.01820) *(Hubinger et al. 2019) with a focus on deceptive alignment. Watching* [*The OTHER AI Alignment Problem: Mesa-Op... | https://www.lesswrong.com/posts/u256AQr2xiNAgPftG/deception-as-the-optimal-mesa-optimizers-and-inner-alignment |
# Against population ethics
*The following is an excerpt from some comments I wrote to Will MacAskill about a pre-publication draft of* What We Owe the Future. *It is in response to the chapter on population ethics.*
Chapter 8 presented some interesting ideas and did so clearly, I learned a lot from it.
That said, I... | https://www.lesswrong.com/posts/2yW6DkhEaoKMFbNED/against-population-ethics |
# An introduction to signalling theory
**Signalling theory** as a formal concept originated in the field of animal behaviour (ethology) in the Dutch ethologist Tinbergen’s book “Inleiding tot de diersociologie“ (1946) and later found purchase in economics, where it was introduced in the context of job markets by Micha... | https://www.lesswrong.com/posts/vzyhBN2SZpvqER4ao/an-introduction-to-signalling-theory |
# The Open Society and Its Enemies: Summary and Thoughts
In trying to make sense of the world, sometimes the most rewarding questions to pursue are the ones that seem too obvious to ask.
When Russia invaded Ukraine last February, my faith in democracy was shaken. One country decided to start blowing up citizens of an... | https://www.lesswrong.com/posts/qWoTasqcqnoFit28P/the-open-society-and-its-enemies-summary-and-thoughts-1 |
# My thoughts on direct work (and joining LessWrong)
*Epistemic status: mostly a description of my personal timeline, and some of my models (without detailed justification). If you just want the second, skip the Timeline section.*
My name is Robert. By trade, I'm a software engineer. For my entire life, I've lived... | https://www.lesswrong.com/posts/snQLXdGHSSL2AjCF8/my-thoughts-on-direct-work-and-joining-lesswrong |
# Autonomy as taking responsibility for reference maintenance
I think semantics – specifically, maintaining reference relationships – is a core component of intelligent behaviour. Consequently, I think a better understanding of semantics would enable a better understanding of what machine intelligence that is “trying ... | https://www.lesswrong.com/posts/KH8fcM6SK8EkcKcZr/autonomy-as-taking-responsibility-for-reference-maintenance |
# Spoiler-Free Review: Across the Obelisk
Life requires time for a good game now and then.
Across the Obelisk is a roguelike deckbuilder where a party of four adventurers goes on a quest. It is [Slay the Spire](https://thezvi.wordpress.com/2018/12/09/review-slay-the-spire/) meets Dungeons and Dragons.
I played Acros... | https://www.lesswrong.com/posts/u3w2h5HerCFAwJJNj/spoiler-free-review-across-the-obelisk |
# The longest training run
**In short:** Training runs of large Machine Learning systems are likely to last less than 14-15 months. This is because longer runs will be outcompeted by runs that start later and therefore use better hardware and better algorithms. \[Edited 2022/09/22 to fix an error in the hardware impro... | https://www.lesswrong.com/posts/RihYwmskuJT9Rkbjq/the-longest-training-run |
# Thoughts on 'List of Lethalities'
I read through Eliezer’s “AGI Ruin: A List of Lethalities” [post](https://www.lesswrong.com/posts/uMQ3cqWDPHhjtiesc/agi-ruin-a-list-of-lethalities), and wrote down my reactions* as I went.
I tried to track my internal responses, \*before\* adjusting for my perception of the overall... | https://www.lesswrong.com/posts/ZrFBNdD69xPz7Yhnp/thoughts-on-list-of-lethalities |
# Human Mimicry Mainly Works When We’re Already Close
What if we just simulate a bunch of alignment researchers, and have them solve the problem for us?
Of all the Dumb Alignment Ideas, this one is easily the best. Simple argument in favor: well, it’s not going to do any *worse* than the researchers would have done. ... | https://www.lesswrong.com/posts/TLvbXNHNvppNEkXYj/human-mimicry-mainly-works-when-we-re-already-close |
# Interpretability Tools Are an Attack Channel
This seems like a fairly trivial point, but I haven't seen it properly elaborated-on before, so...
When we talk about interpretability failure modes, we often worry about a [deceptively-aligned](https://www.lesswrong.com/s/r9tYkB2a8Fp4DN8yB/p/zthDPAjh9w6Ytbeks) mesa-opti... | https://www.lesswrong.com/posts/rytFP2zRYNK85rFyX/interpretability-tools-are-an-attack-channel |
# The Core of the Alignment Problem is...
*Produced As Part Of The SERI ML Alignment Theory Scholars Program 2022 Under* [*John Wentworth*](https://www.lesswrong.com/users/johnswentworth)
Introduction
============
When trying to tackle a hard problem, a generally effective opening tactic is to[ Hold Off On Propos... | https://www.lesswrong.com/posts/vMM6HmSQaKmKadvBi/the-core-of-the-alignment-problem-is-1 |
# Conditioning, Prompts, and Fine-Tuning
*(Thanks to Evan Hubinger and Nicholas Schiefer for comments on these ideas.)*
These are some notes on the relation between conditioning language models, prompting, and fine-tuning. The key takeaways are:
1. Prompting and fine-tuning can both be used to condition language mo... | https://www.lesswrong.com/posts/chevXfQmRYrTZnj8r/conditioning-prompts-and-fine-tuning |
# Progress links and tweets, 2022-08-17
**Opportunities**
-----------------
* [Long Now is looking for a Development Director](https://longnow.org/jobs/#development-director) (via [@nicholaspaul26](https://twitter.com/nicholaspaul26/status/1558120994723676161))
* [Open Phil is hiring global health R&D Program Off... | https://www.lesswrong.com/posts/4rrPt3nyjLvnC7Ngx/progress-links-and-tweets-2022-08-17 |
# Concrete Advice for Forming Inside Views on AI Safety
A lot of people want to form inside views on AI Safety, and this is a post about concrete advice on how to get started. I have a lot of hot takes on inside views and ways I think people misframe them, so I'll begin with some thoughts on this - if you just want to... | https://www.lesswrong.com/posts/LbSWMfbAHQnFpApQ7/concrete-advice-for-forming-inside-views-on-ai-safety |
# Matt Yglesias on AI Policy
Yglesias is a widely read center-left journalist. Co-founder of Vox, ex-NYT. Note the implicit invitation: “I also try to make it clear to people who are closer to the object-level work that I’m interested in writing columns on AI policy if they have ideas for me, but they mostly don’t.”
... | https://www.lesswrong.com/posts/SyNQ6LaTuntWpaQJu/matt-yglesias-on-ai-policy |
# Announcing Encultured AI:
Building a Video Game
*Also available on the* [*EA Forum.*](https://forum.effectivealtruism.org/posts/SbMfbYnMyFYX6acZ2/announcing-encultured-ai-building-a-video-game)
*Preceded By:* [Encultured AI Pre-planning, Part 2: Providing a Service](https://www.lesswrong.com/posts/2vxoTfuScspraSJ... | https://www.lesswrong.com/posts/ALkH4o53ofm862vxc/announcing-encultured-ai-building-a-video-game |
# What's up with the bad Meta projects?
What's going on with Meta? The other day Mark Zuckerberg posted a screenshot from Horizon Worlds, a supposedly central Metaverse project [which is currently being dunked on by the internet like there's no tomorrow](https://kotaku.com/metaverse-meta-facebook-selfie-vr-zuckerberg-... | https://www.lesswrong.com/posts/YXhymbmNmPeGW3X7d/what-s-up-with-the-bad-meta-projects |
# Playing Without Affordances
Play structures are so easy that even children can figure out how to use them correctly. There is a wheel. Spin it! There is a slide. Go down it! There is a staircase. Climb it! There is a tunnel. Crawl through it!
, Jonathan Richens, Matt MacDermott (Imperial) and Tom Everitt.*
Our DeepMind Alignment team researches ways to avoid AGI systems that knowingly act against the wishes of their designers. We’re particularly concerned about agents which may ... | https://www.lesswrong.com/posts/XxX2CAoFskuQNkBDy/discovering-agents |
# A conversation about progress and safety
A while ago I did a long [interview](https://rootsofprogress.org/interview-hear-this-idea-with-fin-moorhouse-and-luca-righetti) with Fin Moorhouse and Luca Righetti on their podcast *Hear This Idea*. Multiple people commented to me that they found our discussion of safety par... | https://www.lesswrong.com/posts/sKRts4bY7Fo9fXnmQ/a-conversation-about-progress-and-safety |
# Announcing the Distillation for Alignment Practicum (DAP)
Hey, fellow LessWrongers!
Around the time of John Wentworth's post [A Call for Distillers,](https://www.lesswrong.com/posts/zo9zKcz47JxDErFzQ/call-for-distillers) we started to work on a distillation course which now has turned into a practicum. (A practicum... | https://www.lesswrong.com/posts/y2CMx3DAJ86HGrjue/announcing-the-distillation-for-alignment-practicum-dap-1 |
# Alignment's phlogiston
*Crossposted from the EA Forum:* [*https://forum.effectivealtruism.org/posts/DCtqgsywCRakLvHn6/alignment-s-phlogiston*](https://forum.effectivealtruism.org/posts/DCtqgsywCRakLvHn6/alignment-s-phlogiston)
*Epistemic status: quick and dirty reaction to the claim that alignment research is like... | https://www.lesswrong.com/posts/epQFRuu78LqiboGkd/alignment-s-phlogiston |
# What does moral progress consist of?
There’s something that strikes me as odd about the way I hear “moral progress” discussed in the EA/longtermist community in general and in *What We Owe the Future* in particular.
The issues that are generally discussed under this heading are things like: animal welfare, existent... | https://www.lesswrong.com/posts/j7CKL5CQevXwhgxxx/what-does-moral-progress-consist-of |
# Introducing the Existential Risks Introductory Course (ERIC)
*Crossposted to the* [*EA Forum*](https://forum.effectivealtruism.org/posts/KrjyrsRky5JL4P5MF/introducing-the-existential-risks-introductory-course-eric-1)*.*
Introduction
------------
We (the Cambridge Existential Risks Initiative) ran an Existential Ri... | https://www.lesswrong.com/posts/oaYfCjJXtHGxXckwg/introducing-the-existential-risks-introductory-course-eric |
# Epistemic Artefacts of (conceptual) AI alignment research
*The fact that this post is seeing the light of day now rather than in some undefined number of weeks is, not in small part, due to participating in the second* [*Refine*](https://www.lesswrong.com/posts/5uiQkyKdejX3aEHLM/how-to-diversify-conceptual-alignment... | https://www.lesswrong.com/posts/CewHdaAjEvG3bpc6C/epistemic-artefacts-of-conceptual-ai-alignment-research |
# How to do theoretical research, a personal perspective
>
> _"Where do new [algorithms] come from? I keep reading about someone who invented [an algorithm] to do something-or-other but there's no mention of how."_
>
> _A shrug of robed shoulders. "Where do new books come from, Mr. Potter? Those who read many books so... | https://www.lesswrong.com/posts/zbFGRRWPwxBwjwknY/how-to-do-theoretical-research-a-personal-perspective-1 |
# Novelty Generation - The Art of Good Ideas
Recently I set out to understand how new ideas are formed in the mind. My process consisted of poking at my own mind, poking at other people's minds, and then organizing group sessions of poking at each other's minds.
 two [posts](https://www.lesswrong.com/posts/RZNmNwc9SxdKayeQh/unifying-bargaining-notions-2-2), we went over various notions of bargaining. The Nash bargaining solut... | https://www.lesswrong.com/posts/WvHpsqmwQopnaMh46/less-threat-dependent-bargaining-solutions-3-2 |
# Reducing Goodhart: Announcement, Executive Summary
**I - Release announcement**
I have just posted a *very* thorough edit-slash-rewrite of my [Reducing Goodhart sequence](https://www.alignmentforum.org/s/aJvgWxkCBWpHpXti4). Many thanks to people who gave me feedback or chatted with me about related topics, and many... | https://www.lesswrong.com/posts/2drtQcFoyRCjYaJFe/reducing-goodhart-announcement-executive-summary |
# Benchmarking Proposals on Risk Scenarios
*This post has been written for the second* [*Refine*](https://www.alignmentforum.org/posts/5uiQkyKdejX3aEHLM/how-to-diversify-conceptual-alignment-the-model-behind) *blog post day, at the end of the first week of iterating on ideas and concretely aiming at the alignment prob... | https://www.lesswrong.com/posts/DsYe7TKc4NhyJuPEy/benchmarking-proposals-on-risk-scenarios |
# Vague concepts, family resemblance and cluster properties
My definition of a "vague concept" (**A**):
* It creates a group of things that share [family resemblance](https://en.wikipedia.org/wiki/Family_resemblance) or *"family difference"*.
* It doesn't have any meaning outside of context. But in a specific co... | https://www.lesswrong.com/posts/yHjMgGJupzvAvADhn/vague-concepts-family-resemblance-and-cluster-properties |
# What if we approach AI safety like a technical engineering safety problem
*This post has been written for the second *[*Refine*](https://www.alignmentforum.org/posts/5uiQkyKdejX3aEHLM/how-to-diversify-conceptual-alignment-the-model-behind) *blog post day, at the end of the first week of iterating on ideas and concre... | https://www.lesswrong.com/posts/zNYmbFwgrxiNtayMm/what-if-we-approach-ai-safety-like-a-technical-engineering |
# No One-Size-Fit-All Epistemic Strategy
*This post is part of the work done at* [*Conjecture*](https://conjecture.dev)*.*
*This post has been written for the second* [*Refine*](https://www.alignmentforum.org/posts/5uiQkyKdejX3aEHLM/how-to-diversify-conceptual-alignment-the-model-behind) *blog post day, at the end of... | https://www.lesswrong.com/posts/du92yeHQn9iE5vorj/no-one-size-fit-all-epistemic-strategy |
# Refine's Second Blog Post Day
Yesterday was the second blog post day at Refine. It came after the first week of research and iteration on incubatees ideas and questions.
It still went well. There are less posts than the first day, but even the incubatees that didn't write a post ended up working on interesting rese... | https://www.lesswrong.com/posts/EXkY7drHc6jdNxCmB/refine-s-second-blog-post-day |
# The Loire Is Not Dry
Several of my friends have shared this paste (or shares of it) on Facebook:
> This is the current state of the Loire, the longest river in France. This has not happened before in at least the 2000 years since literate people inhabited France. The Romans would have written about this. The mediev... | https://www.lesswrong.com/posts/AfocZvFSmaXpHDYxi/the-loire-is-not-dry |
# Hamburg, Germany – ACX Meetups Everywhere 2022
This year's ACX Meetup everywhere in Hamburg, Germany.
Location: Kleine Wallanlagen on the lawn near Memorial Holstenglacis. Look for pink blankets; I will also have an ACX sign. [Here is an Open Street Map Link](https://routing.openstreetmap.de/?z=19¢er=53.556405%2C9.... | https://www.lesswrong.com/events/5ZrDt9df7hAJMzdr3/hamburg-germany-acx-meetups-everywhere-2022 |
# Broad Picture of Human Values
I have two strong intuitions about human values that'd seemed utterly irreconcilable to me, up until recently.
* On the one hand, human values are clearly inchoate, unstable messes of niche heuristics and preferences that often contradict each other and can change on a dime. A twenty... | https://www.lesswrong.com/posts/kQKcEkzKmpPe6qxDb/broad-picture-of-human-values-1 |
# What's the Least Impressive Thing GPT-4 Won't be Able to Do
It seems like GPT-4 is going to be coming out soon and, so I've heard, it will be [awesome](https://twitter.com/Scobleizer/status/1560843951287898112). Now, we don't know anything about its architecture or its size or how it was trained. If it were only tra... | https://www.lesswrong.com/posts/HNqTzxjvaczguH2uJ/what-s-the-least-impressive-thing-gpt-4-won-t-be-able-to-do |
# Paper is published! 100,000 lumens to treat seasonal affective disorder
Let’s give people with winter depression (seasonal affective disorder, SAD) LOTS OF LIGHT and see what happens!
This is an update to my previous post that said that our preprint is out. :) The paper was accepted with very minor changes, so you ... | https://www.lesswrong.com/posts/spcShtqP7J4medJAD/paper-is-published-100-000-lumens-to-treat-seasonal |
# Review: Amusing Ourselves to Death
*This is a linkpost that I'm posting in response to a request to share more of the archive of my* [*personal blog*](https://www.strataoftheworld.com/) *on LessWrong.*
At first glance, *Amusing Ourselves to Death* seems neither interesting nor relevant. It is a short book, publi... | https://www.lesswrong.com/posts/Pix8haJTbkuxPtxpb/review-amusing-ourselves-to-death |
# How evolution succeeds and fails at value alignment
Disclaimer: I don't have a background in alignment research or reinforcement learning, and I don't think any of the ideas discussed here are new, but they might be interesting to some.
A recent post suggested that [humans provide an untapped wealth of evidence abo... | https://www.lesswrong.com/posts/tGehLsMHSdyZohpuL/how-evolution-succeeds-and-fails-at-value-alignment |
# Embracing the Opposition's Point
Part of why I like making fun of debate bros is because I used to be one. It’s embarrassing to admit – but here we are. The tension, the adrenalin rush, and the righteous wrath felt invigorating. “Winning” an argument was satisfying. I have worked hard on changing that for the last c... | https://www.lesswrong.com/posts/DWxAqp7YHtLtHJtBt/embracing-the-opposition-s-point |
# My Plan to Build Aligned Superintelligence
*2ish months ago, I realized making sure superintelligent AI doesn’t kill everyone was the biggest problem period. I resolved to understand and maybe help solve the problem. A friend of mine who does alignment research said that, in order to clarify my thinking, I should si... | https://www.lesswrong.com/posts/EaDN9grmTfzXbM652/my-plan-to-build-aligned-superintelligence |
# Pivotal acts using an unaligned AGI?
Introduction
============
A recent post ([AGI Ruin: A List of Lethalities](https://www.lesswrong.com/posts/uMQ3cqWDPHhjtiesc/agi-ruin-a-list-of-lethalities)) gives an argument for humanity’s likely inability to survive the invention of superintelligent AGI. The argument goes ver... | https://www.lesswrong.com/posts/FsEDu6CvzyzJkrQd8/pivotal-acts-using-an-unaligned-agi |
# The Bunny: An EA Short Story
*This is cross-posted from* [*my blog*](https://www.johncgreer.com/the-bunny/)*.*
“Eight villages in our district have had significant to total structural damage from mudslides. Furthermore, the weather conditions conducive to mudslides have only been increasing in the past decade.”
Pr... | https://www.lesswrong.com/posts/8y5RapenxvuoXwuhy/the-bunny-an-ea-short-story |
# Volunteer to host a meetup!
Once (or twice) a year, Scott Alexander promotes a bunch of meetups on [his blog](https://astralcodexten.substack.com/), and it's that time of year again! If you've ever wanted to start a meetup, this is a great time — these events get a ton of visibility. If you feel like you're 'not qua... | https://www.lesswrong.com/posts/CBWwHcekJ7JNWJzyH/volunteer-to-host-a-meetup |
# AXRP Episode 17 - Training for Very High Reliability with Daniel Ziegler
[YouTube link](https://www.youtube.com/watch?v=S4HmgrXievg&list=PLmjaTS1-AiDeqUuaJjasfrM6fjSszm9pK&index=17)
Sometimes, people talk about making AI systems safe by taking examples where they fail and training them to do well on those. But how ... | https://www.lesswrong.com/posts/3hJhAsZHXP44Ljah5/axrp-episode-17-training-for-very-high-reliability-with |
# What if we solve AI Safety but no one cares
Suppose that next year, AI Safety is solved, the solution is approved by Eliezer Yudkowsky, etc.
How do we actually get people to follow this solution?
It seems to me a lot of people or companies will ignore any AI Safety solutions for the same reasons they are currently... | https://www.lesswrong.com/posts/4rgDink5LirgzwyqF/what-if-we-solve-ai-safety-but-no-one-cares |
# AI art isn't "about to shake things up". It's already here.
For a while, I've been seeing people commenting about how AI art is on the cusp of shaking up the art world. Quite frankly, at this point I consider such sentiments to be behind the times. AI art is not "on the cusp" of disrupting such things. It is *alread... | https://www.lesswrong.com/posts/DhDAXQw4PsWXnmwPS/ai-art-isn-t-about-to-shake-things-up-it-s-already-here |
# ACX Meetup Madrid
As part of the [ACX Meetups Everywhere 2022](https://astralcodexten.substack.com/p/meetups-everywhere-2022-call-for), we'll be hosting an ACX meetup!
[This](https://plus.codes/8CGRC897+F8C) is the exact location. There will be an ACX sign to make us easy to find. Feel free to come or leave at any ... | https://www.lesswrong.com/events/GxucDGFJFPtxpGziN/acx-meetup-madrid |
# On Car Seats as Contraception
(Note: This post takes the perspective of strongly preferring higher American fertility to lower American fertility on anything close to current margins. I endorse this view and consider it very overdetermined, but this post does not defend or justify it.)
(Note: This post has been edi... | https://www.lesswrong.com/posts/c4daDbQD398BmW5k5/on-car-seats-as-contraception |
# Please (re)explain your personal jargon
Technical jargon is quite useful. It allows us to put sharp definitions onto natural phenomena useful for precise scientific discussion. However, it can also be confusing, since you are creating or re-using words to mean this new specific precise thing. So, whenever you write ... | https://www.lesswrong.com/posts/Qigz6JsEQsAnRddhG/please-re-explain-your-personal-jargon |
# Finding Goals in the World Model
### *Produced As Part Of The SERI ML Alignment Theory Scholars Program 2022 Under* [*John Wentworth*](https://www.lesswrong.com/users/johnswentworth)
Introduction
============
This post works off the assumption that the first AGI comes relatively soon, and has an architecture which... | https://www.lesswrong.com/posts/jnyTqPRHwcieAXgrA/finding-goals-in-the-world-model |
# Stable Diffusion has been released
It's an open-source text-to-image model capable of producing NSFW content. I, for one, I'm very excited to see the consequences this has on society, which I expect to be mostly positive (except if it leads to an increase in AI funding).
Any thoughts?
Other relevant links:
https:/... | https://www.lesswrong.com/posts/BjHjktmqYN8nfbuaz/stable-diffusion-has-been-released |
# ACX Montreal Meetups Everywhere September 24 2022
Remember: **Please check this page the day of, as I will update it in the event of rain.**
**Venue: **
Jeanne-Mance Park, at the corner of Duluth and Esplanade. Rough location here: [https://plus.codes/87Q8GC89+37](https://plus.codes/87Q8GC89+37)
Look for a guy w... | https://www.lesswrong.com/events/oSwZHswrEnudpYzBh/acx-montreal-meetups-everywhere-september-24-2022 |
# Digital Dinner Signup
For [house dinner](https://www.jefftk.com/p/shared-house-setup) we need a way of coordinating who is going to cook what days and marking who will/won't be at dinner. For the past ~7y we've used a piece of paper on the fridge with a row for each day. This is not a bad system, and it's generally ... | https://www.lesswrong.com/posts/633zrPiJDhFRqfSCR/digital-dinner-signup |
# Discussion on utilizing AI for alignment
_Summarized/re-worded from a discussion I had with [John Wentworth](https://www.lesswrong.com/users/johnswentworth). John notes that everything he said was off-the-cuff, and he doesn’t always endorse such things on reflection._
Eli: I have a question about your post on [Godz... | https://www.lesswrong.com/posts/vMkvx3ZPNKxffXeLr/discussion-on-utilizing-ai-for-alignment-1 |
# Nate Soares' Life Advice
Disclaimer: Nate gave me some life advice at EA Global; I thought it was pretty good, but it may or may not be useful for other people. If you think any of this would be actively harmful for you to apply, you probably shouldn't.
Notice subtle things in yourself
=============================... | https://www.lesswrong.com/posts/aan3jPEEwPhrcGZjj/nate-soares-life-advice |
# AI alignment as “navigating the space of intelligent behaviour”
Tl;dr
=====
In this post, I introduce a conceptual tool for thinking about the epistemic landscape of AI alignment and then describe three epistemic strategies for making progress on the alignment problem: 1) tinkering, 2) idealisation and 3) intellige... | https://www.lesswrong.com/posts/FuToH2KHxKmJLGk2B/ai-alignment-as-navigating-the-space-of-intelligent |
# Why does AGI need a utility function?
My intuition says that a narrow AI like DALL-E would not blow up the world, no matter how much smarter it became. It would just get really good at making pictures.
This is clearly a form of superintelligence we would all prefer, and the difference seems to me to be that DALL-E ... | https://www.lesswrong.com/posts/aAn6g2bRsccBx8cei/why-does-agi-need-a-utility-function |
# AGI Timelines Are Mostly Not Strategically Relevant To Alignment
First things first: this post is about [timelines](https://www.lesswrong.com/tag/ai-timelines) (i.e. what year we first get transformative AGI), not [takeoff speeds](https://www.lesswrong.com/tag/ai-takeoff) (i.e. foom vs gradual takeoff).
Claim: time... | https://www.lesswrong.com/posts/d7RFGPMkbKurQjW6o/agi-timelines-are-mostly-not-strategically-relevant-to |
# Timelines ARE relevant to alignment research (timelines 2 of ?)
This post is written in part in reaction to John Wentworth's post [AGI Timelines Are Mostly Not Strategically Relevant To Alignment](https://www.lesswrong.com/posts/d7RFGPMkbKurQjW6o/agi-timelines-are-mostly-not-strategically-relevant-to)
Oddly, I most... | https://www.lesswrong.com/posts/wgcFStYwacRB8y3Yp/timelines-are-relevant-to-alignment-research-timelines-2-of |
# Survey advice
Things I believe about making surveys, [after](https://aiimpacts.org/2016-expert-survey-on-progress-in-ai/) [making](https://aiimpacts.org/2022-expert-survey-on-progress-in-ai/) [some](https://worldspiritsockpuppet.com/2022/01/18/covid-survey.html) [surveys](https://worldspiritsockpuppet.com/2020/11/06... | https://www.lesswrong.com/posts/oyKzz7bvcZMEPaDs6/survey-advice |
# Thoughts about OOD alignment
We may have one example of realized out-of-distribution alignment: maternal attachment. Evolution has been able to create an architecture that seems to take care of something reliably enough that modern humans, with access to unlimited food, drugs, and VR headsets, do not seek to feed a ... | https://www.lesswrong.com/posts/7kcDTJmEPAnSbyeYh/thoughts-about-ood-alignment |
# Ethan Perez on the Inverse Scaling Prize, Language Feedback and Red Teaming
I talked to [Ethan Perez](https://www.lesswrong.com/users/ethan-perez) about the [Inverse Scaling Prize](https://www.lesswrong.com/posts/eqxqgFxymP8hXDTt5/announcing-the-inverse-scaling-prize-usd250k-prize-pool) (deadline August 27!), [Trai... | https://www.lesswrong.com/posts/xPrtsLAyMcZkkszvw/ethan-perez-on-the-inverse-scaling-prize-language-feedback |
# Adversarial epistemology
This is the first article in my *[Bah-Humbug Sequence](https://www.lesswrong.com/s/9ctJpFJvf8h5ivT6Q)* a.k.a. "Everything I Don't Like Is A Defect/Defect Equilibrium". Epistemic status: strong opinion weakly held, somewhat exaggerated for dramatic effect; I'm posting this here so that the en... | https://www.lesswrong.com/posts/ezTA3fSPCWZieBfjS/adversarial-epistemology |
# Why are some problems Super Hard?
The problem
===========
I have been wondering a lot about a problem lately, that I believe has far reaching consequences.
Some context: There are many theoretical problems that are considered to be obviously far far outside our problem solving ability. Two examples of those wo... | https://www.lesswrong.com/posts/w26TY2tFHRTdposre/why-are-some-problems-super-hard |
# Beliefs and Disagreements about Automating Alignment Research
*Epistemic status: Mostly organizing and summarizing the views of others.*
*Thanks to those whose views I summarized in this post, and to Tamera Lanham, Nicholas Kees Dupuis, Daniel Kokotajlo, Peter Barnett, Eli Lifland, and Logan Smith for reviewing a d... | https://www.lesswrong.com/posts/JKgGvJCzNoBQss2bq/beliefs-and-disagreements-about-automating-alignment |
# OpenAI's Alignment Plans
Our alignment research aims to make artificial general intelligence (AGI) aligned with human values and follow human intent. We take an iterative, empirical approach: by attempting to align highly capable AI systems, we can learn what works and what doesn’t, thus refining our ability to make... | https://www.lesswrong.com/posts/28sEs97ehEo8WZYb8/openai-s-alignment-plans |
# Vingean Agency
I've been involved with several discussions about different notions of agency (and their importance/relationships) lately, especially with the PIBBSS group including myself, Daniel, Josiah, and Ramana; see [here](https://www.lesswrong.com/posts/rQDYQrDjPGqjrf8Mk/bridging-expected-utility-maximization-... | https://www.lesswrong.com/posts/fWiYdyicEaCHCcAKx/vingean-agency |
# History of Zero-based Months?
JavaScript is a language of many silly things, and one of them is:
\> new Date()
Wed Aug 24 2022 ...
\> new Date().getFullYear()
2022
\> new Date().getMonth()
7
\> new Date().getDate()
24
It represents 2022-08-24 as (2022, 7, 24). One-based indexing for the year and day, but zero-base... | https://www.lesswrong.com/posts/8D4x46qsvYabjPc89/history-of-zero-based-months |
# Google AI integrates PaLM with robotics: SayCan update [Linkpost]
Back in April, Google AI announced [SayCan](https://say-can.github.io/), a project which integrated a language model (FLAN) with robotics in order to produce a robot which could follow instructions. For example, cleaning up a mess in a kitchen. ([Less... | https://www.lesswrong.com/posts/CuBKm8bkfWhYegcw8/google-ai-integrates-palm-with-robotics-saycan-update |
# What Makes A Good Measurement Device?
In some sense, any new measurement device automates a part of research. A thermometer automates the task of sticking one's finger in something to check how hot it is, a scale automates the task of holding something to check how heavy it is, etc. The automated version is not only... | https://www.lesswrong.com/posts/aKoeYyHZTRu8izMwA/what-makes-a-good-measurement-device |
# The Shard Theory Alignment Scheme
*Generated as part of SERI MATS, Team Shard, under John Wentworth.*
*Thanks to Logan Smith, Garrett Baker, Michael Einhorn, Quintin Pope, and Thomas Kwa for chatting about related topics. All mistakes are my own.*
The Alignment Scheme
====================
The shard theory alignme... | https://www.lesswrong.com/posts/ZNXDRGshgoq3cmxhB/the-shard-theory-alignment-scheme |
# [Review] The Problem of Political Authority by Michael Huemer
> The state is often ascribed a special sort of authority, one that obliges citizens to obey its commands and entitles the state to enforce those commands through threats of violence. This book argues that this notion is a moral illusion: no one has ever ... | https://www.lesswrong.com/posts/dhuDBKxaHCvyDnwf8/review-the-problem-of-political-authority-by-michael-huemer |
# The Solomonoff prior is malign. It's not a big deal.
*Epistemic status: Endorsed at ~85% probability. In particular, there might be clever but hard-to-think-of encodings of observer-centered laws of physics that tilt the balance in favor of physics. Also, this isn't that different from Mark Xu's post.*
[Previously]... | https://www.lesswrong.com/posts/f7qcAS4DMKsMoxTmK/the-solomonoff-prior-is-malign-it-s-not-a-big-deal |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.