id stringlengths 36 36 | source stringclasses 15
values | formatted_source stringclasses 13
values | text stringlengths 2 7.55M |
|---|---|---|---|
4199f1e0-1e0e-4273-b686-4159d5207ff8 | trentmkelly/LessWrong-43k | LessWrong | (Cryonics) can I be frozen before being near-death?
For a number of reasons, I'd like to wake up in the future instead of sticking around in the present, and I'd like to know about the options. Is there a cryonics company which will freeze me without me first being near-death, and without qualifying for euthenasia (wh... |
1df67e6c-ff77-4304-ac20-3d968b4786fe | trentmkelly/LessWrong-43k | LessWrong | Has anyone written stories happening in Hanson's em world?
EtA: Robin Hanson complains that while most hard sci-fi may get the hard tech right, they often get the economics wrong. I'm interested in stories that get both right. |
eb4aa4ab-288a-4f0d-ad5a-28b01187c9cd | trentmkelly/LessWrong-43k | LessWrong | Meetup : Rationality Reading Group (71-75)
Discussion article for the meetup : Rationality Reading Group (71-75)
WHEN: 10 August 2015 06:30:00PM (-0700)
WHERE: Paul G. Allen Center (185 Stevens Way, Seattle, WA) Room 503
Reading group for Yudkowsky's "Rationality: AI to Zombies", which is basically an organized and... |
f05ddfaf-0310-48c0-93e1-89aabb9ef3e8 | trentmkelly/LessWrong-43k | LessWrong | Mechanistic interpretability through clustering
How are activations in a transformer clustered together and what can we learn?
There has been a lot of progress using unsupervised methods (such as sparse autoencoders) to find monosemantic features in LLMs. However, is there a way that we can interpret activations with... |
712ec2fa-c931-45aa-a7db-126b80031083 | trentmkelly/LessWrong-43k | LessWrong | Humans as a truth channel
Crossposted at Intelligence Agents Forum.
Defining truth and accuracy is tricky, so when I've proposed designs for things like Oracles, I've either used a very specific and formal question, or and indirect criteria for truth.
Here I'll try and get a more direct system so that an AI will tel... |
1444df2c-93f2-49aa-978b-70abf7b0cd12 | StampyAI/alignment-research-dataset/blogs | Blogs | Model Mis-specification and Inverse Reinforcement Learning
In my previous post, “[Latent Variables and Model Mis-specification](https://jsteinhardt.wordpress.com/2017/01/10/latent-variables-and-model-mis-specification/)”, I argued that while machine learning is good at optimizing accuracy on observed signals, it has l... |
eb8c1ae4-6675-4da1-bf2c-561dcb713028 | trentmkelly/LessWrong-43k | LessWrong | Evaluating Synthetic Activations composed of SAE Latents in GPT-2
This research was completed for London AI Safety Research (LASR) Labs 2024. The team was supervised by Stefan Heimershiem (Apollo Research). Find out more about the programme and express interest in upcoming iterations here.
TL;DR
We find that real ac... |
0367ba47-fa66-41f5-a9d3-4ae8cb464deb | trentmkelly/LessWrong-43k | LessWrong | Welcome to Oxford Rationalish
Oxford Rationalish is a social group for people interested in things like cognitive biases (ways we might be wrong about what we know or how we think), critical thought (how we might become "less wrong" about the world), and good-faith debate (hoping to discover truth, rather than just wi... |
074b71ac-b364-4041-ae3e-6d4d285157cf | trentmkelly/LessWrong-43k | LessWrong | In the beginning, Dartmouth created the AI and the hype
I've just been through the proposal for the Dartmouth AI conference of 1956, and it's a surprising read. All I really knew about it was its absurd optimism, as typified by the quote:
An attempt will be made to find how to make machines use language, form abstrac... |
2ff581ba-691a-4743-b758-84b2ef028cf7 | trentmkelly/LessWrong-43k | LessWrong | An alternative way to browse LessWrong 2.0
This is something I’ve been tinkering with for a while, but I think it’s now complete enough to be generally useful. It’s an alternative frontend for LessWrong 2.0, using the GraphQL API.
Features:
* Fast, even on low-end computers and phones
* Quickly jump to new c... |
1049affa-c080-4901-ab5f-25c85e1e25f3 | trentmkelly/LessWrong-43k | LessWrong | Sufficiently many Godzillas as an alignment strategy
Assuming that alignment by default happens with nontrivial probability, one way to produce human-aligned AGIs would be to simultaneously create sufficiently many (different) AGI. This leads to a multipolar scenario where there are a few AGI aligned with humans, and ... |
f2125aa3-7ff3-45c3-9b94-ab29d9d22d4b | trentmkelly/LessWrong-43k | LessWrong | Reasons to believe
I've been thinking recently that I believe in the Theory of Evolution on about the same level as in the Theory of Plate Tectonics. I have grown up being taught that both are true, and I am capable of doing research in either field, or at least reading the literature to examine them for myself. I hav... |
2f1477ff-8943-4432-bb41-7a5212f5a734 | trentmkelly/LessWrong-43k | LessWrong | Validity>Soundness in Creative Writing
Apologies for the clickbaity title. A question that I've often come across is how much freedom should science fiction writers have in world building. Is it permissible to take liberty with the laws of physics? Is it OK to depict anthropomorphic AIs?
I hold the opinion that fanta... |
eeb4d129-f628-4453-ad8a-61943f85c745 | trentmkelly/LessWrong-43k | LessWrong | Glossary Dance Game
Harris recently laser-cut some tokens for walking through contra dances (presumably in preparation for calling BIDA this Sunday) and remarked that they looked like game pieces:
At dinner we talked some about trying to make this into a proper game, and made a bit of progress but didn't get to a f... |
06a6b374-f954-46c8-a845-1012a10b6221 | trentmkelly/LessWrong-43k | LessWrong | Hyperreals in a Nutshell
Epistemic status: Vaguely confused and probably lacking a sufficient technical background to get all the terms right. Is very cool though, so I figured I'd write this.
> And what are these Fluxions? The Velocities of evanescent Increments? And what are these same evanescent Increments? They a... |
cf8c1b8c-1417-400f-bbe4-0f39b0ebd027 | trentmkelly/LessWrong-43k | LessWrong | Scott Alexander 2021 Predictions: Market Prices
Scott Alexander has posted some predictions for 2021. Taking Zvi's approach from here and rather than making my own adjustments, making my best estimate of what various prediction (and financial) markets are saying the odds are.
If anyone has seen a market for any of t... |
bb69ff81-b577-4c12-b168-4a85a4a9bec8 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Oracles: reject all deals - break superrationality, with superrationality
.mjx-chtml {display: inline-block; line-height: 0; text-indent: 0; text-align: left; text-transform: none; font-style: normal; font-weight: normal; font-size: 100%; font-size-adjust: none; letter-spacing: normal; word-wrap: normal; word-spacing:... |
cd87d484-7de7-4487-bc1d-eac903cc89d7 | trentmkelly/LessWrong-43k | LessWrong | A social norm against unjustified opinions?
A currently existing social norm basically says that everyone has the right to an opinion on anything, no matter how little they happen to know about the subject.
But what if we had a social norm saying that by default, people do not have the right to an opinion on anything... |
caa4062e-dfde-42f5-8a70-4e85490e5e89 | trentmkelly/LessWrong-43k | LessWrong | Dating Roundup #1: This is Why You’re Single
Developments around relationships and dating have a relatively small speed premium, so I figured I would wait until I had a full post worth of them.
Indeed I now present such a post, in which I present several theories as to why so many of you might still be single.
While... |
b4ee90a9-467b-4bb7-88a7-e533c8d28429 | StampyAI/alignment-research-dataset/blogs | Blogs | diversity vs novelty
diversity vs novelty
--------------------
[a friend of mine](https://ouroborista.neocities.org/) likes to say "novelty is inherently good"; as for myself, i've made known how much i care about cultural diversity ([1](from-above-fine-grain-diversity.html), [2](systems-and-diversity.html), [3](alb... |
74caddd5-3829-4910-979d-089b13288906 | trentmkelly/LessWrong-43k | LessWrong | When did Eliezer Yudkowsky change his mind about neural networks?
In 2008, Eliezer Yudkowsky was strongly critical of neural networks. From his post "Logical or Connectionist AI?":
> Not to mention that neural networks have also been "failing" (i.e., not yet succeeding) to produce real AI for 30 years now. I don't t... |
01de151f-afee-4d9a-a3a5-80a630086426 | trentmkelly/LessWrong-43k | LessWrong | Double Cruxing the AI Foom debate
This post is generally in response to both Paul Christiano’s post and the original foom debate. I am trying to find the exact circumstances and conditions which would point to fast or slow AI takeoff.
Epistemic Status: I have changed my mind a few times about this. Currently thinking... |
37e586a5-1ea5-49d0-8a91-114b22c8b16b | trentmkelly/LessWrong-43k | LessWrong | Questions for a Friendly AI FAQ
I've begun work (with a few others) on a somewhat comprehensive Friendly AI F.A.Q. The answers will be much longer and more detailed than in the Singularity FAQ. I'd appreciate feedback on which questions should be added.
1. Friendly AI: History and Concepts
1. What is Friendly... |
1b0fd074-9d46-4a31-8715-0d3b657fdae1 | trentmkelly/LessWrong-43k | LessWrong | LW Team Updates - October 2019
Like last month, this is a once-monthly updates for LessWrong team activities and announcements.
Please also feel free to use the comments section on this post as a Schelling point to give feedback, file bug reports, or ask questions you have about the site. (You can also email us, ask ... |
a8d05fef-acf4-4351-8a81-6fa361a81ba6 | StampyAI/alignment-research-dataset/lesswrong | LessWrong | Clarifying how misalignment can arise from scaling LLMs
I'm trying to understand how the classical case for AI safety (Bostrom/Yudkowsky argument) best works for LLMs.
My impression is that LLMs have passed an important threshold, they have learnt to learn (for instance, GPT 3 understands translation and can learn a ... |
01fd5131-5489-4577-bc64-f87ad0b201d3 | trentmkelly/LessWrong-43k | LessWrong | There is no IQ for AI
Most disagreement about AI Safety strategy and regulation stems from our inability to forecast how dangerous future systems will be. This inability means that even the best minds are operating on a vibe when discussing AI, AGI, SuperIntelligence, Godlike-AI and similar endgame scenarios. The trou... |
e3d5f75e-bf4c-4929-86d3-b3d54db45862 | trentmkelly/LessWrong-43k | LessWrong | Harry Potter and the Methods of Rationality discussion thread, March 2015, chapter 120
This is a new thread to discuss Eliezer Yudkowsky’s Harry Potter and the Methods of Rationality and anything related to it. This thread is intended for discussing chapter 120.
Plans for next chapter release:
> The next update will... |
9057be37-bf21-4d09-9aec-3a49705ca6d6 | StampyAI/alignment-research-dataset/youtube | Youtube Transcripts | Reinforcement Learning 2: Exploration and Exploitation
let's get started welcome to the second
lecture on the reinforcement learning
this one's about exploration and
exploitation and I'll go into depth
about what I mean when I say those words
I said something about that last lecture
in terms of logistics just to make ... |
cda04f48-0ab7-4545-ac1f-b543a100fe6d | trentmkelly/LessWrong-43k | LessWrong | Synthesis of subagents: exercise
What topics is the exercise related to:
1. Mitigation of cognitive distortions such as the stereotype of physical attractiveness
2. Sequence "Multiagent models of mind", article “Why Subagents?”
3. article "Being the (Pareto) Best in the World"
Two sources are the basis for this ex... |
af3caaa7-fe61-4a7b-8ca6-229b79e314d9 | trentmkelly/LessWrong-43k | LessWrong | Asking for a name for a symptom of rationalization
What is the most specific word for the phenomenon of when someone holds an opinion very strongly on an emotional level but whose reasons for holding that opinion that seem very weak and don't evoke a strong emotional response? |
dfffdddb-b188-40a3-8ea8-5fd089576c82 | trentmkelly/LessWrong-43k | LessWrong | Why Psychologists Are Wrong About The Illusion Of Explanatory Depth
The Khan Academy did a video where they asked random people in the streets what the internet is, a question to which almost all gave some ridiculous answers —confidently. A young girl said it is the WiFi because, according to her, "no WiFi, no interne... |
efcdfca8-3a84-4186-89b8-f14805726219 | trentmkelly/LessWrong-43k | LessWrong | Meetup : The first meetup in Kyiv, Ukraine
Discussion article for the meetup : The first meetup in Kyiv, Ukraine
WHEN: 02 August 2015 06:30:00PM (+0300)
WHERE: Kyiv, Ukraine
Hello =)
I glad to announce the first meetup in Kyiv.
Our google group: https://groups.google.com/forum/?#!forum/lesswrong-kyiv
Please join... |
383adb18-a10a-4b41-abfe-7577384e1024 | trentmkelly/LessWrong-43k | LessWrong | Should we push for requiring AI training data to be licensed?
We frequently speak about AI capability gain being bad because it shortens the timeframe for AI safety research. In that logic, taking steps to decrease AI capability would be worthwhile.
At the moment the large language models are trained with a lot of da... |
10ccd302-d6d3-454c-b9af-69210909e528 | trentmkelly/LessWrong-43k | LessWrong | Catalogue of POLITICO Reports and Other Cited Articles on Effective Altruism and AI Safety Connections in Washington, DC
"Brendan Bordelon is a reporter for POLITICO on lobbying and influence in Washington, DC, by the tech industry. In late 2023, Bordelon began writing a series of articles investigating the influence ... |
db0a095d-01b6-48e1-bdcf-9dbaca56c701 | trentmkelly/LessWrong-43k | LessWrong | High Failure-Rate Solutions
Short interview, doesn't go into too much depth, but makes an interesting point relevant to LW:
"When you’re mitigating a complex problem, you become a bad picker. It’s too complicated, so you can’t separate the ideas that are going to work and the ones that won’t. [...] But if we were gon... |
2932fd3b-dbc2-4085-844c-cee04467a6dc | trentmkelly/LessWrong-43k | LessWrong | Payor's Lemma in Natural Language
Preceded by: Modal Fixpoint Cooperation without Löb's Theorem
It turns out Payor's Lemma and its proof can be explained in natural language even more easily than Löb's Theorem. Here's how.
Imagine a group of people, and let x denote the statement "everyone in the group cooperates"... |
64b5ca6d-3a37-4b53-8f1d-01050faa9922 | trentmkelly/LessWrong-43k | LessWrong | Meetup : Frankfurt meetup
Discussion article for the meetup : Frankfurt meetup
WHEN: 09 February 2014 02:00:00PM (+0100)
WHERE: Frankfurt
We are meeting again! If you have any special requirements to attend or need some help, please try me in advance, we'll try to accommodate your needs.
Location will be published... |
27c177a5-7ce8-428f-992d-d422c5be98fc | trentmkelly/LessWrong-43k | LessWrong | Why We Need More Shovel-Ready AI Notkilleveryoneism Megaproject Proposals
A lot of people within AI safety and adjacent ecosystems have discussed AI notkilleveryoneism megaproject ideas, but there appears to be a key problem with these previous discussions. These discussions presuppose that the funding landscape will ... |
17df4d77-39fd-40c7-bf13-58572598faa9 | trentmkelly/LessWrong-43k | LessWrong | AXRP Episode 16 - Preparing for Debate AI with Geoffrey Irving
YouTube link
Many people in the AI alignment space have heard of AI safety via debate - check out episode 6 of AXRP if you need a primer. But how do we get language models to the stage where they can usefully implement debate? In this episode, I talk to G... |
16df7bc7-1239-479d-8535-9af2ca2cb9d1 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | EAI Alignment Speaker Series #1: Challenges for Safe & Beneficial Brain-Like Artificial General Intelligence with Steve Byrnes
A couple months ago EleutherAI started an alignment speaker series, some of these talks have been recorded. This is the first instalment in the series. The following is a transcript generated ... |
4326c00e-79bb-4dbe-a2ec-9f638a65df65 | trentmkelly/LessWrong-43k | LessWrong | Commenting Patterns by Platform
How does the growth of Bluesky compare to Mastodon in 2022? What about Google Plus in 2011? I can't answer this globally, but (now that my comment archive is up to date) I can look at it for my immediate social network by looking at a proxy: how many of the replies people write to my po... |
d6821476-acf5-461c-90c6-3bf5c07e63ab | trentmkelly/LessWrong-43k | LessWrong | Bayeswatch 3: A Study in Scarlet
Miriam and Vi brunched at the Dusty Knuckle. Miriam's cell phone rang.
"Hello?" said Miriam.
"Someone painted the roof of the Glasgow Weather Prediction Center scarlet," said the voice.
Miriam hung up her phone. She activated voice obfuscation. Miriam called another number.
"This i... |
ab42f01c-1cc1-4291-9d7f-25595f830060 | trentmkelly/LessWrong-43k | LessWrong | Meetup : Portland, OR: Improv for Rationalists
Discussion article for the meetup : Portland, OR: Improv for Rationalists
WHEN: 18 October 2014 01:00:00PM (-0700)
WHERE: 2945 NE 64th Ave, Portland, OR
Last meetup we discussed using techniques from acting improv to improve social skills and impose personal habits. Th... |
c16c9d75-5b9a-4112-b9fc-969fe3b6a594 | trentmkelly/LessWrong-43k | LessWrong | AI alignment with humans... but with which humans?
Note: I'm cross-posting this from EA Forum (where I posted it on Sept 8, 2022), in case anybody on LessWrong or the AI Alignment Forum is interested in commenting; note that there were some very helpful suggested readings in the replies to this: https://forum.effecti... |
01f7d573-2f18-4edf-8c5d-85e909d65154 | trentmkelly/LessWrong-43k | LessWrong | Scaling of AI training runs will slow down after GPT-5
My credence: 33% confidence in the claim that the growth in the number of GPUs used for training SOTA AI will slow down significantly directly after GPT-5. It is not higher because of (1) decentralized training is possible, and (2) GPT-5 may be able to increase ha... |
57f91a63-ecbf-4996-bae3-750eeda7dcb6 | trentmkelly/LessWrong-43k | LessWrong | What causes a decision theory to be used?
Most (all?) of the decision theory discussion I've seen has been about prediction or contradiction of choice, mostly about how to model the fact that decisions have causes, and those causes could be legible to other agents, or correlated with future experiences in ways that af... |
999a05b8-e8bd-4d4d-afa4-ed66b254503b | trentmkelly/LessWrong-43k | LessWrong | "Evil" decision problems in provability logic
A while ago, drnickbone showed in a LessWrong post that for any decision theory, there is a "fair" decision problem on which that decision theory fails---that is, gets a low reward, even though a simple alternative decision theory gets a high reward. Although drnickbone's ... |
03b5f575-963b-49f2-8191-919ab118c95c | trentmkelly/LessWrong-43k | LessWrong | AI models inherently alter "human values." So, alignment-based AI safety approaches must better account for value drift
Hi all. This post outlines my concerns over the effectiveness of alignment-based solutions to AI safety problems. My basic worry is that any rearticulation of human values, whether by a human or an A... |
3de10d38-3c26-4b32-8047-e39188b8c457 | trentmkelly/LessWrong-43k | LessWrong | How "honest" is GPT-3?
Edited to add: There is now a 1k bounty for a closely related question.
In Inaccessible Information, Paul writes about the strategy "BAD", in which an AI system uses its understanding of the world to give answers that humans will find very plausible and rate highly, in contrast to a more honest... |
8e1c7902-9575-4f87-b551-db25ce28639e | trentmkelly/LessWrong-43k | LessWrong | When not to know?
Jeff at Cheap Talk reports on Andrew Caplin’s good point: making tests less informative can make people better off, because often they don’t want that much information, but may still want a bit.
This reminded me of a more basic question: what makes people want to avoid getting information?
That is,... |
12014fb9-6afc-4158-9f70-24011300eff2 | trentmkelly/LessWrong-43k | LessWrong | Book Review: Orality and Literacy: The Technologizing of the Word
Thanks to Diarmuid Morgan for reviewing an earlier version of this post.
> Dream cancels dream in this new realm of fact
> From which we wake into the dream of act[1]
Review of Orality and Literacy: The Technologizing of the Word, by Walter J. Ong, 1... |
0f71036f-4e4a-48cf-b43a-8dc990e71428 | trentmkelly/LessWrong-43k | LessWrong | How to Know Yourself
Cross-posted from https://regkith.substack.com/p/how-to-know-yourself
> “To thine own self be true.”
While not the best father figure, at least Polonius got this one right.
Or did he? It’s definitely appealing, but what does it mean to know yourself? You might think you know yourself, no doubt ... |
b78dbd2c-2be0-4e59-b5d6-6dda37aade93 | trentmkelly/LessWrong-43k | LessWrong | Native Russian speakers wanted (for proofreading LW texts)
As part of an effort to familiarize Russian-speaking audience with Less Wrong materials, I am participating in an ongoing translation of a certain sequence of texts to Russian.
Now I need volunteers who are willing to read the text and give feedback on its qu... |
62d5f7b6-24f4-44b6-a0bc-1b8ed12d92a2 | trentmkelly/LessWrong-43k | LessWrong | Suggest best book as an introduction to computational neuroscience
Im trying to find a best place to start learning the field. I have no special math background. Im very eager to learn. Thanks alot!
|
5dfd7c82-0c91-4c85-9f38-ecc89c28839a | trentmkelly/LessWrong-43k | LessWrong | The Risk of Orbital Debris and One (Cheap) Way to Mitigate It
I worked as a researcher on orbital debris during university and now work full time on this and adjacent problems. My full write-up on the approach for using cheap radar data to assess orbital debris is here on my blog.
Orbital Debris
There has been a con... |
5f96337f-b9dc-4b26-bbff-cf9ea29304cd | StampyAI/alignment-research-dataset/special_docs | Other | A Note on the Existence of Ratifiable Acts.
A Note on the Existence of Rati able Acts
Joseph Y. Halpern
Cornell University
Computer Science Department
Ithaca, NY 14853
halpern@cs.cornell.edu
http://www.cs.cornell.edu/home/halpern
August 15, 2018
Abstract
Sucient conditions are given under which rati able acts exist.... |
e595e355-9584-49d4-b386-1d5453197569 | trentmkelly/LessWrong-43k | LessWrong | Connections between Mind-Body Problem & Civilizations
Recently I got into the “mind-body problem,” a philosophical term for the question of how our mental consciousness/awareness is related to our physical body/world. Take my thoughts with a grain of salt.
What I noticed is that a civilization’s institutions influenc... |
dc864d55-0f07-49eb-a42d-4fdeec046694 | StampyAI/alignment-research-dataset/lesswrong | LessWrong | Decision Theories: A Less Wrong Primer
**Summary:** *If you've been wondering why people keep going on about decision theory on Less Wrong, I wrote you this post as an answer. I explain what decision t... |
ef4c130d-5e01-4bbc-ac61-fd87365b6c28 | trentmkelly/LessWrong-43k | LessWrong | Cambridge Less Wrong meetup this Sunday, May 16
Last month's Cambridge, Massachusetts meetup was a success, with six people in attendance and some very interesting discussions. Meetups will recur on the third Sunday of each month, so the next one is this week, at the same time and place: 4pm at the Clear Conscience Ca... |
c4043f41-2757-4582-901e-bbd72712368d | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Training Trace Priors
I’m worried about scenarios involving deceptive models. We’ve failed at inner alignment so the model has goals that are not aligned with ours. It can somehow detect when it’s in training, and during training it pretends to share our goals. During deployment, surprise! The model paperclips the uni... |
60a859d9-7529-4d88-88a2-13f876940ddb | trentmkelly/LessWrong-43k | LessWrong | Preregistration: Air Conditioner Test
Warning: None of the participants in the Great Air Conditioner Debate Of 2022 have endorsed my summaries of their positions in this post. Including me.
Background
In Everything I Need To Know About Takeoff Speeds I Learned From Air Conditioner Ratings On Amazon, I complained abo... |
64f8ebaf-683d-480b-a1d5-e27211488e30 | trentmkelly/LessWrong-43k | LessWrong | Variadic functions in Hindley Milner
Note: math didn't transfer to LW. You might find this easier to read on my blog.
I previously mentioned an interest in adding variadic functions to Haskenthetical, the "Haskell with a Lisp syntax" I'm vaguely working on.
It sounds difficult. This is an attempt to figure out just ... |
4e9fcc19-7daa-40d2-92ec-3c33ddef7c6c | trentmkelly/LessWrong-43k | LessWrong | Spamming Micro-Intentions to Generate Willpower
The purpose of this post is to outline a potentially general trick for getting yourself out of a rut or motivating yourself to do a small task that you don't feel like doing. Part (i) describes the source of the idea. Part (ii) outlines an example for testing the techniq... |
52d6ee4c-7966-4cce-bb98-171cc5689252 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Discussion on the machine learning approach to AI safety
(Cross-posted from [personal blog](https://vkrakovna.wordpress.com/2018/11/01/discussion-on-the-machine-learning-approach-to-ai-safety/).)
At this year's [EA Global London](http://london.eaglobal.org/) conference, Jan Leike and I ran a [discussion session](http... |
7c060454-c44e-4c81-afc3-d4bd9400318c | trentmkelly/LessWrong-43k | LessWrong | Frontier Model Security
> As the capabilities of frontier artificial intelligence models continue to increase rapidly, ensuring the security of these systems has become a critical priority. In our previous posts, we’ve focused on Anthropic’s approach to safety, and Claude’s capabilities and applications. In this post,... |
831fbbc8-a062-4f1b-b4a2-421d29010f0e | StampyAI/alignment-research-dataset/arxiv | Arxiv | Training a Resilient Q-Network against Observational Interference
Introduction
------------
Deep reinforcement learning (DRL) methods have shown enhanced performance, gained widespread applications (Mnih et al. [2015](#bib.bib48), [2016](#bib.bib47); Silver et al. [2017](#bib.bib67)), and improved robot learning (G... |
a8c37172-6a0f-4bee-a486-181015f9f996 | trentmkelly/LessWrong-43k | LessWrong | Finding the Central Limit Theorem in Bayes' rule
I feel like there's something in the basic Bayesian update formula that suggests that, as the amount of input data grows, almost every Bayesian posterior will tend toward a Gaussian. This post is me trying to scrape at that vague intuition and get a handle on it. I'm go... |
214142ce-882e-4d16-b908-102d13b3d6f4 | trentmkelly/LessWrong-43k | LessWrong | Quick Thoughts on Our First Sampling Run
Cross-posted from my NAO Notebook
While the NAO has primarily focused on wastewater, we're now spinning up a swab sampling effort. The idea is to go to busy public places, ask people to swab their noses, pool the swabs, sequence them, and look for novel pathogens. We did our f... |
06a8dea2-867a-41fb-a037-f7d46dde7e26 | trentmkelly/LessWrong-43k | LessWrong | Cynical About Cynicism
I'm cynical about cynicism. I don't believe that most cynicism is really about knowing better. When I see someone being cynical, my first thought is that they're trying to show off their sophistication and assert superiority over the naive. As opposed to, say, sharing their uncommon insight a... |
d2d040f4-c782-49b7-a2fc-4ad35b7bd97d | trentmkelly/LessWrong-43k | LessWrong | Technologies and Terminology: AI isn't Software, it's... Deepware?
A few weeks ago, I (David) tried to argue that AI wasn’t software. In retrospect, I think I misstated the case by skipping some inferential steps, and based on some feedback from that article, and on an earlier version of this article, with a large ass... |
8439f9f2-dfae-4524-b0a2-862404577732 | trentmkelly/LessWrong-43k | LessWrong | Ambiguity in Prediction Market Resolution is Harmful
(Disclaimers: I work in the financial industry, though not in a way related to prediction markets. Anything I write here is my opinion and not that of my employer.)
SOMEBODY SET US UP THE BOMB
You've heard of this new 'prediction market' fad, so you decide to try... |
3fa43f7f-0316-4fa6-9bc3-f2b7f4ca2340 | trentmkelly/LessWrong-43k | LessWrong | [Link] Digital Democracy Is Within Reach
EPISODE SUMMARY
Imagine a world where every country has a digital minister and technologically-enabled legislative bodies. Votes are completely transparent and audio and video of all conversations between lawmakers and lobbyists are available to the public immediately. Conspir... |
b0dfef77-acbb-48a0-a243-ab26c9a51d33 | StampyAI/alignment-research-dataset/blogs | Blogs | GPTs are GPTs: An early look at the labor market impact potential of large language models
WORKING PAPER
GPTs are GPTs: An Early Look at the Labor Market Impact
Potential of Large Language Models
Tyna Eloundou1, Sam Manning1,2, Pamela Mishkin∗1, and Daniel Rock3
1OpenAI
2OpenResearch
3University of Pennsylvania
March ... |
836a99f1-52a6-4f6c-89b9-54b3bb841761 | StampyAI/alignment-research-dataset/eaforum | Effective Altruism Forum | Introducing the Mental Health Roadmap Series
According to the United States’ Center for Disease Control (ᴄᴅᴄ), **over half of all people will be diagnosed with a mental illness or disorder at some point in their lifetime.**[[1]](#fnvxc9ehtdmlo) This is an undercount since not everyone who has one receives a diagnosis.... |
ca8fccc8-6ce8-4ea6-b540-237da9fbf412 | trentmkelly/LessWrong-43k | LessWrong | To determine alignment difficulty, we need to know the absolute difficulty of alignment generalization
A core insight from Eliezer is that AI "capabilities generalize further than alignment once capabilities start to generalize far".
This seems quite true to me, but doesn't on its own make me conclude that alignment ... |
97d48612-f33c-44b5-94a4-b07a488fde01 | trentmkelly/LessWrong-43k | LessWrong | The ELK Framing I’ve Used
Epistemic Status: Uncertain.
This post assumes familiarity with ARC’s Eliciting Latent Knowledge paper.
After working on ELK for a little while, I arrived at a slightly differently problem. I feel like this problem still captures the core difficulty of ELK.
The purpose of this post is not ... |
9864e9f4-5705-486b-9e4c-2ffd1f4ed8aa | trentmkelly/LessWrong-43k | LessWrong | Group Rationality Diary, January 1-15
Happy new year to those who observe the Gregorian calendar! This is the public group instrumental rationality diary for January 1-15.
> 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
> * Ob... |
c2b2f342-7b19-4fb7-9b3a-643c2558941a | trentmkelly/LessWrong-43k | LessWrong | Why is rhetoric taboo among rationalists?
We acknowledge that sometimes type1 solutions are better that type2 solutions and the only way to change type1 processes is with rhetoric. But improving type1 systems is a cardinal sin. |
10041c77-0655-4ee1-830c-5eac183ccfc8 | trentmkelly/LessWrong-43k | LessWrong | Link - total body transplant
Amazing video. This is really "out there" medical research that actually worked, and not surprisingly, was closed down.
http://www.vbs.tv/newsroom/dr-white-s-total-body-transplant-2-of-2--2 |
072ff910-ed86-46c9-9e10-7bfb2e90468c | trentmkelly/LessWrong-43k | LessWrong | Could a single alien message destroy us?
Merely listening to alien messages might pose an extinction risk, perhaps even more so than sending messages into outer space. Our new video explores the threat posed by passive SETI and potential mitigation strategies.
Below, you can find the script of the video. Matthew Bar... |
12cec813-2027-422d-a96a-811999645c29 | trentmkelly/LessWrong-43k | LessWrong | Anthropics and Embedded Agency
I am not very familiar with the AI field, which is probably why I have only recently come across Scott Garrabrant's posts about embedded agency. To my surprise, I found it connected with anthropics - something I have been interested in for a long time. In particular, my solution to the p... |
dd5b3a69-8f0c-48b7-bb2d-17cca78c8f0b | trentmkelly/LessWrong-43k | LessWrong | Towards equilibria-breaking methods
All of this is vague intuition.
Much has been said about the problems of Nash equilibria. I've been wondering about how to quantify these sorts of problems. From Moloch's Toolbox:
> 3. Systems that are broken in multiple places so that no one actor can make them better, even thou... |
6a625ac7-fe8e-4d20-97c7-1a93f02f76a3 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | AI that shouldn't work, yet kind of does
There are some things that work surprisingly well in AI. For example, AI that transfers [the style of one image to the content of another.](https://www.tensorflow.org/tutorials/generative/style_transfer) Why is the approach described here a hack. It starts with a neural net tra... |
6141da76-fab7-47e9-b933-4da6c1d1b926 | trentmkelly/LessWrong-43k | LessWrong | Catastrophic Risks from AI #3: AI Race
This is the third post in a sequence of posts giving an overview of catastrophic AI risks.
3 AI Race
The immense potential of AIs has created competitive pressures among global players contending for power and influence. This “AI race” is driven by nations and corporations who ... |
d371dc9b-363d-4ecf-b1f1-32d5169ce709 | trentmkelly/LessWrong-43k | LessWrong | Many of us *are* hit with a baseball once a month.
Watching the video of Eliezer's Singularity Summit 2010 talk, I thought once more about the 'baseball' argument. Here's a text version from How to Seem (and Be) Deep:
> [...] given human nature, if people got hit on the head by a baseball bat every week, pretty soon... |
98bc9b31-ed08-4c19-8999-0482e65564f3 | trentmkelly/LessWrong-43k | LessWrong | Help Build a Landing Page for Existential Risk?
The Big Orange Donate Button
Traditional charities, like Oxfam, Greenpeace, and Amnesty International, almost all have a big orange button marked "Donate" right on the very first page that loads when you go to their websites. The landing page for a major charity usually ... |
662c12fc-5eb9-48d1-8017-120f127c37c9 | trentmkelly/LessWrong-43k | LessWrong | [SEQ RERUN] Fairness vs. Goodness
Today's post, Fairness vs. Goodness was originally published on 22 February 2009. A summary (taken from the LW wiki):
> An experiment in which two unprepared subjects play an asymmetric version of the Prisoner's Dilemma. Is the best outcome the one where each player gets as many p... |
781acc7d-5a1c-4c2e-8ae8-e3f55bfbcaae | trentmkelly/LessWrong-43k | LessWrong | How Efficient is the Charitable Market?
When I talk about the poor distribution of funds in charity, people in the effective altruism movement sometimes say, "Didn't Holden Karnofsky show that charity is an efficient market in his post Broad Market Efficiency?"
My reply is "No. Holden never said, and doesn't believe,... |
58dc6791-b07d-4b30-bb64-8e333f155325 | trentmkelly/LessWrong-43k | LessWrong | Distinguishing ways AI can be "concentrated"
Richard Ngo recently wrote,
> Instead of analyzing whether AI takeoff will be “fast” or “slow”, I now prefer to think about the spectrum from concentrated takeoff (within one organization in one country) to distributed takeoff (involving many organizations and countries).... |
76bf0f14-bb10-4ab4-b8fc-cdfa68c96865 | trentmkelly/LessWrong-43k | LessWrong | Bay Solstice 2017: Thoughts
To me, this year's Solstice was a flop. While it's still somewhat fresh in my mind, I would like to say why.
(A few of these thoughts are ones I shared with others and found basic agreement. The rest were not discussed and may be entirely idiosyncratic.)
At the design level, it was almo... |
4ca02c38-59f2-4a65-9862-f212cdb27ddd | trentmkelly/LessWrong-43k | LessWrong | Meetup : Sydney Rationality Dojo - September
Discussion article for the meetup : Sydney Rationality Dojo - September
WHEN: 06 September 2015 04:00:00PM (+1000)
WHERE: Humanist House, 10 Shepherd St Chippendale
Join us for another fantastic rationality dojo. Afterwards we'll meet up somewhere nearby for dinner & con... |
ce298deb-af1d-4e72-b73e-e6c9edea6f41 | trentmkelly/LessWrong-43k | LessWrong | A dataset of questions on decision-theoretic reasoning in Newcomb-like problems
I’ve spent a lot of the last few years working on issues related to acausal cooperation. With LLMs being clearly dominant over recent years, I’ve now led a team to make a benchmark to figure out how good LLMs are at decision theory and whe... |
decc540d-ced2-4780-ae72-3a0f56d984d9 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Stable self-improvement as a research problem
“Stable self-improvement" seems to be a primary focus of MIRI’s work. As I understand it, the problem is "How do we build an agent which rationally pursues some goal, is willing to modify itself, and with very high probability continues to pursue the same goal after modifi... |
0f2056ba-00b3-4ae6-bd24-57f8de795319 | StampyAI/alignment-research-dataset/arxiv | Arxiv | Human-Level Reinforcement Learning through Theory-Based Modeling, Exploration, and Planning
Acknowledgments
---------------
The authors would like to acknowledge Aritro Biswas, Jacqueline Xu, Kevin Wu, Eric Wu, Zhenglong Zhou, Michael Janner, Kris Brewer, Hector Geffner, Marty Tenenbaum, and Alex Kell. The work was... |
54841a92-5d8a-4b38-a07a-a4774cfbe539 | trentmkelly/LessWrong-43k | LessWrong | Security services relationship to social movements
As there has been talk of the necessities of a security mindset for forming an agi research project, I thought it worthwhile to share this so that rationalists have some information about the types of forces that may be at play, if someone decided to form such a group... |
e30b2ade-fdfb-4857-a52e-18e5e2be288a | trentmkelly/LessWrong-43k | LessWrong | My workflow
Over the last 6 months I've started doing a lot of things differently. Some of these changes seem to have increased my work output a good bit and made me happier. I normally hesitate to share habits, but I'm pretty happy with these in particular, and even if they will work for only a few people I think ... |
7fdeb70c-2d24-4060-bfc3-2995efec3ea7 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Active Inference as a formalisation of instrumental convergence
The goal of this post is mainly to increase the exposure of the AI alignment community to Active Inference theory, which seems to be highly relevant to the problem but is seldom mentioned on the forum.
This post links to a freely available book about Act... |
606d2829-aab2-4855-9f2a-5b12de6bba1a | trentmkelly/LessWrong-43k | LessWrong | Young Cryonicist Gathering Warning
Edit 1: context: Cryonics is a common topic on Less Wrong. There are a lot of Less Wrong cryonicists... and a good portion of people to whom this warning is relevant to are Less Wrongers. So this seems like as good a place as any to post this, since I'm just trying to get some potent... |
38acf305-a84f-4cf4-9422-34a5581d446c | trentmkelly/LessWrong-43k | LessWrong | Dopamine-supervised learning in mammals & fruit flies
(Last revised: December 2024. See changelog at the bottom.)
The backstory: Dopamine-supervised learning in mammals
According to a well-known theory from the 1990s, dopamine signals in mammal brains can serve as Reward Prediction Error signals that drive reinforce... |
f48627b2-6d38-4e8a-9dd7-c21f1c262517 | trentmkelly/LessWrong-43k | LessWrong | Beyond a better world
As I take man’s last step from the surface, back home for some time to come — but we believe not too long into the future — I’d like to just say what I believe history will record: that America’s challenge of today has forged man’s destiny of tomorrow. And, as we leave the Moon at Taurus–Littrow,... |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.