id
stringlengths
36
36
source
stringclasses
15 values
formatted_source
stringclasses
13 values
text
stringlengths
2
7.55M
2dabc5a9-fbbe-40de-bf93-c985a3836dab
LDJnr/LessWrong-Amplify-Instruct
LessWrong
"UPDATE: Offer of mentorship is closed, since I received sufficiently many candidates for now. Offer of collaboration remains open for experienced researchers (i.e. researchers that (i) have some track record of original math / theoretical compsci research, and (ii) are able to take on concrete open problems without mu...
568cb06b-8bfa-458a-8b91-4401bc2d04fb
trentmkelly/LessWrong-43k
LessWrong
AISN #27: Defensive Accelerationism, A Retrospective On The OpenAI Board Saga, And A New AI Bill From Senators Thune And Klobuchar Welcome to the AI Safety Newsletter by the Center for AI Safety. We discuss developments in AI and AI safety. No technical background required. Subscribe here to receive future versions. ...
233cfc26-68f4-4c51-b767-d9e7efd405b8
trentmkelly/LessWrong-43k
LessWrong
Guilt: Another Gift Nobody Wants Evolutionary psychology has made impressive progress in understanding the origins of morality. Along with the many posts about these origins on Less Wrong I recommend Robert Wright's The Moral Animal for an excellent introduction to the subject. Guilt does not naturally fall out of th...
d21b5451-54d4-4608-bcf2-9e11a9083181
trentmkelly/LessWrong-43k
LessWrong
Meetup : Atlanta: Rationalist Movie Night! Discussion article for the meetup : Atlanta: Rationalist Movie Night! WHEN: 26 October 2013 07:00:00PM (-0400) WHERE: 1314 Hosea L Williams Dr. NE, Atlanta Woo! Let's do rationalist cinema, and have popcorn, then discussion, games, or going out for coffee or whatever we l...
383dd4a4-7754-4481-b9a7-88286888e11f
trentmkelly/LessWrong-43k
LessWrong
Recommended reading on the ethics of Animal Cognitive Enhancement Does anyone have recommendations of good things to read on the ethics of animal cognitive enhancement? By this I mean applying various methods of human cognitive enhancement (pharmacological, technological, etc) to animals such as the Great Apes. I've a...
338496d9-808a-4475-b5ae-52ece1a58297
trentmkelly/LessWrong-43k
LessWrong
Hedonic vs Preference Utilitarianism in the Context of Wireheading We're probably going to develop the technology to directly produce pleasure in the brain with electrical stimulation. We already do this to some extent, though with the goal of restoring people to normal function. This poses a similar question to drugs...
856ee57a-6932-426b-a6ac-fd70456e0f00
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
How I think about alignment *This was written as part of the first*[*Refine*](https://www.lesswrong.com/posts/5uiQkyKdejX3aEHLM/how-to-diversify-conceptual-alignment-the-model-behind) *blog post day. Thanks for comments by Chin Ze Shen, Tamsin Leake, Paul Bricman, Adam Shimi.* Magic agentic fluid/force ==============...
0763f60d-78d2-4be0-8f7d-1b268e37e4cd
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
The Artificial Intentional Stance Another post in the same incremental vein. Still hoarding the wild speculation for later. **I** The idea of the "intentional stance" comes from [Dan Dennett](https://www.cs.tufts.edu/comp/150AAA/DennettTrueBelievers.pdf), who wanted to explain why it makes sense that we should thin...
2878fc97-163d-4a7a-aa70-5d260d190360
trentmkelly/LessWrong-43k
LessWrong
Change And Identity: a Story and Discussion on the Evolving Self This post is the first in a sequence, which also appears on my Substack page, where I'm working to clarify my views on personal identity and consciousness. It begins with a short piece of fiction, 'Change,' which aims to illustrate the ways our identiti...
becf1233-c9f3-478d-897c-dc3707e2d8b2
trentmkelly/LessWrong-43k
LessWrong
Book: Psychiatry and the Human Condition I'm about half-way through this fascinating book, conveniently available for free online, which is at the intersection of psychiatry and evolutionary psychology.  I don't have the time to do it justice, so I'm going to post a few choice excerpts here in the hope that those who ...
da144cc7-1d5e-410a-b360-a8be1b493e21
StampyAI/alignment-research-dataset/lesswrong
LessWrong
Memory bandwidth constraints imply economies of scale in AI inference Contemporary GPUs often have very imbalanced memory vs arithmetic operation capabilities. For instance, an [H100](https://www.nvidia.com/en-us/data-center/h100/) can do around 3e15 8-bit FLOP/s, but the speed at which information can move between th...
53b7639e-a454-47c7-ac41-7d7256b10063
trentmkelly/LessWrong-43k
LessWrong
A big Singularity-themed Hollywood movie out in April offers many opportunities to talk about AI risk There's a big Hollywood movie coming out with an apocalyptic Singularity-like story, called Transcendence. (IMDB, Wiki, official site) With an A-list cast and big budget, I contend this movie is the front-runner to be...
2e3a8c63-14be-4d5a-818c-9929b95a04fb
StampyAI/alignment-research-dataset/eaforum
Effective Altruism Forum
40,000 reasons to worry about AI safety There is an argument that AI will have a Sputnik moment, an event that triggers a calamitous race towards dangerous AGI. This Sputnik moment may have already happened. OpenAI's decision to develop and release various forms of GPT has triggered [Google to take on more risk](htt...
cea87ba4-74cd-457c-820b-57e728adbc37
trentmkelly/LessWrong-43k
LessWrong
The Office of Science and Technology Policy put out a request for information on A.I. This request for information does cover some of the possible existential risks, in the first section.  I am going to submit a few responses of my own, and I am hopeful others will do the same.   1. What specific measures – such as s...
8350914d-5b40-408e-9ac5-846c162f6f10
trentmkelly/LessWrong-43k
LessWrong
Meetup : Vancouver Minicamp Report-Back Discussion article for the meetup : Vancouver Minicamp Report-Back WHEN: 18 May 2012 05:00:00PM (-0700) WHERE: Vancouver Minicamp was totally awesome and I learned so much. I think I can put it best like this: I no longer have "a sense that more is possible" it is now "More i...
0cbea593-3d36-4cbc-9b2b-67762fc12464
trentmkelly/LessWrong-43k
LessWrong
Meetup : Yale: Initial Meetup Discussion article for the meetup : Yale: Initial Meetup WHEN: 16 February 2014 02:00:00PM (-0500) WHERE: Bass Library Cafe, 110 Wall St, New Haven, CT 06511 Hi. If anyone who goes to Yale is interested in meeting up, I'll be in Bass Cafe on Sunday, February 16, from 2 to 4 pm. I'll br...
9a5ca6b5-6087-485b-919d-5c45879436e7
trentmkelly/LessWrong-43k
LessWrong
Belief in Self-Deception I spoke yesterday of my conversation with a nominally Orthodox Jewish woman who vigorously defended the assertion that she believed in God, while seeming not to actually believe in God at all. While I was questioning her about the benefits that she thought came from believing in God, I introd...
40325f0f-df93-4b7e-8f21-aaaf71bba556
trentmkelly/LessWrong-43k
LessWrong
CFAR: Progress Report & Future Plans Context: This is the first in a series of year-end updates from CFAR. This post mainly describes what CFAR did over the past two years, and what it plans to do next year; other posts in the series will describe more about its mission, financials, and impact. (12/29/19: Edited to m...
b6a5225d-1fec-40a7-88a2-e9a734dd88ec
trentmkelly/LessWrong-43k
LessWrong
Welcome to LessWrong (January 2016) If you've recently joined the Less Wrong community, please leave a comment here and introduce yourself. We'd love to know who you are, what you're doing, what you value, how you came to identify as a rationalist or how you found us. You can skip right to that if you like; the rest o...
6ca14f58-8950-447f-b47b-b309a6478cda
trentmkelly/LessWrong-43k
LessWrong
Meetup : San Francisco Meetup: Fermi Estimates Discussion article for the meetup : San Francisco Meetup: Fermi Estimates WHEN: 27 April 2015 06:15:00PM (-0700) WHERE: 1061 Market St #4, San Francisco, CA 94103 We'll be meeting to practice Fermi estimates, a technique for quick order of magnitude estimates. If you ...
11cee5cb-b337-41b8-9d82-f7efbcf065e4
StampyAI/alignment-research-dataset/arxiv
Arxiv
Actionable Guidance for High-Consequence AI Risk Management: Towards Standards Addressing AI Catastrophic Risks. Actionable Guidance for High-Consequence AI Risk Management: Towards Standards Addressing AI Catastrophic Risks Anthony M. Barrett † *, Dan Hendrycks † , Jessica Newman † , Brandie Nonnecke † † UC Berkeley ...
909797b9-aaa1-4777-ad71-784b7995431e
StampyAI/alignment-research-dataset/eaforum
Effective Altruism Forum
“AI Risk Discussions” website: Exploring interviews from 97 AI Researchers *Tldr:* Check out the new [AI Risk Discussions website](https://ai-risk-discussions.org)! It contains 83 transcripts of AI researcher [interviews](https://ai-risk-discussions.org/interviews), and a [quantitative analysis](https://forum.effectiv...
b74707de-389b-462c-8292-50528b52f8c9
trentmkelly/LessWrong-43k
LessWrong
Harry Potter and the Methods of Rationality discussion thread, part 6 Update: Discussion has moved on to a new thread. After 61 chapters of Harry Potter and the Methods of Rationality and 5 discussion threads with over 500 comments each, HPMOR discussion has graduated from the main page and moved into the Less Wrong ...
c1528863-f908-4e52-b234-39a0d5f95773
trentmkelly/LessWrong-43k
LessWrong
Prepare for COVID-19 Human Challenge Trials — A Petition in Canada (and soon, the UK) Canadians— sign the petition here. TL;DR COVID-19 human challenge trials could save tens of thousands of lives by quickly narrowing the field of promising candidates, and there is strong reason to believe that signaling clear publi...
d9afe52b-8b80-4403-a083-1e34d3c27397
trentmkelly/LessWrong-43k
LessWrong
GPT-4 for personal productivity: online distraction blocker There are many apps for blocking distracting websites: freedom.to, leechblock, selfcontrol, coldturkey, just to name a few. They are useful for maintaining focus, avoiding procrastination, and curbing addictive web surfing. They work well for blocking a list...
d58538b4-8fc0-46d6-8f61-b9957991aa9c
StampyAI/alignment-research-dataset/arxiv
Arxiv
Multi-agent Hierarchical Reinforcement Learning with Dynamic Termination 1 Introduction --------------- Many important real-world tasks are multi-agent by nature, such as taxi coordination [[10](#bib.bib20 "Efficient large-scale fleet management via multi-agent deep reinforcement learning")], supply chain managemen...
8b2bfe0f-4844-41c4-8294-5f5c93e085f2
LDJnr/LessWrong-Amplify-Instruct
LessWrong
"Epistemic Status: Common knowledge, just not to me The Evolution of Trust is a deceptively friendly little interactive game. Near the end, there’s a “sandbox” evolutionary game theory simulator. It’s pretty flexible. You can do quick experiments in it without writing code. I highly recommend playing around. One of the...
6b863b8a-8dd1-4800-aeaa-57f860959107
trentmkelly/LessWrong-43k
LessWrong
"Zero Sum" is a misnomer. This could have been a relatively short note about why "zero sum" is a misnomer, but I decided to elaborate some consequences. This post benefited from discussion with Sam Eisenstat. "Zero Sum" is a misnomer. The term intuitively suggests that an interaction is transferring resources from o...
1ecd79e6-757a-47f3-b7e2-a5e5b7b4d1ba
StampyAI/alignment-research-dataset/distill
Distill Scientific Journal
Understanding RL Vision ### Contents [Introduction](#introduction) [Our CoinRun model](#coinrun) [Model analysis](#analysis) * [Dissecting failure](#dissecting-failure) * [Hallucinations](#hallucinations) * [Model editing](#model-editing) [The diversity hypothesis](#diversity-hypothesis) [Feature visualization](#f...
38b20946-920c-4ec5-8117-ed9b7a8f9b9a
trentmkelly/LessWrong-43k
LessWrong
For what do we need Superintelligent AI? Most of human problems could be solved by humans or slightly above human AIs. Which practically useful tasks – which we could define now – require very high level of superintelligence? I see two main such tasks: 1) Prevention of the creation of other potentially dangerous supe...
a964b93a-df63-4be3-ac5b-7e875c08460a
trentmkelly/LessWrong-43k
LessWrong
On the etiology of religious belief From "Trust in testimony and miracles": > I have been of late fascinated by the research of the developmental psychologist Paul L. Harris, who has investigated how young children acquire information through testimony. Harris gauges two psychological hypotheses. The first, which he ...
b6e64f52-2e56-4e82-9f6c-828da774a801
trentmkelly/LessWrong-43k
LessWrong
CDC Changes Isolation Guidelines The CDC has changed its isolation guidelines. Previously, if you tested positive for Covid, you needed to isolate for ten days. Now you only need to isolate for five. The official announcement is here. Read-through first, with my reactions, then reactions from others. > Given what w...
895ccd66-891d-4555-9fda-12de881b8f75
trentmkelly/LessWrong-43k
LessWrong
Is Legal Reasoning Fundamentally Irrational? And, if so, is this a bad thing? This is a short essay on modern law; specifically, how it is possibly irrational, and how this irrationality could in fact be necessary for large legal systems to operate. I conclude with some observations on potential wider significance o...
d6cb5408-dae1-4a05-8378-1e78e979ca43
trentmkelly/LessWrong-43k
LessWrong
Chess as a case study in hidden capabilities in ChatGPT (Edit: This post is out of date, and understates ChatGPT's chess abilities. gpt-3.5-turbo-instruct, when prompted correctly, plays consistently legal moves at around the 1800-2000 Elo level. Thanks to commenter GoteNoSente for pointing this out and to ClevCode Lt...
d2971989-bdaf-4dd2-a9d4-aa554f2a5267
trentmkelly/LessWrong-43k
LessWrong
Expected utility of control vs expected utility of actions
d1a8d8a4-b2c3-41bd-a204-69a82433bc97
StampyAI/alignment-research-dataset/lesswrong
LessWrong
Activation adding experiments with llama-7b *Produced as part of the* [*SERI ML Alignment Theory Scholars Program*](https://serimats.org/) *- Summer 2023 Cohort* Following my initial investigation with [activation adding experiments with FLAN-T5](https://www.lesswrong.com/posts/c38nAg23YTCzd7m8P/activation-adding-exp...
7d0f8594-6d93-40e7-b6a1-60837de9bdfa
StampyAI/alignment-research-dataset/lesswrong
LessWrong
On the possibility of impossibility of AGI Long-Term Safety **Preface** ----------- The following article is a layman’s summary of researcher and philosopher Forrest Landry’s work titled [No People as Pets: A Dialogue on the Complete Failure of Exogenous AGI/APS Alignment](https://mflb.com/ai_alignment_1/no_people_as...
3110da6c-eb7d-48e7-a461-7e544c16a83c
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
Committing, Assuming, Externalizing, and Internalizing This is the tenth post in the [Cartesian frames](https://www.lesswrong.com/posts/BSpdshJWGAW6TuNzZ/introduction-to-cartesian-frames) sequence. Here, we define a bunch of ways to construct new (additive/multiplicative) (sub/super)-(agents/environments) from a give...
361d8da7-bc4e-48e0-9fe3-8f3928c62a50
trentmkelly/LessWrong-43k
LessWrong
Babble challenge: 50 ways to stop torture Content warning: torture. Torture is deliberate infliction of severe pain or suffering on someone with the purpose of punishment, extracting a confession, or getting information out of them. Even though torture is illegal under international law, many governments continue to ...
d01023da-f1f9-4fc3-b1f8-020b7ba6de2a
trentmkelly/LessWrong-43k
LessWrong
Automating the Breath Pulse One of the pieces of my rhythm stage setup is a breath controller, which measures how much I blow. It's a good fit for anything where I want to have some flexible imprecise control, but one thing I've ended up using it for a lot is just a pulse: Here's it by itself: Here's it live, laye...
f5bf2f47-7824-41ae-9237-7750a79aa8d8
trentmkelly/LessWrong-43k
LessWrong
Book Review: Decisive by Chip and Dan Heath I first heard of the Heath brothers (Chip and Dan) years ago when I saw Chip speak about their first two books, Made to Stick and Switch. So I was pleased to discover that they’ve written a book specifically about decision-making called Decisive. I like Decisive primarily a...
75e2f8f9-c136-4289-bcd4-298d88108a13
trentmkelly/LessWrong-43k
LessWrong
Towards a Formalisation of Logical Counterfactuals Introduction This is an attempt to build a basis for formalising logical counterfactuals. It's not workable as is, but I think it can potentially solve the problem of "outthinking the counterfactual". For causal counterfactuals, we build a directed acyclic causal gr...
92e17196-7cb5-4d91-af37-70aa0b38cd2d
trentmkelly/LessWrong-43k
LessWrong
AI Alignment, Philosophical Pluralism, and the Relevance of Non-Western Philosophy This is an extended transcript of the talk I gave at EAGxAsiaPacific 2020. In the talk, I present a somewhat critical take on how AI alignment has grown as a field, and how, from my perspective, it deserves considerably more philosophic...
5b752f54-444c-4e5f-bd28-8d9825f791cb
trentmkelly/LessWrong-43k
LessWrong
When life was literally full of crap If you want a single, vivid, and frankly disgusting example to hold in mind to remember how much our lives have improved over the last ~150 years… Consider shit. Literally, excrement. How much previous generations had to think about it, be around it, even handle it. Before the a...
854359bc-b2c9-4f04-8a66-1ba56ee491b5
StampyAI/alignment-research-dataset/lesswrong
LessWrong
Coherent decisions imply consistent utilities (*Written for Arbital in 2017.*) --- Introduction to the introduction: Why expected utility? ======================================================= So we're talking about how to make good decisions, or the idea of 'bounded rationality', or what sufficiently advanced...
8b454528-bd85-45aa-a78b-e3fd02566c6d
StampyAI/alignment-research-dataset/special_docs
Other
Multiparty Dynamics and Failure Modes for Machine Learning and Artificial Intelligence Abstract -------- \*\*:\*\* An important challenge for safety in machine learning and artificial intelligence systems is a set of related failures involving specification gaming, reward hacking, fragility to distributional shifts, ...
1a8fb00f-777b-442e-abbe-7ea54b5188a2
trentmkelly/LessWrong-43k
LessWrong
An Introduction to Representation Engineering - an activation-based paradigm for controlling LLMs Representation Engineering (aka Activation Steering/Engineering) is a new paradigm for understanding and controlling the behaviour of LLMs. Instead of changing the prompt or weights of the LLM, it does this by directly in...
bd6a6a24-485e-4310-9d31-af8048f2c752
trentmkelly/LessWrong-43k
LessWrong
Group rationality diary, 11/13/12 This is the public group instrumental rationality diary for the week of October 29th. It's a place to record and chat about it if you have done, or are actively doing, things like: * Established a useful new habit * Obtained new evidence that made you change your mind about some be...
8f868a4d-c482-4feb-a3ea-28f5fea0b8e1
StampyAI/alignment-research-dataset/blogs
Blogs
Seeking beta readers I've found it really helpful to run early drafts of Cold Takes posts by people who haven't talked with me about the relevant topics, don't have much background on them, etc. I often learn about things that are "off" about the way I'm trying to present things, and I think it ends up improving the p...
c58d0195-da39-4101-bea4-c84b772bfa42
StampyAI/alignment-research-dataset/aisafety.info
AI Safety Info
Could we program an AI to automatically shut down if it starts doing things we don’t want it to? One thing that might make your AI system safer is to include an off switch. If it ever does anything we don’t like, we can turn it off. This implicitly assumes that we’ll be able to turn it off before things get bad, which...
ac81df47-6881-42f2-9eb6-05db83dc6234
trentmkelly/LessWrong-43k
LessWrong
Censorship in LLMs is here to stay because it mirrors how our own intelligence is structured Much has been made of how increasing levels of censorship (roughly by increasing the impact of RLHF) is nerfing LLMs down. Frustrated users can't see why OpenAI is watering down the very sherbert that hit them gold. Surely, th...
bac31136-2223-4e84-96b2-0392eb4863f2
trentmkelly/LessWrong-43k
LessWrong
Basics of Rationalist Discourse Introduction This post is meant to be a linkable resource. Its core is a short list of guidelines (you can link directly to the list) that are intended to be fairly straightforward and uncontroversial, for the purpose of nurturing and strengthening a culture of clear thinking, clear com...
7a0d579c-dd7d-496a-9272-0ccbc534d8eb
trentmkelly/LessWrong-43k
LessWrong
Placebo effect report: chiropractic adjustment Content warning: I accidentally got chiropractic treatment and felt better afterward, and I think talking too much about the placebo effect might be a mild cognitohazard, but less hazardous than not talking about it.   This is a story about being treated by a medical pr...
ad983387-614c-4503-b3de-bf1fcbf2c35a
trentmkelly/LessWrong-43k
LessWrong
Ethicality Behind Breaking the Glass Ceiling I am a woman who works in finance. When I originally studied finance I only went into it because I liked playing games with numbers and through some courses I had taken I recognized that there was a way to make money by solving the same math problems for other people. I wa...
adc2242d-86c1-48d3-a6e2-7a95c46632c0
trentmkelly/LessWrong-43k
LessWrong
[SEQ RERUN] Beautiful Math Today's post, Beautiful Math was originally published on 10 January 2008. A summary (taken from the LW wiki):   > The joy of mathematics is inventing mathematical objects, and then noticing that the mathematical objects that you just created have all sorts of wonderful properties that you ...
007e930b-b777-4153-a612-759557465faf
trentmkelly/LessWrong-43k
LessWrong
Doing medical research with (a lot) of personal money [link] http://www.bloomberg.com/news/2011-09-07/mogul-using-own-100-million-in-race-to-cure-daughter-prompts-novartis-aid.html I think this an interesting data point (though obviously a single one) of what money can do when directed with a lot of drive to a partic...
a9f96a4f-56f6-43d4-a81d-9f80839ed8e3
trentmkelly/LessWrong-43k
LessWrong
Probability theory implies Occam's razor Occam's razor follows from the rules of probability theory. It need not be assumed, or taken as a prior. Looking at what it means to spread probability mass across a larger space, in the absence of any information on how to distribute that probability, shows why. Spread-out d...
3a6cc559-95de-4217-9151-af4774725459
trentmkelly/LessWrong-43k
LessWrong
Should questions be called "questions" or "confusions" (or "other")? Or, what is the best way to think about "what is a question?" on LW? The LW Team just had a retreat where we thought through a lot of high level strategy. We have a lot of ideas building off of the "questions" feature. One thing that struck us is ...
52d5892a-3a55-4bed-9f18-0e99b527f981
StampyAI/alignment-research-dataset/eaforum
Effective Altruism Forum
A grand strategy to recruit AI capabilities researchers into AI safety research *Edit*: I was told privately that the grand strategy was unclear. The proposed grand strategy is to: 1. Convince everyone. Recruit all AI capabilities researchers to transition into AI safety until the alignment problem is solved. 2. Scal...
4e069d7a-be6a-4a58-acd3-64d70298cc2a
trentmkelly/LessWrong-43k
LessWrong
Five Ways To Prioritize Better This piece is cross-posted on my blog here. I’m going to let you in on a secret of productivity.   Those people you admire, the ones who make you wonder how on earth they accomplish so much? Those people might work more hours than you or be more talented or more passionate. Or they mig...
08d24d75-91f4-401f-aa33-e48aff78413b
trentmkelly/LessWrong-43k
LessWrong
Why it feels like everything is a trade-off Epistemic status: A cute insight that explains why it might feel like you always have to make sacrifices along one metric to get improvements along another. Seems tautological once you understand it. Might be obvious to everyone. Meta-epistemic status or something: My first...
47999c4e-c34f-4981-8c8a-220c0660dab6
trentmkelly/LessWrong-43k
LessWrong
Linear infra-Bayesian Bandits Linked is my MSc thesis, where I do regret analysis for an infra-Bayesian[1] generalization of stochastic linear bandits. The main significance that I see in this work is: * Expanding our understanding of infra-Bayesian regret bounds, and solidifying our confidence that infra-Bayesiani...
3c9bb7b6-9482-481c-893d-1375883c7b66
StampyAI/alignment-research-dataset/eaforum
Effective Altruism Forum
G7 Summit—Cooperation on AI Policy TL;DR—While the specifics available thus far are underwhelming, it seems notable and a net-positive that cooperation on AI policy is—for the first time—on the G7 Summit agenda this weekend.  The readout below is from an April meeting that was convened to lay the groundwork for AI pol...
c1d447fd-8bbf-440c-ba4f-0f18e83a18c8
trentmkelly/LessWrong-43k
LessWrong
How Should We Critique Research? A Decision Perspective None
d1642c1c-6d87-4553-88a0-cc35ed5d38a5
trentmkelly/LessWrong-43k
LessWrong
The Gears of Argmax In the last post, I made the high-level claim that optimization can sometimes be done safely. And went on to claim that a key factor enabling safe optimization is sufficiently understanding the optimization process, optimization metric, and domain of candidate solutions.  In this post, I'll dig in...
897cdf18-8ad7-4776-8491-a542698872ee
trentmkelly/LessWrong-43k
LessWrong
Modular empirical science None
25a8910d-03c8-4e62-9f50-d2c287ca0717
trentmkelly/LessWrong-43k
LessWrong
What Value Hermeneutics? I often take for granted that there is something to be said about how we know things. For example: * In high school I took a class titled Theory of Knowledge that was a combination intro to philosophy course and applied epistemology workshop. * My friends are concerned with questions about ...
d4ef887b-cfc0-4494-9869-fb4ef53db16b
trentmkelly/LessWrong-43k
LessWrong
You Can Now Embed Flashcard Quizzes in Your LessWrong posts! With the help of the LessWrong.com team, we've set up a way for you to embed flashcard quizzes directly in your LessWrong posts! This means that you can write flashcards for any of your LessWrong posts and either:  (a) quiz people as they read your article ...
bf1dde3b-fcb4-4542-b594-45d1a55ea912
StampyAI/alignment-research-dataset/lesswrong
LessWrong
Cruxes on US lead for some domestic AI regulation *Written quickly. Suggestions welcome.* A possible risk of some US AI regulation is that US regulation would differentially slow US AI progress and that would be bad. This post explores the factors that determine *how much US regulation would differentially slow US AI...
ea17519a-07b4-4950-8bb0-dc93cd1a3f89
StampyAI/alignment-research-dataset/lesswrong
LessWrong
Key Questions for Digital Minds *Last revised on June 14, 2023.* ChatGPT, Sydney, LaMDA, and other large language models (LLMs) have inspired an unprecedented wave of interest in artificial intelligence (AI). Many are asking if these LLMs are not merely problem-solving tools but have mental faculties such as [emotion...
4226df46-c5b5-441f-8c1e-6b34d9f93036
trentmkelly/LessWrong-43k
LessWrong
Dying in Many Worlds I feel extremely embarrassed about asking for help with this, but I have a philosophical quandary that has been eating at me for days.  I'm sure that many of you already have it figured out.  I would appreciate it if you would lend your cached thoughts to me, because I can't seem to resolve it. W...
83d939bb-c1c0-485f-b868-2f98091e6ebf
trentmkelly/LessWrong-43k
LessWrong
[Book Review] "Houdini on Magic" by Harry Houdini Houdini on Magic is a collection of the writings of Harry Houdini. It mostly covers three things. 1. Houdini's greatest tricks. (Mostly escapes.) 2. Houdini's work discrediting spiritualists. Spiritualists are people who pretend to talk to the dead. These aren't wel...
5b86978d-5e41-4872-8eb2-f518b11c9743
awestover/filtering-for-misalignment
Redwood Research: Alek's Filtering Results
id: post208 Say an LLM agent behaves innocuously in some context A, but in some sense “knows” that there is some related context B such that it would have behaved maliciously (inserted a backdoor in code, ignored a security bug, lied, etc.). For example, in the recent alignment faking paper Claude Opus chooses to say ...
4c96972e-8d72-48f6-b3d3-b2c8e9822a98
StampyAI/alignment-research-dataset/special_docs
Other
Where Do You Think You're Going?: Inferring Beliefs about Dynamics from Behavior Where Do YouThink You’re Going?: Inferring Beliefs about Dynamics from Behavior Siddharth Reddy, Anca D. Dragan, Sergey Levine Department of Electrical Engineering and Computer Science University of California, Berkeley {sgr,anca,svlevine...
62d987b0-1a19-4c84-a750-6aff6e41315f
trentmkelly/LessWrong-43k
LessWrong
AISN #16: White House Secures Voluntary Commitments from Leading AI Labs and Lessons from Oppenheimer Welcome to the AI Safety Newsletter by the Center for AI Safety. We discuss developments in AI and AI safety. No technical background required. Subscribe here to receive future versions. ----------------------------...
edbd60ce-ea54-4646-b32e-17dafbf9e42a
trentmkelly/LessWrong-43k
LessWrong
a space habitat design This post is mainly about a design concept for far-future large space habitats. some proposed designs As you can see on Wikipedia, many space habitat designs have been proposed. Below are some that I thought were worth mentioning. ---------------------------------------- current space stati...
4d6e0d4c-8587-4cdb-8c59-2b413a92cfb7
trentmkelly/LessWrong-43k
LessWrong
A Parable On Obsolete Ideologies Followup to:  Yudkowsky and Frank on Religious Experience, Yudkowksy and Frank On Religious Experience Pt 2 With sincere apologies to: Mike Godwin You are General Eisenhower. It is 1945. The Allies have just triumphantly liberated Berlin. As the remaining leaders of the old regime are...
3faf2007-ba45-4b17-8360-4063fc6f3e91
trentmkelly/LessWrong-43k
LessWrong
OpenAI releases functional Dota 5v5 bot, aims to beat world champions by August > Our team of five neural networks, OpenAI Five, has started to defeat amateur human teams at Dota 2. While today we play with restrictions, we aim to beat a team of top professionals at The International in August subject only to a limite...
9a04fa5a-50e1-4cde-a8dc-876365457166
StampyAI/alignment-research-dataset/lesswrong
LessWrong
on “learning to summarize” This post is a much extended version of an [LW comment](https://www.lesswrong.com/posts/mWJbYebezFdhoFHP6/learning-to-summarize-with-human-feedback-openai-1?commentId=DRPygZwXL7KjWE4z3) I made about OpenAI’s new paper, [“Learning to summarize from human feedback.”](https://arxiv.org/abs/2009...
371c74a8-6efe-4f07-aca6-b5ce5b693c3d
trentmkelly/LessWrong-43k
LessWrong
LLM Psychometrics and Prompt-Induced Psychopathy This post contains experimental results and personal takes from my participation in the July 2024 edition of the BlueDot Impact AI Safety Fundamentals course. TL;DR: * Psychopaths are willing to manipulate and deceive. Psychometrics try to measure this with standardi...
423cd975-4326-4cb2-a330-9c7010478350
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
Multiplicative Operations on Cartesian Frames This is the seventh post in the [Cartesian frames](https://www.lesswrong.com/posts/BSpdshJWGAW6TuNzZ/introduction-to-cartesian-frames) sequence. Here, we introduce three new binary operations on Cartesian frames, and discuss their properties.   1. Tensor --------- Our ...
ffa2a217-23f4-4b2e-a851-84f10b38c97b
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
Why Not Just Outsource Alignment Research To An AI? Warmup: The Expert ------------------ If you haven’t seen “The Expert” before, I recommend it as a warmup for this post: > The Client: “We need you to draw seven red lines, all strictly perpendicular. Some with green ink, some with transparent. Can you do that?” >...
cddb5751-da30-4707-8062-9b15f58cbe6a
trentmkelly/LessWrong-43k
LessWrong
[AN #122]: Arguing for AGI-driven existential risk from first principles Alignment Newsletter is a weekly publication with recent content relevant to AI alignment around the world. Find all Alignment Newsletter resources here. In particular, you can look through this spreadsheet of all summaries that have ever been in...
ced332e2-b498-4435-893b-9402118c42d1
trentmkelly/LessWrong-43k
LessWrong
Convincing my dad to sign up for Alcor: Advice? Since it came to my attention that signing up for cryonics is not as pointless as I'd once thought, I've been pondering how to sell my dad on the idea. This is somewhat urgent for a couple of reasons. First, he's already pushing sixty and would meet increased resistance...
03cbd005-ec1b-44da-bde4-7cf6a7c47e99
trentmkelly/LessWrong-43k
LessWrong
How should vegans think about Methionine needs? Non-animal based protein sources mostly have a different amino acid profile than animal-based protein sources.  Different plants also have a different composition.  From looking a bit at the data myself it seems that if you mix different plant protein sources, you can g...
26f608b5-7358-4128-b6ea-5eaaf2ac4f16
trentmkelly/LessWrong-43k
LessWrong
Meetup : Melbourne Social Southern Summer Solstice Celebration Discussion article for the meetup : Melbourne Social Southern Summer Solstice Celebration WHEN: 20 December 2013 07:00:00PM (+1100) WHERE: 54 Scenic Cres, Eltham North Welcome to the end of 2013, a time of feasts and of plenty, and a time to celebrate t...
f8614ca7-45d2-4e51-8903-dcc10d8872ba
trentmkelly/LessWrong-43k
LessWrong
Dominant characters on the left From Psyblog: > Research finds that people or objects moving from left to right are perceived as having greater power (Maass et al., 2007): > > * Soccer goals are rated as stronger, faster, even more beautiful when the movement of the scorer is from left to right, rather than right t...
f0814484-891f-4efa-a498-62aa2bbc92a2
trentmkelly/LessWrong-43k
LessWrong
Introduction To The Infra-Bayesianism Sequence TLDR: Infra-Bayesianism is a new approach to epistemology / decision theory / reinforcement learning theory, which builds on "imprecise probability" to solve the problem of prior misspecification / grain-of-truth / nonrealizability which plagues Bayesianism and Bayesian r...
32d32283-c3ca-42e1-ab78-bcbc41ec7fd2
StampyAI/alignment-research-dataset/blogs
Blogs
2013 in Review: Outreach This is the 2nd part of my personal and qualitative [self-review of MIRI in 2013](http://intelligence.org/2013/12/20/2013-in-review-operations/). By “outreach” I refer to *general* outreach efforts, rather than e.g. *outreach to specific researchers*, which will be discussed in the post abou...
b2537eba-9d5f-4c31-aa89-d69748d1b01b
trentmkelly/LessWrong-43k
LessWrong
Which meat to eat: CO₂ vs Animal suffering Animal agriculture generates an ungodly amount of animal suffering and greenhouse gas emissions. Ideally, everyone would adopt a vegan diet[1], but since this appears to be too much to ask for most people, the next question becomes: which food is the least bad? Here we run i...
870f38da-5a2d-40ed-b8d5-4a39d5b3eb98
StampyAI/alignment-research-dataset/arxiv
Arxiv
Grokking: Generalization Beyond Overfitting on Small Algorithmic Datasets 1 Introduction --------------- The generalization of overparameterized neural networks has long been a source of interest to the machine learning community since it defies intuitions derived from classical learning theory. In this paper we sh...
c9e5da5e-6e97-4f52-8c89-62786d33d260
StampyAI/alignment-research-dataset/youtube
Youtube Transcripts
Let's get physical - DeepMind: The Podcast (S2, Ep4) hello and welcome back to deepmind the podcast [Music] over the last two episodes we've been exploring deepmind's goal of solving intelligence asking what that actually means and traveling along some of the roads that could take us there this time it's all about the...
8d7ca155-65fd-4592-9878-e12cfc16c83b
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
Notes on Learning the Prior *This post was written to fulfil requirements of the SERI MATS Training Program.* One of the goals in writing this post is to arrive at an explanation of a particular set of issues raised by Irving in the [Learning the smooth prior](https://www.alignmentforum.org/posts/9x5mtYjHYfr4T7KLj/le...
dc3c6c75-dfe2-4a04-ab0d-872df54f327e
trentmkelly/LessWrong-43k
LessWrong
Weekly LW Meetups This summary was posted to LW main on September 27th. The following week's summary is here. Irregularly scheduled Less Wrong meetups are taking place in: * Atlanta Pre-Meetup Museum Field Trip: 28 September 2013 03:00PM * Atlanta Lesswrong September Meetup (2nd of 2): 28 September 2013 06:00PM *...
13decad3-6fa8-4f58-8d8e-a80bd78a9f9c
StampyAI/alignment-research-dataset/arxiv
Arxiv
AugMix: A Simple Data Processing Method to Improve Robustness and Uncertainty. 1 Introduction --------------- Current machine learning models depend on the ability of training data to faithfully represent the data encountered during deployment. In practice, data distributions evolve (Lipton et al., [2018](#bib.bib2...
f9b9b05a-f344-4757-9ba3-c162d1d2bc35
trentmkelly/LessWrong-43k
LessWrong
My Superpower: OODA Loops I've pushed the value of OODA loops in various posts over the years, but I don't recall ever writing a post specifically about them. Which is weird, because they're a superpower. So much so that I think they are at least 60% responsible for what enables me to lead a life I enjoy (the rest is ...
90525e55-bf14-430b-86d9-67689ffab006
trentmkelly/LessWrong-43k
LessWrong
Dialectic algorithm - For calculating if an argument is sustained or refuted
820e8279-5ad9-42c5-b038-fc44bb557518
trentmkelly/LessWrong-43k
LessWrong
The (partial) fallacy of dumb superintelligence In her opening statement for the Munk debate on AI risk, Melanie Mitchell addressed a proposed example of AGI risk: an AGI tasked with fixing climate change might decide to eliminate humans as the source of carbon emissions. She says: > This is an example of what's call...
f510bcf4-993c-4fa3-a83c-b54605f961e2
trentmkelly/LessWrong-43k
LessWrong
Weekly LW Meetups This summary was posted to LW Main on October 28th. The following week's summary is here. New meetups (or meetups with a hiatus of more than a year) are happening in: * Zürich social Rationality meetup: 29 October 2016 05:00PM Irregularly scheduled Less Wrong meetups are taking place in: * Muni...
8a2bc0fe-e759-4b03-a9fe-17e1a79dd6a2
trentmkelly/LessWrong-43k
LessWrong
Notes on Notes on the Synthesis of Form Christopher Alexander was an architect and designer; I've written about some of his later works in another post. This, one of his earliest books, focuses on design as a formal process. What is design, and how does it work? I'll split things up slightly differently than he does, ...