id
stringlengths
36
36
source
stringclasses
15 values
formatted_source
stringclasses
13 values
text
stringlengths
2
7.55M
b8dc5d8d-192a-4dbc-950f-5c99943957f3
trentmkelly/LessWrong-43k
LessWrong
Any significant updates on long covid risk analysis? [ETA: This question is specifically about risk analysis of long covid, given current knowledge. Discussion should be about things such as: * base rates * priors and posteriors for LC risk * evidence that would make one update in favor of certain new policies wit...
741b3449-a3f6-4b80-8550-1581622acc57
LDJnr/LessWrong-Amplify-Instruct
LessWrong
"I just heard a comment by Braddock of Lovesystems that was brilliant: All that your brain does when you ask it a question is hit "search" and return the first hit it finds. So be careful how you phrase your question. Say you just arrived at work, and realized you once again left your security pass at home. You ask you...
bddd6882-8e34-4135-a6a0-17133a07cafe
StampyAI/alignment-research-dataset/special_docs
Other
Antitrust-Compliant AI Industry Self-Regulation A n titr u s t- C o m p l i a n t A I I n d u s tr y   S e l f - R e g u l a tio n   Cullen O’Keefe Research Scientist (Policy), OpenAI Research Affiliate, Centre for the Governance of AI, Future of Humanity Institute Research...
086d82dd-e9da-4b87-ac95-9f04623473e3
StampyAI/alignment-research-dataset/lesswrong
LessWrong
AI Safety Newsletter #7: Disinformation, Governance Recommendations for AI labs, and Senate Hearings on AI Welcome to the AI Safety Newsletter by the [Center for AI Safety](https://www.safe.ai/). We discuss developments in AI and AI safety. No technical background required. Subscribe [here](https://newsletter.safe.ai...
c9e583c6-f5e6-4ee9-842e-1e4424eb0fbf
trentmkelly/LessWrong-43k
LessWrong
Moral gauge theory: A speculative suggestion for AI alignment [Epistemic status: Speculative. I've written this post mostly to clarify and distill my own thoughts and have posted it in an effort to say more wrong things.] Introduction The goal of this post is to discuss a theoretical strategy for AI alignment, part...
01260a22-2c94-47e1-92b4-fc8b72f6e358
StampyAI/alignment-research-dataset/lesswrong
LessWrong
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.” Fu...
779bd11f-6989-43e5-8e60-b54d548e4ad0
trentmkelly/LessWrong-43k
LessWrong
Monthly Roundup #14: January 2024 There’s always lots of stuff going on. The backlog of other roundups keeps growing rather than shrinking. I have also decided to hold back a few things to turn them into their own posts instead. BAD NEWS I wonder if it is meaningful that most of the bad news is about technology? I ...
d6849a54-ea20-4d35-b2e9-8c2454dbbed2
trentmkelly/LessWrong-43k
LessWrong
Dangerous capability tests should be harder Note: I am cross-posting a previous blog post from Planned Obsolescence that I wrote in August Imagine you’re the CEO of an AI company and you want to know if the latest model you’re developing is dangerous. Some people have argued that since AIs know a lot of biology now —...
14212938-7b5b-49f1-aff4-410a2008c62e
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
Conditioning Predictive Models: Open problems, Conclusion, and Appendix *This is the final of seven posts in the [Conditioning Predictive Models Sequence](https://www.alignmentforum.org/s/n3utvGrgC2SGi9xQX) based on the paper “[Conditioning Predictive Models: Risks and Strategies](https://arxiv.org/abs/2302.00805)” by...
ca450d62-647b-493a-81cd-a0b142f05d39
trentmkelly/LessWrong-43k
LessWrong
A My Little Pony fanfic allegedly but not mainly about immortality My Little Pony (generation 4) has 2 immortal characters, who get a lot of sympathy from the bronies.  "How sad!  Poor Celestia and Luna must see everyone they know grow old and die.  How much better to die yourself!" I tried to write a fanfic saying t...
fa96924d-f244-4843-953f-bfc77b043b14
trentmkelly/LessWrong-43k
LessWrong
Concave Utility Question This post will just be a concrete math question. I am interested in this question because I have recently come tor reject the independence axiom of VNM, and am thus playing with some weaker versions. Let Ω be a finite set of deterministic outcomes. Let L be the space of all lotteries over the...
7fa964ed-6f47-414f-8352-14f937a0b7da
trentmkelly/LessWrong-43k
LessWrong
ARENA 6.0 - Call for Applicants TL;DR: We're excited to announce the sixth iteration of ARENA (Alignment Research Engineer Accelerator), a 4-5 week ML bootcamp with a focus on AI safety! Our mission is to provide talented individuals with the ML engineering skills, community, and confidence to contribute directly to t...
ff94c1e0-24f9-453f-996c-2ba5c1e73996
trentmkelly/LessWrong-43k
LessWrong
Overcoming Paradox of Choice Through Random Selection Epistemic Status: Personal anecdote with reasonable sounding theory. Background I love to read. There is no activity I like more than it, or one I want to do as often. I’ve been blessed with an intensely vivid imagination, and twenty minutes spent reading a well-...
242af5e1-2693-4a53-8729-56aa42df2be8
LDJnr/LessWrong-Amplify-Instruct
LessWrong
"This is a post about my own confusions. It seems likely that other people have discussed these issues at length somewhere, and that I am not up with current thoughts on them, because I don’t keep good track of even everything great that everyone writes. I welcome anyone kindly directing me to the most relevant things,...
41ccd623-c991-468b-a46f-cb2bc3e1d7af
StampyAI/alignment-research-dataset/blogs
Blogs
continue working on hard alignment! don't give up! continue working on hard alignment! don't give up! -------------------------------------------------- let's call "hard alignment" the (["orthodox"](https://scottaaronson.blog/?p=6821)) problem, historically worked on by MIRI, of preventing [strong agentic AIs](stron...
d6050f85-5f35-472b-afed-221ceaa3be8b
trentmkelly/LessWrong-43k
LessWrong
Apocalypse, corrupted Epistemic status: don't take it seriously In a post apocalyptic setting, the world would be run by the socially skilled and the well connected, with corruption and nepotism ruling. I say that at the start, because I've been trying to analyse the attraction of post-apocalyptic settings: why do w...
c65f46b0-f2ab-4489-803f-2fc2f6714b17
StampyAI/alignment-research-dataset/lesswrong
LessWrong
Solving the two envelopes problem Suppose you are presented with a game. You are given a red and a blue envelope with some money in each. You are allowed to ask an independent party to open both envelopes, and tell you the ratio of blue:red amounts (but not the actual amounts). If you do, the game master replaces the ...
dd7a4fa1-7526-4993-99d5-4491c4bd3ef6
trentmkelly/LessWrong-43k
LessWrong
Mike Darwin on the Less Wrong intelligentsia He has resumed posting at his blog Chronopause and he is essential reading for those interested in cryonics and, more generally, rational decision-making in an uncertain world.  In response to a comment by a LW user named Alexander, he writes: > I have no objection to you...
ffa81e54-90a6-403a-b909-770cac59e333
trentmkelly/LessWrong-43k
LessWrong
Exploring the Multiverse of Large Language Models [Crossposted from Medium.com.] Credit: Alexas_Fotos from Pixabay Large language models can generate texts under different tasks, for example, question answering, machine translation, document summarization, etc. Beyond being right and wrong, large language models can ...
a6466b02-fa98-4174-a85f-488aca1b9a79
trentmkelly/LessWrong-43k
LessWrong
The Ultimatum Game l The Ultimatum Game is a simple game in which two players attempts to split a $100 reward.  They can communicate with each other for 10 minutes, after which: * Player 1 proposes an integer split (e.g. $75 for Player 1, $25 for Player 2) * Player 2 may Accept or Reject this split.  If Player 2 r...
46eb4476-1243-4557-94cd-cbd6ff006ccb
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
Tracr: Compiled Transformers as a Laboratory for Interpretability | DeepMind Abstract -------- Interpretability research aims to build tools for understanding machine learning (ML) models. However, such tools are inherently hard to evaluate because we do not have ground truth information about how ML models actually...
69b19018-f8ca-4868-8153-2787820e277c
trentmkelly/LessWrong-43k
LessWrong
Social Impact, Effective Altruism, and Motivated Cognition Money is one measure of social status. People compare themselves favorably or unfavorably to others in their social circles based on their wealth and their earning power, and signals thereof, and compare their social circles favorably or unfavorably with other...
9d23b831-e24e-4a8a-91f8-c4d574c6f53b
trentmkelly/LessWrong-43k
LessWrong
A plan for spam I'm getting tired of banning tons of similar articles about jewelry etc. in the discussion section. Our situation looks like a textbook-perfect use case for a Bayesian spam filter (ahem). Or just implement the 5 karma limit that was discussed earlier, that would help too.
e1bbf39d-f849-4505-bdb7-3815cf2f9f90
trentmkelly/LessWrong-43k
LessWrong
The US Already Has A Wealth Tax In Modeling A Wealth Tax, Paul Graham writes: > Even a 0.5% wealth tax would start to keep founders away from a state or country that imposed it. The US, however, effectively already has a ~0.3%-0.4% wealth tax. Specifically, the long-term capital gains rate in the US is 20% for hig...
8dbb4461-6e11-424d-8bda-c72848b1ad12
trentmkelly/LessWrong-43k
LessWrong
An early stage prioritisation model What do you care about in this world? I claim that this doesn't matter much, due to convergent instrumental goals of rational agents (for a more technical account, see here). For any set of values humans are likely to have, there are a set of instrumental goals that will be necessar...
249f56f0-6b99-4157-bff4-219b6c508dcc
trentmkelly/LessWrong-43k
LessWrong
Why Bigger Models Generalize Better There is still a lingering belief from classical machine learning that bigger models overfit and thus don't generalize well. This is described by the bias-variance trade-off, but this no longer holds in the new age of machine learning. This is empirically shown by phenomena like dou...
c5bb9bd1-9b1e-4e80-b831-dfcb7ee8f8ed
trentmkelly/LessWrong-43k
LessWrong
[SEQ RERUN] GAZP vs. GLUT Today's post, GAZP vs. GLUT was originally published on 07 April 2008. A summary (taken from the LW wiki):   > Fleshes out the generalized anti-zombie principle a bit more, and describes the game "follow-the-improbability". Discuss the post here (rather than in the comments to the origina...
1fae8397-9328-41b5-9379-6958d448dc09
StampyAI/alignment-research-dataset/lesswrong
LessWrong
A Proposed Test to Determine the Extent to Which Large Language Models Understand the Real World Large Language Models such as GPT-3 and PaLM have demonstrated an ability to generate plausible-sounding, grammatically-correct text in response to a wide range of prompts. And some Language-Model-based applications such a...
c3541b31-8127-4a94-acde-0c8c99a0e54f
trentmkelly/LessWrong-43k
LessWrong
If I were a well-intentioned AI... II: Acting in a world Classifying images is one thing. But what if I'm an agent that is actually active in some setting? The previous approach still applies: detecting when I'm out of distribution, and trying to keep my behaviour compatible with the various reward function that coul...
4c5d3b7d-3a2e-4927-9f84-13f8d9a7df60
trentmkelly/LessWrong-43k
LessWrong
[SEQ RERUN] That Tiny Note of Discord Today's post, That Tiny Note of Discord was originally published on 23 September 2008. A summary (taken from the LW wiki):   > Eliezer started to dig himself out of his philosophical hole when he noticed a tiny inconsistency. Discuss the post here (rather than in the comments ...
13078b96-7032-4dde-b1e0-5a188e6e84b4
trentmkelly/LessWrong-43k
LessWrong
Preventing discussion from being watered down by an "endless September" user influx.   In the thread "LessWrong could grow a lot, but we're doing it wrong.", I explained why LessWrong has the potential to grow quite a lot faster in my opinion, and volunteered to help LessWrong grow.  Of course, a lot of people were co...
84e5758b-4314-4ed4-932b-e84479c73499
StampyAI/alignment-research-dataset/eaforum
Effective Altruism Forum
AI Ethical Committee Is there any AI ethical committee within EA that individuals can privately and securely consult regarding potentially sensitive AI safety-related issues? Consider the following two cases: * #whistleblowing A ML researcher in a big tech company comes upon information regarding potentially dangero...
59f1e13e-81d7-43a6-a8d8-3b8575fc2e19
trentmkelly/LessWrong-43k
LessWrong
Notes on Claude 4 System Card Anthropic released Claude 4. I've read the accompanying system card, and noted down some of my remarks. Alignment assessment: system prompt mix-ups There's a worrying theme that's repeated in the alignment evaluations: > The most concerning issues that we observed in our testing of Cla...
56fa156d-7234-4fc1-9300-04499997ee16
trentmkelly/LessWrong-43k
LessWrong
The Good Place has a line that defines the human dilemma First off, Kudos to Less Wrong which had some comments about the TV show The Good Place, so I watched it on Netflix and enjoyed it a lot. There is one line from the show that encapsulates the human dilemma ``There is a chicken sandwich that, if you eat it, you...
812096b3-d4b0-4a2c-a0d3-02d198639ebd
trentmkelly/LessWrong-43k
LessWrong
[Link] Sparrows may be vulnerable to sunk-cost-like effects You walk into a store that sells two identical candies. Would you buy the candy that ordinarily costs $2 at 50% off (for one dollar), or the candy that ordinarily costs 10$ for 80% off (for one dollar)?  In an isomorphic situation, sparrows preferred the lat...
6d69c834-0cae-4fe1-a98c-118e22e88694
trentmkelly/LessWrong-43k
LessWrong
Meaningful things are those the universe possesses a semantics for Logic courses are very risk-averse to talking about philosophy. This is bad, because philosophy is the motivation/intuition for logic & TCS. This essay gives a scout's view. Epistemic status: A computer scientist would nod through Chapters 2-5, at lea...
6c54bc36-ba4d-4a2b-9dfa-d6a1a15afb27
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
AI coordination needs clear wins *Thanks to Kate Woolverton and Richard Ngo for useful conversations, comments, and feedback.* EA and AI safety have invested a lot of resources into building our ability to get coordination and cooperation between big AI labs. So far, however, despite that investment, it doesn’t seem...
fab5ebbd-d9cd-4fd7-bd8d-cd6299af2c3d
trentmkelly/LessWrong-43k
LessWrong
Meetup : Tortuga Rationalists Meetup Discussion article for the meetup : Tortuga Rationalists Meetup WHEN: 14 July 2011 02:34:01PM (-0700) WHERE: 850 Williams Way, Mountain View, CA We are a South Bay LessWrong Rationality group who meets every Thursday at the Tortuga (http://tortuga.coop) lounge in Mountain View f...
e08f2cdf-1621-4bb6-8d53-cabe1ae10c3d
LDJnr/LessWrong-Amplify-Instruct
LessWrong
"So perhaps you're reading all this, and asking: "Yes, but what does this have to do with reductionism?" Partially, it's a matter of leaving a line of retreat. It's not easy to take something important apart into components, when you're convinced that this removes magic from the world, unweaves the rainbow. I do plan t...
e34cd81d-10f4-41aa-85a3-c57b2c511198
trentmkelly/LessWrong-43k
LessWrong
Nanosystems are Poorly Abstracted To do something, you need energy, and you need information. Energy to do the thing, and information to define what's being done. The mental image I have of a "robot", has inputs of both continuously. The input of energy comes as electricity, and the input of information comes in the f...
f6e1f69c-ca83-418b-a841-dad373786177
trentmkelly/LessWrong-43k
LessWrong
Meditation Retreat: Immoral Mazes Sequence Introduction I just got home from a six day meditation retreat and began writing. The catch is that I arrived at the retreat yesterday. I knew going in that it was a high variance operation. All who had experience with such things warned us we would hate the first few day...
6b1660b8-6130-4dec-9350-a1e106554101
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
Test Cases for Impact Regularisation Methods *Epistemic status: I’ve spent a while thinking about and collecting these test cases, and talked about them with other researchers, but couldn’t bear to revise or ask for feedback after writing the first draft for this post, so here you are.* A motivating concern in AI al...
b433c58d-3b62-4640-ac2c-7aff110063e6
trentmkelly/LessWrong-43k
LessWrong
Actually updating Actually updating can be harder than it seems. Hearing the same advice from other people and only really understanding it the third time (though internally you felt like you really understood the first time) seems inefficient. Having to give yourself the same advice or have the same conversation with...
b109bf8a-3aeb-49b0-b04a-fbb47230a9bc
trentmkelly/LessWrong-43k
LessWrong
AI prediction case study 4: Kurzweil's spiritual machines Myself, Kaj Sotala and Seán ÓhÉigeartaigh recently submitted a paper entitled "The errors, insights and lessons of famous AI predictions and what they mean for the future" to the conference proceedings of the AGI12/AGI Impacts Winter Intelligenceconference. S...
45d0e476-3c73-4283-9cd0-c7aedb53ea86
trentmkelly/LessWrong-43k
LessWrong
[SEQ RERUN] Well-Kept Gardens Die By Pacifism Today's post, Well-Kept Gardens Die By Pacifism was originally published on 21 April 2009. A summary (taken from the LW wiki):   > Good online communities die primarily by refusing to defend themselves, and so it has been since the days of Eternal September. Anyone accul...
ba7919ae-342f-4afd-8816-b598d1eccd80
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
Environments as a bottleneck in AGI development Given a training environment or dataset, a training algorithm, an optimiser, and a model class capable of implementing an AGI (with the right parameters), there are two interesting questions we might ask about how conducive that environment is for training an AGI. The fi...
783dfb08-d8c1-4d63-9b00-08958073ba5e
trentmkelly/LessWrong-43k
LessWrong
Organ donation versus cryonics Simultaneously signing up for organ donation and cryonics versus only cryonics. Does having less organs decrease the likelihood of cryonics (including revival) working? Is it a good idea to have only your head frozen anyway, to save on electricity and storage? Do the benefits of organ do...
b365e575-7736-4202-ace1-dfa97f9aacb3
trentmkelly/LessWrong-43k
LessWrong
AI Constitutions are a tool to reduce societal scale risk This work was funded by Polaris Ventures As AI systems become more integrated into society, we face potential societal-scale risks that current regulations fail to address. These risks include cooperation failures, structural failures from opaque decision-maki...
857695fb-611b-4d56-ab9f-ae2b23a1e90b
trentmkelly/LessWrong-43k
LessWrong
Masculinity—A Case For Courage HOW THIS POST CAME ABOUT When I put a call out for non-zero-sum topics from subscribers, it was my brother who first replied with masculinity—pointing me to the work of author Richard Reeves. > “I think the danger of zero-sum thinking is present … men are suffering because women are...
b1392fc1-856c-435f-9e59-8e35ab1c8f42
trentmkelly/LessWrong-43k
LessWrong
A Fate Worse Than Death The claim has been made that, all things being equal, it is better to be alive than dead.  I dissent.   It is much more complicated than this.  If I knew somehow that I would spend the next fifty years of my life in Guantanamo bay, I would rather kill myself than suffer that fate.  If a fortun...
7324472a-16d4-4fec-93a9-756085b81a59
trentmkelly/LessWrong-43k
LessWrong
Top li’l pots I am a connoisseur of a certain genre of mostly-snack food, combining aesthetics, convenience and the idiosyncratic types of actual food that I like to eat (central examples of food whose substance I like, ignoring form, are parsley, farro, black lentils, pecans, craisins, dark chocolate, feta cheese, sw...
cd5c50fd-3671-470b-8e5f-45f7c1eb57a1
trentmkelly/LessWrong-43k
LessWrong
AXRP Episode 13 - First Principles of AGI Safety with Richard Ngo YouTube link This podcast is called AXRP, pronounced axe-urp and short for the AI X-risk Research Podcast. Here, I (Daniel Filan) have conversations with researchers about their papers. We discuss the paper and hopefully get a sense of why it’s been wr...
3e10acba-7d97-4841-8b46-5edeb7202d20
trentmkelly/LessWrong-43k
LessWrong
An approach to logical counterfactuals inspired by the Demski prior This is a probabilistic approach to logical counterfactuals inspired by the Demski prior. Given a sentence ϕ, we wish to ask what other logical sentences are true in the world C(ϕ) where we correctly counterfactually assume ϕ. I spoke about this befor...
e888ab34-c589-4e5d-8be7-9e1795668d6a
StampyAI/alignment-research-dataset/lesswrong
LessWrong
OpenAI: "Scaling Laws for Transfer", Hernandez et al. OpenAI has released a paper on scaling laws for transfer learning from general English text to python code. [Author's tweet summary](https://twitter.com/Hernandez_Danny/status/1357035777406853128). Abstract: "We study empirical scaling laws for transfer learning b...
71f37391-be89-4e6d-bebc-5f6169341845
trentmkelly/LessWrong-43k
LessWrong
Forecasting Thread: Existential Risk This is a thread for displaying your probabilities of an existential catastrophe that causes extinction or the destruction of humanity’s long-term potential. Every answer to this post should be a forecast showing your probability of an existential catastrophe happening at any give...
bcd2b745-7d57-4536-9b65-cf2d95703117
trentmkelly/LessWrong-43k
LessWrong
Community bias in threat evaluation If I were to ask the question "What threat poses the greatest risk to society/humanity?" to several communities I would expect to get some answers that follow a predictable trend: If I asked the question on an HBD blog I'd probably get one of the answers demographic disaster/dysgen...
c3b723f8-a954-4574-a14b-d5cfbc01c22e
trentmkelly/LessWrong-43k
LessWrong
The 2023 LessWrong Review: The Basic Ask Each December, the LessWrong community reflects on the best blogposts of yesteryear, to decide which posts stood the tests of time.  In this post, I aim to: * Explain some subtleties of what I'm hoping we get out of the Annual Review * Make some explicit asks (even of busy ...
7b2989f3-1fc7-4e70-9fbe-6b26397a3a8d
StampyAI/alignment-research-dataset/arxiv
Arxiv
Machine learning 2.0 : Engineering Data Driven AI Products 1 Introduction --------------- Attempts to embed machine learning-based predictive models into products and services in order to make them smarter, faster, cheaper, and more personalized will dominate the technology industry for the foreseeable future. Curr...
5d22d6d0-00c7-4119-b429-31d203ebb36c
StampyAI/alignment-research-dataset/lesswrong
LessWrong
Why the beliefs/values dichotomy? I'd like to suggest that the fact that human preferences can be decomposed into beliefs and values is one that deserves greater scrutiny and explanation. It seems intuitively obvious to us that rational preferences must decompose like that (even if not exactly into a probability distr...
ffe43a40-85cd-4b36-ac04-9d4a8891a7fd
trentmkelly/LessWrong-43k
LessWrong
Using Prompt Evaluation to Combat Bio-Weapon Research With many thanks to Sasha Frangulov for comments and editing Before publishing their o1-preview model system card on Sep 12, 2024, OpenAI tested the model on various safety benchmarks which they had constructed. These included benchmarks which aimed to evaluate wh...
5764d66d-660e-4f0e-bfd4-435c4e0a5929
trentmkelly/LessWrong-43k
LessWrong
In Search of Slack Original article: Studies on Slack 1 Proposition: If you have enough free resources you can evolve irreducibly complex features. Imagine a bunch of yeast cells added to a barrel of malt when beer is being brewed. They have an ocean of free resources at their disposal. Surely, the evolutionary pre...
ed9a9f9a-a9ed-447d-9767-db99153d4c84
trentmkelly/LessWrong-43k
LessWrong
Thoughts on reward engineering Note: This is the first post from part five: possible approaches of the sequence on iterated amplification. The fifth section of the sequence breaks down some of these problems further and describes some possible approaches. ---------------------------------------- Suppose that I woul...
d611d19a-1dd5-4a31-af84-09581ec61ca8
trentmkelly/LessWrong-43k
LessWrong
Life Has a Cruel Symmetry t ≈ 0s I am born into knowledge. I know the world is dangerous, but not malign. If I were unshielded, I would quickly be killed, torn apart by eddies in spacetime that grow and shrink and combine and split, turning all they encounter to ash. But I know that I am shielded. I have been given...
f432e528-3e73-47af-bae2-d1c1151e2279
StampyAI/alignment-research-dataset/lesswrong
LessWrong
The Domain of Your Utility Function **Unofficial Followup to:** [Fake Selfishness](/lw/kx/fake_selfishness/), [Post Your Utility Function](/lw/zv/post_your_utility_function/) A perception-determined utility function is one which is determined only by the perceptual signals your mind receives from the world; for inst...
89c8b88a-9833-4754-a528-a166daa71a46
trentmkelly/LessWrong-43k
LessWrong
Human-AI Complementarity: A Goal for Amplified Oversight By Sophie Bridgers, Rishub Jain, Rory Greig, and Rohin Shah Based on work by the Google DeepMind Rater Assist Team: Vladimir Mikulik, Sophie Bridgers, Tian Huey Teh, Rishub Jain, Rory Greig, Lili Janzer (randomized order, equal contributions)   Human oversight...
12257344-56f1-49f4-9786-08aaf0c7b301
trentmkelly/LessWrong-43k
LessWrong
Don't Think About the Thing Behind the Curtain. Foreword This is a submissions to the rational animation writing contest this script is Halloween themed and inspired by their most recent video (Please imagine Yann LeCun in his Pjs/HJ). I wanted to get it out ASAP. I have left out any types of directions for animation...
7ae578d8-11a6-4274-939c-d3a4ee1dcfc8
trentmkelly/LessWrong-43k
LessWrong
Meditations On Moloch [Content note: Visions! omens! hallucinations! miracles! ecstasies! dreams! adorations! illuminations! religions!] I. Allan Ginsberg’s famous poem, Moloch: > What sphinx of cement and aluminum bashed open their skulls and ate up their brains and imagination? > > Moloch! Solitude! Filth! Uglin...
9880af56-86d8-4d82-bff7-61f1810877c7
trentmkelly/LessWrong-43k
LessWrong
An Argument For Prioritizing: "Positively Shaping the Development of Crypto-assets" None
5ad3be2e-1004-4740-8bdb-9b08ea5bbbe3
trentmkelly/LessWrong-43k
LessWrong
AI Safety Camp 2024 AI Safety Camp connects you with a research lead to collaborate on a project – to see where your work could help ensure future AI is safe. Apply before December 1, to collaborate online from January to April 2024.   We value diverse backgrounds. Many roles but definitely not all require some kno...
c82e4fae-6368-4132-86a0-612a9dd8420b
trentmkelly/LessWrong-43k
LessWrong
Best non-textbooks on every subject The best way to learn a subject is undoubtedly by reading a textbook on it. But I find textbooks a drudgery, and tend to give up after a couple of chapters. On the other hand I don't need a deep broad formal knowledge in every subject. I often just want to know enough that I know w...
7d95e9eb-8cb2-40b5-8bbd-18eb97e532f3
trentmkelly/LessWrong-43k
LessWrong
Meetup : First meetup of Rationality Zürich Discussion article for the meetup : First meetup of Rationality Zürich WHEN: 01 November 2015 03:00:00PM (+0200) WHERE: Sphères, Hardturmstrasse 66, 8005 Zürich Purposes of this meetup: - Get the community together and talking to each other - Find out what people are inte...
9274f41d-b60f-416c-90ab-726d934eff21
StampyAI/alignment-research-dataset/aisafety.info
AI Safety Info
What is adversarial training? Adversarial training is a safety technique that pits two (or more) models against each other. This is helpful for training and safety because it enables us to explore and correct harmful inputs that could otherwise go unexplored due to limited training data or limited human oversight. “Ad...
6f811b9b-f530-47a8-a324-ebd9f9256e0e
trentmkelly/LessWrong-43k
LessWrong
If we're in a sim... My father, who is home recovering from surgery, emailed the following web page to me and a few other members of my family, and expressed interest in reading interesting responses. > If We Are In A Computer Simulation > > Is the universe just a big computer simulation running in another universe?...
1c65015c-5b97-4a3f-8b13-f329a4eaee68
trentmkelly/LessWrong-43k
LessWrong
For Policy’s Sake: Why We Must Distinguish AI Safety from AI Security in Regulatory Governance Over the past few months, I’ve been conducting independent policy research focusing on understanding how AI safety concepts can be translated into AI regulation / policy governance. I am a laywer working in AI Governance fo...
3cfb4498-e8af-442e-a90b-cd9914ff9548
trentmkelly/LessWrong-43k
LessWrong
Open thread, Feb. 23 - Mar. 1, 2015 If it's worth saying, but not worth its own post (even in Discussion), then it goes here. ---------------------------------------- Notes for future OT posters: 1. Please add the 'open_thread' tag. 2. Check if there is an active Open Thread before posting a new one. (Immediately ...
7adbfbd8-8a6a-4ffc-bd3d-66229283043e
trentmkelly/LessWrong-43k
LessWrong
Singularity Institute Strategic Plan 2011 Thanks to the hard work and cooperation of Singularity Institute staff and volunteers, especially Louie Helm and Luke Muehlhauser (lukeprog), we now have a Strategic Plan, which outlines the near-term goals and vision of the Institute, and concrete actions we can take to fulfi...
63b26656-854d-4fe4-8ae6-9201e0d661f0
trentmkelly/LessWrong-43k
LessWrong
Semantic Disagreement of Sleeping Beauty Problem This is the tenth post in my series on Anthropics. The previous one is Beauty and the Bets. Introduction In my previous posts I've been talking about the actual object-level disagreement between halfers and thirders - which of the answers formally is correct and which...
57d2c93f-8b85-4ad3-9482-8424198662f2
trentmkelly/LessWrong-43k
LessWrong
Examples of governments doing good in house (or contracted) technical research An open question is the degree to which governments can do good technical research, either in-house or commissioned. Some possible ways in which this could be relevant: * Evals (although this likely could be contracted in a successful wa...
5693015e-4eee-4e88-8336-bd4f8219f727
trentmkelly/LessWrong-43k
LessWrong
Civil Law and Political Drama The purpose of this essay is to clarify to YOU, the mind reading this essay, your relation to law, your characteristic blind spot around adversaries, and what sorts of processes are competing for attention in the humans you’re instantiated through. For contracts and property not to just ...
98149c22-614e-452a-921b-444a731ea9b0
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
What I talk about when I talk about AI x-risk: 3 core claims I want machine learning researchers to address. Recently, as PCSOCMLx, I (co-)hosted a session with the goal of explaining, debating, and discussing what I view as "the case for AI x-risk". Specifically, my goal was/is to make the case for the "out-of-contro...
9ddf760e-f127-47ee-8164-ba8e0a6a9a60
StampyAI/alignment-research-dataset/special_docs
Other
Introduction to STAMP Aeronautics and Astronautics MIT Introduction to STAMP: Part 1 Prof. Nancy G. Leveson 1 2 3 Some Current Uses Aviation Maritime Automobiles Dams Defense Manufacturing Transportation Communication Healthcare/Hospitals Rail Workplace Safety Robotics Petrochemicals Unmanned vehicles (air, ground, wa...
061301ec-0621-4119-89ff-d407700fccee
trentmkelly/LessWrong-43k
LessWrong
How would the Scaling Hypothesis change things? The Scaling Hypothesis roughly says that current Deep Learning techniques, given ever more computing power, data, and perhaps some relatively minor improvements, will scale all the way to human-level AI and beyond. Let's suppose for the sake of argument that the Scaling ...
c281ba05-9aa2-48c7-b595-034f55af5afa
trentmkelly/LessWrong-43k
LessWrong
Book Review: Mathematics for Computer Science (Suggestion for MIRI Research Guide) tl;dr: I read Mathematics for Computer Science (MCS) and found it excellent. I sampled Discrete Mathematics and Its Applications (Rosen)—currently recommended in MIRI's research guide—as well as Concrete Mathematics and Discrete Mathema...
e3a9ace2-b8bf-4b4c-80ee-719c05ac3333
trentmkelly/LessWrong-43k
LessWrong
Risk Premiums vs Prediction Markets aka Why (some) market forecasts are biased and what can we do about them? tldr; When you care about an outcome, it will skew the odds you'll bet at. (To insure against bad outcomes). When everyone cares about an outcome, it skews the market odds. People (including myself) often po...
519e3829-f75c-4ac8-be88-4c30f2205237
trentmkelly/LessWrong-43k
LessWrong
How to Not Lose an Argument Related to: Leave a Line of Retreat Followup to: Talking Snakes: A Cautionary Tale, The Skeptic's Trilemma "I argue very well. Ask any of my remaining friends. I can win an argument on any topic, against any opponent. People know this, and steer clear of me at parties. Often, as a sign of...
e1fb5dc0-8b25-4f04-a9c7-481ca91a50ff
trentmkelly/LessWrong-43k
LessWrong
Turing Test and Machine Intelligence The recent news of Eugene Goostman passing the turing test has raised all kinds of debate around what counts as passing the turing tests and whether or not the chatbots are attempting to do it properly. The Turing Test is not a good criterion for machine intelligence. The compute...
4ce64a1c-6125-47e9-b64a-cf5ea2d0e1b4
trentmkelly/LessWrong-43k
LessWrong
(In)appropriate (De)reification So obviously there’s a 2x2 here: appropriate/inappropriate vs dereification/reification. I think the basic concepts here are unfortunately not that familiar to everyone, so I'll lay them out at least a bit. Reification I've already discussed, and is perhaps worth reading if the word is...
700c7837-6388-43e9-9401-6a0d00bdf096
trentmkelly/LessWrong-43k
LessWrong
Cooperative model knob-turning (Cross-posted from my medium channel) When you are trying to understand something by yourself, a useful skill to check your grasp on the subject is to try out the moving parts of your mental model and see if you can simulate the resulting changes. Suppose you want to learn how a rocket...
904b1aa0-1849-41f2-86b6-f6ace6f2fee2
StampyAI/alignment-research-dataset/special_docs
Other
A Citizen's Guide to Artificial Intelligence [] [] © 2021 Massachusetts Institute of Technology All rights reserved. No part of this book may be reproduced in any form by any electronic or mechanical means (including photocopying, recording, or information storage and retrieval) without permission in writing from t...
49531de0-1297-465d-ae56-ef7b29d0abf7
trentmkelly/LessWrong-43k
LessWrong
FAI, FIA, and singularity politics In discussing scenarios of the future, I speak of "slow futures" and "fast futures". A fast future is exemplified by what is now called a hard takeoff singularity: something bootstraps its way to superhuman intelligence in a short time. A slow future is a continuation of history as w...
c285a43e-dc40-4df1-8031-cd4198cc7d8c
trentmkelly/LessWrong-43k
LessWrong
Improved regret bound for DRL % operators that are separated from the operand by a space % autosize deliminaters % operators that require brackets % operators that require parentheses % Paper specific There is a simple way to somewhat improve the regret bound for DRL we derived befor...
46e6c3d0-f4b0-499c-991e-edc77e64c9e9
trentmkelly/LessWrong-43k
LessWrong
Luna Lovegood and the Chamber of Secrets - Part 5 "What do you mean 'bureaucratically impossible'?" said Harry James Potter-Evans-Verres. "It just is," said Mad-Eye Moody. "Nothing 'just is'," Harry said, "Everything happens for a material reason. What precisely is preventing you from acquiring the permits?" "I sus...
b9932d85-b300-46b6-82e8-7ce2e509b255
trentmkelly/LessWrong-43k
LessWrong
A tentative dialogue with a Friendly-boxed-super-AGI on brain uploads [Unnecessary explanation: Some people asked me why I thought the world of Friendship is optimal is dystopic… During the discussion, I inferred that what they saw as a “happy story” in AI safety is something like this: we’ll first solve something lik...
d2ecfbd8-2f4f-41c7-912e-a2f25015c0c7
StampyAI/alignment-research-dataset/arxiv
Arxiv
Scaling up psychology via Scientific Regret Minimization. 1 Introduction --------------- The standard methodology in psychological research is to identify a real-world behavior, create a laboratory paradigm that can induce that behavior, and then test a variety of hypotheses on a group of participants. This methodo...
fa7aeedd-19c0-4533-98f1-d043887ca226
trentmkelly/LessWrong-43k
LessWrong
The Singularity in the Zeitgeist As a part of public relations, I think it's important to keep tabs on how the Singularity and related topics (GAI, FAI, life-extension, etc.) are presented in the culture at large.  I've posted links to such things in the past, but I think there should be a central clearinghouse, and a...
60ce5270-71fb-4f93-84d6-160d3d8483c6
trentmkelly/LessWrong-43k
LessWrong
Dremeling In high school, I was on a robotics team (FRC) who, in honesty, were basically incompetent. (Me included.) We were bad at building a robot, bad at programming it, bad at design, but most relevantly for this post bad at planning ahead. The particular instance which added a term to my internal monologue was t...
0645a360-14dc-4838-92bd-e6ff68ccae88
trentmkelly/LessWrong-43k
LessWrong
Using the universal prior for logical uncertainty
ebf666bc-7772-4893-b0ac-81e14e5928af
trentmkelly/LessWrong-43k
LessWrong
Priors and Prejudice I Imagine an alternate version of the Effective Altruism movement, whose early influences came from socialist intellectual communities such as the Fabian Society, as opposed to the rationalist diaspora. Let’s name this hypothetical movement the Effective Samaritans. Like the EA movement of today,...
2976534e-aa20-49b0-9521-d41eef33b448
trentmkelly/LessWrong-43k
LessWrong
Conditioning Predictive Models: Open problems, Conclusion, and Appendix This is the final of seven posts in the Conditioning Predictive Models Sequence based on the paper “Conditioning Predictive Models: Risks and Strategies” by Evan Hubinger, Adam Jermyn, Johannes Treutlein, Rubi Hudson, and Kate Woolverton. Each pos...
4f17a269-0e5d-4f86-8049-9227a4f18781
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
Proper scoring rules don’t guarantee predicting fixed points *Johannes Treutlein and Rubi Hudson worked on this post while participating in* [*SERI MATS*](https://www.serimats.org/)*, under Evan Hubinger's and Leo Gao's mentorship respectively. We are grateful to Marius Hobbahn, Erik Jenner, and Adam Jermyn for useful...