id
stringlengths
36
36
source
stringclasses
15 values
formatted_source
stringclasses
13 values
text
stringlengths
2
7.55M
5d244ae6-f400-4d0a-aec8-8de91dd43281
trentmkelly/LessWrong-43k
LessWrong
Rational Terrorism or Why shouldn't we burn down tobacco fields? Related: Taking ideas seriously Let us say hypothetically you care about stopping people smoking.  You were going to donate $1000 dollars to givewell to save a life, instead you learn about an anti-tobacco campaign that is better. So you chose to donat...
a8025cc1-62b6-4872-8fe5-eb1d12b07166
trentmkelly/LessWrong-43k
LessWrong
STRUCTURE: A Crash Course in Your Brain This post is part of my Hazardous Guide To Rationality. I don't expect this to be new or exciting to frequent LW people, and I would super appreciate comments and feedback in light of intents for the sequence, as outlined in the above link. Also, note this is a STRUCTURE post, a...
4401cd0c-00fd-47a9-8df0-2624e2bc6c18
trentmkelly/LessWrong-43k
LessWrong
The national security dimension of OpenAI's leadership struggle As the very public custody battle over OpenAI's artificial intelligences winds down, I would like to point out a few facts, and then comment briefly on their possible significance. It has already been noticed that "at least two of the board members, Tash...
e9b8ee30-de61-4c30-88aa-846896a07162
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
The Pragmascope Idea Pragma (Greek): thing, object. A “pragmascope”, then, would be some kind of measurement or visualization device which shows the “things” or “objects” present. I currently see the pragmascope as *the* major practical objective of [work on natural abstractions](https://www.lesswrong.com/posts/cy3B...
884969a4-0c62-4fc3-833e-c8100825d25d
trentmkelly/LessWrong-43k
LessWrong
You are Underestimating The Likelihood That Convergent Instrumental Subgoals Lead to Aligned AGI This post is an argument for the Future Fund's "AI Worldview" prize.   Namely, I claim that the estimates given for the following probability are too high: > P(misalignment x-risk|AGI)”: Conditional on AGI being developed...
5d488a63-5d33-4219-95df-e83d1698ed8f
trentmkelly/LessWrong-43k
LessWrong
Reputation bets People don’t often put their money where their mouth is, but they do put their reputation where their mouth is all the time. If I say ‘The Strategy of Conflict is pretty good’ I am betting some reputation on you liking it if you look at it. If you do like it, you will think better of me, and if you don...
c805136b-ab96-44b1-adf5-edb21e2bbfb0
trentmkelly/LessWrong-43k
LessWrong
Are extrapolation-based AIs alignable? (This is an account of my checking a certain alignment idea and finding that it doesn't work. Also my thinking is pretty naive and could easily be wrong.) When thinking about AIs that are trained on some dataset and learn to extrapolate it, like the current crop of LLMs, I asked...
b96512fd-8295-47ab-9d95-f4c3b3856d8e
StampyAI/alignment-research-dataset/arbital
Arbital
Harmless supernova fallacy Harmless supernova fallacies are a class of arguments, usually a subspecies of false dichotomy or continuum fallacy, which [can equally be used to argue](https://arbital.com/p/3tc) that almost any physically real phenomenon--including a supernova--is harmless / manageable / safe / unimportan...
adfe86f6-d5b5-44b2-bcf6-e9a39545aaa1
StampyAI/alignment-research-dataset/eaforum
Effective Altruism Forum
Technical AGI safety research outside AI I think there are many questions whose answers would be useful for technical AGI safety research, but which will probably require expertise outside AI to answer. In this post I list 30 of them, divided into four categories. Feel free to get in touch if you’d like to discuss the...
e248c3b4-fc70-452b-9921-4463b6d6b903
trentmkelly/LessWrong-43k
LessWrong
Meetup : DC Meetup Discussion article for the meetup : DC Meetup WHEN: 20 May 2012 03:00:00PM (-0400) WHERE: National Portrait Gallery, Washington, DC 20001, USA If the SIAI response to the Givewell critique has been posted, that will be the meetup topic. Discussion article for the meetup : DC Meetup
809a3585-2386-46a6-a5f1-0c0b197146eb
StampyAI/alignment-research-dataset/eaforum
Effective Altruism Forum
Refer the Cooperative AI Foundation’s New COO, Receive $5000 **TL;DR: The** [**Cooperative AI Foundation**](https://www.cooperativeai.com/foundation) **(CAIF) is a new AI safety organisation and we're hiring for a** [**Chief Operating Officer**](https://www.cooperativeai.com/job-listing/chief-operating-officer) **(COO...
d7603973-6774-4bb5-8115-7307a1540bf9
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
Alignment Newsletter #42 Cooperative IRL as a definition of human-AI group rationality, and an empirical evaluation of theory of mind vs. model learning in HRI Find all Alignment Newsletter resources [here](http://rohinshah.com/alignment-newsletter/). In particular, you can [sign up](http://eepurl.com/dqMSZj), or loo...
b692083a-d059-4389-94d3-a861b4f28bbc
trentmkelly/LessWrong-43k
LessWrong
Why do you need the story? * Boy: Why are you washing your hands? * Shaman: Because this root is poisonous. * Boy: Then why pull it out of the ground? * Shaman: Because the pulp within it can cure the mosquito disease. * Boy: But, I thought you said it was poisonous. * Shaman: It is, but the outside is more so t...
f82e2a66-e327-4ac7-9ba0-49506b27eb1c
trentmkelly/LessWrong-43k
LessWrong
Analysis of Algorithms and Partial Algorithms
30a4d42d-7152-4928-bf80-816d516b9b37
trentmkelly/LessWrong-43k
LessWrong
Maze-solving agents: Add a top-right vector, make the agent go to the top-right Overview: We modify the goal-directed behavior of a trained network, without any gradients or finetuning. We simply add or subtract "motivational vectors" which we compute in a straightforward fashion. In the original post, we defined a "...
4f09a2af-611f-4875-bab7-77b6a0468cdf
trentmkelly/LessWrong-43k
LessWrong
Genetically Modified Humans Born (Allegedly) I realize normally we don't talk about the news or hot-button issues, but this is of sufficiently high importance I am posting anyway. There is a link from Nature News on Monday here. There is a link from MIT Technology Review discussing the documents uploaded by the team ...
7b75a4f0-b12e-4e5e-88d6-552aa253c60f
trentmkelly/LessWrong-43k
LessWrong
Meetup : Less Wrong in Dublin Discussion article for the meetup : Less Wrong in Dublin WHEN: 02 February 2013 04:30:25PM (+0000) WHERE: 28 Dame St Dublin, Co. Dublin (The Mercantile) A meeting of minds. Discussion article for the meetup : Less Wrong in Dublin
d8933ac0-a901-45bb-8443-e9c6284c4b0f
trentmkelly/LessWrong-43k
LessWrong
The Joan of Arc Challenge For Objective List Theory Introduction The Joan of Arc challenge to objective list theory, as I shall argue, shows that only happiness is of intrinsic value. Pluralists—people who say that there are many things of intrinsic value—otherwise known as objective list theorists (maybe there’s som...
430f0bff-b81f-4415-8685-8bb56ce24e1c
StampyAI/alignment-research-dataset/blogs
Blogs
August 2018 Newsletter #### Updates * New posts to the new [AI Alignment Forum](https://www.alignmentforum.org): [Buridan’s Ass in Coordination Games](https://www.alignmentforum.org/posts/4xpDnGaKz472qB4LY/buridan-s-ass-in-coordination-games); [Probability is Real, and Value is Complex](https://www.alignmentforum.or...
ae3b4690-a04c-430f-b2c8-601ba9f2e48d
trentmkelly/LessWrong-43k
LessWrong
Plan for mediocre alignment of brain-like [model-based RL] AGI (This post is a more simple, self-contained, and pedagogical version of Post #14 of Intro to Brain-Like AGI Safety.) (Vaguely related to this Alex Turner post and this John Wentworth post.) I would like to have a technical plan for which there is a stron...
e61976b8-e546-4472-b737-9b94268fe2bc
trentmkelly/LessWrong-43k
LessWrong
Causal confusion as an argument against the scaling hypothesis Abstract We discuss the possibility that causal confusion will be a significant alignment and/or capabilities limitation for current approaches based on "the scaling paradigm": unsupervised offline training of increasingly large neural nets with empirical ...
699cf16f-5e64-4247-bd83-169b8e6b2989
StampyAI/alignment-research-dataset/eaforum
Effective Altruism Forum
My highly personal skepticism braindump on existential risk from artificial intelligence. **Summary** ----------- This document seeks to outline why I feel uneasy about high existential risk estimates from AGI (e.g., 80% doom by 2070). When I try to verbalize this, I view considerations like  * selection effects at ...
b12b85b5-3435-401d-a712-249064fc2b8e
trentmkelly/LessWrong-43k
LessWrong
Exploiting the Typical Mind Fallacy for more accurate questioning? I was reading Yvain's Generalizing from One Example, which talks about the typical mind fallacy.  Basically, it describes how humans assume that all other humans are like them.  If a person doesn't cheat on tests, they are more likely to assume others ...
75f5142a-8842-495f-b840-f1495efc68e0
trentmkelly/LessWrong-43k
LessWrong
In Defense of a Butlerian Jihad [Epistemic Status: internally strongly convinced that it is centrally correct, but only from armchair reasoning, with only weak links to actual going-out-in-the-territory, so beware: outside view tells it is mostly wrong] I have been binge-watching the excellent Dwarkesh Patel during...
af85d921-77e4-4a95-8fd7-ff69dbeead61
trentmkelly/LessWrong-43k
LessWrong
Hard-Coding Neural Computation Previously: Teaser: Hard-coding Transformer Models Introduction Transformer models are incredibly powerful for natural language tasks (and they are starting to find uses in many other fields of machine learning). Unfortunately, it is nigh-impossible to interpret what goes on inside the...
68f9623e-e96f-4f2a-992c-f905a091a4bb
StampyAI/alignment-research-dataset/arxiv
Arxiv
Turing's Red Flag
63cdc040-5208-455d-b527-f37cc220cd0c
trentmkelly/LessWrong-43k
LessWrong
Permission for mind uploading via online files Giulio Prisco made a blog post giving permission to use the data in his Gmail account to reconstruct an uploaded copy of him. > To whom it may concern: > > I am writing this in 2010. My Gmail account has more than 5GB of data, which contain some information about me and...
baa2f372-66e0-4419-a4f1-d80a73fdef3e
trentmkelly/LessWrong-43k
LessWrong
Can ChatGPT count? A few month back there was a question on lesswrong about what the least impressive thing is going to be that GPT4 can't do. At the time I thought maybe counting could be a good candidate, because I had noticed that this seems to be a problem for all big transformer models.  Dall-E and other image g...
746f4947-c080-49cd-a040-6378199a1b7d
trentmkelly/LessWrong-43k
LessWrong
Suggestions for net positive LLM research I am starting a PhD in computer science, focusing on agent foundations so far, which is great. I intend to continue devoting at least half my time to agent foundations. However, for several reasons, it seems to be important for me to do some applied work, particularly with LL...
0f52f09c-2469-4cf2-8a96-7bbf90d0e7b2
trentmkelly/LessWrong-43k
LessWrong
Why you ask the significance question why Hello. (I am already doing 75% better) I used to think that keeping your identity small is laudable. The reasoning goes that if "me" and "you" are interchangable instead of being only applicable for "me" there is a wider scope of applicability. And there is an effect. At som...
59d7e8f6-dd8b-46bd-9f4f-83ec20d2b4d4
StampyAI/alignment-research-dataset/arxiv
Arxiv
Trust-Region Method with Deep Reinforcement Learning in Analog Design Space Exploration I Introduction --------------- The annual increment of computing power described by Moore’s law is pioneering unprecedented possibilities. This remarkable progress has been accompanied by a collinearity with tremendous increases...
e8210d5a-ee54-4206-9cdd-675dc6d03a8f
trentmkelly/LessWrong-43k
LessWrong
Meetup : Atlanta Meetup, Topic: How to Become Immortal Discussion article for the meetup : Atlanta Meetup, Topic: How to Become Immortal WHEN: 26 January 2014 07:00:00PM (-0500) WHERE: 491 Lindbergh Place NE Apt 618 Atlanta, GA 30324 Come join us! Our topic this time is how to become immortal. We'll be bring up the...
6d44222d-c03e-4bc7-893c-417b3da84857
trentmkelly/LessWrong-43k
LessWrong
[SEQ RERUN] Beware Stephen J. Gould Today's post, Beware of Stephen J. Gould was originally published on 06 November 2007. A summary (taken from the LW wiki):   > A lot of people have gotten their grasp of evolutionary theory from Stephen J. Gould, a man who committed the moral equivalent of fraud in a way that is d...
efe17d14-5ba7-4cfc-8ab6-976c1c2b851c
trentmkelly/LessWrong-43k
LessWrong
Just letting alcoholics drink "Wet houses"-- subsidized housing for alcoholics (they need to get most of their own money for alcohol, but their other expenses are covered) might actually be a good idea. It's cheaper than trying to get them to stop drinking, arguably kinder than trying to get people to take on a very h...
9f567dc1-ac93-4939-8867-9c010aafa84f
StampyAI/alignment-research-dataset/aisafety.info
AI Safety Info
How might things go wrong with AI even without an agentic superintelligence? Failures can happen with narrow non-agentic systems, mostly from humans not anticipating safety-relevant decisions made too quickly to react, much like in the [2010 flash crash](https://en.wikipedia.org/wiki/2010_flash_crash). A helpful meta...
d14e7171-821d-4d57-bf8a-4ec90f0ab1c2
trentmkelly/LessWrong-43k
LessWrong
What if AI is “IT”, and we don’t know about this? I am using IT character from that "IT" movie as a metaphor for describing a problem in AGI research. AI can potentially learn (or have already learnt) about all our irresistible pleasures and agonizing fears, similar to what IT could do in the film. AGI initially “kno...
a8bd07c7-54f7-4a14-88c5-3d38107b0e54
trentmkelly/LessWrong-43k
LessWrong
Shift Resources to Advocacy Now (Post 4 of 7 on AI Governance) In my previous post in this series, I estimated that we have 3 researchers for every advocate working on US AI governance, and I argued that this ratio is backwards. When allocating staff, you almost always want to have more people working on the more cent...
d4069a84-ef83-4e74-95ed-d307d7024369
trentmkelly/LessWrong-43k
LessWrong
Who are the people who are currently profiting from inflation? When prices increase we pay more for goods and services. Cost of living is higher because of inflation - we have to work more to earn enough money to buy the same goods and services. Who is at the other end of that transaction? Who is it that gets paid mor...
0a09ab2f-6185-44eb-9ae8-6b16fc1eef70
trentmkelly/LessWrong-43k
LessWrong
The Low-Hanging Fruit Prior and sloped valleys in the loss landscape You can find code for the referenced experiments in this GitHub repository Many have postulated that training large neural networks will enforce a simplicity, or Solomonoff prior. This is grounded in the idea that simpler solutions occupy expansive ...
87b59df1-bea5-4217-8ae0-0a91f4a70861
StampyAI/alignment-research-dataset/lesswrong
LessWrong
Many important technologies start out as science fiction before becoming real I was pretty impressed by [AI Risk is like Terminator; Stop Saying it's Not](https://forum.effectivealtruism.org/posts/zsFCj2mfnYZmSW2FF/ai-risk-is-like-terminator-stop-saying-it-s-not-1) and the [Followup](https://forum.effectivealtruism.o...
37049a19-69fc-4800-94ae-dd0863c1e09d
trentmkelly/LessWrong-43k
LessWrong
Helsinki LW Meetup - Sat March 5th According to the statistics, out of all the cities in the world without a Less Wrong meetup yet, Helsinki has the third largest LW readership. I know I would certainly like to get to personally know more people interested in rationality topics, and I'm sure I'm not alone in that. My ...
a8610dac-4eb0-406c-b389-f6d028db291b
StampyAI/alignment-research-dataset/lesswrong
LessWrong
Selfish preferences and self-modification One question I've had recently is "Are agents acting on selfish preferences doomed to having conflicts with other versions of themselves?" A major motivation of TDT and UDT was the ability to just do the right thing without having to be tied up with precommitments made by your...
6fc160e0-5244-4a49-b4e7-1353ec24b352
trentmkelly/LessWrong-43k
LessWrong
Infant Mortality and the Argument from Life History Many people argue that suffering predominates in nature. A really simple form of the argument, supported by people like Brian Tomasik, is what one might call the argument from life history. In general, in most species, females produce many more offspring than can sur...
0ab62371-238c-4796-ba4a-01d17cef4b2d
trentmkelly/LessWrong-43k
LessWrong
Less Wrong Sequences+Website feed app for Android I use my Android phone much more than my computer, and reading the Sequences on a mobile device is a pain. I needed an easy way to access the Sequences, but since there are no apps for this website I had to create one myself. Since I'm no app developer, I used the IBui...
208fd70c-7219-4b24-9deb-a2bbecc8cf11
trentmkelly/LessWrong-43k
LessWrong
A note about differential technological development Quick note: I occasionally run into arguments of the form "my research advances capabilities, but it advances alignment more than it advances capabilities, so it's good on net". I do not buy this argument, and think that in most such cases, this sort of research does...
64c618b3-61ce-4ca2-a1d4-34812dc20449
trentmkelly/LessWrong-43k
LessWrong
Making Nanobots isn't a one-shot process, even for an artificial superintelligance Summary: Yudkowsky argues that an unaligned AI will figure out a way to create self-replicating nanobots, and merely having internet access is enough to bring them into existence. Because of this, it can very quickly replace all human ...
1c5869c9-3232-4ae0-896f-404d1b8651b7
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
Agency in Conway’s Game of Life *Financial status: This is independent research. I welcome* [*financial support*](https://www.alexflint.io/donate.html) *to make further posts like this possible.* *Epistemic status: I have been thinking about these ideas for years but still have not clarified them to my satisfaction....
bfee08b8-e2a3-4088-80a3-6641a10be25e
trentmkelly/LessWrong-43k
LessWrong
Orienting to 3 year AGI timelines My median expectation is that AGI[1] will be created 3 years from now. This has implications on how to behave, and I will share some useful thoughts I and others have had on how to orient to short timelines. I’ve led multiple small workshops on orienting to short AGI timelines and co...
2f5f04dc-f159-4277-aea0-e1df882be61d
trentmkelly/LessWrong-43k
LessWrong
Science eats its young Let's start by talking about scientific literacy. I'm going to use a weak definition of scientific literacy, one that simply requires familiarity with the Baconian method of inquiry. I don't want to place an exact number on this issue, but I'd wager the vast majority of the population of "educa...
0a614306-a733-4261-a2ec-12d8997dde77
trentmkelly/LessWrong-43k
LessWrong
Gauging of interest: LW stock picking? EDIT: Based on criticism below, I am reconsidering how to proceed with this idea (or something in the neighbourhood). A topic that has been on my mind recently is where, in our complicated lives, there might be low-hanging fruit ready to be picked by a motivated rationalist. Act...
0c6eb876-ac82-4107-9f9e-a279dd9e0013
trentmkelly/LessWrong-43k
LessWrong
Relying on Future Creativity I used to write songs for a rock band. Sometimes I'd have a song written, thinking it was my best work ever, then when it came time to rehearse, we'd realize it wasn't going to work out. Maybe we didn't have the instruments to do it justice (we were a three piece), or it was out of my comf...
dd2a9868-9a59-4da8-a222-3d12ac19cb71
trentmkelly/LessWrong-43k
LessWrong
Ngo and Yudkowsky on AI capability gains This is the second post in a series of transcribed conversations about AGI forecasting and alignment. See the first post for prefaces and more information about the format. Color key:   Chat by Richard Ngo and Eliezer Yudkowsky     Other chat      Inline comments      5. ...
d1e9570a-c9d3-47a3-94e3-36c4017116ac
trentmkelly/LessWrong-43k
LessWrong
How can I reduce existential risk from AI? Suppose you think that reducing the risk of human extinction is the highest-value thing you can do. Or maybe you want to reduce "x-risk" because you're already a comfortable First-Worlder like me and so you might as well do something epic and cool, or because you like the com...
3c0c5fdd-d1f6-46a9-b5df-ff37abc1d011
trentmkelly/LessWrong-43k
LessWrong
Is this a better way to do matchmaking? When friends ask me to matchmake them, they usually list traits they want: "Kind, curious, growth-minded..." (Sometimes they even write date-me docs!) But I find it really hard to think of the right people from these descriptions. Here's what I'm exploring, and suspect works be...
f4c18e50-70f1-48a6-8069-aa4c597c2126
StampyAI/alignment-research-dataset/lesswrong
LessWrong
Is the Star Trek Federation really incapable of building AI? In the Star Trek universe, we are told that it’s really hard to make genuine artificial intelligence, and that Data is so special because he’s a rare example of someone having managed to create one. But this doesn’t seem to be the best hypothesis for explai...
c151bd5c-da8c-40fc-83d4-6b041ddba1e3
trentmkelly/LessWrong-43k
LessWrong
Some Thoughts on Singularity Strategies Followup to: Outline of possible Singularity scenarios (that are not completely disastrous) Given that the Singularity and being strategic are popular topics around here, it's surprising there hasn't been more discussion on how to answer the question "In what direction should w...
f0d1ebc5-4889-4733-93c9-2c8eee255c4e
StampyAI/alignment-research-dataset/arbital
Arbital
Immediate goods One of the potential views on 'value' in the value alignment problem is that what we should want from an AI is a list of immediate goods or outcome features like 'a cure for cancer' or 'letting humans make their own decisions' or 'preventing the world from being wiped out by a paperclip maximizer'. (...
552fd40b-57b0-4a71-86cc-4239a2a55c86
trentmkelly/LessWrong-43k
LessWrong
Meetup : Moscow: Words of estimative probability, transparency illusion, belief investigation, rational games Discussion article for the meetup : Moscow: Words of estimative probability, transparency illusion, belief investigation, rational games WHEN: 26 June 2016 02:00:00PM (+0300) WHERE: Москва, ул. Большая Дорог...
905c3560-ee9b-4e31-9e81-eafceab5f204
trentmkelly/LessWrong-43k
LessWrong
Anthropic Paradoxes and Self Reference In this post I will explain how anthropic paradoxes are connected with self-reference.    Sleeping Beauty Problem The contention is how to treat the fact that "I am awake now/today". To briefly summarize the debate at the cost of oversimplification: SIA suggests treating toda...
216bab80-921a-469f-9dfb-1c5e490b2036
trentmkelly/LessWrong-43k
LessWrong
Why I hang out at LessWrong and why you should check-in there every now and then As the title indicates, this post, which in cross-posting from my home blog, New Savanna, is not directed at a LessWrong audience. However,, some of you might find in useful in thinking about LessWrong's place in the world. The first tw...
389a2729-9e99-49cd-817a-daf8f4da90b3
trentmkelly/LessWrong-43k
LessWrong
[LINK] How Do Top Students Study? I found this Quora discussion very informative. > 2. Develop the ability to become an active reader. Don't just passively read material you are given. But pose questions, develop hypotheses and actively test them as you read through the material. I think this is what another poster r...
fd753063-e56d-4827-b873-edcee3f50ad8
trentmkelly/LessWrong-43k
LessWrong
Lesswrong Meetup in Barcelona? Perhaps the Spaniard and/or Catalan tropers, and those others who simply feel like visiting this wonderful city that is Barcelona, could meet here during Semana Santa, the Spanish Spring Break, which lasts from April 15 to April 24. We could decide on a date within that interval. I'm not...
f339f611-ec6d-4e5a-8262-edfefc17cdcf
trentmkelly/LessWrong-43k
LessWrong
Hypothesis: gradient descent prefers general circuits Summary: I discuss a potential mechanistic explanation for why SGD might prefer general circuits for generating model outputs. I use this preference to explain how models can learn to generalize even after overfitting to near zero training error (i.e., grokking). I...
afb1a12d-abae-41f0-8089-1a3093ab37a6
trentmkelly/LessWrong-43k
LessWrong
[SEQ RERUN] Doublethink (Choosing to be Biased) Today's post, Doublethink (Choosing to be Biased) was originally published on 14 September 2007. A summary :   > George Orwell wrote about what he called "doublethink", where a person was able to hold two contradictory thoughts in their mind simultaneously. While some ...
eff868d1-8329-4ce0-8e5c-d9b3910f5336
trentmkelly/LessWrong-43k
LessWrong
Make an Extraordinary Effort > "It is essential for a man to strive with all his heart, and to understand that it is difficult even to reach the average if he does not have the intention of surpassing others in whatever he does." >         —Budo Shoshinshu > > "In important matters, a 'strong' effort usually results ...
495d4a11-a50b-4b97-b336-212a8e934982
trentmkelly/LessWrong-43k
LessWrong
MIRI's 2017 Fundraiser Update 2017-12-27: We've blown past our 3rd and final target, and reached the matching cap of $300,000 for the $2 million Matching Challenge! Thanks so much to everyone who supported us! All donations made before 23:59 PST on Dec 31st will continue to be counted towards our fundraiser total. Th...
277fc8d4-c324-4b53-a9a3-612175ec9282
trentmkelly/LessWrong-43k
LessWrong
Paper: In-context Reinforcement Learning with Algorithm Distillation [Deepmind] Authors train transformers to imitate the trajectory of reinforcement learning (RL) algorithms. Find that the transformers learn to do in-context RL (that is, the transformers implement an RL algorithm)---the authors check this by having t...
aa170321-19b4-4876-998b-f744d6f41f02
trentmkelly/LessWrong-43k
LessWrong
Why are the websites of major companies so bad at core functionality? Intuitively, you might think that Amazon cares about people being able to enter their banking details as easily as possible. In reality at least Amazon.de doesn't seem to care.  Instead of instantly validating with Javascript that a entered IBAN is ...
9c395b47-fc8a-40cf-b7d0-d6a5859eef65
trentmkelly/LessWrong-43k
LessWrong
Might humans not be the most intelligent animals? The idea that humans are the most intelligent animals on Earth appears patently obvious to a lot of people. And to a large extent, I agree with this intuition. Humans clearly dominate the world in technological innovation, control, communication, and coordination. How...
e242219e-71e0-4b10-b52b-1b1ea091a5f8
trentmkelly/LessWrong-43k
LessWrong
[Productivity] How not to use "Important // Not Urgent" Epistemic status: Personal experience. This is an Eisenhower matrix, named after President Dwight Eisenhower. Eisenhower matrices have a very useful, very specific role in my life these days. I usually bust one out on paper when I'm feeling overwhelmed by my o...
dc9769b3-e837-4fe3-8e07-c97a95f2562e
trentmkelly/LessWrong-43k
LessWrong
I think Michael Bailey's dismissal of my autogynephilia questions for Scott Alexander and Aella makes very little sense I am autogynephilic, and there's been a lot of autogynephilia talk lately. One subject that sometimes comes up, but hasn't been discussed much on LessWrong, is how common autogynephilia is in cis wom...
c87bfa9f-8304-4d0e-921e-1057f59c4138
trentmkelly/LessWrong-43k
LessWrong
LLM Basics: Embedding Spaces - Transformer Token Vectors Are Not Points in Space This post is written as an explanation of a misconception I had with transformer embedding when I was getting started. Thanks to Stephen Fowler for the discussion last August that made me realise the misconception, and others for helping ...
31ade82f-570e-4d98-9450-e40efdaab1d8
trentmkelly/LessWrong-43k
LessWrong
[LINK] The Mathematics of Gamification - Application of Bayes Rule to Voting Fresh from slashdot: A smart application of Bayes' rule to web-voting. http://engineering.foursquare.com/2014/01/03/the-mathematics-of-gamification/ > [The results] are exactly the equations for voting you would expect. But now, they’re der...
c89f71b6-a023-4ad4-98a6-f6b33ca4fe2f
trentmkelly/LessWrong-43k
LessWrong
How accurate was my "Altered Traits" book review? 4.4 years ago, I posted a review of Altered Traits, a book about the science of meditation. At the time, I was a noob. I hadn't hit any important checkpoints yet. Since then, I have sat quietly. In this post, I will review whether the claims in my original post are con...
d838a074-cebe-44cd-8531-2da1b605795c
trentmkelly/LessWrong-43k
LessWrong
Meetup : Madison: Rough Numbers Discussion article for the meetup : Madison: Rough Numbers WHEN: 22 July 2012 07:00:00PM (-0500) WHERE: 302 S Mills St, Apt 5, Madison, WI Note: If it's convenient, and you have a smartphone, laptop, or tablet, bring it. We'll want to be able to all look up data efficiently, and I wa...
6a1b8bfd-d68b-4cc0-8fe8-a6f03a036355
trentmkelly/LessWrong-43k
LessWrong
[SEQ RERUN] Sorting Pebbles Into Correct Heaps Today's post, Sorting Pebbles Into Correct Heaps was originally published on 10 August 2008. A summary (taken from the LW wiki):   > A parable about an imaginary society that don't understand what their values actually are. Discuss the post here (rather than in the co...
5c748bbb-422a-4f13-a32d-b5f669969c69
StampyAI/alignment-research-dataset/special_docs
Other
CHAI Newsletter 2018 5/4/22, 2:39 PM Center for Human-Compatible AI (CHAI) Newsletter https://mailchi.mp/c8b632a14601/center -for-human-compatible-ai-chai-newsletter 1/4Congratulations to our 2017 & 2018 Intern Cohorts "Interning at CHAI has been one of those rare experiences that makes you question beliefs you didn't...
fe289120-05bd-44ba-8687-97ec02693fd1
trentmkelly/LessWrong-43k
LessWrong
Timer Toxicities Follow-up to: Free-to-Play Games: Three Key Trade-Offs The central free-to-play mechanic is to ration action and resources via real world time. This leads to two of the three key trade offs. Players are prevented from having fun because they are time restricted, either unable to play or unable to hav...
90461579-f5e4-471b-a85e-4c73c50fafae
trentmkelly/LessWrong-43k
LessWrong
Source Control for Prototyping and Analysis When I'm doing exploratory work I want to run many analyses. I'm usually optimizing for getting something quick, but I want to document what I'm doing enough that if there are questions about my analysis or I later want to draw on it I can reconstruct what I did. I've taken ...
dc075fdc-2584-433b-8cd5-b9648484772a
StampyAI/alignment-research-dataset/arxiv
Arxiv
Generative Teaching Networks: Accelerating Neural Architecture Search by Learning to Generate Synthetic Training Data 1 Introduction and Related Work -------------------------------- Access to vast training data is now common in machine learning. However, to effectively train neural networks (NNs) does not require ...
812375c2-a1af-4528-a45f-923ad471262e
trentmkelly/LessWrong-43k
LessWrong
Shame as low verbal justification alarm What do you think feeling shame means? You are scared that you might be ostracized from the group But I feel shame about things that other people don’t even care about. You are scared that you should be ostracized from the group. That seems unrelated to the social realities....
7b6b77c3-e260-40c6-9281-d9ef76bd85f1
StampyAI/alignment-research-dataset/lesswrong
LessWrong
How well can the GPT architecture solve the parity task? Suppose I give it pairs of strings and ask it to output 1 if the the number of 1s in the string is even and zero if it's odd. e. g. 0 -> 0 1 -> 1 11 -> 0 101 -> 0 1101-> 1 10101001 -> 0 111000101110 -> 1 How well does it do on this task? What i...
fce5cb06-1176-4e2b-a9a4-4427be44e079
trentmkelly/LessWrong-43k
LessWrong
Book Review: The Elephant in the Brain We don’t only constantly deceive others. In order to better deceive others, we also deceive ourselves. You’d pay to know what you really think. Robin Hanson has worked tirelessly to fill this unmet need. Together with Kevin Simler, he now brings us The Elephant in the Brain. I ...
60b64f70-357c-4552-8599-1a88f7196dfc
trentmkelly/LessWrong-43k
LessWrong
Help Request: Cryonics Policies I’m hoping to sign up for cryonics when I can afford it, and I’m not sure which agency and treatment plan to get. As of this Cryonics institute document, whole-body suspension with Alcor costs $200,000. Neurosuspension costs $80,000. With the cheaper but possibly lower quality Cryonics...
e124528b-6b72-45d8-9846-b323d7f307bf
StampyAI/alignment-research-dataset/lesswrong
LessWrong
A concrete bet offer to those with short AGI timelines [Update 3 (3/16/2023): Matthew has now [conceded](https://www.lesswrong.com/posts/nto7K5W2sNR3Cpmec/conceding-a-short-timelines-bet-early).] [Update 2 (11/4/2022): Matthew Barnett now thinks he will probably lose this bet. You can read a post about how he's updat...
79bbc1f5-0a5b-45be-8231-007fc34c8555
trentmkelly/LessWrong-43k
LessWrong
Five Whys This post appeared first on the EA Coaching blog. Five Whys is a technique I borrowed from Lean methodology for getting to the root cause of a problem. As shown in the example below, I use the method to identify many possible solutions to a particular productivity problem.  The simple steps: 1. Ask “Why ...
40b9e01f-b191-46ba-835d-e38dc78535ab
trentmkelly/LessWrong-43k
LessWrong
Is alignment possible? Is AGI alignment even possible in the long term? Will AGI simply outsmart our best defenses? It would be, after all, superhuman (and by an enormous margin). Isn’t it likely that an AGI will recognize what actions humans took to control it and simply undo those controls? Or just create a novel mo...
47cdbbc2-2326-4a30-be78-59a2861f4dc9
trentmkelly/LessWrong-43k
LessWrong
Transcript: "You Should Read HPMOR" The following is the script of a talk I gave for some current computer science students at my alma mater, Grinnell College. This talk answers "What do I wish I had known while at Grinnell?". Hi, I'm Alex Turner. I’m honored to be here under Sam’s invitation. I'm in the class of 201...
5172129f-366e-4c9f-9875-bf1a61b9e7de
trentmkelly/LessWrong-43k
LessWrong
[Link] Nerds are nuts Related to: Reason as memetic immune disorder, Commentary on compartmentalization On the old old gnxp site site Razib Khan wrote an interesting piece on a failure mode of nerds. This is I think something very important to keep in mind because for better or worse LessWrong is nerdspace. It deals ...
42dcc522-c260-4c25-9afb-3f89bdd00829
trentmkelly/LessWrong-43k
LessWrong
The Evil AI Overlord List Top ten tips for an AI bent on destroying humanity: 1. Humanoid robots? Really? 2. Humans have proven very adept at controlling and killing other humans - surveillance drones, bio-warfare, nuclear weapons, unmanned drones, bombs, mines, chemicals. Try and be at least as competent as that...
d6d0aca9-1311-4b30-99a9-50e1df59b73f
trentmkelly/LessWrong-43k
LessWrong
Meetup : Melbourne, practical rationality Discussion article for the meetup : Melbourne, practical rationality WHEN: 01 February 2013 07:00:00PM (+1100) WHERE: 55 Walsh Street, West Melbourne VIC 3003, Australia Practical rationality. This meetup repeats on the 1st Friday of each month and is distinct from our soci...
691cba8a-b082-40e0-bd6e-331e10686bcb
trentmkelly/LessWrong-43k
LessWrong
[SEQ RERUN] Invisible Frameworks Today's post, Invisible Frameworks was originally published on 22 August 2008. A summary (taken from the LW wiki):   > A particular system of values is analyzed, and is used to demonstrate the idea that anytime you consider changing your morals, you do so using your own current meta-...
9df3f605-75ee-41d2-9bad-3eebe6ac0145
StampyAI/alignment-research-dataset/eaforum
Effective Altruism Forum
Apply for the ML Winter Camp in Cambridge, UK [2-10 Jan] *TL;DR: We are running a UK-based ML upskilling camp from 2-10 January in Cambridge for people with no prior experience in ML who want to work on technical AI safety.*[*Apply here*](https://forms.gle/aDa7JjCb6x8PWEy79) *by 11 December.* We (Nathan Barnard, J...
9d3c72d7-175c-4dfa-b0b3-2869d3d42357
StampyAI/alignment-research-dataset/arxiv
Arxiv
On the Expressivity of Markov Reward. 1 Introduction --------------- How are we to use algorithms for reinforcement learning (RL) to solve problems of relevance in the world? Reward plays a significant role as a general purpose signal: For any desired behavior, task, or other characteristic of agency, there must ex...
95483e22-2042-4a1c-a39e-5ac08e80292b
trentmkelly/LessWrong-43k
LessWrong
Help me understand the rationale of the NIH to not recommend Fluvoxamine for COVID treatment. So the TOGETHER trial signal boosted by Scott of slate star codex found Fluvoxamine to be effective at reducing 30% of COVID hospitalisation and fatality. The NIH looked at the study and found it unconvicing, I am a bit conf...
5528d267-090f-42dd-9849-eba41639c492
trentmkelly/LessWrong-43k
LessWrong
Meetup : Less Wrong Montreal - Easy Lifehacks Discussion article for the meetup : Less Wrong Montreal - Easy Lifehacks WHEN: 28 April 2014 07:00:00PM (-0400) WHERE: 3459 Mctavish, Montréal, QC There are lots of little things you can do to gain massive improvements in your life, and some of these tricks aren't widel...
8b134480-c167-4b59-8614-43ef9857c9d7
trentmkelly/LessWrong-43k
LessWrong
2011 Buhl Lecture, Scott Aaronson on Quantum Complexity I was planning to post this in the main area, but my thoughts are significantly less well-formed than I thought they were. Anyway, I hope that interested parties find it nonetheless. In the Carnegie Mellon 2011 Buhl Lecture, Scott Aaronson gives a remarkably cle...
0d4d79de-0c86-4d0e-b4c7-12adc2af531f
trentmkelly/LessWrong-43k
LessWrong
Timaeus is hiring! TLDR: We’re hiring two research assistants to work on advancing developmental interpretability and other applications of singular learning theory to alignment.  About Us Timaeus’s mission is to empower humanity by making breakthrough scientific progress on alignment. Our research focuses on applic...
03c15bb1-3e53-44ff-88e1-fb954a0668c5
trentmkelly/LessWrong-43k
LessWrong
Retrieval Augmented Genesis a prototype and some thoughts on semantics Before reading this article I strongly encourage you to Checkout the RAGenesis App! Full code available at https://github.com/JoaoRibeiroMedeiros/RAGenesisOSS.  Screenshot of Verse Uni Verse page Project Goals * Send a message of unity and mut...
626c48cd-32a1-44c9-a577-19fe8d82ad97
trentmkelly/LessWrong-43k
LessWrong
Do Corporations Have a Right to Privacy? The link to Bruce Schneier's original post. > This week, the U.S. Supreme Court will hear arguments about whether or not corporations have the same rights to "personal privacy" that individuals do. The Electronic Privacy Information Center (EPIC) has filed a amicus curiae bri...