id
stringlengths
36
36
source
stringclasses
15 values
formatted_source
stringclasses
13 values
text
stringlengths
2
7.55M
89730c2a-e3af-4bae-b0de-cb7565fd1cbb
StampyAI/alignment-research-dataset/arxiv
Arxiv
Open Questions in Creating Safe Open-ended AI: Tensions Between Control and Creativity 1 Introduction --------------- Artificial life (ALife) and artificial intelligence (AI) have largely developed independently as fields. Statistical machine learning (ML), including deep learning, has driven much progress in moder...
91ce5605-9023-4db2-9f97-ba851f5ccb0c
StampyAI/alignment-research-dataset/eaforum
Effective Altruism Forum
Owen Cotton-Barratt: What does (and doesn't) AI mean for effective altruism? --- *In this 2017 talk, The Future of Humanity Institute's* [*Owen Cotton-Barratt*](https://www.fhi.ox.ac.uk/team/owen-cotton-barratt/) *discusses what strategy effective altruists ought to adopt with regards to the development of advanced a...
bf1555e9-ba13-4e0a-b15c-64b270a05735
trentmkelly/LessWrong-43k
LessWrong
Good job opportunities for helping with the most important century Yes, this is my first post in almost a year. I’m no longer prioritizing this blog, but I will still occasionally post something. I wrote ~2 years ago that it was hard to point to concrete opportunities to help the most important century go well. That’...
c9d247d3-1ffb-4f8d-9cc1-75f64c06ca98
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
What happens to variance as neural network training is scaled? What does it imply about "lottery tickets"? Daniel Kokotajlo [asks](https://www.lesswrong.com/posts/wFJqi75y9eW8mf8TR/does-the-lottery-ticket-hypothesis-suggest-the-scaling) whether the lottery ticket hypothesis implies the scaling hypothesis. The way I s...
41ba3268-831d-4a2c-be1a-73dc5ab8ba7c
trentmkelly/LessWrong-43k
LessWrong
The Importance of Goodhart's Law This article introduces Goodhart's law, provides a few examples, tries to explain an origin for the law and lists out a few general mitigations. Goodhart's law states that once a social or economic measure is turned into a target for policy, it will lose any information content that h...
3d9a4c86-bbc4-475b-b84d-8776c22c202d
StampyAI/alignment-research-dataset/arbital
Arbital
Derivative The derivative of $y$ with respect to $x$ describes the [https://arbital.com/p/-rate](https://arbital.com/p/-rate) at which $y$ changes, given a change in $x$. In particular, we consider how tiny changes in one variable affect another variable. To take the derivative of a [https://arbital.com/p/-3jy](https:...
3a380066-f72e-46a5-9968-62ef480eab1e
trentmkelly/LessWrong-43k
LessWrong
Please steelman the accusations of election fraud Trump has accused several states of faking votes to help Joe Biden win the election. Meanwhile, Mitch Mcconnel is allowing Republican congressman to take either position on the issue (heresay). Popular fringe members of the Republican party Ted Cruz and Jim Jordan have...
51a0e809-c2fd-4ee6-86fb-a3318fbd1e62
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
Alignment Newsletter #39 Happy New Year! Find all Alignment Newsletter resources [here](http://rohinshah.com/alignment-newsletter/). In particular, you can [sign up](http://eepurl.com/dqMSZj), or look through this [spreadsheet](https://docs.google.com/spreadsheets/d/1PwWbWZ6FPqAgZWOoOcXM8N_tUCuxpEyMbN1NYYC02aM/edit?u...
3691f5d1-500d-4999-845a-0132d72e6e71
trentmkelly/LessWrong-43k
LessWrong
The Fear of Common Knowledge Followup to:  Belief in Belief One of those insights that made me sit upright and say "Aha!"  From The Uncredible Hallq: > Minor acts of dishonesty are integral to human life, ranging from how we deal with casual acquaintances to writing formal agreements between nation states.  Steven P...
ab1a39fb-3e67-4fd6-85b9-88bd0a5f35f1
trentmkelly/LessWrong-43k
LessWrong
Rationality Quotes Thread February 2016 Another month, another rationality quotes thread. The rules are: * Provide sufficient information (URL, title, date, page number, etc.) to enable a reader to find the place where you read the quote, or its original source if available. Do not quote with only a name. * Post al...
7a8b35b5-9082-47b9-8cee-5194a53e86fd
trentmkelly/LessWrong-43k
LessWrong
debating buying NVDA in 2019 Alice: You saw GPT-2, right? Bob: Of course. Alice: It's running on GPUs using CUDA. OpenAI will keep scaling that up, and other groups will want to do the same thing. Bob: Right. Alice: So, does this mean we should buy Nvidia stock? Bob: I'm not sure. Nvidia makes the hardware used n...
d309004b-b8dc-4b89-a502-7fbe6ac683ce
trentmkelly/LessWrong-43k
LessWrong
A Significant Portion of COVID-19 Transmission Is Presymptomatic Epistemic status: Not quite settled science, but preprints seem to agree. Strong evidence points to presymptomatic sources as a major source of COVID-19 infections, possibly the majority. The exact proportion is environment-dependent; awareness and publ...
c6f3efb4-0ed8-439d-97a2-e9a24f5ec8dc
trentmkelly/LessWrong-43k
LessWrong
How To Do Patching Fast This post outlines an efficient implementation of Edge Patching that massively outperforms common hook-based implementations. This implementation is available to use in my new library, AutoCircuit, and was first introduced by Li et al. (2023). What is activation patching? I introduce new term...
3b0fc070-0814-4497-8d60-6e73df916ce3
trentmkelly/LessWrong-43k
LessWrong
General alignment properties AIXI and the genome are both ways of specifying intelligent agents.  1. Give AIXI a utility function (perhaps over observation histories), and hook it up to an environment, and this pins down a policy.[1]  2. Situate the genome in the embryo within our reality, and this eventually grows...
d8d1a4e0-3c31-40c6-a5d3-91f8aaf19551
trentmkelly/LessWrong-43k
LessWrong
Tulpa References/Discussion There have been a number of discussions here on LessWrong about "tulpas", but it's been scattered about with no central thread for the discussion. So I thought I would put this up here, along with a centralized list of reliable information sources, just so we all stay on the same page. Tul...
8a368698-f9ed-4534-ac7f-f6114a883981
trentmkelly/LessWrong-43k
LessWrong
Petition: Off topic area   Petition: LW should introduce a dedicated off topic area   Why?   1) I want to discuss various topics with people who are both intelligent and rationalist, and i know of no other place where to do it.   2) If find that rationality is getting boring in itself. I need to use it on some...
ebd58a55-6b5c-47ba-afb8-50cc02639950
trentmkelly/LessWrong-43k
LessWrong
Meetup : First Meetup- Cleveland and Akron, Ohio Discussion article for the meetup : First Meetup- Cleveland and Akron, Ohio WHEN: 17 November 2012 08:03:00PM (-0400) WHERE: Cleveland, OH I'm posting on behalf of someone who got in touch with me on the Ohio listserv. He is looking for people active in the Cleveland...
9a06d9ae-2b3d-42a3-a5bd-12cdc5f99b16
trentmkelly/LessWrong-43k
LessWrong
On not diversifying charity A common belief within the Effective Altruism movement that you should not diversify charity donations when your donation is small compared to the size of the charity. This is counter-intuitive, and most people disagree with this. A Mathematical Explanation of Why Charity Donations Shouldn'...
7fdb1366-d4f5-4f3f-bf6e-a0456b7f80b5
trentmkelly/LessWrong-43k
LessWrong
What are you working on? December 2011 This is the sixth bimonthly 'What are you working On?' thread. Previous threads are here. So here's the question: What are you working on?  Here are some guidelines: * Focus on projects that you have recently made progress on, not projects that you're thinking about doing but...
b39aea49-e801-48c7-9379-be379e13cb0b
trentmkelly/LessWrong-43k
LessWrong
LW Women- Crowdsourced research on Cognitive biases and gender In the last LW Women post, it was mentioned, and I agree, that a two-way conversation is more productive, and presents varied viewpoints better than a one-way lecture. To that end, I am making this post an experiment in crowdsourcing research to LW. Instea...
3229ecbe-6ee5-4149-acb1-4e35e25a46f9
trentmkelly/LessWrong-43k
LessWrong
Polarization is Not (Standard) Bayesian TLDR: Standard rational models can explain many aspects of polarization. But not all. Ideological sorting implies that our beliefs often evolve in predictable ways, violating the “martingale property” of Bayesian belief-updating. Rational models need to reckon with this. ------...
501e94c0-6158-45b3-baa8-be9c978618b3
trentmkelly/LessWrong-43k
LessWrong
Preference synthesis illustrated: Star Wars EDIT: this post is intended to be mainly about preference ordering synthesis, not Star Wars. This is my ordering of the core Star Wars movies: * The Empire Strikes Back > Return of the Jedi >= The Last Jedi > Rogue One > A New Hope = The Rise of Skywalker > The Force Awak...
8910975a-aec4-47df-8a7a-df93e83629fe
trentmkelly/LessWrong-43k
LessWrong
Dangerous Virtual Worlds Once, I played a virtual reality game where robots attack you and you shoot them. This was a very fun and compelling experience despite being rather basic. In fact, it was too compelling -- the experience was so intense that I had to stop playing, as I thought that if things got significantly ...
37eb7e7f-fb1f-42b4-8a91-6413d8760f94
trentmkelly/LessWrong-43k
LessWrong
LW Study Group Facebook Page Update: There is now an online sign up to groups with workflowy, based on subject and current ability. You do not have to be signed up to Facebook to join a group, but do add an email address so that the group can contact you:  https://workflowy.com/shared/cf1fd9ca-885f-c1b9-c2e8-e3a315f70...
8ee7a6b3-ced7-45a8-b7e3-4cb1fea8ca98
trentmkelly/LessWrong-43k
LessWrong
Boston Megameetup July 13-14 On the weekend of July 13-14, Harvard High-Impact Philanthropy will be hosting a rationalist megameetup. Everyone who can make the trip is strongly encouraged to come. We'll meet at the Harvard Science Center in room B-10 at noon on both days. Crash space is available; comment or send me a...
ecc3b05f-e141-4cfb-a9ee-0436b2b182bb
StampyAI/alignment-research-dataset/lesswrong
LessWrong
Concerns Surrounding CEV: A case for human friendliness first I am quite new here so please forgive the ignorance (I'm sure there will be some) of these questions, but I am all of about half way through reading CEV and I just simply cannot read any further without formal clarification from the lw community. That being...
6020cf5e-5641-4e7e-a4a5-e7d944e2a3be
trentmkelly/LessWrong-43k
LessWrong
Meetup Tip: Conversation Starters Summary If you're running or attending meetups a lot, it's helpful to have a short list of conversation starters ready. Conversation starters, sometimes called icebreaker questions, are questions that everyone can answer. They aren't deeply personal, but you do learn something new abo...
09df8563-d31f-4e1a-b12c-e393d79d6095
StampyAI/alignment-research-dataset/arxiv
Arxiv
A Low-Cost Ethics Shaping Approach for Designing Reinforcement Learning Agents 1 Introduction --------------- As AI systems become part of our lives and sometimes make decisions related to life-or-death consequences such as clinical decision making [[Bennett and Hauser2013](#bib.bibx9)], awareness should be raised ...
ba316fa2-7285-4c15-bcc9-1e225f6fdfc1
trentmkelly/LessWrong-43k
LessWrong
What (feasible) augmented senses would be useful or interesting? The BrainPort is a device which allows blind people to have their sight restored (partially) by sending signals through electrodes that rest on the tongue.  > The signals are sent to the tongue via a "lollipop," an electrode array about nine square cent...
765b6c44-c6fc-423a-83c9-a02f8aceb616
trentmkelly/LessWrong-43k
LessWrong
What's Your Best AI Safety "Quip"? Motivated by thinking gay rights were advanced by asking "When did you choose to be straight?" Which emphasised that what isn't a choice and doesn't harm others shouldn't be proscribed. Here, we're seeking a memetic way of framing the fact that the alignment problem is unsolved. Aut...
128b769f-9a63-4092-9b5e-53826b7ac398
trentmkelly/LessWrong-43k
LessWrong
Fighting the evil influence of Facebook (but keeping the good bits): a manifesto and how-to guide
b935cac7-39cf-45e0-ad64-75b73785574a
trentmkelly/LessWrong-43k
LessWrong
Lifeguards 1 Many readers will be familiar with Peter Singer’s Drowning Child experiment: > On your way to work, you pass a small pond. On hot days, children sometimes play in the pond, which is only about knee-deep. The weather’s cool today, though, and the hour is early, so you are surprised to see a child splashin...
9bde7544-8930-4d53-941b-0b3b318d85d9
trentmkelly/LessWrong-43k
LessWrong
Meetup : Montreal Less Wrong/Effective Altruism - 8, 760 hours: Setting goals for 2015 Discussion article for the meetup : Montreal Less Wrong/Effective Altruism - 8, 760 hours: Setting goals for 2015 WHEN: 08 January 2015 07:00:00PM (-0500) WHERE: 1720 st denis montreal The end of a year is the perfect time to rev...
38e64646-450b-493d-bdba-0dc0873763e7
trentmkelly/LessWrong-43k
LessWrong
What did you change your mind about in the last year? Seems like a good New Year activity for rationalists, so I thought I'd post it early instead of late. Here are some steps I recommend: * Go looking through old writings and messages.  * If you keep a journal, go look at a few entries from throughout the last ...
4c177ed3-44b1-4f2c-8cd0-1473902247d3
trentmkelly/LessWrong-43k
LessWrong
How do you think about mildly technical people trying to advance science and technology? If you were, say 22, had free time, enough money, and decent curiosity and technical skills (but less than spectacular) what would you do? I could get a PhD. I could try and do something with more upside—maybe startups. But what o...
b6500050-de8a-46fe-bfda-e75d5772cc36
trentmkelly/LessWrong-43k
LessWrong
Meetup : Atlanta: Problems With Friendly AI: A (non-technical) MIRI Book Review Discussion article for the meetup : Atlanta: Problems With Friendly AI: A (non-technical) MIRI Book Review WHEN: 12 April 2014 07:00:00PM (-0400) WHERE: 491 Lindbergh Place NE Apt 618 Atlanta, GA 30324 At this meetup, we'll be looking ...
0cc1576d-4ff6-4805-9d52-50f23ec5f721
trentmkelly/LessWrong-43k
LessWrong
Inconsistent Beliefs and Charitable Giving There is common tendency in human life to act in ways contrary to what we believe. The classic example is the German people under Nazi rule, most of whom likely thought of themselves as good people—the kind of people who would help their neighbors even at risk to themselves,...
1e9fd545-a995-4dcb-9dfb-226df8871799
trentmkelly/LessWrong-43k
LessWrong
21 on 21 This is a linkpost for https://amirbolous.com/posts/21/ Today, I turn 21. Here are 21 reflections for 21 years. 1. The highest ROI thing you can invest in early on is learning to be an independent thinker. The world is full of noise. It's impossible to drown out it all out, but it's possible to pick out th...
70e7d322-79f4-41bb-a86f-45f99daa222f
trentmkelly/LessWrong-43k
LessWrong
[Link] TED Talk on Perceived Value Rory Sutherland, an ad man (who missed his calling as a comedian), gives this talk on perceived value versus "real" value, and comes down in favor of more of the first.  He also dabbles in history, status, behavioral economics, and the importance of user interface design.
8febe736-27e2-4c78-9435-3509f57178db
trentmkelly/LessWrong-43k
LessWrong
Towards Guaranteed Safe AI: A Framework for Ensuring Robust and Reliable AI Systems I want to draw attention to a new paper, written by myself, David "davidad" Dalrymple, Yoshua Bengio, Stuart Russell, Max Tegmark, Sanjit Seshia, Steve Omohundro, Christian Szegedy, Ben Goldhaber, Nora Ammann, Alessandro Abate, Joe Hal...
bd4bc042-68fa-4cfb-87df-d038bd16a74b
trentmkelly/LessWrong-43k
LessWrong
Training Regime Day 21: Executing Intentions Introduction It is written about the first of the Twelve Virtues: > Curiosity seeks to annihilate itself; there is no curiosity that does not want an answer. The main flaw of the Twelve Virtues is that they are epistemic in nature. It is written about the first of the ap...
93610eec-fa25-4e1b-8af4-37f4035ee8c0
trentmkelly/LessWrong-43k
LessWrong
Seeking ethical rules-of-thumb for comparison Rules-of-thumb are handy, in that they let you use a solution you've figured out beforehand without having to take the time and effort to re-derive it in the heat of the moment. They may not apply in all situations, they may not provide the absolutely maximally best answer...
e5cddb7a-570a-4221-b6f2-ed306b119a5a
trentmkelly/LessWrong-43k
LessWrong
Quantilizers maximize expected utility subject to a conservative cost constraint Summary: this post reframes ideas from "Learning a concept using only positive examples" and "The AI, the best human advisor" in terms of a strong constraint for conservative behavior, and shows that quantilizers are optimal under this co...
44cf4e76-b700-49fb-b5f4-46885ac43f9e
trentmkelly/LessWrong-43k
LessWrong
Evolution from distinction difference If we have norms such that each copy of a common behavior must to be a tiny step away from from its parent, rather than a giant step or no step, this would seem to make culture much more amenable to gradient descent via evolution than it otherwise would be. Is the latter somehow ...
b531e15c-cf3d-4c21-9187-d822416d6d8b
StampyAI/alignment-research-dataset/lesswrong
LessWrong
A rough idea for solving ELK: An approach for training generalist agents like GATO to make plans and describe them to humans clearly and honestly. GATO is the most general agent we currently know about. It's a general-purpose model with transformer architecture. GATO can play atari games, speak to humans, and classify...
9ef5406f-8269-4963-9612-f8e684548314
trentmkelly/LessWrong-43k
LessWrong
Wielding civilization (cross-posted from my blog, Sunday Stopwatch) The other day, I was walking home on an empty road and a bus drew past me. It got me thinking about the differences in our speeds. I could never compete with the bus on speed. Not only because the bus has a big engine, but also because of the context...
71f45c69-7a45-4452-8f78-11963e112dc2
trentmkelly/LessWrong-43k
LessWrong
How do the side effects of Novavax compare to other COVID-19 vaccines? I'm curious whether Novavax has the same side-effects of the mRNA vaccines where normal side-effects often involve the person being feeling too sick to do anything and having fever the next day. Unfortunately, most sources I find while googling are...
e4671c33-930d-4f98-b4d4-18a834f4cc26
trentmkelly/LessWrong-43k
LessWrong
Toward Safety Cases For AI Scheming Developers of frontier AI systems will face increasingly challenging decisions about whether their AI systems are safe enough to develop and deploy. One reason why systems may not be safe is if they engage in scheming. In our new report "Towards evaluations-based safety cases for AI...
e67b30c9-09a6-4973-8c40-b4a2d29f5c8d
trentmkelly/LessWrong-43k
LessWrong
AGI in sight: our look at the game board From our point of view, we are now in the end-game for AGI, and we (humans) are losing. When we share this with other people, they reliably get surprised. That’s why we believe it is worth writing down our beliefs on this. 1. AGI is happening soon. Significant probability of ...
61649c22-896e-4649-bb54-f00f98268e00
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
An anthropomorphic AI dilemma *[Metadata: crossposted from <https://tsvibt.blogspot.com/2023/01/the-anthropomorphic-ai-dilemma.html>. First completed January 21, 2023.]* Either generally-human-level AI will work internally like humans work internally, or not. If generally-human-level AI works like humans, then takeo...
ff172a46-2fb9-4482-bd0b-20d380e366e3
trentmkelly/LessWrong-43k
LessWrong
An alternative approach to superbabies (Note: This post might be slightly funnily written but it is not a joke but serious and important.) Me: I think it's possible that we might not need to figure out biotechnological advances for how we can create superbabies through embryo selection or so, but that actually  they ...
88bc7ccd-5400-4951-9859-6a13150fbe18
StampyAI/alignment-research-dataset/lesswrong
LessWrong
[Linkpost] A shared linguistic space for transmitting our thoughts from brain to brain in natural conversations *This is a linkpost for* [*https://www.biorxiv.org/content/10.1101/2023.06.27.546708v1*](https://www.biorxiv.org/content/10.1101/2023.06.27.546708v1)*.*  > Effective communication hinges on a mutual unders...
b2822bef-5028-418f-8957-a625e4509637
trentmkelly/LessWrong-43k
LessWrong
More thoughts on assertions Response to: The "show, don't tell" nature of argument Morendil says not to trust simple assertions. He's right, for the certain class of simple assertions he's talking about. But in order to see why, let's look at different types of assertions  and see how useful it is to believe them. S...
bbcb0d72-36f5-4ed2-ab1d-03f47d586cbb
trentmkelly/LessWrong-43k
LessWrong
Serving LLM on Huawei CloudMatrix AI 2027 Compute Forecast basically completely ignores China for Compute Production section and I don't think it can be justified. This paper from Huawei is a timely reminder. 
2f4d3584-6090-479f-96ad-93689feedccf
trentmkelly/LessWrong-43k
LessWrong
PoMP and Circumstance: Introduction This is the first part of a series about the Principle of Minimal Privilege, also known as the Principle of Least Authority. One technique that is quickly trotted out in discussions about making AI safer is just not giving the AI agents access to certain things. A naive suggestion ...
ad7cb940-fd43-4f1c-9861-064e1c5c0a83
trentmkelly/LessWrong-43k
LessWrong
The Power to Be Emotionally Mature This is Part VII of the Specificity Sequence When Steve unknowingly lacks an understanding of what his claim that "Uber exploits its drivers" means, we can say that he's being intellectually immature. Similarly, when someone unknowingly lacks an understanding of the cause of an emot...
0a4c570f-805b-4080-a824-6a5ce00d49fe
trentmkelly/LessWrong-43k
LessWrong
Rationality Quotes: April 2011 You all know the rules: * Please post all quotes separately, so that they can be voted up/down separately.  (If they are strongly related, reply to your own comments.  If strongly ordered, then go ahead and post them together.) * Do not quote yourself. * Do not quote comments/posts o...
03b4fdfb-8a59-4c63-bca3-60968f763be3
trentmkelly/LessWrong-43k
LessWrong
Beware The Man Of One Study Aquinas famously said: beware the man of one book. I would add: beware the man of one study. For example, take medical research. Suppose a certain drug is weakly effective against a certain disease. After a few years, a bunch of different research groups have gotten their hands on it and d...
cc184f33-3210-4a9c-86b2-ee89e6472bf6
trentmkelly/LessWrong-43k
LessWrong
Upcoming LW Changes Thanks to the reaction to this article and some conversations, I'm convinced that it's worth trying to renovate and restore LW. Eliezer, Nate, and Matt Fallshaw are all on board and have empowered me as an editor to see what we can do about reshaping LW to meet what the community currently needs. T...
b25083f5-e030-4696-82f7-a3a6a4d3f93c
StampyAI/alignment-research-dataset/lesswrong
LessWrong
How teams went about their research at AI Safety Camp edition 8 AI Safety Camp connects new collaborators worldwide to discuss and decide on a concrete research proposal, gear up online as a team, and try their hand at AI safety research during intensive coworking sprints. Fourteen teams formed at this year's virtual...
e9970b68-8e01-4fb8-8c43-d1f5c197432c
StampyAI/alignment-research-dataset/blogs
Blogs
love, not competition love, not competition --------------------- many people are bemoaning that AI is going to replace them. this includes notably artists, but we can expect it to start covering [mathematicians](https://nitter.net/ScienceStanley/status/1584263426750132224) and, as AI advances, eventually every kind...
a83697d7-bf26-44cc-a1dd-d7a59e23cb0f
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
Paper: On measuring situational awareness in LLMs This post is a copy of the introduction of this [paper](https://owainevans.github.io/awareness_berglund.pdf) on situational awareness in LLMs. Authors: Lukas Berglund, Asa Cooper Stickland, Mikita Balesni, Max Kaufmann, Meg Tong, Tomasz Korbak, Daniel Kokotajlo, O...
7c3cce20-5a7d-47fa-ace9-930a70bb9ca1
StampyAI/alignment-research-dataset/blogs
Blogs
December 2018 Newsletter **Edward Kmett** [has joined the MIRI team](https://intelligence.org/2018/11/28/miris-newest-recruit-edward-kmett/)! Edward is a prominent Haskell developer who popularized the use of lenses for functional programming, and currently maintains many of the libraries around the Haskell core libra...
fcdace0c-490c-4f5d-b33f-7172fbcaf164
trentmkelly/LessWrong-43k
LessWrong
Conjoined twins who share a brain/experience? http://news.ycombinator.com/item?id=1874171   http://www2.macleans.ca/2010/11/02/a-piece-of-their-mind/print/
6b3ab03c-404a-46c5-bbbd-d3553028966c
StampyAI/alignment-research-dataset/lesswrong
LessWrong
High level discourse structure in ChatGPT: Part 2 [Quasi-symbolic?] In my previous post I examined a simple pattern of alternation in ChatGPT’s text output, [Of pumpkins, the Falcon Heavy, and Groucho Marx](https://www.lesswrong.com/posts/ZKgDpktMhnyE6Gvcs/of-pumpkins-the-falcon-heavy-and-groucho-marx-high-level). I’v...
c9f8bea8-be2d-4459-b8e1-b25e2c34b24e
StampyAI/alignment-research-dataset/blogs
Blogs
Where's Today's Beethoven? *Click lower right to download or find on Apple Podcasts, Spotify, Stitcher, etc.* This piece kicks off a short [series](https://www.cold-takes.com/tag/innovationstagnation/) inspired by this question: > Say that Beethoven was the greatest musician of all time (at least in some parti...
52232da3-1ab2-4b7c-a836-ff4a84be1276
trentmkelly/LessWrong-43k
LessWrong
Powerful mesa-optimisation is already here Toolformer: Language Models Can Teach Themselves to Use Tools Timo Schick, Jane Dwivedi-Yu, Roberto Dessì, Roberta Raileanu, Maria Lomeli, Luke Zettlemoyer, Nicola Cancedda, Thomas Scialom (Submitted: 9 Feb 2023) > Language models (LMs) exhibit remarkable abilities to solve...
6ba240e7-aa22-4ae2-b744-04996c22722d
trentmkelly/LessWrong-43k
LessWrong
Epistemic Laws of Motion
0f7693e9-e648-4df8-93e6-5bbd2555cfdf
trentmkelly/LessWrong-43k
LessWrong
Rediscovery, the Mind's Curare I'm often apprehensive about writing things that may already have been written, of sharing ideas that may already have been thought of, of saying things that may be cliché or dull or commonplace. I think I can trace this to a fear of being seen as ignorant or uneducated, which itself can...
4ed6ca7e-3f64-46bd-9c05-ed1b25fd170a
trentmkelly/LessWrong-43k
LessWrong
Sunday July 19, 1pm (PDT) — talks by Raemon, ricraz, mr-hire, Jameson Quinn This Sunday at 1pm (PDT), we're running another session of "lightning talks" by curated LessWrong authors (see here for previous weeks' transcripts). * Each talk will be 3-5 minutes followed by discussion. Afterwards, we'll have a hangout in...
b49d4b50-5482-4795-a177-b12cd44ee364
trentmkelly/LessWrong-43k
LessWrong
EA & LW Forum Weekly Summary (27th Feb - 5th Mar 2023) Supported by Rethink Priorities This is part of a weekly series summarizing the top posts on the EA and LW forums - you can see the full collection here. The first post includes some details on purpose and methodology. Feedback, thoughts, and corrections are welc...
b225c309-ba76-45dc-a484-1fac298c59f6
trentmkelly/LessWrong-43k
LessWrong
The Case of the Speluncean Explorers A fictional legal case about a team of explorers who got trapped in a cave and ate one of their own to survive. Summary: > From the testimony of the defendants, which was accepted by the jury, it appears that it was Whetmore who first proposed that they might find the nutriment wi...
3547b96a-dcc5-4c85-b52b-738b3b540fb8
trentmkelly/LessWrong-43k
LessWrong
Meetup : NYC Solstice Discussion article for the meetup : NYC Solstice WHEN: 19 December 2015 05:30:00PM (-0500) WHERE: 2 West 64th Street Our annual, huge East Coast event. Saturday will be the Solstice Proper - a beforeparty starting at 5:30pm, 2 hours of singalong music starting at 6:30, followed by an afterpart...
cf2a3343-ac5d-4208-b9b8-c2fb8c3e1e35
trentmkelly/LessWrong-43k
LessWrong
August gwern.net links None
cd99291d-1fd4-4ce0-8428-227a70b806b7
trentmkelly/LessWrong-43k
LessWrong
My Trial Period as an Independent Alignment Researcher In the past two months, I have tried out what is like to be an independent alignment researcher. My goals were to figure out if this path is something I would like to do, whether I'm a good fit, which research areas are most promising for me, and whether I feel li...
91336f75-8690-4057-8979-38fd01d975b1
StampyAI/alignment-research-dataset/arxiv
Arxiv
Cautious Reinforcement Learning with Logical Constraints 1. Introduction ---------------- Reinforcement Learning (RL) is an algorithm with no supervision that can be used to train an agent to interact with an unknown environment. The dynamics of the interaction are often assumed to be a Markov Decision Process (MDP...
aa195e1e-6edd-4ec9-82a4-125749b2970a
trentmkelly/LessWrong-43k
LessWrong
[Effective Altruism] Promoting Effective Giving at Conferences via Speed Giving Games Conferences provide a high-impact opportunity to promote effective giving. This is the broad take-away from an experiment in promoting effective giving at two conferences in recent months: the Unitarian Universalist (UU) General Asse...
a2242fa0-2ddd-4cf7-84cb-d3494337cff8
StampyAI/alignment-research-dataset/aisafety.info
AI Safety Info
What is John Wentworth's research agenda? [John Wentworth's plan](https://www.alignmentforum.org/posts/3L46WGauGpr7nYubu/the-plan) is to (1) sort out our fundamental confusions about [agency](/?state=5632&question=What%20is%20an%20agent%3F), and (2) do ambitious [value learning](/?state=9RYV&question=What%20is%20value...
fcf6584a-f267-4105-80ed-b02d0a03c224
trentmkelly/LessWrong-43k
LessWrong
Weirdtopias in science fiction There's "Day Million" and The Age of the Pussyfoot, both by Frederik Pohl, and even they might be more like utopias rather than adequately weird. "Day Million" is a very short story, with a narration which puts emphasis on how much people like living in that world even though it would m...
49409229-c920-446c-aaa3-c942065110d9
trentmkelly/LessWrong-43k
LessWrong
Transformative AI is a process In discussions, transformative AI is often seen as a certain point in time when an AI system achieves a certain level of performance or capabilities. For example, some people believe(d) that transformative AI will occur when an AI system can pass the Turing test, or when an AI system ca...
eb7c8712-ba9b-422e-aaec-e02cc4073cfc
trentmkelly/LessWrong-43k
LessWrong
Meetup : Yale rationality group meeting. Discussion article for the meetup : Yale rationality group meeting. WHEN: 05 October 2014 02:00:00PM (-0400) WHERE: Seminar room K22, Jonathan Edwards College, 68 High Street New Haven, CT 06511 The Yale rationality group will be holding a meeting this Sunday, October 5 at 2...
f8aa3fd8-67e7-4db6-8f69-f23ec6658699
trentmkelly/LessWrong-43k
LessWrong
The Catastrophic Convergence Conjecture Overfitting the AU landscape When we act, and others act upon us, we aren’t just changing our ability to do things – we’re shaping the local environment towards certain goals, and away from others.[1] We’re fitting the world to our purposes. What happens to the AU landscape[2] ...
182dc163-eeae-4412-8dd7-1c4b0a1e7f37
trentmkelly/LessWrong-43k
LessWrong
AI Safety Chatbot Hello World! The AISafety.info team is launching a prototype of the AI Safety Chatbot. The chatbot uses a dataset of alignment literature to answer any questions related to AI safety that you might have, while also citing established sources. Please keep in mind that this is a very early prototype a...
8afc9d8a-614e-4b34-912a-f5cefbaf3706
trentmkelly/LessWrong-43k
LessWrong
Logical uncertainty and mathematical uncertainty
2cbc0e09-b73b-4568-8248-e32c5d7dd4a1
trentmkelly/LessWrong-43k
LessWrong
A&I (Rihanna 'S&M' parody lyrics) Hi all, I wrote lyrics to the tune of Rihanna - S&M that reference a few themes of AI risk, from the point of view of an 'A'-maximizer. I hope you like it — comments welcome :) [edit 2023-05-23: added this intro] A! A! A! One more! A! A! A! One more! A! A! A! A! A! One more! A! A! A!...
ae0ec4f8-e266-47dd-bda7-cd6a2bde1343
trentmkelly/LessWrong-43k
LessWrong
Notion template for personal predictions I created a simple Notion template for creating personal predictions. You can set multiple reminders and even embed line charts showcasing the likelihood across time (though this does rely on an external service). This was created pretty quickly, and the predictions in the tem...
6ad926b9-3615-49cf-8c92-882a9bf76eb2
trentmkelly/LessWrong-43k
LessWrong
LW Philosophers versus Analytics By and large, I would bet money that the devoted, experienced, and properly sequenced LWer, is a better philosopher than the average current philosophy majors concentrating in the analytic tradition. I say this because I have regular philosophical conversations with both populations, a...
b846a7a0-6f64-4999-9c08-e76f36f7f0f6
trentmkelly/LessWrong-43k
LessWrong
Misalignment and misuse: whose values are manifest? Crossposted from world spirit sock puppet. AI related disasters are often categorized as involving misaligned AI, or misuse, or accident. Where: * misuse means the bad outcomes were wanted by the people involved, * misalignment means the bad outcomes were wanted ...
603a3bdc-e274-4772-9fad-1a02ff22f4f6
StampyAI/alignment-research-dataset/arxiv
Arxiv
A New Formalism, Method and Open Issues for Zero-Shot Coordination ### 1 Introduction In multi-agent reinforcement learning (MARL), variations of the *self-play* (SP) regime ([ ])[]tesauro1994td have been successful in producing superhuman policies for two-player zero-sum games such as chess, go, and poker (campbe...
a9ca5866-d633-417d-bac3-c1a439deabea
StampyAI/alignment-research-dataset/eaforum
Effective Altruism Forum
TAI Safety Bibliographic Database Authors: Jess Riedel and Angelica Deibel [Cross-posted to LessWrong](https://www.lesswrong.com/posts/4DegbDJJiMX2b3EKm/tai-safety-bibliographic-database)   In this post we present the first public version of our bibliographic database of research on the safety of transformative ar...
85206a47-5cff-4db9-8344-253ac6fb9e77
trentmkelly/LessWrong-43k
LessWrong
Aggregative Principles of Social Justice 1. Introduction 1.1. Three aggregative principles This article examines aggregative principles of social justice. These principles state that a social planner should make decisions as if they will face the aggregated personal outcomes of every individual in the population. Di...
0f849b83-13bb-44fa-9321-49a935337d47
trentmkelly/LessWrong-43k
LessWrong
What is a definition, how can it be extrapolated? What is a definition? Philosophy has, ironically, a large number of definitions of definitions, but three of them are especially relevant to ML and AI safety. There is the intensional definition, where concepts are defined logically in terms of other concepts (“bachel...
da3beea9-4fe9-48b2-af5a-f1492690c2be
trentmkelly/LessWrong-43k
LessWrong
Meetup : Less Wrong Israel Meetup: Social and Board Games Discussion article for the meetup : Less Wrong Israel Meetup: Social and Board Games WHEN: 14 April 2015 07:00:00PM (+0300) WHERE: Google, Electra Tower, 98 Yigal Alon Street, Tel Aviv This time we're going to have a social meetup! It's going to be a game n...
a13ee505-39cb-42d5-a142-f4c1df440318
trentmkelly/LessWrong-43k
LessWrong
Randomized Zettelkasten Test Motivation and Method I read the book How to Take Smart Notes a couple years ago and have tried to keep my notes together in a Zettelkasten ever since. I currently use Obsidian, which contains 2500 individual notes collected over two or three years - some abstract to encourage being hooked...
3467fb26-47d0-4c93-9d84-8201fb58d30e
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
Notes on the Mathematics of LLM Architectures > *From a mathematical point of view, the building and training of a large transformer* > *language model (LLM) is the construction of a certain function, from some euclidean space to another, that has certain interesting properties. And it may therefore be surprising to...
01dedf14-d50e-425f-957c-4fd293a70cde
StampyAI/alignment-research-dataset/lesswrong
LessWrong
A method for empirical back-testing of AI's ability to self-improve *TL;DR; Train an LLM/AI with text only until 2016, and ask it to come up with AI architecture ideas based on ML research until then to see whether it could come up with the idea of transformers.* One of the biggest promises of A(G)I is automated and ...
f9049cf1-d9ea-4b7c-86e7-f5fe8449569c
trentmkelly/LessWrong-43k
LessWrong
Muddling Along Is More Likely Than Dystopia Summary: There are historical precedents where bans or crushing regulations stop the progress of technology in one industry, while progress in the rest of society continues. This is a plausible future for AI. Epistemic Status: My intuition strongly disagrees with other peop...
d5df0af6-39d0-4ffc-836a-48ec02f08dc0
trentmkelly/LessWrong-43k
LessWrong
Nitric Oxide Spray... a cure for COVID19?? According to https://www.biospace.com/article/releases/uk-clinical-trial-confirms-sanotize-s-breakthrough-treatment-for-covid-19/   > Patients with a self-administered nasal spray application found to have reduced SARS-CoV-2 log viral load by more than 95% in infected partic...
325ea74a-b47f-4f22-a516-54861970f294
trentmkelly/LessWrong-43k
LessWrong
Conditional on the first AGI being aligned correctly, is a good outcome even still likely? Or, what stops someone else from making their own AGI 1-6 months later that doesn’t play nice and ends the world?
597cb620-0a8d-4cf2-8f50-0ef2cd3dc284
trentmkelly/LessWrong-43k
LessWrong
It's 'Well, actually...' all the way down Some people (the “Boubas”) don’t like “chemicals” in their food. But other people (the “Kikis”) are like, “uh, everything is chemicals, what do you even mean?” The Boubas are using the word “chemical” differently than the Kikis, and the way they’re using it is simultaneously ...