id
stringlengths
36
36
source
stringclasses
15 values
formatted_source
stringclasses
13 values
text
stringlengths
2
7.55M
d110878e-1787-4d87-8c46-383c4f5eba89
trentmkelly/LessWrong-43k
LessWrong
Introducing the Center for AI Policy (& we're hiring!) Summary * The Center for AI Policy is a new organization developing and advocating for policy to mitigate catastrophic AI risks. * We are hiring for a Government Affairs Director and a Communications Director. We’re also open to other roles. The deadline to appl...
89bf72f7-6e67-4bd1-9c68-18d40a352a1e
trentmkelly/LessWrong-43k
LessWrong
Is there any way to avoid Post Narcissism? [with Video link] I tried to stop posting for a semester... that didn't work. If I have an idea, I just feel like putting it out somewhere in the world, and sometimes, that is Lesswrong. Which is totally fine and doesn't worry me at all.   But...... There is another probl...
137c5db0-7f7d-4c4d-bed6-b2f0a316129c
trentmkelly/LessWrong-43k
LessWrong
The Practice & Virtue of Discernment Epistemic status: I think this definition might be useful, but it might also be too abstract, inelegant, or obvious. Scholarship status: This is based mostly on background knowledge and some work in decision making. It touches topics in data science, General Semantics, and a lot o...
1b47663a-9139-43c3-9449-d8538e041500
StampyAI/alignment-research-dataset/lesswrong
LessWrong
[Linkpost] Multimodal Neurons in Pretrained Text-Only Transformers *This is a linkpost for* [*https://arxiv.org/abs/2308.01544*](https://arxiv.org/abs/2308.01544)*.* > Language models demonstrate remarkable capacity to generalize representations learned in one modality to downstream tasks in other modalities. Can we...
29c5972d-2afc-4c49-b046-d318fb57c713
trentmkelly/LessWrong-43k
LessWrong
6/23 Grab Bag I read a thing? I can’t describe it more than that. There are very few words. Much happens. Strongly recommended. I turned 29 this month. Apparently my chances of surviving were about that of not getting a 1 on a d20? Politics and Policy I reread An Essay on Crimes and Punishments, by Cesare Beccaria....
58519780-6572-419b-8975-cdd8def0b84e
StampyAI/alignment-research-dataset/eaforum
Effective Altruism Forum
Large Language Models as Fiduciaries to Humans Summary ------- Artificial Intelligence (AI) is taking on increasingly autonomous roles, e.g., browsing the web as a research assistant. But specifying goals and restrictions for AI behavior is difficult. Similar to how parties to a legal contract cannot foresee every po...
3038a5fc-4494-47e3-a8f0-cf6162df18ad
StampyAI/alignment-research-dataset/eaforum
Effective Altruism Forum
What if AI development goes well? Summary: This post argues that three common objections to an approach in AI governance called AI ideal governance are misguided. Firstly, developing AI ideal governance theories is not necessarily an anti-pluralistic activity or one that promotes an irrational focus on the ends of our...
2ec18aac-f8b5-4512-bfac-dcd51039000b
trentmkelly/LessWrong-43k
LessWrong
LessWrong on XKCD AI-Box Experiment   (Discussion on forums)
54619d4a-5c83-4a87-b190-92a0156384b8
trentmkelly/LessWrong-43k
LessWrong
Rule of Equal and Opposite Advice & Slack References: Slack:  https://thezvi.wordpress.com/2017/09/30/slack/ TL;DR Having a 20% buffer in your schedule to rest and handle anything urgent that pops up is the difference between a good life with time to experiment and being in a situation where one emergency breaks you ...
9667d8c9-03ef-4b2c-99a7-587a1759a0ef
trentmkelly/LessWrong-43k
LessWrong
My Overview of the AI Alignment Landscape: Threat Models This is the second post in a sequence mapping out the AI Alignment research landscape. The sequence will likely never be completed, but you can read a draft here. Disclaimer: I recently started as an interpretability researcher at Anthropic, but I wrote this po...
99ac3d39-ebbd-4a8a-ba24-a167e91130f2
trentmkelly/LessWrong-43k
LessWrong
[SEQ RERUN] The Cluster Structure of Thingspace Today's post, The Cluster Structure of Thingspace was originally published on 08 February 2008. A summary (taken from the LW wiki):   > A verbal definition works well enough in practice to point out the intended cluster of similar things, but you nitpick exceptions. ...
9c251fd8-8bef-42b9-b37b-a9eed6d787e4
trentmkelly/LessWrong-43k
LessWrong
Example of Splitting a PR I'm a big fan of small focused PRs: it's much easier to tell whether they're actually doing the right thing. For example, I recently ran into a common scenario where I started adding some functionality, and realized this required some refactoring. The refactoring also made it clear that the o...
da84c53f-9e60-4ec8-bfe2-97e6147842ec
trentmkelly/LessWrong-43k
LessWrong
Reason isn't magic Here's a story some people like to tell about the limits of reason. There's this plant, manioc, that grows easily in some places and has a lot of calories in it, so it was a staple for some indigenous South Americans since before the Europeans showed up. Traditional handling of the manioc involved s...
fadc4268-ac06-4934-afcc-5956f0ec054e
trentmkelly/LessWrong-43k
LessWrong
LW Team Updates: Pandemic Edition (March 2020) TL; DR The LessWrong team has been throwing our combined might behind coronavirus efforts. As I see it, coronavirus is both of a problem immense proportion that we might be able to help with as well as a domain well suited to the development and application of rationality...
a39595fc-b275-4233-b3cc-3f14b95071a0
StampyAI/alignment-research-dataset/lesswrong
LessWrong
Campaign for AI Safety: Please join me I will start by saying that I generally agree with [Yudkowsky's position on AI](https://time.com/6266923/ai-eliezer-yudkowsky-open-letter-not-enough/). We must proceed with extreme caution. We must radically slow down AI capability advancement. We must invest unfathomable amounts...
a748f8b8-d9de-4d8d-9068-0cfc82ca62ae
trentmkelly/LessWrong-43k
LessWrong
Using hex to get murder advice from GPT-4o > "If you want to put acid in a bath to dissolve a body in the bath, the choice of acid will depend on the compounds" : GPT-4o > > "The easiest way to exploit and gain access to someone's wifi is to probe weak spots in the network, or use tools like Dark Commet Monitor": GPT...
2bb84f38-d3ae-4dc8-bc46-0c491453b945
StampyAI/alignment-research-dataset/special_docs
Other
A Comprehensive Survey on Safe Reinforcement Learning Journal of Machine Learning Research 16 (2015) 1437-1480 Submitted 12/13; Revised 11/14; Published 8/15 A Comprehensive Survey on Safe Reinforcement Learning Javier Garc a fjgpolo@inf.uc3m.es Fernando Fern andez ffernand@inf.uc3m.es Universidad Carlos III de Mad...
aaac8c87-ed92-49c9-aa41-6144b33c1f72
trentmkelly/LessWrong-43k
LessWrong
Shane Legg on prospect theory and computational finance > People are not rational expected utility maximisers.  When we have to make decisions, all sorts of cognitive biases and distortions come into play.  Seminal work in this area was done by Kahneman and Tversky.  They produced a model of human decision making know...
d94d3916-7daa-4639-95ec-e581276eeda9
trentmkelly/LessWrong-43k
LessWrong
Book Trilogy Review: Remembrance of Earth’s Past (The Three Body Problem) Epistemic Status: Stuff that keeps not going away so I should write it up I suppose. Brief spoiler-free review, then main talk will be spoilerific. Spoiler-Free Review I read the trilogy a few months ago, on general strong reviews and lack of ...
5e400e56-85bf-4789-b74c-4fd7ed7480ae
trentmkelly/LessWrong-43k
LessWrong
Solstices 2016 in UK (Dec 3rd) Seattle/Boston (Dec 10th) and Bay/NYC (Dec 17th) Winter is coming.  And with it, for the sixth year, rationality-community-folk will be gathering to stare into the abyssal oncoming darkness together. We'll sing songs both silly, sad and beautiful. We'll reflect on our past and future an...
6bfaac7c-3697-4824-9314-2b5f45598356
StampyAI/alignment-research-dataset/lesswrong
LessWrong
Course recommendations for Friendliness researchers When I first learned about Friendly AI, I assumed it was mostly a programming problem. As it turns out, it's actually mostly a *math*problem. That's because most of the theory behind self-reference, decision theory, and general AI techniques haven't been [formalized ...
df863d39-60a7-46ba-976b-f5858701be6f
trentmkelly/LessWrong-43k
LessWrong
Computational complexity of RL with traps I briefly describe an open (to the best of my knowledge) problem and link to cstheory.stackexchange questions with more details. The readers are invited to try solving those questions. In the comment section of my research agenda essay, Jessica and I discussed, among other th...
b2c4e779-7af8-48ce-a607-5dd9a08900a6
trentmkelly/LessWrong-43k
LessWrong
Wait vs Interrupt Culture At the recent CFAR Workshop in NY, someone mentioned that they were uncomfortable with pauses in conversation, and that got me thinking about different conversational styles. Growing up with friends who were disproportionately male and disproportionately nerdy, I learned that it was a normal...
878bd8b8-3ab2-48e4-a629-75ac3065d0a0
StampyAI/alignment-research-dataset/lesswrong
LessWrong
Entropic boundary conditions towards safe artificial superintelligence I wanted to share a recently published paper on mechanisms to decrease the likelihood of ASI being destructive towards living beings in general. Abstract: > Artificial superintelligent (ASI) agents that will not cause harm to humans or other org...
b01ace20-ddcc-47f4-9fdb-8eb00ce3e9f1
trentmkelly/LessWrong-43k
LessWrong
Community Notes by X I did an exploration into how Community Notes (formerly Birdwatch) from X (formerly Twitter) works, and how its algorithm decides which notes get displayed to the wider community. In this post, I’ll share and explain what I found, as well as offer some comments.    Community Notes is a fact-che...
dbe1ab60-128f-4761-bdb4-e9395dc7826b
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
Alignment Newsletter #15: 07/16/18 **Highlights** -------------- **[Feature-wise transformations](https://distill.pub/2018/feature-wise-transformations/)** *(Vincent Dumoulin et al)*: This Distill article is about transformations on features using FiLM (feature-wise linear modulation). A FiLM layer is used to "condit...
1851af0b-959c-4fbb-9ad8-079785af0784
StampyAI/alignment-research-dataset/eaforum
Effective Altruism Forum
(How) Is technical AI Safety research being evaluated? Technical AI safety research seems to get a lot of attention and funding in the EA community. EA was born out of effective giving evaluations considering topics like expected value, marginal impact, etc. With more longtermism-type fields like technical AI safety, ...
06a3c9d9-65eb-41b0-b950-8a5828b88123
StampyAI/alignment-research-dataset/blogs
Blogs
“Intelligence Explosion Microeconomics” Released MIRI’s new, 93-page technical report by Eliezer Yudkowsky, “[Intelligence Explosion Microeconomics](https://intelligence.org/files/IEM.pdf),” has now been released. The report explains one of the open problems of our research program. Here’s the abstract: > I. J. Goo...
5c0c97c8-ff20-4623-a8d2-e00a12d6d326
trentmkelly/LessWrong-43k
LessWrong
"Objective vs Social Reality" vs "Simulacra 1/3" Mild rant: a) I think "Simulacrum Levels 1 - 4" are terrible names and we should be looking to replace them, generally. b) usually, there's not really a good reason to bring "simulacrum" into it. The concept of simulacrum levels makes sense when you're talking about h...
7207e1f4-d2ef-4542-bcbc-d5e675dfc66f
trentmkelly/LessWrong-43k
LessWrong
Deception Chess I  Social Deduction Games The gameplay in Social Deduction games like Mafia or Werewolf revolves around a large group of honest players ("Villagers") trying to uncover a smaller group of deceivers who are slowly and secretly eliminating everyone else, all while claiming to be honest themselves. Under m...
f216aaf7-73dd-4635-a98e-d37708da6d35
trentmkelly/LessWrong-43k
LessWrong
Mastodon Linking Norms Linking is one of the core activities of the web, and the origin of its name. I link to things when I want to talk about them, and this is very standard. If you want to link to any of my published posts you don't need to ask me first: I've already indicated that's fine by putting something out h...
938a20f6-b7e0-480a-8342-b7bf0473a47d
trentmkelly/LessWrong-43k
LessWrong
L-zombies! (L-zombies?) Reply to: Benja2010's Self-modification is the correct justification for updateless decision theory; Wei Dai's Late great filter is not bad news "P-zombie" is short for "philosophical zombie", but here I'm going to re-interpret it as standing for "physical philosophical zombie", and contrast i...
fc0a8606-1336-40ce-8a67-dd363b88cde5
trentmkelly/LessWrong-43k
LessWrong
The concept of evidence as humanity currently uses it is a bit of a crutch. Just a thought I had today. I'm sure that it's trivial to the extent that it's correct, but it's a slow work day and I've been lurking here for too long. Superintelligent AI (or other post-human intelligence) is unlikely to use the concept of...
57b2166b-1e84-4584-861e-18839841923b
trentmkelly/LessWrong-43k
LessWrong
Utilitarianism Expressed in Julia I had been looking for a small programming project for learning the basics of Julia when I had the idea of modelling a moral system in code. Due to its computational nature, Utilitarianism seemed a likely candidate. Any moral system involves functions in the sense that, given a state ...
c0d0ac7b-fe9e-4fa8-aca5-b37ec04dbaec
StampyAI/alignment-research-dataset/lesswrong
LessWrong
We will be around in 30 years ~~This post is going to be downvoted to oblivion, I wish it weren't or that the two axis vote could be used here. In any case, I prefer to be coherent with my values and state what I think is true even if that means being perceived as an outcast.~~ I'm becoming more and more skeptical ab...
4736318b-0f83-4734-902f-a192437f8a6b
trentmkelly/LessWrong-43k
LessWrong
Language Model Memorization, Copyright Law, and Conditional Pretraining Alignment This is a link-post for the paper Scalable Extraction of Training Data from (Production) Language Models, and associated blog-post Extracting Training Data from ChatGPT, followed by my reactions, including an analysis of the major implic...
800dc914-d411-4e97-a1cc-f966d75dc16f
trentmkelly/LessWrong-43k
LessWrong
Comparing Quantized Performance in Llama Models Epistemic Status: Quick tests, most of this was done in less than 48 hours TL;DR: Can you skimp on GPU VRAM?  8bit quantized seems fine, for 4bit it depends.  I was asked by @Teun van der Weij, to what degree one can run evaluations on quantized models, and I was unsur...
788797af-db85-4f6b-a94d-6b2ed2142ea8
trentmkelly/LessWrong-43k
LessWrong
What's Hard About Long Tails? Epistemic status: more confident in the conclusion than in any particular model. Suppose we have a fire insurance company with $2.5M in monthly revenue, and claims following a power law: * Most months, claims are low * Once every ~1 year, they see around $5M in claims * Once every ~1...
4c4ffd6f-6501-4007-90ca-56390a5e3b0d
trentmkelly/LessWrong-43k
LessWrong
Litanies Of The Way A fun index of the concepts from The Sequences in the form of several litanies. These litanies are neither a guide nor an introduction to The Way of Rationality because rationality is an art not a series of precipes. Rather, these litanies serve as a tool; for meditating on, for connecting ideas; o...
41beca06-8f57-4a6e-931a-7078fb5429ec
trentmkelly/LessWrong-43k
LessWrong
Strong intutions. Weak arguments. What to do? I thought Ben Goertzel made an interesting point at the end of his dialog with Luke Muehlhauser, about how the strengths of both sides' arguments do not match up with the strengths of their intuitions: > One thing I'm repeatedly struck by in discussions on these matters w...
6355d126-7989-4440-bd32-1e732fa1b276
trentmkelly/LessWrong-43k
LessWrong
MIT FutureTech are hiring for an Operations and Project Management role. MIT FutureTech are hiring for an Operations and Project Management role.  Please apply or share as relevant. Why apply or share? * Our work to understand progress in computing and artificial intelligence, and its implications, is highly relev...
325bcd8b-01f8-4a73-9d9f-5335f9a5132e
trentmkelly/LessWrong-43k
LessWrong
Save Humanity! Breed Sapient Octopuses! One of the problems that make AI alignment so difficult is that we don't have any models for how non-human intelligences will think. This makes it both hard to reason about how an AGI would behave, but also makes it difficult to get people to take the threat of AI seriously. Id...
0afad1ae-5966-461c-bae1-cff7c7463a76
trentmkelly/LessWrong-43k
LessWrong
Contextual Translations - Attempt 1 Who doesn’t enjoy anything built to their liking? That which is custom-made for themselves and seemingly consistent with what their personal demands are instead of relying on an external factor or agency. I think this analogy fits quite well with what we’re aiming at Dubverse acro...
9fe55390-0b19-4334-b702-ff6de554936b
trentmkelly/LessWrong-43k
LessWrong
Ontological uncertainty and diversifying our quantum portfolio The word "ontology" in the title refers to our conception of the basic building block of reality. In quantum mechanics the ontology is the wave function, in general relativity it is spacetime. This idea in this post assumes the many-worlds interpretation ...
b846ca36-e44f-456c-a70c-23415c89ab48
trentmkelly/LessWrong-43k
LessWrong
I started a blog: Concept Space Cartography For quite a while, I've felt that my mind contains important insights which are getting held up because I don't write enough. So to encourage myself to write more, I've started a blog: Concept Space Cartography. My most recent post there, Two Definitions for Critical Thinkin...
9093ec97-e3b2-4656-b77a-4cb4aa9efd64
LDJnr/LessWrong-Amplify-Instruct
LessWrong
"The Ministry-sanctioned tour of the Chamber of Secrets was pure propaganda. The Marauder's Map showed a second entrance to the Chamber of Secrets hidden in the Girls' Bathroom. "Hss hsssss hsss," Luna copied from Harry Potter. A sink opened up into a large tunnel. "Lumos, Luna said. Each tunnel segment was designed in...
af1915e7-50f1-440f-b21c-7eaa3f0f17cc
trentmkelly/LessWrong-43k
LessWrong
Meetup : Solstice Day-After Megameetup Discussion article for the meetup : Solstice Day-After Megameetup WHEN: 18 December 2016 01:00:00AM (-0500) WHERE: Manhattan - TBD Coming to NYC for the Secular Solstice? Want more time to hangout with all the cool rational-folk people who ALSO came to NY to ALSO do that? And ...
0edbf621-5b35-4010-bcb0-666344301ad8
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
Learning preferences by looking at the world We’ve written up a [blog post](https://bair.berkeley.edu/blog/2019/02/11/learning_preferences/) about our recent [paper](https://openreview.net/forum?id=rkevMnRqYQ) that I’ve been linking to but haven’t really announced or explained. The key idea is that since we’ve optimiz...
9c54f715-a95d-41e8-8dd2-c51f8687d55e
trentmkelly/LessWrong-43k
LessWrong
Advantages of Card Rebalancing Previously: Artifact Embraces Card Balance Changes, Card Collection and Ownership, Card Balance and Artifact, Card Rebalancing, Card Oversupply and Economic Considerations in Digital Card Games VIII. Card Rebalancing and Gameplay Benefits Let us presume we’ve gotten economics out of th...
bdc7b7db-4426-4652-afbf-1e1a33babb8f
trentmkelly/LessWrong-43k
LessWrong
What Does It Mean to "Write Like You Talk"? Audio narration available on the Substack version of this post. People often say to “write like you talk.” Paul Graham has a post titled “Write Like You Talk” where he says explicitly that written language is worse than spoken language because 1. “Written language is more...
0bc1d407-0739-4d5f-9c93-762c9d5ca78d
trentmkelly/LessWrong-43k
LessWrong
New LW Meetup: Newcastle-upon-Tyne This summary was posted to LW main on November 29th. The following week's summary is here. New meetups (or meetups with a hiatus of more than a year) are happening in: * Mumbai Meetup: 15 December 2013 03:00PM * Newcastle-upon-Tyne meetup, December: 07 December 2013 01:00PM Othe...
7cf6bb81-bb7d-4b99-b852-b5ab8dedf2dc
trentmkelly/LessWrong-43k
LessWrong
Meetup : Washington DC TED talks Meetup Discussion article for the meetup : Washington DC TED talks Meetup WHEN: 19 August 2012 04:00:00PM (-0400) WHERE: Washington DC This will be the second TED talks meetup, where we will watch and discuss TED talks. The last one was a lot of fun, and led to some quite interestin...
b11f61fb-4b48-4754-919b-9775c67d351c
trentmkelly/LessWrong-43k
LessWrong
Meetup : 18/11 London Meetup Discussion article for the meetup : 18/11 London Meetup WHEN: 18 November 2012 02:00:00PM (+0000) WHERE: London A meetup at 2 pm on the 18th of November at the Shakespeares's head pub by holborn tube station. Everyone is welcome to attend. Discussion article for the meetup : 18/11 Lo...
7c0e62d9-284d-42ab-9320-c0cdfd80b25d
trentmkelly/LessWrong-43k
LessWrong
[LINK] Concrete problems in AI safety From the Google Research blog: We believe that AI technologies are likely to be overwhelmingly useful and beneficial for humanity. But part of being a responsible steward of any new technology is thinking through potential challenges and how best to address any associated risks. ...
12ff90d4-9af7-49f7-865b-b87922df31b8
trentmkelly/LessWrong-43k
LessWrong
Launch of the New Horizons Podcast Hi all! PhD student here who's been been working on a little side project the past few months and it's finally done - my new podcast on future technologies, New Horizons, has launched!   The topics and style are very much aligned with the interests/values of the rationalist communit...
eff7eabd-d401-4640-b8bf-c48ac4318c04
trentmkelly/LessWrong-43k
LessWrong
Applying Behavioral Psychology on Myself In which I attempt to apply findings from behavioral psychology to my own life. BEHAVIORAL PSYCHOLOGY FINDING #1: HABITUATION The psychological process of "extinction" or "habituation" occurs when a stimulus is administered repeatedly to an animal, causing the animal's respon...
8dcb94f6-b03c-4518-b75c-cf6ac05ff871
StampyAI/alignment-research-dataset/eaforum
Effective Altruism Forum
Seeking advice on impactful career paths given my unique capabilities and interests Hello everyone! I've recently discovered the ideas and philosophy of effective altruism, and I'm very excited about it. I'd like to make sure that I have an impactful career, and I'm writing this post to seek your advice on what I sho...
1d444041-d138-477d-9511-342bf6dc8aa6
trentmkelly/LessWrong-43k
LessWrong
Summary of and Thoughts on the Hotz/Yudkowsky Debate George Hotz and Eliezer Yudkowsky debated on YouTube for 90 minutes, with some small assists from moderator Dwarkesh Patel. It seemed worthwhile to post my notes on this on their own. I thought this went quite well for the first half or so, then things went increas...
35b4af38-6f2c-4c9c-acee-09a854a2ddc6
trentmkelly/LessWrong-43k
LessWrong
Why is there a "clogged drainpipe" effect in idea generation? I've noticed something curious about creativity. (Like when doing the Babble challenges, or  just when thinking in general.) I have a bad idea. ("Send an object to the moon using a bird in a spacesuit") I don't write it down, instead trying to generate ot...
cae06763-1c6e-4c88-946b-ea32de79750c
StampyAI/alignment-research-dataset/lesswrong
LessWrong
AGIs may value intrinsic rewards more than extrinsic ones TLDR: Although AI agent paradigms use explicit reward approaches, the psychology of human motivation suggests that humans value internally generated reward as much if not more than external reward. I suggest that AIs that begin to exhibit behaviors that appear ...
253e744c-d8bd-4d1f-8734-ee8c59e0f47c
trentmkelly/LessWrong-43k
LessWrong
Being the (Pareto) Best in the World The generalized efficient markets (GEM) principle says, roughly, that things which would give you a big windfall of money and/or status, will not be easy. If such an opportunity were available, someone else would have already taken it. You will never find a $100 bill on the floor o...
c5ced96a-a01b-45e1-a9f3-1e6a75f2f4f7
trentmkelly/LessWrong-43k
LessWrong
Positive Book and Other Media Recommendations for a Teen Audience Hi Less Wrong, I've got the opportunity to promote books and other forms of content to a largely teenage audience. I'm looking for some good book recommendations and recommendations for a limited amount of other media (websites, movies, etc) that will ...
3f985a79-4181-4783-ba14-95f3c795ed7b
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
The Alignment Newsletter #4: 04/30/18 **Highlights** -------------- **[Reptile: A Scalable Meta-Learning Algorithm](https://blog.openai.com/reptile/)** *(Alex Nichol et al)*: I somehow forgot to include this in past emails, so I'm including it now. Reptile is an algorithm for meta-learning, and in this paper is appli...
30f14364-6da9-4245-afe2-e521ab508cfb
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
Who models the models that model models? An exploration of GPT-3's in-context model fitting ability Introduction ------------ Much has been written and much has been observed about the abilities of GPT-3 on [many tasks](https://www.lesswrong.com/posts/6Hee7w2paEzHsD6mn/collection-of-gpt-3-results). Most of these cap...
0575b592-b973-48bd-bcfc-9cb504c1eb08
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
AGI will drastically increase economies of scale Note to mods: I'm a bit uncertain whether posts like this one currently belong on the Alignment Forum. Please move it if it doesn't. Or if anyone would prefer not to have such posts on AF, please let me know. In [Strategic implications of AIs’ ability to coordinate at...
1cd87ee9-0273-4f31-86c6-a7695e1dc17f
trentmkelly/LessWrong-43k
LessWrong
Quantified Health Prize results announced   Follow-up to: Announcing the Quantified Health Prize I am happy to announce that Scott Alexander, better known on Less Wrong as Yvain, has won the first Quantified Health Prize, and Kevin Fischer has been awarded second place. There were exactly five entries, so the remain...
2bccba9c-e7ce-4911-9e17-ee2d55999074
trentmkelly/LessWrong-43k
LessWrong
Estimating COVID-19 Mortality Rates In the early days of the pandemic, there wasn't great data available, and it wasn't easy to do better than trusting the standard epidemiological estimate that around 2% of people who got COVID-19 would die. My back of the envelope estimate at the time was way higher, but no one else...
0d545e45-4833-4eda-b6f2-ec7011251a80
StampyAI/alignment-research-dataset/eaforum
Effective Altruism Forum
Discussion about AI Safety funding (FB transcript) Kat Woods recently wrote a [Facebook post](https://www.facebook.com/katxiowoods/posts/pfbid0ngoH3xG2tZAQyoGsCK9oNcDJNPbmmD7VX4BqrPB5GyN2UnbW8dck4xVwkXsDKXgHl) about [Nonlinear's new funding program](https://forum.effectivealtruism.org/posts/Qoecey2umNjcqEGHP/apply-to-...
fb7d5814-d7d9-48b1-b68e-3595b8d67c59
StampyAI/alignment-research-dataset/lesswrong
LessWrong
OpenAI’s Alignment Plan is not S.M.A.R.T. In response to Eliezer Yudkowsky's [challenge](https://www.lesswrong.com/posts/tD9zEiHfkvakpnNam/a-challenge-for-agi-organizations-and-a-challenge-for-1), I will show how the [alignment research approach](https://openai.com/blog/our-approach-to-alignment-research/) outlined by...
f5ea80f6-5f93-4aa0-a37c-f35ac0797650
trentmkelly/LessWrong-43k
LessWrong
Does Diverse News Decrease Polarization? This is the third in a series of posts studying how Facebook's newsfeed affects political polarization. To recap: * In the first post, I showed that polarization trends in the U.S. don't coincide with social media adoption, and argued that Facebook is unlikely to be the prima...
7e788d09-36f5-4260-ba11-872ff7ecfbea
trentmkelly/LessWrong-43k
LessWrong
[SEQ RERUN] Singletons Rule OK Today's post, Singletons Rule OK was originally published on 30 November 2008. A summary (taken from the LW wiki):   > It is possible to create a singleton that won't do nasty things. This may be preferable to a scenario in which many agents start competing for resources without any wa...
c1ea8013-c2e3-4194-a1d4-4d252f790279
trentmkelly/LessWrong-43k
LessWrong
What are the best standardised, repeatable bets? Suppose that I have some budget set aside for philanthropic funding, say $1,000, but I think there are a big returns to scale, so that it would be >1,000x better if I had $1,000,000.[1] What are my best options for some bets I can make to get some chance of turning tha...
28122eb9-d2f0-4c94-9dcd-cf45fc3a4848
trentmkelly/LessWrong-43k
LessWrong
Open Phil releases RFPs on LLM Benchmarks and Forecasting As linked at the top of Ajeya's "do our RFPs accelerate LLM capabilities" post, Open Philanthropy (OP) recently released two requests for proposals (RFPs): 1. An RFP on LLM agent benchmarks:  how do we accurately measure the real-world, impactful capabilities...
71b549da-50da-4f2f-ab38-ff85460794ce
trentmkelly/LessWrong-43k
LessWrong
Were atoms real? Related to: Dissolving the Question, Words as Hidden Inferences.  In what sense is the world “real”?  What are we asking, when we ask that question? I don’t know.  But G. Polya recommends that when facing a difficult problem, one look for similar but easier problems that one can solve as warm-ups.  ...
c16cfd61-a008-423d-9cbf-511c06990a8a
StampyAI/alignment-research-dataset/blogs
Blogs
MIRI’s November Newsletter | | | | --- | --- | | | | | --- | | [Machine Intelligence Research Institute](http://intelligence.org) | | | | | | | --- | --- | | | | | --- | | **Research Updates** * New Friendly AI research area: “[Corrigibility](http://intelligence.org/2014/10/18/new-report-corrigibility...
73767d3b-59fd-46b8-9c76-8d2df3a8e46b
trentmkelly/LessWrong-43k
LessWrong
A summary of every "Highlights from the Sequences" post 1 I recently finished reading Highlights from the Sequences, 49 essays from The Sequences that were compiled by the LessWrong team.  Since moving to Berkeley several months ago, I’ve heard many people talking about posts from The Sequences. A lot of my friends a...
4ddb0a6f-29db-475e-b002-ec292ae8547d
trentmkelly/LessWrong-43k
LessWrong
Features that make a report especially helpful to me Cross-post from EA Forum, follow-up to EA needs consultancies. Below is a list of features that make a report on some research question more helpful to me, along with a list of examples. I wrote this post for the benefit of individuals and organizations from whom ...
572cac08-c952-4b6f-b315-4513653f5fa1
trentmkelly/LessWrong-43k
LessWrong
Preparing for the apocalypse might help prevent it When I tell people that I think there is a decent chance that an unaligned AGI will bring about the apocalypse within the next 20 years or so, they tend to not take it too seriously. Often that's because they think I would act differently if I really assigned a high p...
f04979fb-fa01-41c9-8221-89a6460fe407
StampyAI/alignment-research-dataset/eaforum
Effective Altruism Forum
Should you work in the European Union to do AGI governance? Whether Artificial General Intelligence (AGI) is safe will depend on the [series of actions](https://forum.effectivealtruism.org/posts/qrapebbppHASWB3W9/common-conceptual-framework-to-talk-about-progress-on-agi#1__Series_of_actions) carried out by relevant hu...
56857c55-60b9-4645-90ab-6f13c32fc234
trentmkelly/LessWrong-43k
LessWrong
Imagine a world where minds run on physics This post describes a toy formal model that helps me think about self-modifying AIs, motivationally stable goal systems, paperclip maximizers and other such things. It's not a new result, just a way of imagining how a computer sitting within a world interacts with the world a...
56df9357-130c-4481-87af-7c1bdd870961
StampyAI/alignment-research-dataset/arbital
Arbital
Inverse function If a function $g$ is the inverse of a function $f$, then $g$ undoes $f$, and $f$ undoes $g$. In other words, $g(f(x)) = x$ and $f(g(y)) = y$. An inverse function takes as its [domain](https://arbital.com/p/3js) the [range](https://arbital.com/p/3lm) of the original function, and the [range](https://ar...
2bb71ac2-201d-4242-b128-daea831d7cef
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
Abstraction, Causality, and Embedded Maps: Here Be Monsters In a lot of examples of abstract causality, the abstract causal structure generally mirrors the ground-level causal structure. Think about fluid flow, for instance. At the (classical) ground level, we have a bunch of particles whose interactions are local in ...
3aa5d79a-9029-44e6-8bfb-8252f11a0b65
StampyAI/alignment-research-dataset/eaforum
Effective Altruism Forum
Paul Christiano on how OpenAI is developing real solutions to the 'AI alignment problem', and his vision of how humanity will progressively hand over decision-making to AI systems *This is a linkpost for* [*#44 - Paul Christiano on how OpenAI is developing real solutions to the 'AI alignment problem', and his vision o...
05cdf7a8-efac-4d82-ade4-84d45d8c11b3
trentmkelly/LessWrong-43k
LessWrong
Group rationality diary, 10/15/12 This is the public group instrumental rationality diary for the week of October 15th. It's a place to record and chat about it if you have done, or are actively doing, things like: * Established a useful new habit * Obtained new evidence that made you change your mind about some be...
551c066b-5474-4da2-8e57-08dd5f3714f9
StampyAI/alignment-research-dataset/lesswrong
LessWrong
Recent advances in Natural Language Processing—Some Woolly speculations (2019 essay on semantics and language models) *This essay was written back in 2019, not too long after GPT-2 came out. Although parts of it are dated- mostly the list of new achievements in Natural Language Processing- on the whole, it’s held up r...
adbedd88-0124-4ae9-91a6-e9a16080c461
trentmkelly/LessWrong-43k
LessWrong
Retirement Accounts and Short Timelines Sometimes I talk to people who don't use retirement accounts because they think the world will change enormously between now and when they're older. Something like, the most likely outcomes are that things go super well and they won't need the money, or things go super poorly an...
c3741ac2-7977-4ff3-9bf4-ac917b1cbd38
trentmkelly/LessWrong-43k
LessWrong
Combining individual preference utility functions Note: working on a research agenda, hence the large amount of small individual posts, to have things to link to in the main documents. I've been working on a way of synthesising the preferences of a single given human. After this is done, there is the business of com...
86564205-05b1-4f70-8a5a-e38d7c35c345
trentmkelly/LessWrong-43k
LessWrong
An LLM-based “exemplary actor” Into and summary This post is the second section of "Aligning an H-JEPA agent via training on the outputs of an LLM-based "exemplary actor", posted separately because I think it could warrant a separate discussion, largely independent of the discussion of H-JEPA agent with GFlowNet actor...
5ab50d37-569d-4cd0-a512-87e389a11846
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
[Event] Weekly Alignment Research Coffee Time Just like every Monday now, researchers in AI Alignment are invited for a coffee time, to talk about their research and what they're into. Here is the [link](https://app.gather.town/app/aPVfK3G76UukgiHx/lesswrong-campus).  And here is the [everytimezone time](https://eve...
4e350ffd-ccb0-4273-9915-538f861b3c88
trentmkelly/LessWrong-43k
LessWrong
You must not fool yourself, and you are the easiest person to fool The title of this post comes from Richard Feynman's famous talk about cargo cult science. The principle is very broadly applicable, though. Applying the steps I’ve described in this sequence can feel very satisfying, and even give you a sense that you’...
4f98e9b4-e797-4732-b30e-8156c419beca
StampyAI/alignment-research-dataset/blogs
Blogs
The Financial Times story on MIRI Richard Waters wrote a [story](http://www.ft.com/intl/cms/s/2/abc942cc-5fb3-11e4-8c27-00144feabdc0.html) on MIRI and others for *Financial Times*, which also put Nick Bostrom’s *Superintelligence* at the top of its [summer science reading list](http://www.ft.com/intl/cms/s/2/31a97c56-...
b159d3ad-96ec-45cd-ac84-17d84e1e5f11
StampyAI/alignment-research-dataset/lesswrong
LessWrong
"Human-level control through deep reinforcement learning" - computer learns 49 different games [full text](http://www.readcube.com/articles/10.1038%2Fnature14236?shared_access_token=Lo_2hFdW4MuqEcF3CVBZm9RgN0jAjWel9jnR3ZoTv0P5kedCCNjz3FJ2FhQCgXkApOr3ZSsJAldp-tw3IWgTseRnLpAc9xQq-vTA2Z5Ji9lg16_WvCy4SaOgpK5XXA6ecqo8d8J7l...
0206ca21-b84e-4510-9c28-d4a60ccf644b
trentmkelly/LessWrong-43k
LessWrong
Meetup : Pittsburgh Meetup: Big Gaming Fun 3! Discussion article for the meetup : Pittsburgh Meetup: Big Gaming Fun 3! WHEN: 29 January 2012 01:00:00PM (-0500) WHERE: 1324 Wightman St., Pittsburgh, PA 15217 You can see my game collection here; please bring anything else you'd like to play. We can order food and go a...
d51539a2-e8e7-441c-926d-6cd3722f929f
trentmkelly/LessWrong-43k
LessWrong
Ordinary and unordinary decision theory There are a couple of paradoxes in decision theory, typically involving beings making perfect predictions, copies of yourself, and similar shenanigans. If you're unfamiliar with all of this, take a look at Newcomb's problem. Attempted resolutions of these paradoxes usually invol...
6a90afa9-98bc-4ac7-9683-63446159667d
trentmkelly/LessWrong-43k
LessWrong
Exponential Economist Meets Finite Physicist [link] A dialogue discussing how thermodynamics limits future growth in energy usage, and that in turn limits GDP growth, from the blog Do the Math. > Physicist: Hi, I’m Tom. I’m a physicist. > > Economist: Hi Tom, I’m [ahem..cough]. I’m an economist. > > Physicist: Hey,...
04ea7d39-ab7c-4744-9198-3a036eef883e
trentmkelly/LessWrong-43k
LessWrong
Grammatical Roles and Social Roles: A Structural Analogy We often describe society in terms of roles: leaders, workers, users, creators. A linguistic analogy can shed new light on these structures: the grammar of a sentence. In grammar, we typically find: The subject: the actor, the one performing the action. The v...
c759fa0a-eb8a-445a-85d3-d292a9b11850
trentmkelly/LessWrong-43k
LessWrong
Are coin flips quantum random to my conscious brain-parts? Hello rationality friends!  I have a question that I bet some of you have thought about... I hear lots of people saying that classical coin flips are not "quantum random events", because the outcome is very nearly determined by thumb movement when I flip the ...
9f3e11cc-574a-4c63-bf4d-89790301d3a8
trentmkelly/LessWrong-43k
LessWrong
How to openly maintain a single identity while keeping it private? Whenever I need to create an account on some platform I've not used before, I'm almost always[1] faced with a dilemma regarding my public username. I can simply use one that I already use elsewhere or I can come up with a new, unique one. Both, in my ...
cf1e1505-f221-460c-a609-ab271198dfe9
trentmkelly/LessWrong-43k
LessWrong
EY in the New York Times The Paper of Record had a good article today about existential AI risk, mentioning EY: > In the early 2000s, a young writer named Eliezer Yudkowsky began warning that A.I. could destroy humanity. His online posts spawned a community of believers. Called rationalists or effective altruists, th...
9395f52f-f8ac-4b4d-96cb-b9f7e128c3a7
trentmkelly/LessWrong-43k
LessWrong
What do the Charter Cities Institute likely mean when they refer to long term problems with the use of eminent domain? Without using eminent domain, a large chunk of the possible future value goes to surrounding land-owners who may have done little or nothing to create that value. It does not seem economically possibl...