id
stringlengths
36
36
source
stringclasses
15 values
formatted_source
stringclasses
13 values
text
stringlengths
2
7.55M
6cf7c9e1-76ca-4600-aa6e-09627b41515d
trentmkelly/LessWrong-43k
LessWrong
[LINK] OpenAI doing an AMA today The OpenAI research team is doing a Reddit AMA today! A good opportunity to ask them questions about AI safety and machine learning. 
8120c8ff-0a8b-40fb-ab51-11af090e2962
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
What are the best precedents for industries failing to invest in valuable AI research? The efficient market hypothesis applied to AI is an important variable for timelines. The idea is: If AGI (or TAI, or whatever) was close, the big corporations would be spending a lot more money trying to get to it first. Half of th...
5f364c9d-439d-475a-89b0-5d52b70f7933
trentmkelly/LessWrong-43k
LessWrong
Meetup : April Rationality Discussion: Life planning vs. the coming technology explosion. Discussion article for the meetup : April Rationality Discussion: Life planning vs. the coming technology explosion. WHEN: 05 April 2015 04:30:00PM (+1100) WHERE: Ross House Association, 247-251 Flinders Lane, Melbourne WH...
d576981f-99c2-495a-a9af-98516f31e792
trentmkelly/LessWrong-43k
LessWrong
[SEQ RERUN] Rationality: Common Interest of Many Causes Today's post, Rationality: Common Interest of Many Causes was originally published on 29 March 2009. A summary (taken from the LW wiki):   > Many causes benefit particularly from the spread of rationality - because it takes a little more rationality than usual ...
bbfb1c9e-1687-4df6-b9c9-6359dc220f74
trentmkelly/LessWrong-43k
LessWrong
[SEQ RERUN] What Do We Mean By "Rationality"? Today's post, What Do We Mean By "Rationality"? was originally published on 16 March 2009. A summary (taken from the LW wiki):   > When we talk about rationality, we're generally talking about either epistemic rationality (systematic methods of finding out the truth) or ...
98ed6a97-74a5-4de7-901a-3339676e6ae1
trentmkelly/LessWrong-43k
LessWrong
Program good ethics into artificial intelligence
bbafa893-b8a3-4bac-b137-780df5135a77
trentmkelly/LessWrong-43k
LessWrong
Beeminding Sin (This is something that I originally posted to the CFAR Alumni list, and then fleshed out a bit and posted to the Beeminder blog. It has been well received and there have been suggestions that I post it here, so here it is.) For a long time I found that I was spending too much time on certain unproduct...
51074e11-7b28-4634-ad94-55013fa3d670
trentmkelly/LessWrong-43k
LessWrong
Meetup : Chicago Meetup Discussion article for the meetup : Chicago Meetup WHEN: 09 July 2011 02:30:00PM (-0500) WHERE: 1121 North State Street, Chicago, IL Airedale and I will be hosting the next meetup at the Corner Bakery on 1121 North State Street. Look for a sign saying LW. We hope to see you there! Discussi...
44fc8ee6-1373-4ec5-a89c-88b99fac4797
trentmkelly/LessWrong-43k
LessWrong
Final Day to Order LW Books by Christmas for US Today is the last day we are guaranteeing delivery-before-Christmas to those who preorder the LessWrong books to North America (i.e. US, Canada, Mexico). Order them at lesswrong.com/books by 12pm midnight Pacific time to still be included. Thank you so much to everyone...
dd91211d-169b-4cc7-9be6-109c04074520
trentmkelly/LessWrong-43k
LessWrong
AXRP Episode 30 - AI Security with Jeffrey Ladish YouTube link Top labs use various forms of “safety training” on models before their release to make sure they don’t do nasty stuff - but how robust is that? How can we ensure that the weights of powerful AIs don’t get leaked or stolen? And what can AI even do these da...
f4e1c2be-2b5f-414b-b3ca-8087110f22da
trentmkelly/LessWrong-43k
LessWrong
How well defined is ADHD? I've long had attention and focus problems, but never explored the possibility that I have ADHD till recently. I understand that it's a standard term, but I'm still a bit suspicious; Psychiatry doesn't seem like the most reliable field. Are there good reasons for picking out the behaviors as...
d49b9a1c-b297-4377-a17b-d3150b1f9c03
trentmkelly/LessWrong-43k
LessWrong
Link: NY Times covers Bayesian statistics http://www.nytimes.com/2011/01/11/science/11esp.html?_r=1&src=me&ref=homepage It's disguised as an article about ESP to fool their editors; scroll down two paragraphs and it goes on for quite a while about what Bayesian statistics are and why Bayesian analysis is important. ...
5f2f0e99-bc97-474f-91af-b2253186148c
trentmkelly/LessWrong-43k
LessWrong
[Aspiration-based designs] 1. Informal introduction Sequence Summary. This sequence documents research by SatisfIA, an ongoing project on non-maximizing, aspiration-based designs for AI agents that fulfill goals specified by constraints ("aspirations") rather than maximizing an objective function . We aim to contribu...
0fda7e1d-a907-4601-9107-31d1bc234e52
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
SolidGoldMagikarp (plus, prompt generation) UPDATE (14th Feb 2023): ChatGPT appears to have been patched! However, very strange behaviour can still be elicited in the [OpenAI playground](https://platform.openai.com/playground), particularly with the davinci-instruct model. More technical details [here](https://www.le...
e6ad2608-83d5-4525-899d-a32d6264d26e
trentmkelly/LessWrong-43k
LessWrong
[New Feature] Your Subscribed Feed tl;dr LessWrong now has a Subscribed tab (next to the Latest tab and Enriched tab[1]). You can now "follow" users, which means their posts and comments will show up in your Subscribed tab[2].  You can subscribe to users that are suggested to you, or search for other users using the ...
82bb0956-6835-4a6c-bc31-cb09b0fcabff
trentmkelly/LessWrong-43k
LessWrong
Each Llama3-8b text uses a different "random" subspace of the activation space This is kind of a null result (or WIP research) I got with a few days of fiddling, so don't get too excited. Also, because it's a null result, it's always conceivable that there's just some slight change in the approach which could suddenly...
e40ec24f-eedb-4d6a-9b5c-0b4e6bb6bda9
trentmkelly/LessWrong-43k
LessWrong
A Disneyland Without Children The spaceship swung into orbit around the blue-grey planet with a final burn of its engines. Compared to the distance they had travelled, the world, now only some four hundred kilometres below and filling up one hemisphere of the sky, was practically within reach. But Alice was no less co...
2a1da93b-9f67-41d7-8fed-f6bbf430fecf
trentmkelly/LessWrong-43k
LessWrong
Is value amendment a convergent instrumental goal? Goals such as resource acquisition and self-preservation are convergent in that they occur for a superintelligent AI for a wide range of final goals. Is the tendency for an AI to amend its values also convergent? I'm thinking that through introspection the AI would ...
48db3b4a-96ec-4770-a02a-39ba88927540
trentmkelly/LessWrong-43k
LessWrong
Is it Legal to Maintain Turing Tests using Data Poisoning, and would it work? As AI gets more advanced, it is getting harder and harder to tell them apart from humans. AI being indistinguishable from humans is a problem both because of near term harms and because it is an important step along the way to total human di...
6c5eb4b8-a2b5-4d3b-b946-6294f57e12ef
trentmkelly/LessWrong-43k
LessWrong
Pascal's Mugging: The Word Wars A Long Time From Now, in a Galaxy Far, Far Away: Episode 1: The Phantom Mugger One night, on your way home from work, you find yourself face to face with a mysterious shadowy figure. “Your money or your life,” he hisses through his teeth. The tip of a knife pokes uncomfortably into yo...
a10dbaef-c685-4e29-af70-d6f8e09a74b7
StampyAI/alignment-research-dataset/blogs
Blogs
Transformer Circuit Videos d-byline {display: none;} As an experiment, we recorded a couple [videos](https://www.youtube.com/playlist?list=PLoyGOS2WIonajhAVqKUgEMNmeq3nEeM51) discussing our early stage thinking on trying to reverse engineer neural networks. We made them to share our very informal thoughts with colleag...
036d401c-83e7-434b-9dbd-6bdfb1e49d57
StampyAI/alignment-research-dataset/eaforum
Effective Altruism Forum
First call for EA Data Science/ML/AI Hi there!  For the past years I’ve been involved with the Italian Effective Altruism community while working as data science and involved in many data initiatives/trainings for the private and public sectors, mentoring non-profit to be more data and evidence-driven, and contributi...
ec26cd8a-2123-43d3-a6b0-4b1d207359e5
trentmkelly/LessWrong-43k
LessWrong
Embedding safety in ML development This post was written as part of Refine. Thanks to Adam Shimi, Alexander Gietelink Oldenziel, and Vanessa Kosoy for helpful discussion and feedback.  Summary This post aims to: * Advocate for embedding safety into development of machine learning models * Propose a framing on how...
92a5f8ae-c284-4cf4-9e7e-9b619335e097
StampyAI/alignment-research-dataset/arbital
Arbital
Disjoint cycle notation is unique
71aacb60-03c2-4c41-8023-6cc841bbeac3
trentmkelly/LessWrong-43k
LessWrong
Concept Safety: What are concepts for, and how to deal with alien concepts In The Problem of Alien Concepts, I posed the following question: if your concepts (defined as either multimodal representations or as areas in a psychological space) previously had N dimensions and then they suddenly have N+1, how does that af...
cb7aa204-8ef4-404c-aa59-b0a8b66ae0f2
trentmkelly/LessWrong-43k
LessWrong
[SEQ RERUN] A Priori Today's post, A Priori was originally published on 08 October 2007. A summary (taken from the LW wiki):   > The facts that philosophers call a priori arrived in your brain by a physical process. Thoughts are existent in the universe; they are identical to the operation of brains. The "a priori" ...
1805e4b7-5884-4e6c-a62a-e31177020d3a
trentmkelly/LessWrong-43k
LessWrong
Rock-Paper-Scissors Can Be Weird Let's consider a symmetrical, two-player, zero-sum game, where each player must simultaneously choose one of three moves: A, B, or C. After each player choses their move, a winner is randomly selected with odds based on what the two players selected. (Note that because its symmetrical,...
6f72617f-c4d6-4f11-b216-6fe269afb35d
trentmkelly/LessWrong-43k
LessWrong
A test for symbol grounding methods: true zero-sum games Imagine there are two AIs playing a debate game. The game is zero-sum; at the end of the debate, the human judge assigns the winner, and that AI gets a +1 reward, while the other one gets a −1. Except the game, as described, is not truly zero-sum. That is becau...
2eaced6f-30ef-472b-9f96-3f2cfdbc74f4
trentmkelly/LessWrong-43k
LessWrong
When does an AI become intelligent enough to become self-aware and power-seeking? Do AIs currently display goal-directed behavior? What do we expect happens in a superintelligence which makes it display goal-directed behavior?
3512afd6-5a96-4e98-bc07-33e3e33e797a
trentmkelly/LessWrong-43k
LessWrong
Property as Coordination Minimization A friend recently noted that they were in favor of private property, but the best defense they had to link was instead a defense of finance. So I thought I’d give it a try. In light of a distinction people often draw between ‘private property’ and ‘personal property’, I’m going to...
8f25fa44-6e7f-4c66-8e4f-6d4e3c76f62e
trentmkelly/LessWrong-43k
LessWrong
Meetup : Canberra: Liar's Dice! Discussion article for the meetup : Canberra: Liar's Dice! WHEN: 28 November 2014 06:00:00PM (+1100) WHERE: 108 North Road, Acton, ACT We will be practising our practical rationality skills by playing (some variant of) Liar's Dice. As always, vegan snacks will be provided. General m...
d69d3c1c-3896-474a-aeb0-d8f6745d1e9d
trentmkelly/LessWrong-43k
LessWrong
Instrumental Rationality 6: Attractor Theory [Instrumental Rationality Sequence 6/7] [Attractor Theory is a hybrid model that tries to reconcile the effects of internal and external factors of motivation. It makes the claim that an important additional consideration in decision-making is how the action affects your a...
52d6aed4-ede1-494a-8bd0-e4dfdab5f0a7
StampyAI/alignment-research-dataset/eaforum
Effective Altruism Forum
AGI and Lock-In The long-term future of intelligent life is currently unpredictable and undetermined. In the linked document, we argue that the invention of artificial general intelligence (AGI) could change this by making extreme types of lock-in technologically feasible. In particular, we argue that AGI would make i...
88548078-6caa-41d5-a727-cdeedcc6b691
StampyAI/alignment-research-dataset/lesswrong
LessWrong
A gentle primer on caring, including in strange senses, with applications Introduction ============ Firstly, I will define a straightforward model of caring about other people. I think it is a good model for understanding ([and](https://www.lesswrong.com/posts/ex63DPisEjomutkCw/msg-len) predicting) people's actions. ...
e416f4ab-2e93-418a-aa76-07031e124295
trentmkelly/LessWrong-43k
LessWrong
Blackpool Applied Rationality Unconference 2025 TL;DR Join us for 4 days of applied rationality workshops and activities at the EA Hotel this April. Apply here by  22nd February. In April we’re hosting an intimate 4-day applied rationality unconference/retreat hosted at CEEALAR (aka the EA Hotel). Come along and make...
74936b22-2661-43e7-9932-d8779c25c269
trentmkelly/LessWrong-43k
LessWrong
Meetup Cookbook My spouse and I have been running LessWrong meetups in two different cities for the past six years. Over time, we've gotten lazier and lazier about organizing, while still maintaining similar results, mostly by coming up with a bunch of simple recipes and scripts for running meetups. Here I have docume...
4268e2e3-ed78-4629-b00c-04443f6a595a
trentmkelly/LessWrong-43k
LessWrong
EndeavorOTC legit? Curious to hear anyone's experience with EndeavorOTC, or opinion on it's effectiveness for treating ADHD.
1f89b84d-a071-4f8c-8519-fb827ae6bc2a
StampyAI/alignment-research-dataset/blogs
Blogs
What do ML researchers think you are wrong about? *By Katja Grace, 25 September 2017* So, maybe you are concerned about AI risk. And maybe you are concerned that many people making AI are not concerned enough about it. Or not concerned about the right things. But if so, do you know why they disagree with you? We d...
806338d3-a824-405c-ae0c-88f18366830f
trentmkelly/LessWrong-43k
LessWrong
AI safety reading group in Madrid We are going to set up a reading group in Madrid (Spain), focused on (semi-)technical papers from MIRI and others about AI safety. The goal is to gain a deeper understanding of current work in this area, and perhaps make some contributions in the future. We have tentatively set our f...
460bb2b4-5640-40ec-845a-871dd84945e1
StampyAI/alignment-research-dataset/special_docs
Other
X-Risk Motivations for Safety Research Directions Open Problems in AI X-Risk ========================== Dan Hendrycks (an author of this post), Nicholas Carlini, John Schulman, and Jacob Steinhardt previously wrote [Unsolved Problems in ML Safety](https://www.google.com/url?q=https://arxiv.org/abs/2109.13916&sa=D&sou...
c8f53242-9379-4e45-8259-e543c5073413
trentmkelly/LessWrong-43k
LessWrong
Thoughts on "Ontological Crises" (Originally posted at https://plus.google.com/+ThomasColthurst/posts/5fcw5wgVrpj in 2013) I love the premise of Peter de Blanc's "Ontological Crises in Artificial Agents’ Value Systems":  you are robot who used to think the universe was described by some set of states O1, only to wake...
b47dd494-7021-4d95-9a65-66c3ba039177
trentmkelly/LessWrong-43k
LessWrong
Meetup : Sydney Rationality Dojo - Intuitive Bayes Discussion article for the meetup : Sydney Rationality Dojo - Intuitive Bayes WHEN: 07 June 2015 04:00:00PM (+1000) WHERE: Humanist House, 10 Shepherd St Chippendale This month's dojo will involve running through exercises aiming to help you gain an intuitive sense...
8d96db7c-84dd-4713-9432-4322269cd12e
StampyAI/alignment-research-dataset/blogs
Blogs
AGI in a vulnerable world *By Asya Bergal, 25 March 2020* I’ve been thinking about a class of AI-takeoff scenarios where a very large number of people can build dangerous, unsafe AGI before anyone can build safe AGI. This seems particularly likely if: * It is considerably more difficult to build safe AGI than it i...
f5645eb6-12b2-46b5-96c1-80de8328d8c1
trentmkelly/LessWrong-43k
LessWrong
A Library and Tutorial for Factored Cognition with Language Models We want to advance process-based supervision for language models. To make it easier for others to contribute to that goal, we're sharing code for writing compositional language model programs, and a tutorial that explains how to get started: * The In...
dcb65d1c-67a2-4667-9ebc-4885f518293f
trentmkelly/LessWrong-43k
LessWrong
[Link] An argument for Low-hanging fruit in Medicine Those of us who have found the arguments for stagnation in our near future by Peter Thiel and Tyler Cowen pretty convincing, usually look only to the information and computer industries as something that is and perhaps even can keep us afloat. On the excellent West...
f1b2e261-00fd-4082-a15d-51229dcee4f6
trentmkelly/LessWrong-43k
LessWrong
Debugging Writer's Block The hardest part of writing for me is dealing with writer's block. It appears almost every time I sit down to write–I'll have everything ready: the time, the space, and the super exciting idea I want to share. But right then, the words won't come! And when they eventually do, they come out in ...
12769948-291f-43a0-b2f7-dd715d5d596f
trentmkelly/LessWrong-43k
LessWrong
A simple technique to reduce negative rumination I have faced challenges with negative rumination and anxiety, especially when I am trying to relax or my mind is idle. However, I have recently discovered a simple and effective technique to overcome this. I practice enumerating as many positive future events as I can,...
e632dc4d-8268-444e-8385-a4362fd3f5c6
trentmkelly/LessWrong-43k
LessWrong
Impact measurement and value-neutrality verification Recently, I've been reading and enjoying Alex Turner's Reframing Impact sequence, but I realized that I have some rather idiosyncratic views regarding impact measures that I haven't really written up much yet. This post is my attempt at trying to communicate those v...
f5802bd6-de34-4658-8c26-ee85c0bedc11
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
Extracting and Evaluating Causal Direction in LLMs' Activations *This post was written by Fabien at SaferAI*[[1]](#fnw9a0hn8g8b)*. Simeon has prompted Fabien in relevant directions and has provided valuable feedback.* *Thanks to Jean-Stanislas Denain, Alexandre Variengien, Charbel-Raphael Segerie, and Nicole Nohemi f...
a2b26d4e-5e07-4069-8841-225b527d73d3
trentmkelly/LessWrong-43k
LessWrong
Epistemic Luck Who we learn from and with can profoundly influence our beliefs. There's no obvious way to compensate.  Is it time to panic? During one of my epistemology classes, my professor admitted (I can't recall the context) that his opinions on the topic would probably be different had he attended a different g...
e87892e2-2d8d-45f5-ad9d-e01174e52ed1
trentmkelly/LessWrong-43k
LessWrong
The generalized Sierpinski-Mazurkiewicz theorem. Suppose any uncountable set S and any countable set of bijective transformations T={T1,T2,...} ∀i  Ti:S→S bijective. Then assuming that for all pairs of distinct finite sequences of transformations [p1,p2,...pm]  and [q1,q2...qn] that the set {s∈S:p1∘p2..pm(s)=q1∘q2..q...
9a17f432-0679-4608-a5c6-9c0a3e48cc09
trentmkelly/LessWrong-43k
LessWrong
Meetup : MIRIxAtlanta - MIRI Research Guide + Corrigibility Discussion article for the meetup : MIRIxAtlanta - MIRI Research Guide + Corrigibility WHEN: 22 November 2014 06:00:00PM (-0500) WHERE: 2388 Lawrenceville Hwy. Unit L. Decatur, GA 30033 We'll go over the new research guide http://intelligence.org/research-...
afdb3055-34cf-492b-90c6-360c358224fc
StampyAI/alignment-research-dataset/eaforum
Effective Altruism Forum
Thoughts on AGI organizations and capabilities work (*Note: This essay was largely written by Rob, based on notes from Nate. It’s formatted as Rob-paraphrasing-Nate because (a) Nate didn’t have time to rephrase everything into his own words, and (b) most of the impetus for this post came from Eliezer wanting MIRI to* ...
a30f1bba-535c-46d5-9f71-6f2c266dd17c
trentmkelly/LessWrong-43k
LessWrong
Are the social sciences challenging because of fundamental difficulties or because of imposed ones? Scholarship Status: I'm not an expert in the social sciences. I've taken a few in-person and online courses and read several books, but I’m very much speculating here. I also haven’t done much investigation for this sp...
3f692eb2-8aa8-4f45-9cd9-87bcecb3364f
StampyAI/alignment-research-dataset/agisf
AGI Safety Fund
International Institutions for Advanced AI Executive Summary ----------------- Recent advances in AI have highlighted the potentially transformative impacts of advanced systems.111By “advanced AI systems” we mean systems that are highly capable and general purpose. International institutions may have an important r...
f49c50be-d348-4bb1-96c3-8c5fc105c884
trentmkelly/LessWrong-43k
LessWrong
Instead of technical research, more people should focus on buying time This post is the first in a sequence of posts about AI strategy. In the next post, we’ll provide more examples of “buying time” interventions that we’re excited about.  We’re grateful to Ajeya Cotra, Daniel Kokotajlo, Ashwin Acharya, and Andrea Mi...
91523b89-bcaf-4584-8982-ffe75a104256
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
Japan AI Alignment Conference [Conjecture](https://www.conjecture.dev/) and [ARAYA](https://www.araya.org/en/) are hosting and organizing the first Japan AI Alignment Conference. The conference will take place in Tokyo, Japan on March 11 and 12. Details about the event can be found [here](https://jac2023.ai/). This ev...
f519f9a5-c189-4a7e-8506-2d87325a5a9b
StampyAI/alignment-research-dataset/lesswrong
LessWrong
Applications of logical uncertainty Lately I've been reading a lot about the problem of logical uncertainty. The problem is that there are logical consequences of your beliefs that you're uncertain about. So, for statements like "the billionth digit of pi is even", it's provable from your beliefs. But you're still u...
27f61ebc-a802-4684-8ef4-ddfcc0cfa788
trentmkelly/LessWrong-43k
LessWrong
Another Non-Anthropic Paradox: The Unsurprising Rareness of Rare Events This is the fifth post in my series on Anthropics. The previous one is Anthropical Paradoxes are Paradoxes of Probability Theory. The next one is Why Two Valid Answers Approach is not Enough for Sleeping Beauty. Intoduction As all anthropic prob...
86ccff4c-dd1b-4278-805a-c737a106170f
trentmkelly/LessWrong-43k
LessWrong
Safety in Machine Learning In this week's AI alignment newsletter was a link to a paper titled "Safe Policy Learning from Observations". The paper uses the word "safe" or "safety" to describe ML systems 34 times, but reading through I didn't get a very clear sense of what the authors mean by "safe" or why "safe" was a...
2b1716de-5413-4dbd-a6c9-4c6fbedd2629
StampyAI/alignment-research-dataset/arxiv
Arxiv
Emergent Complexity and Zero-shot Transfer via Unsupervised Environment Design. 1 Introduction --------------- Many reinforcement learning problems require designing a distribution of tasks and environments that can be used to evaluate and train effective policies. This is true for a diverse array of methods includ...
6e5d634e-58d3-4632-a36a-2f092bdebf21
trentmkelly/LessWrong-43k
LessWrong
Unethical Human Behavior Incentivised by Existence of AGI and Mind-Uploading I need help getting out of a logical trap I've found myself in after reading The Age of Em.   Some statements needed to set the trap: If mind-uploading is possible, then a mind can theoretically exist for an arbitrary length of time. If a...
e2667eb5-5049-4939-894c-2b190e83cf90
trentmkelly/LessWrong-43k
LessWrong
Therapy vs emotional support There's a misconception that therapy is like paying someone to be your friend. While most therapists would disagree with that statement, there is a kernel of truth; therapy can often focus on emotional support, and getting those needs met. I sometimes joke that when friends need help, we s...
56c3ad7b-63cb-4f1b-a57d-3f69f9a35203
trentmkelly/LessWrong-43k
LessWrong
Moral enhancement In offline conversations, several Less Wrongers have expressed interest in 'benevolence training'. For those who are interested, here's a head start in the research... The technical term for this is 'moral enhancement.' The key name here is Thomas Douglas, who wrote his Ph.D. thesis on the topic, an...
3a704f5f-b1ae-472e-9a14-82eb091b353f
trentmkelly/LessWrong-43k
LessWrong
Book Review: Scientific Freedom by Don Braben What if tech stagnation, declining growth rates, and the near-inevitable seeming collapse of the West are all because we got worried a few scientists would run off with our tax dollars?  That’s the broad thesis behind Scientific Freedom: The Elixir of Civilization. Publis...
3a5f5fa0-f2a3-41d5-b3cf-74fa867b3bd4
StampyAI/alignment-research-dataset/eaforum
Effective Altruism Forum
How might a herd of interns help with AI or biosecurity research tasks/questions? **Tl;dr:** Are there any research tasks/projects in AI or biosecurity that could potentially benefit from having a large group of lightly-supervised (but preferably paid) interns work on the problem? (E.g., collecting and codifying data ...
6c0e2411-9d4f-4756-af2f-9240811a260f
trentmkelly/LessWrong-43k
LessWrong
The Singularity Institute is hiring remote HTML / WordPress workers! The Singularity Institute needs to hire remote HTML & WordPress workers. (For example, we need to move dozens of Singularity Summit talk transcripts to our new WordPress site and fix any formatting errors that result from that move.) This requires ba...
97e82169-3ae9-45fb-9d64-e7a910003932
trentmkelly/LessWrong-43k
LessWrong
100 Dinners And A Workshop: Information Preservation And Goals Produced on a grant from the LTFF. Thanks to Justis for proofreading and Paul Colognese, David Udell, Alex Altair, Nicky Pochinkov, and Jessica Rumbelow for discussion and feedback.  I would like to reason about the emergence and impact of an AGI on our l...
ff1ee6f8-2013-418d-b1b8-32e2f7cd97e4
trentmkelly/LessWrong-43k
LessWrong
Internal music player: phenomenology of earworms Earworms consist of songs mentally playing back beyond when I'm hearing them, and they have long plagued me, to the point that they were the main factor in my judgment that music is harmful. More recently, I've noticed patterns in how these earworms work, and some power...
1c1aebc1-8df1-4b7b-9f76-cecc5258e462
trentmkelly/LessWrong-43k
LessWrong
You See The Territory or Nothing at All Crossposted from here. This post is about the philosophy of perception, what makes a philosophical debate substantive, and cognitive qualia. Consider two views about the nature of perception and the world: on indirect realism we perceive a veil, we only indirectly get at the w...
95ff8738-3bbc-4465-8226-1b71304770ca
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
Garrabrant and Shah on human modeling in AGI This is an edited transcript of a conversation between Scott Garrabrant (MIRI) and Rohin Shah (DeepMind) about whether researchers should focus more on approaches to AI alignment that don’t require highly capable AI systems to do much human modeling. CFAR’s Eli Tyre facilit...
6345a77a-3ec3-4921-9a6c-7612d4d9b8fe
trentmkelly/LessWrong-43k
LessWrong
3 entirely different things we call “Time” Confusion about the reality of time comes from the fact that there are three entirely different things we call “time.” When we ask a physicist if time is real, we would get an answer about an entirely different phenomenon compared to the answer to the same question of a philo...
01a56c71-4912-463e-b24d-82e0e5464f35
trentmkelly/LessWrong-43k
LessWrong
Concept Safety: World-models as tools The AI in the quantum box In the previous post, I discussed the example of an AI whose concept space and goals were defined in terms of classical physics, which then learned about quantum mechanics. Let's elaborate on that scenario a little more. I wish to zoom in on a certain a...
1945e016-a0b2-45dc-ac1c-b2ff69c13cff
trentmkelly/LessWrong-43k
LessWrong
My AGI safety research—2022 review, ’23 plans “Our greatest fear should not be of failure, but of succeeding at something that doesn't really matter.”  –DL Moody (allegedly) The short version: In this post I’m briefly summarizing how I spent my work-time in 2022, and what I’m planning for 2023. * The first half of ...
3a9615f6-eff2-450c-b883-d6cf424680a5
trentmkelly/LessWrong-43k
LessWrong
Fix simple mistakes in ARC-AGI, etc. ARC-AGI is a diverse artificial dataset that aims to test general intelligence. It's sort of like an IQ test that's played out on rectangular grids. Last month, @ryan_greenblatt proposed an approach that used GPT-4o to generate about 8000 Python programs per task. It then selected...
7a85d8f4-75fb-4f95-b760-c7691d520058
trentmkelly/LessWrong-43k
LessWrong
Anyone with the medical knowledge to evaluate an extraordinary claim? In a different forum I frequent ( The Ornery American ), a regular member there (LetterRip) has recently been making an extraordinary claim - a new theory of medicine he has devised that relates and can contribute in the cure of several neurological...
fb7fd2b3-032d-4f78-a7e3-e4d28fef34d8
StampyAI/alignment-research-dataset/agisf
AGI Safety Fund
Understanding intermediate layers using linear classifier probes 1 Introduction --------------- The recent history of deep neural networks features an impressive number of new methods and technological improvements to allow the training of deeper and more powerful networks. The model themselves had a reputation...
dbef02e9-0337-4f5e-a3f0-6c66efac16cb
trentmkelly/LessWrong-43k
LessWrong
Singularity-oriented Sci-Fi collection to be published [link] Sorry for the blogspammy link, it's the best summary I found. http://upcoming4.me/media-news/book-news/item/8768-contents-of-digital-rapture-the-singularity-anthology-announced Of interest is apparent inclusion of Nick Bostrom in the collection (not sure ...
266ec4ad-4065-4b04-b423-e4e16b6e3652
trentmkelly/LessWrong-43k
LessWrong
Meetup : West LA: Linguistic Relativity Discussion article for the meetup : West LA: Linguistic Relativity WHEN: 19 November 2014 07:00:00PM (-0800) WHERE: 11066 Santa Monica blvd, LA, CA, 90025 How to Find Us: Go into this Del Taco. We will be in the back room if possible. Parking is free in the lot out front or ...
2b75edbc-960b-4964-ac5d-ac764094c679
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
We Found An Neuron in GPT-2 We started out with the question: How does GPT-2 know when to use the word `"an"` over `"a"`? The choice depends on whether the word that comes after starts with a vowel or not, but GPT-2 can only output one word at a time. We still don’t have a full answer, but we did find a single MLP ne...
da08f0de-2925-442c-8b3f-a66f196e7175
StampyAI/alignment-research-dataset/lesswrong
LessWrong
Pausing AI Developments Isn't Enough. We Need to Shut it All Down by Eliezer Yudkowsky [New article](https://time.com/6266923/ai-eliezer-yudkowsky-open-letter-not-enough/) in [Time Ideas](https://time.com/section/ideas/) by Eliezer Yudkowsky. Here’s some selected quotes. In reference to the [letter](https://futureof...
8eea1edf-5a56-4f35-bbee-679cb16bb390
trentmkelly/LessWrong-43k
LessWrong
Call for Volunteers I've recently been appointed Program Coordinator of Humanity+ (http://www.humanityplus.org), which is another big organization in the SIAI/FHI/Kurzweilian/futurist space. If you're interested in volunteering to help us out, you can contact me at pphysics141@gmail.com. If you're interested in volunt...
e024d13b-d316-4e57-a14e-84b9a5f1da0f
trentmkelly/LessWrong-43k
LessWrong
Appeal to Consequence, Value Tensions, And Robust Organizations Epistemic Status: Strong opinions weakly held. Mostly trying to bring some things into the discourse that I think are too often ignored. Some updates I've made based on the discussion in this post are here . Introduction Jessicata's Dialogue on Appeals...
db98d94b-ed0f-455a-9222-0731b82bab95
StampyAI/alignment-research-dataset/arxiv
Arxiv
Predicting human decisions with behavioral theories and machine learning 1 Predicting human decisions with behavioral theories and machine learning Plonsky, Ori1a; Apel, Reuta; Ert, Eyalb; Tennenholtz, Moshea; Bourgin, Davidc; Peterson , Joshua , C.c; Reichman, Danielc; Griffiths , Thomas L.c; Russell , Stu...
b8b22de2-e2ab-4538-b0e5-30c6e36dd674
trentmkelly/LessWrong-43k
LessWrong
Mistakes with Conservation of Expected Evidence Epistemic Status: I've really spent some time wrestling with this one. I am highly confident in most of what I say. However, this differs from section to section. I'll put more specific epistemic statuses at the end of each section. Some of this post is generated from m...
119da778-4840-4dad-8912-e6ad78f9ca0f
trentmkelly/LessWrong-43k
LessWrong
10/50/90% chance of GPT-N Transformative AI? In Developmental Stages of GPTs, orthonormal explains why we might need "zero more cognitive breakthroughs" to reach transformative AI, aka "an AI capable of making an Industrial Revolution sized impact". More specifically, he says that "basically, GPT-6 or GPT-7 might do i...
9254f65b-1ed9-4dc1-bb7c-3f23c998045a
trentmkelly/LessWrong-43k
LessWrong
Meetup : Tel Aviv Meetup: Assorted LW mini-talks Discussion article for the meetup : Tel Aviv Meetup: Assorted LW mini-talks WHEN: 23 June 2015 07:00:00PM (+0300) WHERE: Yigal Alon Street 98, Tel Aviv-Yafo, Israel We're going to have a meetup on Tuesday, June 23 at Google Tel Aviv offices, Electra Tower, 98 Yigal A...
fe98525a-f8d0-49ee-b3e6-237ac0b3b7ba
trentmkelly/LessWrong-43k
LessWrong
Computational Morality (Part 3) - Similarity of Proposals In part 1 ( https://www.lesswrong.com/posts/Lug4n6RyG7nJSH2k9/computational-morality ) I set out a proposal for a system of machine ethics to govern the behaviour of AGI (in which you simply imagine that you are all the people and other sentiences involved in a...
6ccec8d8-151e-4a4d-8138-bfe23f744557
trentmkelly/LessWrong-43k
LessWrong
Amanda Knox: post mortem Continuing my interest in tracking real-world predictions, I notice that the recent acquittal of Knox & Sollecito offers an interesting opportunity - specifically, many LessWrongers gave probabilities for guilt back in 2009 in komponisto’s 2 articles: * “You Be the Jury: Survey on a Current ...
e35dd68e-96f3-465d-9aab-ca23dea0c184
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
Classical symbol grounding and causal graphs [I've been using the term symbol-grounding](https://www.lesswrong.com/posts/ns95FHkkzpjXh4x5Q/what-does-gpt-3-understand-symbol-grounding-and-chinese) in a slightly different way to how people often use it. I've been using it to mean: * "This agent's symbol is reliable en...
3fe23855-2aff-4751-9c6a-c35a1bb5fa8d
trentmkelly/LessWrong-43k
LessWrong
Losing the root for the tree 1 You know that being healthy is important. And that there's a lot of stuff you could do to improve your health: getting enough sleep, eating well, reducing stress, and exercising, to name a few. There’s various things to hit on when it comes to exercising too. Strength, obviously. But ...
f208afc1-a414-4b3e-bae4-39b4603a3277
trentmkelly/LessWrong-43k
LessWrong
Failed Utopia #4-2     Shock after shock after shock—     First, the awakening adrenaline jolt, the thought that he was falling.  His body tried to sit up in automatic adjustment, and his hands hit the floor to steady himself.  It launched him into the air, and he fell back to the floor too slowly.     Second shock.  ...
97db6e55-b970-46bb-a956-ec4cd5a4bee8
StampyAI/alignment-research-dataset/arxiv
Arxiv
What Matters in Learning from Offline Human Demonstrations for Robot Manipulation ### 1 Introduction Human supervision has been at the heart of the most significant recent advances in several domains such as computer vision [deng2009imagenet, krizhevsky2012imagenet, Redmon2016YouOL, Ronneberger2015UNetCN] and natu...
f9e10c6d-8ecc-4921-b98c-44295b0ca0bd
trentmkelly/LessWrong-43k
LessWrong
The elephant in the room, AMA Hello fellow Less Wrongians! Given your comments on my organizing communities series, I get the feeling that many of you are wondering why: * a returned Mormon missionary would even come to Less Wrong in the first place. * why I find religion plausible at all * why I would identify w...
01d7f4bb-931d-4382-b49d-054cf284e031
trentmkelly/LessWrong-43k
LessWrong
How To Prevent a Dystopia (Just edited this first post of mine,  now it's obvious I don't propose to build any dystopias. I'm actually trying to prevent all the permanent enforced ones. This is the result of three years of thinking and modeling hyper‑futuristic and current ethical systems. It's the intro to the serie...
5b3d224a-73fd-4536-b914-4714f29248a0
trentmkelly/LessWrong-43k
LessWrong
Effective altruisms large and small Here is one way the world could be. By far the best opportunities for making the world better can be supported by philanthropic money. They last for years. They can be invested in a vast number of times. They can be justified using widely available information and widely verifiable ...
d4cd5953-816c-4683-a840-57878f4d9332
StampyAI/alignment-research-dataset/blogs
Blogs
How undesired goals can arise with correct rewards #### Exploring examples of goal misgeneralisation – where an AI system's capabilities generalise but its goal doesn't As we build increasingly advanced artificial intelligence (AI) systems, we want to make sure they don’t pursue undesired goals. Such behaviour in an ...
f82ee9ec-c6f9-49f3-ba9e-ac77862e1ebd
trentmkelly/LessWrong-43k
LessWrong
Planned Power Outages With the dubiously motivated PG&E blackouts in California there are many stories about how lack of power is a serious problem, especially for people with medical dependencies on electricity. Examples they give include people who: * Have severe sleep apnea, and can't safely sleep without a CPA...
1d49cc2e-bed7-45b4-96d2-a95257602cfb
trentmkelly/LessWrong-43k
LessWrong
List of techniques to help you remember names Name are very important. Everyone has one; everyone likes to know when you know their name. Everyone knows them to be a part of social interaction. You can't avoid names (well you can, but it gets tricky). In becoming more awesome at names, here is a bunch of suggestions t...
f4b785a6-2497-477a-873a-d1cf2f57b20f
trentmkelly/LessWrong-43k
LessWrong
SBF's recent live interview at the DealBook Summit [PS: If you're hitting a paywall, you can use this link] Sam Bankman-Fried was recently live interviewed by Andrew Ross Sorkin from the New York Times as part of the DealBook summit, an interview which was scheduled before the FTX collapse and remained so. The summa...