id
stringlengths
36
36
source
stringclasses
15 values
formatted_source
stringclasses
13 values
text
stringlengths
2
7.55M
ab218a25-4cc4-4f88-8b10-3f388c928069
trentmkelly/LessWrong-43k
LessWrong
[Speech] Worlds That Never Were This is the speech I gave at the 2018 Bay Area Secular Solstice. It was written to fit the bill of 'concrete visions of the future.' There is one small change here from the version I gave : the Solstice organizers had me change the second-to-last sentence of the speech, because the even...
a8eb6a96-5ce4-4135-8526-84cfafe154f3
StampyAI/alignment-research-dataset/eaforum
Effective Altruism Forum
What Does a Marginal Grant at LTFF Look Like? Funding Priorities and Grantmaking Thresholds at the Long-Term Future Fund The Long-Term Future Fund (LTFF) makes small, targeted grants with the aim of improving the long-term trajectory of humanity. We are currently fundraising to cover our grantmaking budget for the nex...
44aea485-5dfc-4179-bbb4-e4ce73644bed
trentmkelly/LessWrong-43k
LessWrong
Philadelphia SSC Meetup Philadelphia SSC typically meets the last Wednesday of every month at 6:30pm, locations TBA. To avoid upcoming holidays, we've moved up some of our dates; below are our scheduled meetups through the end of 2019: Wednesday, Oct. 30, 6:30pm - Tir Na Nog Irish Pub, 1600 Arch St., Philadelphia We...
c506dabc-6231-4170-a584-1c1262966f2d
StampyAI/alignment-research-dataset/arxiv
Arxiv
Decoupling Exploration and Exploitation for Meta-Reinforcement Learning without Sacrifices 1 Introduction --------------- A general-purpose agent should be able to perform multiple related tasks across multiple related environments. Our goal is to develop agents that can perform a variety of tasks in novel environm...
d60d3733-1318-475f-a967-baf43b58efab
trentmkelly/LessWrong-43k
LessWrong
Video & transcript: Challenges for Safe & Beneficial Brain-Like AGI (Lightly edited) transcript starts here (FYI: This has a decent amount of overlap with the version I posted here two years ago, but it’s accumulated enough edits over time that I wanted to re-transcribe the latest version.) Thank you for inviting me!...
ccc1c8b7-7a7c-4411-a2ea-07512959c0f9
trentmkelly/LessWrong-43k
LessWrong
How to Grow a Mind (video) From the recent NIPS conference, here's a talk by MIT cognitive scientist Josh Tenenbaum on what he calls "rich" machine learning.  Should be of some relevance for people who are interested in AI or human cognitive development.  I found it really interesting. http://videolectures.net/nips20...
fe984d8b-4524-47f6-b85e-a411e8ec896b
StampyAI/alignment-research-dataset/lesswrong
LessWrong
Will working here advance AGI? Help us not destroy the world! I often talk to developers who prefer not destroying the world by accident (specifically by accelerating AGI risk), but neither them nor me can decide if specific companies qualify for this. Could someone knowledgable help? A few short replies could probab...
09704b82-4c7a-44b3-91db-f2bbd6f5ed5f
trentmkelly/LessWrong-43k
LessWrong
Nonverbal thinkers: how do you experience your inner critic? This is something I've been puzzled by for a long time, for myself. The "inner critic" is generally described as a literal voice that people hear, saying things like "wow, I'm such an idiot", "nobody likes me", etc. My problem here is that I generally don't...
57eb7bb3-eb9f-46d2-904c-32514e17ba5e
StampyAI/alignment-research-dataset/arxiv
Arxiv
Multi-Issue Negotiation with Deadlines
58a37a68-ec5d-4b2b-bdd1-f640cc016097
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
Mundane solutions to exotic problems I’m looking for alignment techniques that are [indefinitely scalable](https://ai-alignment.com/directions-and-desiderata-for-ai-control-b60fca0da8f4) and that [work in any situation we can dream up](https://ai-alignment.com/my-research-methodology-b94f2751cb2c). That means I spend ...
33df98fe-007f-4c3d-85bf-52a4b33c16c0
StampyAI/alignment-research-dataset/youtube
Youtube Transcripts
Safe Exploration: Concrete Problems in AI Safety Part 6 hi this is the latest video in a series about the paper concrete problems in AI safety you don't need to have seen the previous videos for this but I'd recommend checking them out anyway there's a link in the description today we're going to talk about safe explo...
d70d2089-45da-4b1c-8bc9-4493361db623
trentmkelly/LessWrong-43k
LessWrong
Dimensionalizing Forecast Value tl;dr In quantitative finance, the value of forecasting is obvious. No decisions are made without forecasts, and every forecast is created for some decision. On Metaculus1, the signal is just as sharp, but the market for the signal is fuzzier. The gap isn’t quality: it’s that the value...
298a3507-6da5-4e31-b021-a886d961a30b
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
Ultra-simplified research agenda This is an ultra-condensed version of the [research agenda](https://www.lesswrong.com/posts/CSEdLLEkap2pubjof/research-agenda-v0-9-synthesising-a-human-s-preferences-into) on synthesising human preferences (video version [here](https://www.youtube.com/watch?v=1M9CvESSeVc)): In order ...
3ff8eba4-231e-4892-a52b-afd7e4207c3d
trentmkelly/LessWrong-43k
LessWrong
Experience diets Imagine you have a part of your mind that just keeps track of which visceral experiences you have how often, and then expects more experiences in that ratio. So if you look at pictures of crocodiles, it feels like crocodiles are a bigger part of what is going on in the world. And then if you watch ten...
be3085e0-2ae9-46b6-b9ae-1365014afaa9
trentmkelly/LessWrong-43k
LessWrong
. .
895fe413-0494-4ae4-9503-d9a3d5a25c72
trentmkelly/LessWrong-43k
LessWrong
Why I hate the "accident vs. misuse" AI x-risk dichotomy (quick thoughts on "structural risk") * I think the large majority of AI x-risk is "structural".  Like climate change. * Here's a good primer on structural risk (note that structural risk is not a synonym for "not caused by out-of-control AI"): https://forum.e...
dd80ed1a-2504-462d-862f-8ca333b2ce39
trentmkelly/LessWrong-43k
LessWrong
Logical Updatelessness as a Robust Delegation Problem (Cross-posted an Less Wrong) There is a cluster of problems in understanding naturalized agency which I call Robust Delegation (EDIT: When I first wrote this, I called it "subagent alignment"). It refers to the way in which a limited agent might direct a more powe...
801e9719-184f-4fec-8263-4cd0449d43de
trentmkelly/LessWrong-43k
LessWrong
We ran an AI safety conference in Tokyo. It went really well. Come next year! Abstract Technical AI Safety 2024 (TAIS 2024) was a conference organised by AI Safety 東京 and Noeon Research, in collaboration with Reaktor Japan, AI Alignment Network and AI Industry Foundation. You may have heard of us through ACX. The goa...
68e5be4f-4060-4d0e-88cb-6ab076b6809f
trentmkelly/LessWrong-43k
LessWrong
[META] Recent Posts for Discussion and Main This link http://lesswrong.com/r/all/recentposts gives a page which lists all the recent posts in both the Main and Discussion sections. I've posted it in the comments section before, but I decided to put it in a discussion post because it's a really handy way of accessing...
069fe342-a6fc-4c09-b66d-9fd8989cf58f
trentmkelly/LessWrong-43k
LessWrong
Constantly ask what you're currently doing that is irrational At today's meetup in Tortuga, we were supposed to discuss something we're currently being irrational about.  In retrospect I could probably have done better than the item I picked (for example, it now occurs to me that I'm probably currently being irrationa...
d234be42-ab5b-42dc-8f20-3f67505846d9
trentmkelly/LessWrong-43k
LessWrong
Meetup : Sydney Rationality Dojo - May Discussion article for the meetup : Sydney Rationality Dojo - May WHEN: 01 May 2016 04:00:00PM (+1000) WHERE: 10 Shepherd Street, Chippendale Join us for the May dojo at Humanist House! As per usual, we will be running different exercises for two hours, followed by socialising...
ce211bf5-b200-443d-94e5-13f37f597cec
trentmkelly/LessWrong-43k
LessWrong
Fractional progress estimates for AI timelines and implied resource requirements This post was written by Mark Xu based on interviews with Carl Shulman. It was paid for by Open Philanthropy but is not representative of their views. A draft was sent to Robin Hanson for review but received no response. Summary * Robi...
ac787963-7e48-45d0-a95e-39abfb87cd22
trentmkelly/LessWrong-43k
LessWrong
Applications of logical uncertainty Lately I've been reading a lot about the problem of logical uncertainty. The problem is that there are logical consequences of your beliefs that you're uncertain about. So, for statements like "the billionth digit of pi is even", it's provable from your beliefs. But you're still un...
eb5e5347-c677-4929-a2da-99481789028e
trentmkelly/LessWrong-43k
LessWrong
Bayesians Commit the Gambler's Fallacy TLDR: Rational people who start out uncertain about an (in fact independent) causal process and then learn from unbiased data will rule out “streaky” hypotheses more quickly than “switchy” hypotheses. As a result, they’ll commit the gambler’s fallacy: expecting the process to swi...
724c0e24-422d-4386-9e61-9b22ed8cafe8
StampyAI/alignment-research-dataset/eaforum
Effective Altruism Forum
Concrete actions to improve AI governance: the behaviour science approach **Summary** =========== Human decision-making and concrete, observable actions about how to develop and use AI now will influence how successfully we navigate the transitions to a world with advanced AI systems. In this article, I apply behavio...
665e1cfd-fa0e-4522-9d80-e839c706fcaa
trentmkelly/LessWrong-43k
LessWrong
"Robot scientists can think for themselves" I recently saw this Reuters article on Yahoo News. In typical science reporting fashion, the headline seems to be pure hyperbole - does anyone here know enough to clarify what the groups referenced have actually achieved? This links represent what I could find: Homepage of...
cd3f1812-0da7-4a56-ae7a-5a7c47fd6f61
trentmkelly/LessWrong-43k
LessWrong
Is recursive self-alignment possible? Suppose an emulation of Eliezer Yudkowsky, as of January 2023, discovered how to self-modify. It's possible that the returns of capability per unit effort would start increasing, which would make its intelligence FOOM. The emulation starts with an interest in bringing about utopi...
0ce871d7-9c1c-4c7e-8e9f-0f1d2359b5a1
StampyAI/alignment-research-dataset/arxiv
Arxiv
Interpretable Multi-Objective Reinforcement Learning through Policy Orchestration 1 Introduction --------------- Concerns about the ways in which cyber-physical and/or autonomous decision making systems behave when deployed in the real world are growing: what various stakeholder are worried about is that the system...
17ba5fae-a192-42f4-a1fb-ec96cc5fdd82
trentmkelly/LessWrong-43k
LessWrong
Why keep a diary, and why wish for large language models Inspired by a dream I just woke up from, where I did not keep a diary One of the people with whom I have the most intimate of connections is my past self - in particular, my child self. We share a large number of commonalities: much of our basic outlook, our pe...
234b2247-85ba-4f5c-81f5-e0d87e520a71
trentmkelly/LessWrong-43k
LessWrong
Announcing The Effective Altruism Forum The Effective Altruism Forum will be launched at effective-altruism.com on September 10, British time. Now seems like a good time time to discuss why we might need an Effective Altruism Forum, and how it might compare to LessWrong. About the Effective Altruism Forum The motiv...
f3d160e7-8389-4438-9e90-7f9146761d30
trentmkelly/LessWrong-43k
LessWrong
Cooperation with and between AGI\'s [This post is primarily aimed at a Foresight Institute audience. I don't know whether the average LW reader will learn much from it.] This post is mostly a response to the Foresight Institute's book Gaming the Future, which is very optimistic about AI's being cooperative. They expe...
ca349dfd-4c2e-441c-a63e-c9be00eb934a
trentmkelly/LessWrong-43k
LessWrong
What Jason has been reading, April 2023 A monthly feature. Note that I generally don’t include very recent writing here, such as the latest blog posts (for those, see my Twitter digests); this is for my deeper research. AI First, various historical perspectives on AI, many of which were quite prescient: Alan Turing...
f5af3d98-7e10-4905-8ad1-6b207b83a23d
trentmkelly/LessWrong-43k
LessWrong
Joy in the Here and Real Summary: What do you do for simple, quick fun? Do you mind sharing with others? Tags: Repeatable Purpose: “Have fun” and “be happy” are common goals people have. If you can’t turn a few bucks and a few hours into a good time, you have a glaring gap in your abilities and I believe this is a ...
a003a8ec-ef08-4a6d-a3fb-e5180fc322f5
StampyAI/alignment-research-dataset/youtube
Youtube Transcripts
Interactions between the AI Control Problem and the Governance Problem | Nick Bostrom [Music] so the first one was the question if human-level AI is developed then what are the likely outcomes and I'm going to try to go over this fairly quickly to spend more time on the second question but we heard about them is talki...
08bc6c1e-39fb-4837-8066-f4fa26026e2b
StampyAI/alignment-research-dataset/arxiv
Arxiv
Aggregating incoherent agents who disagree 1 Aggregating credences ------------------------ Let’s begin by considering two different methods for aggregating a collection of credence functions that are defined over the same set of propositions ℱ={X1,…,Xm}ℱsubscript𝑋1…subscript𝑋𝑚\mathcal{F}=\{X\_{1},\ldots,X\_{m}\...
9f36918e-861e-4ae0-8669-025f98f34d24
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
Disentangling inner alignment failures **TL;DR:** This is an attempt to disentangle some concepts that I used to conflate too much as just "inner alignment". This will be old news to some, but might be helpful for people who feel confused about how deception, distributional shift, and "sharp left turns" are related. I...
ca7eb90b-607d-43f8-a968-a86dcaeb2632
trentmkelly/LessWrong-43k
LessWrong
Stable Pointers to Value II: Environmental Goals Cross-posted. In Stable Pointers to Value, I discussed various ways in which we can try to “robustly point at what we want” (ie, do value learning). I can tidy up the discussion there into three categories: 1. Standard reinforcement learning (RL) frameworks, includin...
3eac9783-5e8b-4dde-9273-fead8f80a1a3
StampyAI/alignment-research-dataset/eaforum
Effective Altruism Forum
EU's importance for AI governance is conditional on AI trajectories - a case study The goal of this post is to show that AI trajectories matter a great deal when evaluating how some institution is expected to influence AI governance. To show this I argue that the importance of the Brussels effect, one of the EU's leve...
a7fe2554-c7f2-4829-ad65-927395f72ef5
StampyAI/alignment-research-dataset/blogs
Blogs
Kurzweil, The Singularity is Near [The Singularity Is Near](http://en.wikipedia.org/wiki/The_Singularity_Is_Near) is a book by [Ray Kurzweil](http://en.wikipedia.org/wiki/Ray_Kurzweil). It argues that a [technological singularity](http://en.wikipedia.org/wiki/Technological_singularity) will occur in around 2045. This ...
65557c42-497e-463d-9af8-5edb77043f85
trentmkelly/LessWrong-43k
LessWrong
Senses of "knowing" a person Saying that you "know" someone is imprecise. What do you know about them? Here's a slightly arbitrary heptachotomy of information that could count as knowledge of a person: 1. trivially-recorded data, like their name, age, or location 2. their interests, hobbies, or occupation 3. thei...
79bc602a-d6db-4915-b8aa-65e99d7b6fd0
StampyAI/alignment-research-dataset/eaforum
Effective Altruism Forum
Intrinsic limitations of GPT-4 and other large language models, and why I'm not (very) worried about GPT-n Introduction ------------ Over my years of involvement in Effective Altruism, I have written [several](https://forum.effectivealtruism.org/posts/j7X8nQ7YvvA7Pi4BX/a-critique-of-ai-takeover-scenarios) forum posts...
bdbe1e30-e006-4e7e-9cd8-90cb2ab69763
trentmkelly/LessWrong-43k
LessWrong
Is there a Schelling point for group house room listings? My rationalist group house near Boston has a room opening (details here), and I'm wondering if there's a Schelling point for listing that sort of thing. I posted the opening here and to the relevant local Facebook groups, but I'm wondering where might be good ...
d4b1616a-e861-4a8e-a892-d78cf60a8431
trentmkelly/LessWrong-43k
LessWrong
Qualitative Strategies of Friendliness Followup to:  Magical Categories What on Earth could someone possibly be thinking, when they propose creating a superintelligence whose behaviors are reinforced by human smiles? Tiny molecular photographs of human smiles - or if you rule that out, then faces ripped off and perma...
bce397bc-2946-468b-aa92-43c6642fb70c
trentmkelly/LessWrong-43k
LessWrong
[Link] The Decision Education Foundation Among Michael Anissimov's recent Delicious links, I spotted The Decision Education Foundation.  It seems to be relevant to our interest in spreading rationality and bringing up young rationalists, but I hadn't seen it shared or discussed before and just thought I'd share it. >...
0db5b419-d3f0-42a1-b980-2bfa5395201d
trentmkelly/LessWrong-43k
LessWrong
Key Decision Analysis - a fundamental rationality technique The technique This post is signal-boosting and recommending a strategy for improving your decision making that I picked up from the entrepreneur Ivan Mazour. He describes the process here, and publishes his own results every year on his blog. In his words.....
f2f3df8f-3807-40ff-87cf-b1fc37500692
trentmkelly/LessWrong-43k
LessWrong
Essay competition on the Automation of Wisdom and Philosophy — $25k in prizes With AI Impacts, we’re pleased to announce an essay competition on the automation of wisdom and philosophy. Submissions are due by July 14th. The first prize is $10,000, and there is a total of $25,000 in prizes available.  Submit an entry ...
e77efbd7-0150-48d2-b208-1c9a8cfd0121
trentmkelly/LessWrong-43k
LessWrong
There have been 3 planes (billionaire donors) and 2 have crashed I'd rather not go into the details about which billionaires are which, so if it's actually 4 and 3 or 6 and 4, then that may or may not be debatable. I'm much more worried about whether MIRI survives the decade. It seems to me like this is a good place ...
9b01f6c7-ab5b-4d2c-9eb6-e41fa4c6d6ca
StampyAI/alignment-research-dataset/blogs
Blogs
Unsupervised deep learning identifies semantic disentanglement in single inferotemporal face patch neurons Our brain has an amazing ability to process visual information. We can take one glance at a complex scene, and within milliseconds be able to parse it into objects and their attributes, like colour or size, and u...
e624d0fd-86d4-4b3d-b834-fdcaeecfe87b
trentmkelly/LessWrong-43k
LessWrong
We're Redwood Research, we do applied alignment research, AMA Redwood Research is a longtermist organization working on AI alignment based in Berkeley, California. We're going to do an AMA this week; we'll answer questions mostly on Wednesday and Thursday this week (6th and 7th of October). Buck Shlegeris, Bill Zito, ...
310efa75-3382-4a7e-afc7-b634528f7185
trentmkelly/LessWrong-43k
LessWrong
Consider tabooing "I think" People say "I think" a lot. Here are some examples: * I think you brought me the wrong order. * I think the numbers in the report are wrong. * I think you need to turn left at the light. * I think we need to replace the whole water heater. * I think iPhones are better than Android pho...
cfccf225-4779-4bc9-8947-d062389fea2b
trentmkelly/LessWrong-43k
LessWrong
Meetup : Canberra: Econ 101 and its Discontents Discussion article for the meetup : Canberra: Econ 101 and its Discontents WHEN: 08 November 2014 06:00:00PM (+1100) WHERE: 108 North Road, Acton, ACT Nick will explain one of the most basic concepts in economics, the demand and supply curves, and how they interact to...
93284fa6-b20d-40f4-9d03-6b561ec58e13
trentmkelly/LessWrong-43k
LessWrong
Why such low detected rates of COVID-19 in children? In China 0.9% of COVID-19 cases were in people <10. In the Chinese population 11.9% of people are <10. In Italy its 0.5% and 8.4%. In South Korea it's 1% and 8.4%. 10-19 year olds are also very underrepresented although less severely so. This seems odd. I've se...
6f240543-81fa-4daf-882b-cf0c3172844b
StampyAI/alignment-research-dataset/lesswrong
LessWrong
The public supports regulating AI for safety A high-quality American public survey on AI, [Artificial Intelligence Use Prompts Concerns](https://www.monmouth.edu/polling-institute/documents/monmouthpoll_us_021523.pdf/), was released yesterday by Monmouth. Some notable results: * 9% say AI[1](https://aiimpacts.org/th...
3ced8ee4-184b-49e3-9db6-3611ae8c61b4
trentmkelly/LessWrong-43k
LessWrong
Dimensionalization TL;DR Dimensionalization is the practice of identifying and using quasi-independent axes or "dials" along which things vary. It’s the mental move behind tradeoff analysis, optimization, and design thinking. You are doing it already! Unlike categorization or decomposition, dimensionalization preserv...
b45ce2e1-1f0e-4d20-a909-e46446327b43
trentmkelly/LessWrong-43k
LessWrong
Opinions on the sleep synaptic homeostasis hypothesis? To a very crude approximation, the brain works via Hebbian learning. If neuron A fires immediately before neuron B then the synapse from A to B gets strengthened. There’s just one problem with Hebbian learning. Carry on long enough and you will eventually end up ...
f224f859-4984-45c6-a36a-e814ba18f072
trentmkelly/LessWrong-43k
LessWrong
Can an agent use interactive proofs to check the alignment of succesors? Some time ago I read an article on the difficulties of agents building succesors. The argument was something like this: Suppose we have solved the ultimate objective of goodness and built an agent A who does only things that he can prove are good...
d4a2efd4-fc48-4be6-9973-f2e4cd6d195c
trentmkelly/LessWrong-43k
LessWrong
MATS Winter 2023-24 Retrospective Co-Authors: @Rocket, @LauraVaughan, @McKennaFitzgerald, @Christian Smith, @Juan Gil, @Henry Sleight, @Matthew Wearden, @Ryan Kidd The ML Alignment & Theory Scholars program (MATS) is an education and research mentorship program for researchers entering the field of AI safety. This wi...
fe38be39-9168-4573-a4be-b950db2111b1
trentmkelly/LessWrong-43k
LessWrong
Hatching the Cosmic Egg (Hymn to Dionysus) Invocation Dionysus, deep-hearted one, granter of words, words heard best by the drunken and the mad, all ecstasies and inspirations are of your lips, I praise and honor you, I thank you for your blessings. Dionysus, rapturous one, god of the vine, who is lord of the far rea...
19b037ad-0c8f-406b-9a26-31b76b9c541a
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
Updates and additions to "Embedded Agency" Abram Demski and Scott Garrabrant's "[Embedded Agency](https://intelligence.org/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](https...
7475cb94-0532-4c61-8ef9-46c61bb284e2
trentmkelly/LessWrong-43k
LessWrong
The "AI Debate" Debate As far as I can tell, I have a disjoint set of concerns to many of the concerns I've heard expressed in conversations about AI Safety via Debate. My main concern with AI Debate is this: each debater has an incentive to trick the operator into running code that takes over the world, replaces the...
e1c4ae4a-736a-49bd-8b1f-13deb807ead6
trentmkelly/LessWrong-43k
LessWrong
Why doesn't China (or didn't anyone) encourage/mandate elastomeric respirators to control COVID? (This post was inspired in part by this comment by Florin, which I think makes some good related points.) Elastomeric respirators are reusable respirators made with some elastic material such as silicone, typically with...
db117e9a-3e9c-47a9-adb8-6d706cbcb9ee
trentmkelly/LessWrong-43k
LessWrong
Should we break up Google DeepMind? Regulators should review the 2014 DeepMind acquisition. When Google bought DeepMind in 2014, no regulator, not the FTC, not the EC's DG COMP, nor the CMA, scrutinized the impact. Why? AI startups have high value but low revenues. And so they avoid regulation (and tax, see below). Bu...
786f8251-c625-4d51-86bc-273e70e02304
StampyAI/alignment-research-dataset/youtube
Youtube Transcripts
AGI & Corporations Seminar - Conclusions so I would love to start the session by briefly kind of everyone summarizing what they thought was kind of like the most interesting part in terms of in the sense of a a I incorporations that is relevant to their own approach to their own standpoint so what's the main takeaway ...
a6b2b72b-760a-4e89-b3e3-cc6cd86645a1
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
What AI Safety Materials Do ML Researchers Find Compelling? I (Vael Gates) recently ran a small pilot study with Collin Burns in which we showed ML researchers (randomly selected NeurIPS / ICML / ICLR 2021 authors) a number of introductory AI safety materials, asking them to answer questions and rate those materials. ...
a9639d17-88b9-4046-9810-f70512844469
trentmkelly/LessWrong-43k
LessWrong
In favour of a selective CEV initial dynamic Note: I appreciate that at this point CEV is just a sketch. However, it’s an interesting topic and I don’t see that there’s any harm in discussing certain details of the concept as it stands. 1. Summary of CEV Eliezer Yudkowsky describes CEV - Coherent Extrapolated Voliti...
162940e0-7d5b-4096-9135-788d3aa3d1aa
StampyAI/alignment-research-dataset/eaforum
Effective Altruism Forum
AI Safety Newsletter #4: AI and Cybersecurity, Persuasive AIs, Weaponization, and Geoffrey Hinton talks AI risks 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.s...
2b868d20-0e1e-4889-8b0d-efe9e4b34ae5
trentmkelly/LessWrong-43k
LessWrong
Community Capital Argument: There is a Thing called Community Capital (closely related to social capital), and because it is based more on How Humans Work, you can get a lot more out of it than just plain old Money. Epistemic Status: Opening a conversation. All these ideas are in sand. --------- I used to compare t...
d3553aea-953c-4943-8c91-89d41e629c87
trentmkelly/LessWrong-43k
LessWrong
Feature proposal: Close comment as resolved There is an article. It contains a typo. Someone mentions it in a comment. The comment is upvoted as helpful. Author fixes the typo. Now there is a no-longer-useful comment in the discussion, sometimes with replies (the author saying "thanks, fixed", etc.). Sometimes it sits...
c903b532-8572-492f-b254-6ace9eb3d064
trentmkelly/LessWrong-43k
LessWrong
Time and Energy Costs to Erase a Bit Landauer derived a limit on the amount of energy dissipated in order to erase a bit. This limit is kTlog2, where k is Boltzmann's constant and T is the temperature of the environment. However, recently some questions have been raised as to what situations this limit applies to. Ma...
694aaa64-7e46-475f-a159-fdf1221d1f40
trentmkelly/LessWrong-43k
LessWrong
People who "don't rationalize"? [Help Rationality Group figure it out] Anna Salamon and I are confused. Both of us notice ourselves rationalizing on pretty much a daily basis and have to apply techniques like the Litany of Tarski pretty regularly. But in several of our test sessions for teaching rationality, a handfu...
6d6e2ad0-2d35-48f7-9797-a97e97bc0483
trentmkelly/LessWrong-43k
LessWrong
Meetup : Vancouver! Discussion article for the meetup : Vancouver! WHEN: 11 August 2013 03:00:00PM (-0700) WHERE: 2505 west broadway, vancouver, bc The usual, come chat and hang out with cool smart people who are into LW. 15:30 on Sunday at Benny's Discussion article for the meetup : Vancouver!
7e25217d-48f1-4e64-88d9-6e52dfe6a671
trentmkelly/LessWrong-43k
LessWrong
Berkeley visit report A few weeks ago, ShannonFriedman, an IFS counselor living in Berkeley, posted Female compatriots stay for a week in Berkeley an offer for rationalist women who are agenty to stay in her spare room for a few days. I replied, as I have a goal of getting to know the people associated with Less Wrong...
aabe152a-69ea-4faa-abbc-d47a75017146
StampyAI/alignment-research-dataset/blogs
Blogs
Local KL Divergence The KL divergence is an important tool for studying the distance between two probability distributions. Formally, given two distributions ![p](https://s0.wp.com/latex.php?latex=p&bg=f0f0f0&fg=555555&s=0&c=20201002) and ![q](https://s0.wp.com/latex.php?latex=q&bg=f0f0f0&fg=555555&s=0&c=20201002), th...
9dc546be-aeab-42d0-9b6a-72c31af4bb24
LDJnr/LessWrong-Amplify-Instruct
LessWrong
"One of your very early philosophers came to the conclusion that a fully competent mind, from a study of one fact or artifact belonging to any given universe, could construct or visualize that universe, from the instant of its creation to its ultimate end . . . —First Lensman If any one of you will concentrate upon one...
6e529138-2dbb-4fc2-ba62-6b89c2e1e4e9
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
Can we efficiently explain model behaviors? ARC’s [current plan](https://ai-alignment.com/mechanistic-anomaly-detection-and-elk-fb84f4c6d0dc) for solving ELK (and maybe also deceptive alignment) involves three major challenges: 1. Formalizing probabilistic heuristic argument as an operationalization of “explanation” ...
d4195975-cb37-4acc-93ef-1ed077aa1dac
trentmkelly/LessWrong-43k
LessWrong
Bayesian Reflection Principles and Ignorance of the Future Hi everyone, New user here. I am learning Bayesian epistemology by working through Mike Titelbaum’s recent Fundamentals of Bayesian Epistemology. I’ve been learning about deference principles — principles concerning deferring to experts, where experts may inc...
ea94836e-6e69-48a1-9204-435d649f71a0
trentmkelly/LessWrong-43k
LessWrong
Beauty and the Beast I remember when I first realized I was hideous. Generally, my autobiographical memory is poor  - but pain is the best anti-amnesic and it was among the more painful days of my life. Before that first day of Kindergarten, my parents kept me isolated. They say they did this to protect me from ridic...
7dec9f78-66a2-4686-96fd-67961a0bf0dc
StampyAI/alignment-research-dataset/arxiv
Arxiv
Learning to Play Against Any Mixture of Opponents 1 Introduction --------------- Reinforcement learning (RL) agents commonly interact in environments with other agents, whose behavior may be uncertain. For any particular probabilistic belief over the behavior of another agent (henceforth, *opponent*), we can learn ...
a56c5363-e58f-4db8-b3f0-cdf0aacbf9a7
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
Deep Learning Systems Are Not Less Interpretable Than Logic/Probability/Etc There’s a common perception that various non-deep-learning ML paradigms - like logic, probability, causality, etc - are very interpretable, whereas neural nets aren’t. I claim this is wrong. It’s easy to see where the idea comes from. Look at...
cf3facb9-a6a7-4a58-bede-8e6d791833f4
trentmkelly/LessWrong-43k
LessWrong
Some Thoughts Are Too Dangerous For Brains to Think [EDIT - While I still support the general premise argued for in this post, the examples provided were fairly terrible. I won't delete this post because the comments contain some interesting and valuable discussions, but please bear in mind that this is not even close...
5a160d79-c45f-4fe9-aa2c-7a4be0431067
trentmkelly/LessWrong-43k
LessWrong
Russia will do a nuclear test A thought occurred to me, and it's so logical, I concluded that it must be true. Russia will not detonate a nuclear weapon in Ukraine, fearing repercussion. Russia will detonate a nuclear weapon in Russia. In other words, Russia will do a nuclear test. Like North Korea did. As far as I...
da6d8d93-f2dd-42d1-97a6-69205ff092e9
StampyAI/alignment-research-dataset/youtube
Youtube Transcripts
229. The Case For Aligning Narrow Superhuman Models and welcome to session 229 in the asa.com reading group tonight we'll be discussing the uh the post on alignment forum called the case for aligning narrowly superhuman models by ajaya kocher we've seen some previously previous work by jayakotra her work on ai timelin...
1d754bf5-c9ab-4cb2-9920-e757d7dc1f53
StampyAI/alignment-research-dataset/eaforum
Effective Altruism Forum
Navigating the Future: A Guide on How to Stay Safe with AI | Emmanuel Katto Uganda Hello my name is Emmanuel Katto and I'm from Uganda. Today I will share how to stay safe with AI. Artificial Intelligence (AI) is transforming the way we live and work, offering incredible opportunities and conveniences. However, as AI ...
9e672078-5e3e-454d-b3b0-9e1727948254
StampyAI/alignment-research-dataset/special_docs
Other
CHAI Newsletter #2 2019 5/4/22, 2:43 PM Center for Human-Compatible AI (CHAI) Newsletter https://mailchi.mp/1e421759290d/jq4ushvzjd 1/5H u ma n - C o mp a t i b l e Now Available for Pre-Order on Amazon Stuart Russell's new book,  Human Compatible: Artificial Intelligence and the Problem of Control , is available for P...
11dea935-98fb-49e3-92ab-ef0ee608fc59
trentmkelly/LessWrong-43k
LessWrong
Transcript: Testing ChatGPT's Performance in Engineering This is a supplement to an article I wrote elsewhere, to provide the complete transcript of an interaction with ChatGPT.  I evaluated ChatGPT's performance by tasking it with real-world problems I encounter in my work as an applications engineer.  Main takeawa...
4e3af9f2-6bf8-4e12-93b9-a0b4813807da
StampyAI/alignment-research-dataset/arbital
Arbital
Bayesian reasoning According to its advocates, Bayesian reasoning is a way of seeing the world, and our beliefs about the world, in the light of [probability theory](https://arbital.com/p/1bv), in particular Bayes's Theorem or [Bayes's Rule](https://arbital.com/p/1lz). This probability-theoretic way of seeing the wor...
3fb48ba2-a4f1-4d65-9a7e-b7724ee42f21
trentmkelly/LessWrong-43k
LessWrong
Meetup : Brussels - all fun and games Discussion article for the meetup : Brussels - all fun and games WHEN: 19 April 2014 12:00:00PM (+0100) WHERE: Rue des Alexiens 55 1000 Bruxelles (Because of a scheduling conflict with the Berlin community meetup, April's meetup will not be on the second Saturday of the month. ...
be78350c-5205-48db-a262-29b3c70264e3
trentmkelly/LessWrong-43k
LessWrong
The Changing Face of Twitter [NOTE: This post has been edited extensively to account for key info I had missed.] As we move to ban TikTok, and as we use that move as an excuse to pass a new supercharged Patriot Act for the entire internet, every other social network including Twitter moves to transform itself into Ti...
545b2309-aea2-443d-813e-3b2878cb4d29
trentmkelly/LessWrong-43k
LessWrong
Prescriptions, Paradoxes, and Perversities [WARNING: I am not a pharmacologist. I am not a researcher. I am not a statistician. This is not medical advice. This is really weird and you should not take it too seriously until it has been confirmed] I. I’ve been playing around with data from Internet databases that agg...
3a61e0f2-ad11-42bb-be69-829b622ae3e4
trentmkelly/LessWrong-43k
LessWrong
[SEQ RERUN] The Uses of Fun (Theory) Today's post, The Uses of Fun (Theory) was originally published on 02 January 2009. A summary (taken from the LW wiki):   > Fun Theory is important for replying to critics of human progress; for inspiring people to keep working on human progress; for refuting religious arguments ...
c2445303-6012-4987-b320-1791a2eeaa6b
trentmkelly/LessWrong-43k
LessWrong
All knowledge is circularly justified Many philosophers have tried to find the foundations of our knowledge, but why do we think there are any? The framing of foundations implies a separate bottom layer of knowledge from which everything is built up. And while this is undoubtedly a useful model in many contexts, why s...
205959e4-1128-468d-a9a8-e55b08ce49cd
trentmkelly/LessWrong-43k
LessWrong
DIY Deliberate Practice In the spirit of growth and self-improvement, I recently attempted to apply Ericsson’s principles of deliberate practice to my own growth goal: speeding up my writing. If you're unfamiliar with the minutia of Ericsson's methods, don't worry, I was in the same boat — and hence my initial goal ha...
2875e479-2b50-4db2-8bd4-6616b05cf1de
StampyAI/alignment-research-dataset/lesswrong
LessWrong
AI Alignment Research Engineer Accelerator (ARENA): call for applicants TL;DR ===== Apply[**here**](https://airtable.com/shrYsHmqcrIurFONE)for the second iteration of ARENA!    Introduction ============ We are excited to announce the second iteration of ARENA (Alignment Research Engineer Accelerator), a 6-week ML ...
c49c6f73-fa9f-40b4-abab-77192d4ebaef
StampyAI/alignment-research-dataset/lesswrong
LessWrong
Reflections on Trusting Trust & AI **Reflections on Trusting Trust & AI** ====================================== The goal of the post is to highlight the necessity for mechanistic interpretability from an information security perspective: ### **Background** In 1984, Ken Thompson authored a seminal paper on informat...
f80fdfbf-3862-4322-afde-5055e473878d
trentmkelly/LessWrong-43k
LessWrong
Zach Weiner on video game violence See the comic here. And the red button below:   What are your thoughts on this?
3883e3e8-bde6-4d8a-b540-7d993f002a8d
trentmkelly/LessWrong-43k
LessWrong
Duels & D.Sci March 2022: It's time for D-d-d-d-d-d-d-d-d-d-d-d-d-d-data! This is an entry in the 'Dungeons & Data Science' series, a set of puzzles where players are given a dataset to analyze and an objective to pursue using information from that dataset.  STORY (skippable) You may have experienced a meteoric rise...
99ac0a76-6878-4bbf-843e-e9a97de36e0e
StampyAI/alignment-research-dataset/blogs
Blogs
Giving Tuesday 2019 **Update January 25, 2020**: $77,325 was donated to MIRI through Facebook on Giving Tuesday. $45,915 of this was donated within 13.5 seconds of the Facebook matching event starting at 5:00AM PT and was matched by Facebook. Thank you to everybody who set their clocks early to support us so generousl...
6a93e91b-e78c-420e-9716-57fdce61ea14
trentmkelly/LessWrong-43k
LessWrong
Are we the Wolves now? Human Eugenics under AI Control For thousands of years, humans have selectively bred animals to enhance desirable traits. Dogs, originally wolves, were bred for hunting, guarding, herding, and companionship. Over generations, we shaped them to be smarter, more obedient, or even just cuter. The m...
900d709e-9f74-491c-a5bb-07a23c437ce7
trentmkelly/LessWrong-43k
LessWrong
[SEQ RERUN] Total Tech Wars Today's post, Total Tech Wars was originally published on November 29, 2008. A summary:   > Unpleasant situations can arise if people believe that the development of some sort of AI is likely to be a big win for the developers, and a big loss for everyone else. Avoid the possibility of a ...
ccb6258c-72e6-4261-9c43-fb3b26516401
StampyAI/alignment-research-dataset/agisf
AGI Safety Fund
Summarizing books with human feedback To safely deploy powerful, general-purpose artificial intelligence in the future, we need to ensure that machine learning models act in accordance with human intentions. This challenge has become known as the *alignment problem*. A scalable solution to the alignment problem needs...