id
stringlengths
36
36
source
stringclasses
15 values
formatted_source
stringclasses
13 values
text
stringlengths
2
7.55M
7b032e3b-25c1-473b-83dc-29ca4e07395e
trentmkelly/LessWrong-43k
LessWrong
Why Parents Should Let Their Children Play More Video Games The gaming industry has a huge and gradually expanding audience. According to Statista, 2.69 billion people enjoyed digital journeys in 2020. Their number jumped to 2.81 billion in 2021. This data covers men, women, and kids. The latter category is especially...
1cc3238c-5b69-451d-9e4c-3d823599589b
trentmkelly/LessWrong-43k
LessWrong
If you're not sure how to sort a list or grid—seriate it! "Getting Things in Order: An Introduction to the R Package seriation", Hahsler et al 2008: > Seriation [or "ordination"], i.e., finding a suitable linear order for a set of objects given data and a loss or merit function, is a basic problem in data analysis. C...
da4aa28d-b70a-4557-bfac-0e6572a1e9b3
trentmkelly/LessWrong-43k
LessWrong
Can startups be impactful in AI safety? With Lakera's strides in securing LLM APIs, Goodfire AI's path to scaling interpretability, and 20+ model evaluations startups among much else, there's a rising number of technical startups attempting to secure the model ecosystem. Of course, they have varying levels of impact ...
f0f618c9-17e0-4d1f-a446-9ddec995fdee
trentmkelly/LessWrong-43k
LessWrong
Meetup : Baltimore Weekly Meetup Discussion article for the meetup : Baltimore Weekly Meetup WHEN: 17 July 2016 08:00:00PM (-0400) WHERE: 1726 reisterstown road pikesville maryland 21208 Pikesville DoubleTree - Hilton, at the restaurant / bar or at one of the tables outside. Discussion article for the meetup : Ba...
bd352c7e-b8ba-45e2-9387-74762abe2d36
trentmkelly/LessWrong-43k
LessWrong
Updates and additions to "Embedded Agency" Abram Demski and Scott Garrabrant's "Embedded Agency" has been updated with quite a bit of new content from Abram. All the changes are live today, and can be found at any of these links: * as a hand-drawn sequence (LW link, AIAF link); * as blog posts (MIRI link, LW link, ...
317384c4-4305-4ebd-b81a-da34c40ad79f
trentmkelly/LessWrong-43k
LessWrong
Emergence Spirals—what Yudkowsky gets wrong Before proposing the following ideas and critiquing those of Yudkowsky, I want to acknowledge that I’m no expert in this area—simply a committed and (somewhat) rigorous speculator. This is not meant as a “smack down” (of a post written 18 years ago) but rather a broadening o...
4f9eec5f-1a43-428f-ad6f-59805f6fa46d
trentmkelly/LessWrong-43k
LessWrong
January 2011 Oxford Intelligence Conference videos online Here. Including, for example, Nick Bostrom's talk "Superintelligence: The Control Problem" and Jurgen Schmidhuber's talk "Universal AI & Formal Theory of Fun."
48332e4e-7598-4f14-9bd4-02daf280a097
trentmkelly/LessWrong-43k
LessWrong
Anthropic Decision Theory IV: Solving Selfish and Average-Utilitarian Sleeping Beauty A near-final version of my Anthropic Decision Theory paper is available on the arXiv. Since anthropics problems have been discussed quite a bit on this list, I'll be presenting its arguments and results in this, subsequent, and previ...
db8f311a-7c44-4cf8-94cc-05189fb2329c
trentmkelly/LessWrong-43k
LessWrong
Decoupling deliberation from competition I view intent alignment as one step towards a broader goal of decoupling deliberation from competition. * Deliberation. Thinking about what we want, learning about the world, talking and learning from each other, resolving our disagreements, figuring out better methodologies ...
c446fe81-b040-4c35-972e-b00e6276d7e5
StampyAI/alignment-research-dataset/eaforum
Effective Altruism Forum
Aim for conditional pauses **TL;DR:** I argue for two main theses: 1. *[Moderate-high confidence]* It would be better to aim for a conditional pause, where a pause is triggered based on evaluations of model ability, rather than an unconditional pause (e.g. a blanket ban on systems more powerful than GPT-4). 2. *[Mode...
7c9ccbe1-f619-46cc-b3ad-a26fb30e79a0
trentmkelly/LessWrong-43k
LessWrong
Manifesto of the Silent Minority Yesterday I asked readers the Thiel Question. 40 people responded. I have combined the responses into a single political platform. You can view the raw responses here. ---------------------------------------- This country is tired of the Democrats, Republicans, Libertarians and human...
00af25c6-7e04-453c-9121-c7e24a854dcd
trentmkelly/LessWrong-43k
LessWrong
Controllables and Observables, Revisited This is the fourth post in the Cartesian frames sequence. Read the first post here. Previously, we defined controllables as the sets of possible worlds an agent can both ensure and prevent, and we defined observables as the sets of possible worlds such that the agent can imple...
105c2c2d-6256-463b-bd13-ed179caca08c
trentmkelly/LessWrong-43k
LessWrong
Is the speed of training large models going to increase significantly in the near future due to Cerebras Andromeda? Cerebras recently unveiled Andromeda - https://www.cerebras.net/andromeda/, an AI supercomputer that enables near linear scaling. Do I understand correctly that this might have a big impact on the large ...
3a2f9bed-ad22-42b4-9109-d93a1a89f881
trentmkelly/LessWrong-43k
LessWrong
Open thread: Language I believe that the discussion of language is currently an underrated component of rationality. The language we use constantly biases us and restricts the way that we see the world. At the same time, many of these influences are so small that they aren't worth writing a separate post about. So wha...
baad13a2-860b-46f2-87e4-14c9f467d258
StampyAI/alignment-research-dataset/blogs
Blogs
October 2019 Newsletter #### Updates * Ben Pace summarizes [a second round of AI Alignment Writing Day posts](https://www.lesswrong.com/posts/DG7asvufKgaqEknKd/ai-alignment-writing-day-roundup-2). * [The Zettelkasten Method](https://www.lesswrong.com/posts/NfdHG6oHBJ8Qxc26s): MIRI researcher Abram Demski describes a...
eff4f2c9-e10e-4e9b-afc4-68fc2d207d1f
trentmkelly/LessWrong-43k
LessWrong
Andrew Gelman on "the rhetorical power of anecdotes" Andrew Gelman has a post up today discussing a particularly illustrative instance of narrative fallacy involving the recent plagiarism discussion surrounding Karl Weick. I think there are also some interesting lessons in there about generalizing from fictional evide...
bd94abb8-d452-449e-9d3e-1249e03d3444
trentmkelly/LessWrong-43k
LessWrong
Post series on "Liability Law for reducing Existential Risk from AI" Gabriel Weil (Assistant Professor of Law, Touro University Law Center) wrote this post series on the role of Liability Law for reducing Existential Risk from AI.  I think this may well be of interest to some people here, so wanted for a linkpost to e...
ecde0e22-84a5-4782-84df-1f0de26176cd
trentmkelly/LessWrong-43k
LessWrong
The Psychology Of Resolute Agents Epistemic Status: Exploratory Consider Parfit's Hitchhiker from the perspective of a completely selfish agent: > "Suppose you're out in the desert, running out of water, and soon to die - when someone in a motor vehicle drives up next to you. Furthermore, the driver of the motor veh...
23b395d4-752b-4820-b991-1da0f2670a11
trentmkelly/LessWrong-43k
LessWrong
Rationality Reading Group: Part J: Death Spirals This is part of a semi-monthly reading group on Eliezer Yudkowsky's ebook, Rationality: From AI to Zombies. For more information about the group, see the announcement post. ---------------------------------------- Welcome to the Rationality reading group. This fortnig...
b8a8f16b-bed3-4da2-9184-980fe6b734f5
trentmkelly/LessWrong-43k
LessWrong
Is LW making progress? Some people have an intuition that with free exchange of ideas, the best ones will eventually come out on top. I'm less optimistic, so I ask if that's really happening. The alternative would be to have the same people talking about the same problems without accumulating anything. People would s...
cc0ba600-5730-4e1a-8324-0f871f6fb98e
StampyAI/alignment-research-dataset/aisafety.info
AI Safety Info
Isn't the real concern AI being misused by terrorists or other bad actors? AI is a tool that could be misused by actors who want to harm you or people like you. Such actors might include [terrorists](https://unicri.it/News/Algorithms-Terrorism-UNICRI-UNOCCT), unfriendly nations, or a [totalitarian state](/?state=6409&...
9ea306ae-32a1-4d02-bcc8-249fc60bcd09
trentmkelly/LessWrong-43k
LessWrong
Where's Your Sense of Mystery? Related to: Joy in the Merely Real, How An Algorithm Feels From Inside, "Science" As Curiosity-Stopper Your friend tells you that a certain rock formation on Mars looks a lot like a pyramid, and that maybe it was built by aliens in the distant past. You scoff, and respond that a lot of...
8fcfc681-2c4d-4029-9036-18ebc38d3eaa
trentmkelly/LessWrong-43k
LessWrong
Timaeus is hiring researchers & engineers TLDR: We're hiring for research & engineering roles across different levels of seniority. Hires will work on applications of singular learning theory to alignment, including developmental interpretability. About Us Timaeus' mission is to empower humanity by making breakthrou...
623f10ed-7455-4235-bf23-b5088fcbd3a7
StampyAI/alignment-research-dataset/youtube
Youtube Transcripts
DeepMind x UCL RL Lecture Series - Planning & models [8/13] hello i'm matthew hassell and i'll be teaching the reinforcement learning module together with hado and diana and today i'll be talking about models and planning in the past few lectures we already covered quite a bit of material so we started with the bandit...
b964b435-8da8-4626-abf3-e5c5c40930dd
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
Emergent modularity and safety Biological neural networks (i.e. brains) and artificial neural networks have sufficient commonalities that it's often reasonable to treat our knowledge about one as a good starting point for reasoning about the other. So one way to predict how the field of neural network interpretability...
35d6fa21-2752-41e0-b2df-597c4fda168a
StampyAI/alignment-research-dataset/aisafety.info
AI Safety Info
Where can I learn more about AI alignment? If you like interactive FAQs, you're in the right place already! Joking aside, some great entry points are the [AI alignment playlist](https://youtu.be/pYXy-A4siMw?list=PLCRVRLd2RhZTpdUdEzJjo3qhmX3y3skWA&t=14) on YouTube, “[The Road to Superintelligence](https://waitbutwhy.co...
444f7068-32da-4cc7-ac69-484a979bbe94
trentmkelly/LessWrong-43k
LessWrong
What's wrong with simplicity of value? In the Wiki article on complexity of value, Eliezer wrote: > The thesis that human values have high Kolmogorov complexity - our preferences, the things we care about, don't compress down to one simple rule, or a few simple rules. > > [...] > > Thou Art Godshatter describes the...
24b417ff-06e7-461a-b6c2-f45cbf0a67a3
trentmkelly/LessWrong-43k
LessWrong
Methods of Phenomenology NB: Originally posted on Map and Territory on Medium, so some of the internal series links go there. In the previous post we looked at phenomenology and the thinking that motivates it. We saw that it is based on taking a naive, skeptical, beginner’s view to asking “why?” and choosing to addre...
2c5d039c-6144-4a38-8f0d-3021b3c28fe3
trentmkelly/LessWrong-43k
LessWrong
Advice for time management as a manager This post was adapted from an internal doc I wrote at Wave. Welcome to being a manager! Your time-management problem just got a lot harder. As an IC, you can often get away with a very simple time-management strategy: 1. Decide what your one most important thing is. 2. Work...
e99bfb86-089d-4db2-9b8c-f99fc8684517
trentmkelly/LessWrong-43k
LessWrong
Beware of WEIRD psychological samples Most of the research on cognitive biases and other psychological phenomena that we draw on here is based on samples of students at US universities.  To what extent are we uncovering human universals, and to what extent facts about these WEIRD (Western, Educated, Industrialized, Ri...
15dd58f3-d49e-4750-8bfb-5276a5d086f7
trentmkelly/LessWrong-43k
LessWrong
Against asking if AIs are conscious People sometimes wonder whether certain AIs or animals are conscious/sentient/sapient/have qualia/etc. I don't think that such questions are coherent. Consciousness is a concept that humans developed for reasoning about humans. It's a useful concept, not because it is ontologically ...
9f7ae3fc-b1dd-4204-acd2-7c63148d9377
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
Anthropic paradoxes transposed into Anthropic Decision Theory .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: normal; whi...
b25b70c9-a040-4abf-ac33-1317a1420e51
StampyAI/alignment-research-dataset/arxiv
Arxiv
Learning When to Advise Human Decision Makers 1 Introduction --------------- Artificial intelligence (AI) is increasingly being used to support human decision making in high-stake settings in which the human operator, rather than the AI algorithm, needs to make the final decision. For example, in the criminal justi...
b2fa3f7e-04d6-434f-8376-e16437ff4530
trentmkelly/LessWrong-43k
LessWrong
Contra papers claiming superhuman AI forecasting [Conflict of interest disclaimer: We are FutureSearch, a company working on AI-powered forecasting and other types of quantitative reasoning. If thin LLM wrappers could achieve superhuman forecasting performance, this would obsolete a lot of our work.] Widespread, mis...
7d29ec6e-718c-496c-81c6-656cd5380420
trentmkelly/LessWrong-43k
LessWrong
Becoming Stronger as Epistemologist: Introduction Introduction: a new hope If you’ve read anything I wrote in the last 6 months, you know that I’m into the epistemology of alignment. More precisely, I’m into figuring out how to leverage epistemic tools and insights for helping both researchers make progress and newcom...
e56124ef-a113-4a64-a380-94a3138f0660
trentmkelly/LessWrong-43k
LessWrong
Berkeley group house, spots open If you are interested in living in a group house of EAs in Berkeley starting as soon as you’d like, please fill out this 1-minute form - we have a few spots available that we’re hoping to fill. Rent durations are up to 8 months (longer if a new lease is signed). The form will close whe...
ae3d6cc6-94eb-4b6e-853a-f292bbf0695f
StampyAI/alignment-research-dataset/eaforum
Effective Altruism Forum
Data Publication for the 2021 Artificial Intelligence, Morality, and Sentience (AIMS) Survey *This blog post was edited by Ali Ladak and Jacy Reese Anthis. The 2021 AIMS survey was designed by Janet Pauketat, Jamie Harris, Ali Ladak, and Jacy Reese Anthis. The data was collected and analyzed by Janet Pauketat, Ali La...
60504001-4baa-4906-9f23-0716571e6938
trentmkelly/LessWrong-43k
LessWrong
Meetup : Sydney Rationality Dojo - February 2016 Discussion article for the meetup : Sydney Rationality Dojo - February 2016 WHEN: 07 February 2016 04:00:00PM (+1100) WHERE: 10 Shepherd Street, Chippendale Join us for the first dojo of the new year. Get a head start on accomplishing everything you want to get done ...
b48330d6-07f2-48b8-a76d-912c4bdb4805
trentmkelly/LessWrong-43k
LessWrong
Expansive translations: considerations and possibilities A crowd probably best served by a wide variety of translations TLDR: Language translation is a decent first step for written works, but the ideal looks more like an empathetic personal tutor. There’s a lot to do in-between, both in the near term with human labor...
56cc320d-06b6-4945-a086-f91fa3cc9b01
trentmkelly/LessWrong-43k
LessWrong
[SEQ RERUN] My Naturalistic Awakening Today's post, My Naturalistic Awakening was originally published on 25 September 2008. A summary (taken from the LW wiki):   > Eliezer actually looked back and realized his mistakes when he imagined the idea of an optimization process. Discuss the post here (rather than in the...
2b960f16-e81f-4332-8814-073207bbee12
trentmkelly/LessWrong-43k
LessWrong
dojo - Bad day contingency plan http://bearlamp.com.au/dojo-bad-day-contingency-plan/ The following is an exercise I composed to be run at the Lesswrong Sydney dojos.  It took an hour and a half but could probably be done faster with some adaptations that I have included in these instructions. In regards to what ...
c2553e56-09dc-4605-be76-4f97d78b8862
trentmkelly/LessWrong-43k
LessWrong
Scoring My AstraZeneca Predictions After a relatively good month or so, as many Western countries have begun to relax their restrictions, COVID-19 cases and deaths are on the rise again, increasing especially in Africa, the Levant and South-East Asia.[1] Over fifty thousand people die each week, and to date over four ...
76bbc9fa-eaf3-4dea-8323-35a0e441f33e
trentmkelly/LessWrong-43k
LessWrong
Housing Roundup #10 There’s more campaign talk about housing. The talk of needing more housing is highly welcome, as one prominent person after another (including Jerome Powell!) talking like a YIMBY. A lot of the concrete proposals are of course terrible, but not all of them. I’ll start off covering all that along w...
f65612a3-3bfc-4a97-b034-e7ee44375496
StampyAI/alignment-research-dataset/lesswrong
LessWrong
[Incomplete] What is Computation Anyway? Disclaimer ========== This post was written three months ago, and I haven't really touched it since then. I never fully completed it[[1]](#fnrxd9a4nyupe), and I'm not satisfied with the current draft. But as is, I'm not on track to complete it anytime in the foreseeable future...
db5a03fd-3e08-497e-ba9d-a393cb240bca
trentmkelly/LessWrong-43k
LessWrong
Stanzas On Power Calculation Effects can be ignorable - oh, real, yes, but fine - So say, worst case scenario, μ's right there on the line.   The power is the modest chance Effects as small as μ's Will have to be significant  In the C.I. I choose.   I pick a null hypothesis and set out to reject. But can I...
5537757a-2f87-47eb-917d-8b21baac3fcb
trentmkelly/LessWrong-43k
LessWrong
[SEQ RERUN] Against Maturity Today's post, Against Maturity was originally published on 18 February 2009. A summary (taken from the LW wiki):   > Dividing the world up into "childish" and "mature" is not a useful way to think. Discuss the post here (rather than in the comments to the original post). This post is ...
e738bef2-be3a-4dc2-9ace-3c13ec20c99d
trentmkelly/LessWrong-43k
LessWrong
Meetup : San Francisco Meetup: Clicker Training/Improv Games Discussion article for the meetup : San Francisco Meetup: Clicker Training/Improv Games WHEN: 25 May 2015 06:15:00PM (-0700) WHERE: 1061 Market St #4, San Francisco, CA 94103 We'll be meeting to play clicker training/other improv games. As usual, call 301...
78ca3e02-33a3-49da-a9f3-eb17bd1681dd
trentmkelly/LessWrong-43k
LessWrong
Altruism is Incomplete Previously (SlateStarCodex) on Effective Altruism: Fear and Loathing at Effective Altruism Global 2017 I A few weeks ago, I listened to a podcast from Freakonomics Radio that purported to ask the question, Are The Rich Less Generous Than The Poor? A clever researcher decided he wanted to meas...
d4539ca9-3416-4c35-a9ed-e863b4aacda7
trentmkelly/LessWrong-43k
LessWrong
How to read a book I'm considering reading the book by the title How to read a book. A friend of mine (his critical thinking is quite good, but certainly not as good as it could be, so I can't trust his opinion too much) said he has read it and that it helped him a lot. He said it had advice on reading comprehension, ...
6e61e4de-7a4c-4eb1-9343-9f4d1db32d4b
trentmkelly/LessWrong-43k
LessWrong
Compute Governance Literature Review This white paper reviews the use of compute governance to mitigate risks from AI, the reasoning behind this approach, the ways in which it has been used and in which it could be further developed. It aims to provide an introduction for individuals who feel that there is a need for...
737ec6fb-5545-4536-82a1-1a3c5f433349
trentmkelly/LessWrong-43k
LessWrong
Agency and Sphexishness: A Second Glance This post is backported from my old personal blog, originally published 11 October, 2014. It's similar in theme but very different in focus than my recent post, On the Nature of Agency, making for a good complement. Epistemic Status: Though I wrote this over four years ago and...
c494b8f1-90ed-4fb1-bb4f-bc46ca903d2d
trentmkelly/LessWrong-43k
LessWrong
Auto-GPT: Open-sourced disaster? Sharing this here doesn't seem like an infohazard at this point. This is all over my YouTube feed anyway. Description from the authors: > Auto-GPT is an experimental open-source application showcasing the capabilities of the GPT-4 language model. This program, driven by GPT-4, autono...
18d22b31-ab09-4ab1-83f8-7a6ba17b1410
trentmkelly/LessWrong-43k
LessWrong
The Journal of (Failed) Replication Studies One of Seed Magazine's "Revolutionary Minds" is Moshe Pritsker, who created the Journal of Visualized Experiments, which to me looks like a very cool idea. I imagine that early on it may have looked somewhat silly ("he can't implant engineered tissue in a rat heart and he ca...
e59e4039-7800-4cac-8740-9324e60b8f99
StampyAI/alignment-research-dataset/arxiv
Arxiv
Identifying Adversarial Attacks on Text Classifiers 1 Introduction --------------- Text classifiers have been under attack ever since spammers started evading spam filters, nearly 20 years ago Hulten et al. ([2004](#bib.bib18)). In recent years, however, attacking classifiers has become much easier to carry out. Ma...
3e2df392-b0c1-436d-bf7f-90c829b78a72
StampyAI/alignment-research-dataset/lesswrong
LessWrong
The Löbian Obstacle, And Why You Should Care In 2013, Eliezer Yudkowsky and Marcello Herreshoff published [Tiling Agents for Self-Modifying AI, and the Löbian Obstacle](https://intelligence.org/files/TilingAgentsDraft.pdf). It is worth comprehending because it: * Is a very well written paper. * Expresses an idea that...
d2fef184-7e78-4e80-ab9d-51ef4cfe3cb2
trentmkelly/LessWrong-43k
LessWrong
Finite Factored Sets: Conditional Orthogonality We now want to extend our notion of orthogonality to conditional orthogonality. This will take a bit of work. In particular, we will have to first extend our notions of partition generation and history to be defined on partitions of subsets of S.   4.1 Generating a Su...
5fc47954-33e8-40cc-9f61-6d86aef950e4
trentmkelly/LessWrong-43k
LessWrong
Overconfident talking down, humble or hostile talking up A Distribution of Knowlege If one were to make a distribution of the amount of knowledge different people have about, say, macroeconomics, I would suspect the distribution to be somewhat lognormal; they would have tails to both ends, but be very skewed to the ri...
70431874-0fdb-4223-82b5-8decd960eb4d
trentmkelly/LessWrong-43k
LessWrong
[SEQ RERUN] Bayesians vs. Barbarians Today's post, Bayesians vs. Barbarians was originally published on 14 April 2009. A summary (taken from the LW wiki):   > Suppose that a country of rationalists is attacked by a country of Evil Barbarians who know nothing of probability theory or decision theory. There's a certai...
b63ca22f-e72e-4fb3-b60b-3773443d9bc9
trentmkelly/LessWrong-43k
LessWrong
Rest Days vs Recovery Days Based on a comment I made on this EA Forum Post on Burnout. Related links: Sabbath hard and go home, Bring Back the Sabbath ---------------------------------------- That comment I made generated more positive feedback than usual (in that people seemed to find it helpful to read and found ...
4d1f80bc-b272-464d-922b-e9197e7a4127
trentmkelly/LessWrong-43k
LessWrong
Humor Reading the recent list of rationality quotes arranged by karma underlines the popularity of funniness, and being funny should probably be included in the pursuit of awesomeness. My best guesses about characteristics of humor: If there's a word which makes the line funny, put it at the end. Phyllis Diller recom...
4729db72-8f18-4fc5-a1b6-56096769b027
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
What Is The True Name of Modularity? TL;DR ===== Modularity seems like an important feature of neural networks, but there is currently no canonical way of properly defining or measuring it which is properly theoretically motivated and doesn’t break down in some cases - in other words, we haven’t yet found a [True Nam...
5aee9b2a-5e7d-4bfd-8579-c0a6112d51d7
trentmkelly/LessWrong-43k
LessWrong
Prereq: Question Substitution A quote from a 2002 paper by Kahneman and Frederick referencing some of Kahneman's earlier work with Tversky: > Early research on the representativeness and availability heuristics was guided by a simple and general hypothesis: when confronted with a difficult question people often answe...
69a4f896-88f1-4d6c-abfb-bd393a37b60f
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
The positional embedding matrix and previous-token heads: how do they actually work? *tl;dr: This post starts with a mystery about positional embeddings in GPT2-small, and from there explains how they relate to previous-token heads, i.e. attention heads whose role is to attend to the previous token. I tried to make th...
93921010-40c8-49c3-bcce-1ebc52c65df2
trentmkelly/LessWrong-43k
LessWrong
Baltimore? Anyone reading this in Baltimore?
3029ea3d-fa08-487e-9ed1-7e0419e8e2c8
trentmkelly/LessWrong-43k
LessWrong
Do the Safety Properties of Powerful AI Systems Need to be Adversarially Robust? Why? Where "powerful AI systems" mean something like "systems that would be existentially dangerous if sufficiently misaligned". Current language models are not "powerful AI systems".   In "Why Agent Foundations? An Overly Abstract Expl...
803ec70f-f1b3-461a-940e-759e57c3ff8d
StampyAI/alignment-research-dataset/lesswrong
LessWrong
Andrew Ng wants to have a conversation about extinction risk from AI Andrew Ng writes: > I'd like to have a real conversation about whether AI is a risk for human extinction. Honestly, I don't get how AI poses this risk. What are your thoughts? And, who do you think has a thoughtful perspective on how AI poses this ...
7cec7510-3803-4cb1-b04e-f908d81d5655
trentmkelly/LessWrong-43k
LessWrong
Do the change you want to see in the world [Epistemic status: old truth, still seemed to deserve to be repeated a zillionth time in different words] Really, I’m just going to state the obvious here. Most people have know and applied this very well for decades. On the other hand, it’s an obvious thing I personally for...
7fa4219c-0d90-4bd1-b087-0a9f41a3c380
trentmkelly/LessWrong-43k
LessWrong
Notes on dynamism, power, & virtue This is very rough — it's functionally a collection of links/notes/excerpts that feel related. I don’t think what I’m sharing is in a great format; if I had more mental energy, I would have chosen a more-linear structure to look at this tangle of ideas. But publishing in the current ...
e11dd62e-7ca8-433d-ad27-15e88af3bc42
trentmkelly/LessWrong-43k
LessWrong
Saying the quiet part out loud: trading off x-risk for personal immortality Statement: I want to deliberately balance the caution and the recklessness in developing AGI, such that it gets created in the last possible moment so that I and my close ones do not die. This Statement confuses me. There are several observat...
8ef70600-06a3-4051-9173-287c56ed1726
trentmkelly/LessWrong-43k
LessWrong
My thoughts on OpenAI's alignment plan Epistemic Status: This is my first attempt at writing up my thoughts on an alignment plan. I spent about a week on it.  I’m grateful to Olivia Jimenez, Thomas Larsen, and Nicholas Dupuis for feedback. A few months ago, OpenAI released its plan for alignment. More recently, Jan ...
4527ecf0-e4f9-4da7-9fe9-f77c50e41ed5
trentmkelly/LessWrong-43k
LessWrong
Some 2-4-6 problems The sequence [2, 4, 6] is valid. Test other sequences to discover what makes a sequence valid. When you think you know, write down your guess, reveal the rule, and see how it compares. (You should try to deduce the truth using as few tests as possible; however, your main priority is getting the ru...
4fd4847c-a175-4bd2-872a-995a1842d682
trentmkelly/LessWrong-43k
LessWrong
Meetup : Washington DC Kennedy Center Meetup with Michael Vassar Discussion article for the meetup : Washington DC Kennedy Center Meetup with Michael Vassar WHEN: 18 April 2013 12:00:00PM (-0400) WHERE: John F. Kennedy Center for the Performing Arts, 2700 F Street, NW, Washington, DC 20566. (This is very short noti...
35065a55-525b-4d23-b448-a3759ef51286
trentmkelly/LessWrong-43k
LessWrong
First Triangle LW Meetup 5/4 7pm (Raleigh/Durham/Chapel Hill) When: Wednesday, May 4th at 7:00 pm Where: Morrisville Outlet Mall Food Court  I'll be wearing a Lego evolution shirt and may or may not have a LW sign. This Google doc contains some preliminary notes as well as my phone number.  I imagine this meeting wi...
efb3d18e-81a3-4fa6-b583-28c20b6edfb2
StampyAI/alignment-research-dataset/youtube
Youtube Transcripts
8 New Ways to Use Bing's Upgraded 8 [now 20] Message Limit (ft. pdfs, quizzes, tables, scenarios...) Bing chat from Microsoft has just raised the conversation limit on messages to eight per turn down from unlimited launch fair enough but up from six yesterday in fact I actually saw this change live on my phone las...
9c2072c9-adc8-479e-a04f-5cb35312fd43
trentmkelly/LessWrong-43k
LessWrong
On Not Having an Advance Abyssal Plan "Even though he could foresee the problem then, we can see it equally well now.  Therefore, if he could foresee the solution then, we should be able to see it now.  After all, Seldon was not a magician.  There are no trick methods of escaping a dilemma that he can see and we can't...
5e9476c1-84a6-48d2-90b1-4e8e08136c1f
StampyAI/alignment-research-dataset/lesswrong
LessWrong
Clarifying Consequentialists in the Solomonoff Prior I have spent a long time being confused about [Paul’s post](https://ordinaryideas.wordpress.com/2016/11/30/what-does-the-universal-prior-actually-look-like/) on consequentialists in the Solomonoff prior. I now think I understand the problem clearly enough to engage ...
3044a435-188b-4a6e-ad97-7271a19c1bcf
trentmkelly/LessWrong-43k
LessWrong
My slack budget: 3 surprise problems per week Previously: Slack In a couple earlier articles I urged people to adopt strategies that reliably maintain a margin of "30% slack." I've seen lots of people burn out badly (myself included), and preserving a margin of resources such that you don't risk burning out seems qui...
68ed5465-6721-45f1-99e7-406465c0edb3
LDJnr/LessWrong-Amplify-Instruct
LessWrong
"Yesterday I talked about the cubes {1, 8, 27, 64, 125, ...} and how their first differences {7, 19, 37, 61, ...} might at first seem to lack an obvious pattern, but taking the second differences {12, 18, 24, ...} takes you down to the simply related level. Taking the third differences {6, 6, ...} brings us to the perf...
4b3385c5-a13c-4fe9-8026-9e1e7fa416c6
StampyAI/alignment-research-dataset/arxiv
Arxiv
Modeling the Mistakes of Boundedly Rational Agents Within a Bayesian Theory of Mind Introduction ------------ A key aspect of human intuitive psychology is our understanding that other agents are fallible: they may possess false beliefs (?, ?), lack knowledge (?, ?), fail to plan ahead (?, ?), or act unintentionall...
098a98c5-b2b5-44b0-a9ca-99184fc98e69
trentmkelly/LessWrong-43k
LessWrong
You won’t solve alignment without agent foundations The following is a part of a response to a Vivek Hebbar's problem (the questions in cursive). I looked at why an alignment proposal doesn't work at an AGI level, in connection to only two of Eliezer's doom arguments. After that, I talked about why I'm generally pessi...
013976da-c3e9-4b7b-b265-a597fd991887
awestover/filtering-for-misalignment
Redwood Research: Alek's Filtering Results
id: post3675 Available on the arXiv, my paper on two types of Oracles (AIs constrained to answering questions only), and how to use them more safely. An Oracle is a design for potentially high power artificial intelligences (AIs), where the AI is made safe by restricting it to only answer questions. Unfortunately most...
a969d74f-cb56-466d-a3a4-f3026cc71fdd
trentmkelly/LessWrong-43k
LessWrong
Meetup : Madison: Generating More Ideas Discussion article for the meetup : Madison: Generating More Ideas WHEN: 23 September 2012 07:00:00PM (-0500) WHERE: 4371 Doncaster Dr, Madison, WI We've discussed some methods and basic principles for idea generation. But, of course, there are more! I'll review what we've ...
79240c41-66c1-4fca-9c38-faa325a65470
StampyAI/alignment-research-dataset/lesswrong
LessWrong
Measures, Risk, Death, and War This is the fourth post of a [sequence on decision analysis](/lw/8xr/decision_analysis_sequence/), preceded by [Compressing Reality to Math](/lw/8uj/compressing_reality_to_math/). It touches on a wide variety of topics which didn't seem to work well as posts of their own, either because ...
ef8b870e-5809-41da-aa64-8bcdad5a09d1
trentmkelly/LessWrong-43k
LessWrong
What is it like doing AI safety work? How do you know if you’ll like AI safety work? What’s the day-to-day work like? What are the best parts of the job? What are the worst? To better answer these questions, we talked to ten AI safety researchers in a variety of organizations, roles, and subfields. If you’re interest...
293dbc38-3491-447e-9958-5ac5d74144ca
LDJnr/LessWrong-Amplify-Instruct
LessWrong
"In L. Sprague de Camp's fantasy story The Incomplete Enchanter (which set the mold for the many imitations that followed), the hero, Harold Shea, is transported from our own universe into the universe of Norse mythology. This world is based on magic rather than technology; so naturally, when Our Hero tries to light a ...
5f545355-633a-4c62-9458-1e888f1697ed
trentmkelly/LessWrong-43k
LessWrong
A Defense of Utilitarianism 1. Definitions   Let’s start by defining some terms. First, we ought to distinguish between what is positive and what is normative. A positive claim states what is, whereas a normative one states what ought to be. Through Hume’s famous guillotine, we cannot derive one from the other. The...
4b9c9482-fab8-4ad1-903d-85c14cc34231
StampyAI/alignment-research-dataset/eaforum
Effective Altruism Forum
[CFP] NeurIPS workshop: AI meets Moral Philosophy and Moral Psychology <https://aipsychphil.github.io/>   **\*About\*** Be it in advice from a chatbot, suggestions on how to administer resources, or which content to highlight, AI systems increasingly make value-laden decisions. However, researchers are becoming in...
1d34e289-4eab-47e1-af0d-95851df1370e
trentmkelly/LessWrong-43k
LessWrong
Announcing the CNN Interpretability Competition TL;DR I am excited to announce the CNN Interpretability Competition, which is part of the competition track of SATML 2024.  Dates: Sept 22, 2023 - Mar 22, 2024 Competition website: https://benchmarking-interpretability.csail.mit.edu/challenges-and-prizes/  Total prize...
aca24732-cd38-4306-bbf3-80706e1f57c3
trentmkelly/LessWrong-43k
LessWrong
If you don't know the name of the game, just tell me what I mean to you Following: Let's split the Cake tl;dr: Both the Nash Bargaining solution (NBS), and the Kalai-Smorodinsky Bargaining Solution (KSBS), though acceptable for one-off games that are fully known in advance, are strictly inferior for independent repea...
2036e132-a51e-40e9-817a-9f5d2821e4a1
StampyAI/alignment-research-dataset/lesswrong
LessWrong
Agency engineering: is AI-alignment "to human intent" enough? TL;DR: We have not solved the "free will" problem: i.e. we do not know to what extent consciousness, culture, biology or a myriad of other factors shape human will and intent. Yet, prediction of future behavior of self and other agent-caused actions (i.e. w...
58bfb568-0b91-493c-9418-e443a7793e4e
StampyAI/alignment-research-dataset/eaforum
Effective Altruism Forum
We should say more than “x-risk is high” | | | --- | | **This is a**[**Draft Amnesty Day**](https://forum.effectivealtruism.org/posts/vHogKj87ZBLgN8TpQ/today-is-draft-amnesty-day-december-16-18) **draft.**That means it’s not polished, it’s probably not up to my standards, the ideas are not thought out, and I haven’t ...
2918fdab-ac03-49d5-b96f-c6dff9c90454
trentmkelly/LessWrong-43k
LessWrong
(Rational) website design and cognitive aesthetics generally- why no uptake? So I'm working for a friend's company at the moment (friend is a small business owner who designs websites and a bit of an entrepreneur) anyway, I've persuaded him that we should research the empirical literature on what makes websites effect...
2955f77f-565c-4d10-bb89-9f6020ce5f60
trentmkelly/LessWrong-43k
LessWrong
Any LWers in UK West Midlands? After seeing LW meetups pop up in London, Edinburgh, Oxford and Cambridge, I thought I'd see if there was any interest in kicking one off somewhere I might reasonably be able to get to. I know it's not something people like to readily admit to, but I've not seen anyone else mention bein...
124a366b-ef7f-4f53-be46-574002555ad0
trentmkelly/LessWrong-43k
LessWrong
Meetup : Helsinki Practical Meetup Discussion article for the meetup : Helsinki Practical Meetup WHEN: 04 August 2013 03:00:00PM (+0300) WHERE: Kirjasto 10, Elielinaukio, Helsinki We’ll tackle the theme of Quantified Self – can measurement and tracking of personal statistics be fun and useful, and how? We’ll go thr...
340fd16d-afed-47d8-b8b6-b916cdbedf6a
StampyAI/alignment-research-dataset/agisf
AGI Safety Fund
Syllabus: Artificial Intelligence and China Jeffrey Ding, Sophie-Charlotte Fischer, Brian Tse, Chris Byrd January 2020 Syllabus: Artificial Intelligence and China 1   In recent years, China’s ambitious development of artificial intelligence (AI) has attracted much attention in policymaking and academic circles...
b0ab497f-1153-42e9-9097-d3da9201e03a
trentmkelly/LessWrong-43k
LessWrong
Has anyone actually tried to convince Terry Tao or other top mathematicians to work on alignment? This has been discussed several times in the past, see: * Have You Tried Hiring People?, a LW post * It talks about this ACX comment thread * Greg Coulbourn’s “Mega-money for mega-smart people to solve AGI Alignment...
530d7caa-d1e7-4911-a75f-c6887e00cc24
trentmkelly/LessWrong-43k
LessWrong
[SEQ RERUN] Epistemic Viciousness Today's post, Epistemic Viciousness was originally published on 13 March 2009. A summary (taken from the LW wiki):   > An essay by Gillian Russell on "Epistemic Viciousness in the Martial Arts" generalizes amazingly to possible and actual problems with building a community around ra...
2d4ffc7b-68c7-4450-9842-1cc5d56d7ee7
trentmkelly/LessWrong-43k
LessWrong
Zeroing Out Related to (Eliezer Yudkowski at Less Wrong): An Equilibrium of No Free Energy Related to (Satvik Beri at Less Wrong): Competitive Truth Seeking Follow-up to: Leaders of Men I: Markets Suppose you have an insight about Google. The efficient market hypothesis says you can’t make a profit. Your insight i...
66f43d82-224e-44d0-ae1f-95a31fa774b5
trentmkelly/LessWrong-43k
LessWrong
Scott Garrabrant's problem on recovering Brouwer as a corollary of Lawvere Edit: Pre-print available here https://arxiv.org/abs/2005.01563 Also link to earlier post here https://www.alignmentforum.org/posts/5bd75cc58225bf06703753a9/formal-open-problem-in-decision-theory?fbclid=IwAR2u-8RVfnfRtEo51vag5vYWAcfkoHDJAr6ayd...
1a045396-b113-4b3c-a86c-1153fcb6b0e9
trentmkelly/LessWrong-43k
LessWrong
Babble Thread On LessWrong, we have a sort of hierarchy of posts: 1. Curated 2. Front Page 3. Personal Blog 4. Open Thread / Short Form Here I'd like to experiment with a fifth: the Babble Thread. What is babbling? Babble is the post that introduced the concept, I believe. The Babble and Prune tag is also worth...