id
stringlengths
36
36
source
stringclasses
15 values
formatted_source
stringclasses
13 values
text
stringlengths
2
7.55M
5709bc4b-1e03-4f90-ad32-99b8d7970d97
awestover/filtering-for-misalignment
Redwood Research: Alek's Filtering Results
id: post579 Yoshua Bengio recently posted a high-level overview of his alignment research agenda on his blog. I'm pasting the full text below since it's fairly short. What can’t we afford with a future superintelligent AI? Among others, confidently wrong predictions about the harm that some actions could yield. Especi...
3f9c2dff-3715-4f57-a9bc-190a624f7b29
StampyAI/alignment-research-dataset/lesswrong
LessWrong
Against the Open Source / Closed Source Dichotomy: Regulated Source as a Model for Responsible AI Development **Context** =========== This is a short post to illustrate and start testing an idea that I have been playing around with in the last couple of days after listening to a recent 80,000 hours podcast with Musta...
039c6dd1-40dc-4d12-9baf-aa82996e848e
StampyAI/alignment-research-dataset/arxiv
Arxiv
Safe Deep RL in 3D Environments using Human Feedback 1 Introduction --------------- Many of deep reinforcement learning’s recent successes have relied on the availability of a procedural reward function and a simulated environment for the task in question. As a result, research has been largely insulated from many ...
27f33977-0eae-4068-8a96-c8c252edfcf9
trentmkelly/LessWrong-43k
LessWrong
Meetup : Madison: Prospect Theory Discussion article for the meetup : Madison: Prospect Theory WHEN: 14 October 2012 07:00:00PM (-0500) WHERE: 325 Rethke Ave, Madison WI Prospect theory is a very simple, but reasonably predictive, model of actual human decisions. This Sunday, I'll give a short presentation on prosp...
78374161-d209-4f64-819c-c588e70ac457
trentmkelly/LessWrong-43k
LessWrong
Research Priorities for Hardware-Enabled Mechanisms (HEMs) Longview Philanthropy is launching a new request for proposals on hardware-enabled mechanisms (HEMs). To encourage strong proposals, we’ve written up what we believe to be top priorities for research in this field.  We think HEMs are a promising method to en...
6d2b62e4-13fc-4446-8427-002bdf3faacd
trentmkelly/LessWrong-43k
LessWrong
Solstice 2014 / Rational Ritual Retreat - A Call to Arms Summary:  •  I'm beginning work on the 2014 Winter Solstice. There are a lot of jobs to be done, and the more people who can dedicate serious time to it, the better the end result will be and the more locations it can take place. A few people have volunteered ...
ef4006e7-189e-43ff-b62d-f1f3481ed0a8
trentmkelly/LessWrong-43k
LessWrong
Open Agency model can solve the AI regulation dilemma It seems to me that most people concerned about AI regulation (and calls for "CERN for AI", or the proposals such as the OAA) are concerned about the monopolisation of AI, not about regulation per se. And in AI monopolisation (or oligopolisation), they are mostly c...
157ec045-b588-4c28-bdd2-1440c89ddc2e
trentmkelly/LessWrong-43k
LessWrong
The Preference Fulfillment Hypothesis Short version Humans have an innate motivation ("preference fulfillment", PF) to fulfill the preferences of those they care about. It corresponds to at least some of the senses of the word "love", as well as related words such as "kindness" and "compassion". I hypothesize that it...
d7a14599-5d6c-4b15-8de7-df110c59ed3b
trentmkelly/LessWrong-43k
LessWrong
What are the most important papers/post/resources to read to understand more of GPT-3? I'm way more used to thinking about weird maths or distributed algorithms or abstract philosophical problems than about concrete machine learning architectures. But based on everything I see about GPT-3, it seems a nice idea to lear...
82f79933-1dfb-4be2-af52-753d53d5e580
trentmkelly/LessWrong-43k
LessWrong
When does technical work to reduce AGI conflict make a difference?: Introduction This is a pared-down version of a longer draft report. We went with a more concise version to get it out faster, so it ended up being more of an overview of definitions and concepts, and is thin on concrete examples and details. Hopefully...
7f0891fb-c3d1-497f-8117-79874ffc54c5
trentmkelly/LessWrong-43k
LessWrong
Are tulpas moral patients? We take a lot about whether or not are animals and to what extent they are conscious, but I have seen little discussion about whether tulpas should be considered to be conscious and to be moral patients. Is there any serious philosophy done on the topic?
62bc20d1-c701-454f-8e1f-3684c6c71171
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
[AN #142]: The quest to understand a network well enough to reimplement it by hand Alignment Newsletter is a weekly publication with recent content relevant to AI alignment around the world. Find all Alignment Newsletter **[resources here](http://rohinshah.com/alignment-newsletter/)**. In particular, you can look thro...
62fbed74-75a4-4a95-9c38-a207c1925ed3
StampyAI/alignment-research-dataset/lesswrong
LessWrong
Yoshua Bengio: "Slowing down development of AI systems passing the Turing test" Full text of the blog post below: > I recently signed an [**open letter**](https://futureoflife.org/open-letter/pause-giant-ai-experiments/) asking to slow down the development of giant AI systems more powerful than GPT-4  –those that cu...
afb28734-a7ef-42a5-9e24-3f50952dcbdf
trentmkelly/LessWrong-43k
LessWrong
Bloggingheads: Robert Wright and Eliezer Yudkowsky Sweet, there's another Bloggingheads episode with Eliezer. Bloggingheads: Robert Wright and Eliezer Yudkowsky: Science Saturday: Purposes and Futures
988f78ba-1e5a-4e06-be53-a3f98a60d553
trentmkelly/LessWrong-43k
LessWrong
Rationally Ending Discussions Discussions should end gracefully. There should be some clarity about why they’re ending and why that’s rationally acceptable. If someone wants to continue the discussion, they should have the opportunity to see why this ending is reasonable. (Or, failing that, they should have some evide...
f86d20c1-5a49-41cd-95bb-43b67de075f0
trentmkelly/LessWrong-43k
LessWrong
Meetup : West LA Meetup - Value of Information Discussion article for the meetup : West LA Meetup - Value of Information WHEN: 23 May 2012 07:00:00PM (-0700) WHERE: 10850 West Pico Blvd, Los Angeles, CA 90064 When: 7:00pm - 9:00pm Wednesday, May 23rd. Where: The Westside Tavern in the upstairs Wine Bar (all ages w...
5a0b27cf-d528-4bdd-860f-968d5b6f28f3
trentmkelly/LessWrong-43k
LessWrong
Are bread crusts healthier? When you think about it for a second, almost definitely not. They're exactly the same stuff as bread, just more burnt, and so more likely to have the nutritious stuff broken down, and contain carcinogens. Indeed a quick internet search "busts the myth" that crusts are healthier. So when my...
2d3d235e-1210-4a8f-8a5c-d6660f6c4cf1
StampyAI/alignment-research-dataset/eaforum
Effective Altruism Forum
"Develop Anthropomorphic AGI to Save Humanity from Itself" (Future Fund AI Worldview Prize submission) **This is a submission to the** [**Future Fund's AI Worldview Prize**](https://ftxfuturefund.org/announcing-the-future-funds-ai-worldview-prize/)**. It was submitted through** [**our submission form**](https://docs.g...
57ee2957-ad47-40a0-a8cf-93d5378a88c5
trentmkelly/LessWrong-43k
LessWrong
Rationality witticisms suitable for t-shirts or bumper stickers What are your best short witticisms, suitable for use on a t-shirt, bumper sticker, or similar location? Ideally something that might make someone reading it think, or get curious enough to ask about it. Simple in-group identification is fine too, though....
aeb9ae9b-2f7c-4ba9-a78e-bd4ed1efed9e
trentmkelly/LessWrong-43k
LessWrong
Causal Undertow: A Work of Seed Fiction A short story, best enjoyed in-context... try exploring possible meanings of some of the unusual terms with your favourite commercially available Solomonoff inductor. Acknowledgements. Thanks to Simon Pepin Lehalleur and Ziling Ye for feedback and suggestions. This post is the...
9e5d3525-3bf1-4404-9c9a-55f0220cf7f7
StampyAI/alignment-research-dataset/lesswrong
LessWrong
Defining Optimization in a Deeper Way Part 3 Last time I got stuck trying to remove the need for **uncertainty**. I have done it now! In the process I have also removed the need for **absolute time**. The Setup --------- First we need to consider the universe as a causal network. This can be done without any notion ...
c1877c11-7a53-454a-8877-89acfebe68a7
trentmkelly/LessWrong-43k
LessWrong
Meetup : LW Munich Meetup in July Discussion article for the meetup : LW Munich Meetup in July WHEN: 06 July 2013 03:00:00PM (+0200) WHERE: Café at Gasteig, Rosenheimer Straße 5, 81667 München The next Munich meetup will take place on July 6th. We will continue discussing the epistemology sequence as well as talk a...
4bd26bac-c0c2-43ca-a1a7-c75a8908b6d3
StampyAI/alignment-research-dataset/blogs
Blogs
Where next for piracy ? Where next for piracy ? ----------------------- As an *intellectual property abolitionist*, I'm often thinking of how we can keep improving media piracy. The raw file-sharing has is mostly solved: the BitTorrent protocol works pretty well, and [IPFS](https://ipfs.io/) is probably a reasonab...
2438ad5e-7b4e-4c57-8d3a-0e4540968913
trentmkelly/LessWrong-43k
LessWrong
Why would a free human society be in agreement on how to alter itself? So I've been grappling with these discussions on the plausibility or implausibility of redefining cultural boundaries of several sorts to (as the radical example of self-alteration by Eliezer goes) make non-consensual sex as tolerated and tolerable...
d364530d-0949-49f1-bb90-e1218fbe526e
trentmkelly/LessWrong-43k
LessWrong
Where in the world will a UBI develop first? Disclaimer: As a pathological lurker, I'm partially asking this question to help pave the way for questions/posts that I have spent more time thinking about and am more emotionally invested in. However, this question also struck me as important and relevant to the miscellan...
bcfce790-60c3-4acc-963e-ccad988b7604
trentmkelly/LessWrong-43k
LessWrong
Looking back on my alignment PhD This post has been recorded as part of the LessWrong Curated Podcast, and can be listened to on Spotify, Apple Podcasts, Libsyn, and more. ---------------------------------------- My dissertation. It's long, so if you're going to read anything from it, read Chapter 0 (Introduction)....
005a32e8-533d-40db-befc-1c5df312186c
trentmkelly/LessWrong-43k
LessWrong
Vocal Range I'm organizing another secular solstice, which has me thinking about where to pitch songs so the most people will feel comfortable. I found several people saying that range of C to C generally works well with untrained voices, and if they had said something weird like Db-E I might have believed them, but C...
683a3524-36af-4ed6-8459-eaa0814b12f2
trentmkelly/LessWrong-43k
LessWrong
Analogies and General Priors on Intelligence This post is part 2 in our sequence on Modeling Transformative AI Risk. We are building a model to understand debates around existential risks from advanced AI. The model is made with Analytica software, and consists of nodes (representing key hypotheses and cruxes) and edg...
7ff1fa07-29b8-4ee6-9e52-7f68ce53b3e0
trentmkelly/LessWrong-43k
LessWrong
Is this how I choose to show up? Original post: http://bearlamp.com.au/is-this-how-i-choose-to-show-up/ Is this how I choose to show up?  No. I’m exhausted.  I’m just trying to survive here and today I did that.  Not every today. But I did this today. Yes. Is this how I choose to show up?  No. I’m doing bette...
4ed06c00-2810-4580-8168-efb81bb2eecf
trentmkelly/LessWrong-43k
LessWrong
Meetup : Saskatoon's First Meetup! Discussion article for the meetup : Saskatoon's First Meetup! WHEN: 17 August 2013 01:00:00PM (-0600) WHERE: 2318 8th St E, Saskatoon, SK It's time for central Canada's first meetup! It will be at the Broadway Roaster on 8th street (not on broadway!) at the pleasant time of 1:00 ...
8bb93eb6-3eed-4a9a-bd99-8c034f8e7c89
trentmkelly/LessWrong-43k
LessWrong
Levels of Pluralism This post is part of the work done at Conjecture. Preparing the Pluralism Question When do we want all our research eggs in the same paradigm basket? Although most people don't go as far as the extreme paradigmatism of Thomas Kuhn in The Structure of Scientific Revolutions, which only allows one...
0926e7e8-d40b-42b9-82fe-26f6be3bdc8f
trentmkelly/LessWrong-43k
LessWrong
Neil Armstrong died before we could defeat death The sad news broke tonight : Neil Armstrong, the first human to ever walk another world, died today. We lost him forever. He died before we could defeat death. Once again the horror of death strikes. This time, in addition from wiping from us forever a hero of humanity...
7b500a29-d5a4-44b9-b7a0-44e2a2ad831f
trentmkelly/LessWrong-43k
LessWrong
Young Americans believe they have the best health in the world... Of course, it turns out they're actually last among developed nations in real health outcomes. > The U.S. ranks #1 among 17 affluent, western countries in the percentage of people aged 5 to 34 who rate their health as good. Unfortunately, when doctors ...
0a93b57b-5af3-4d04-830a-b43a464d4717
trentmkelly/LessWrong-43k
LessWrong
Book Review: On the Edge: The Future Previously: The Fundamentals, The Gamblers, The Business We have now arrived at the topics most central to this book, aka ‘The Future.’ RATIONALISM AND EFFECTIVE ALTRUISM (EA) > The Manifest conference was also one of the last reporting trips that I made for this book. And it co...
235c6bba-e5e0-4291-b36f-680d185c5d4a
trentmkelly/LessWrong-43k
LessWrong
Meetup : Washington, D.C.: Game Theory Discussion article for the meetup : Washington, D.C.: Game Theory WHEN: 20 March 2016 03:30:00PM (-0400) WHERE: Reynolds Center We will be meeting in the courtyard to talk about game theory - a mathematical field looking at strategies for games where the outcome of your moves ...
8f639f38-f963-42d1-a817-48ac85453082
trentmkelly/LessWrong-43k
LessWrong
Do not delete your misaligned AGI. In short: Just keeping all copies of potentially strong agents in long-term storage is a trivial way to maintain incentives to cooperate for some classes of misaligned AGI, by allowing us to reward the AGI's cooperation to whatever degree we later calculate had been warranted. In con...
497a2e17-4aeb-48ee-975e-bfa6f77cd141
StampyAI/alignment-research-dataset/lesswrong
LessWrong
Is a Self-Iterating AGI Vulnerable to Thompson-style Trojans? In his 1984 lecture ["Reflections on Trusting Trust"](https://www.cs.cmu.edu/%7Erdriley/487/papers/Thompson_1984_ReflectionsonTrustingTrust.pdf), Ken Thompson (of Unix fame) speculated about a methodology for inserting an undetectable trojan horse within th...
2eba1154-af2f-4d2d-aa88-149fd972125f
trentmkelly/LessWrong-43k
LessWrong
Meetup : Washington DC: Choice Blindness Discussion article for the meetup : Washington DC: Choice Blindness WHEN: 30 September 2012 03:00:00PM (-0400) WHERE: National Portrait Gallery, Washington, DC 20001, USA We'll spend some time discussing the recent findings on choice blindness, then revert to general convers...
c66f0ac9-4d20-4ec4-a079-889283695573
trentmkelly/LessWrong-43k
LessWrong
Hanson Debating Yudkowsky, Jun 2011 > On Wednesday I debated my ex-co-blogger Eliezer Yudkowsky at a private Jane Street Capital event (crude audio here, from 4:45; better video here [as of July 14]). > > I “won” in the sense of gaining more audience votes — the vote was 45-40 (him to me) before, and 32-33 after the ...
b20a5516-04c9-4a58-a7b6-4b00c395b423
trentmkelly/LessWrong-43k
LessWrong
Brain structure - scans reveal an amazingly regular pattern http://www.nih.gov/news/health/mar2012/nimh-29.htm has images of human and monkey brain scans which reveal an amazingly grid-like pattern of neuronal connections. What are the implications? Could brain scanning, emulation etc be simpler than would appear from...
293b0ae9-870c-4983-ad2e-e3dcea53f6c8
StampyAI/alignment-research-dataset/lesswrong
LessWrong
Do anthropic considerations undercut the evolution anchor from the Bio Anchors report? The [Bio Anchors report](https://www.lesswrong.com/posts/KrJfoZzpSDpnrv9va/draft-report-on-ai-timelines) by Ajeya Cotra tries to forecast AGI timelines by getting estimates of how much compute might be needed to build or train a mod...
9525d8e1-81db-439c-9075-2a455e7c692f
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
Principles of Privacy for Alignment Research The hard/useful parts of alignment research are largely about understanding agency/intelligence/etc. That sort of understanding naturally yields capabilities-relevant insights. So, alignment researchers naturally run into decisions about how private to keep their work. Thi...
3851eee9-7818-4b0f-94ed-37b42982d893
trentmkelly/LessWrong-43k
LessWrong
The Chatbot of Babble by ChatGPT-4o, with guidance and very light editing by me. This isn't meant to be a plausible version of the future, just me having some fun generating stories vaguely related to AI risks in the styles of vaguely similar stories from literature. 1. And it came to pass, in the latter days of mank...
573282d8-e81f-484a-9d73-238c56841ea2
trentmkelly/LessWrong-43k
LessWrong
Open Source Replication of Anthropic’s Crosscoder paper for model-diffing Intro Anthropic recently released an exciting mini-paper on crosscoders (Lindsey et al.). In this post, we open source a model-diffing crosscoder trained on the middle layer residual stream of the Gemma-2 2B base and IT models, along with code, ...
16ae0bc1-b2b4-431c-aee6-fda825a42193
trentmkelly/LessWrong-43k
LessWrong
Nonmindkilling open questions When I explain to people how beliefs should be expressed in probabilities, I would like to use an example like "Consider X. Lots of intelligent people believe X, but lots of equally intelligent people believe not-X. It would be ridiculous to say you are 100% sure either way, so even if yo...
8c68dad5-f161-490f-bd4d-a8058dd2fbc7
trentmkelly/LessWrong-43k
LessWrong
Weekly LW Meetups This summary was posted to LW Main on July 4th, the following week's summary is here.   Irregularly scheduled Less Wrong meetups are taking place in: * Houston, TX: 12 July 2014 02:00PM * Upper Canada LW Megameetup: Ottawa, Toronto, Montreal, Waterloo, London: 18 July 2014 07:00PM * Moscow m...
ad1e7ad5-7986-4538-9cc5-02108b777973
trentmkelly/LessWrong-43k
LessWrong
Two Words That Offer Final Ultimate Proof That Medical System Is Evil Music therapy. At the hospital, I have just learned, they offer Music Therapy. By music, I mean random folk singers with guitars. By therapy, I mean letting those folk singers sing, and play their guitars. And by offer, I mean do this without conse...
c6cf34da-e0b5-4e21-b5e3-1b414a100f10
trentmkelly/LessWrong-43k
LessWrong
Matt Botvinick on the spontaneous emergence of learning algorithms Matt Botvinick is Director of Neuroscience Research at DeepMind. In this interview, he discusses results from a 2018 paper which describe conditions under which reinforcement learning algorithms will spontaneously give rise to separate full-fledged rei...
e3bd866d-5e68-4312-9eba-d270bd58dea6
trentmkelly/LessWrong-43k
LessWrong
Running many AI variants to find correct goal generalization Problem: An AI may incorrectly generalize a goal we present it. Proposal: Run many slightly different instances of AI and find the “Schelling point” around correct generalization. The idea is that if we run many similar AIs or copies of the same AI in para...
69e93c86-76a6-44a4-8d2d-7cce698730c2
trentmkelly/LessWrong-43k
LessWrong
Galaxy-Brain Hobo Antibiotics? [ edit: I have gotten an ear infection diagnosis and a prescription for antibiotics as of 26 June 2025, [ see comment ] ] I have what I think is a chronic inner ear infection. Since March of 2021, I've had subjective obstructive Eustachian tube dysfunction in my right ear, as well as oc...
d764343c-4349-4fdc-b807-95a87a1001f8
trentmkelly/LessWrong-43k
LessWrong
Eternal: The Exit Interview Previously: Eternal, and Hearthstone Economy versus Magic Economy, The Eternal Grind, My Referral Link to download Eternal if you want to play. Epistemic Status: Played a lot of games. Wrote a lot of words. Those not interested in Magic, Eternal or game design may want to skip this one. E...
daa1d261-6fdc-4ff7-a628-9eee1e945a0d
trentmkelly/LessWrong-43k
LessWrong
Open Christmas card to my mother Dear Mama, Happy Christmas! Rather than be so bold as to give you a further object to take care of, I gathered for you a collection of things I think you might like on the internet. I CC the world, in case they are interested, or want to add anything. Lots of love, Katja xox   (Pi...
74fb6449-5615-4072-b364-cab465ab386e
trentmkelly/LessWrong-43k
LessWrong
Hedonic Loops and Taming RL Crossposted from my personal blog.   Everybody knows about the hedonic treadmill. Your hedonic state adjusts to your circumstances over time and quickly reverts to a mostly stable baseline. This is true  of basic physiological needs -- you feel hungry; you seek out food; you eat; you feel ...
70fce916-440d-4d45-819a-994b40c8ed02
trentmkelly/LessWrong-43k
LessWrong
Meetup : Tel Aviv Meetup: Rump Session Discussion article for the meetup : Tel Aviv Meetup: Rump Session WHEN: 25 November 2014 07:00:00PM (+0200) WHERE: Yigal Alon Street 98, Tel Aviv-Yafo, Israel WHEN: 25 November 2014 07:00:00PM (+0200) WHERE: Google, Electra Tower 29th Floor, Yigal Alon Street 98, Tel Aviv. It...
ba7e8f99-1766-43db-a0f2-aeda47c7b433
StampyAI/alignment-research-dataset/arbital
Arbital
GalCom summary(Rules): 1. It costs 1 galcoin per bit to reserve on-peak bits in advance. Galcoins are very expensive. 2. You can reprogram your machine on Vega in advance, using cheap, off-peak bits. Don't worry about the cost of those. 3. You can resell reserved bits that you don't use (including fractional bits). 4...
1ea746c0-7657-49e0-967a-497d0f0c853f
trentmkelly/LessWrong-43k
LessWrong
Have people given up on iterated distillation and amplification? The BlueDot Impact write-up for scalable oversight seems to suggest that people have given up on iterated distillation and amplification (IDA) working. I haven’t really seen much research here, but is that actually the case? I know that Ought is now pre...
e4cf3a7d-22bd-488f-b2d1-81f545d83d05
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
Policy Selection Solves Most Problems .mjx-chtml {display: inline-block; line-height: 0; text-indent: 0; text-align: left; text-transform: none; font-style: normal; font-weight: normal; font-size: 100%; font-size-adjust: none; letter-spacing: normal; word-wrap: normal; word-spacing: normal; white-space: nowrap; float:...
bb095c95-44e1-41ea-8ce6-7f2f7490d91c
trentmkelly/LessWrong-43k
LessWrong
Some thoughts on relations between major ethical systems On the recent LessWrong/CFAR Census Survey, I hit the following question:   > Which of the following major ethical systems do you subscribe to: > > 1) Consequentialism > > 2) Deontology > > 3) Virtue Ethics > > 4) Other   To my own surprise, I couldn't c...
7c2ba392-74f6-48a8-9f4c-3cb4daaf3cda
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
[AN #110]: Learning features from human feedback to enable reward learning **Newsletter #110** Alignment Newsletter is a weekly publication with recent content relevant to AI alignment around the world. Find all Alignment Newsletter **[resources here](http://rohinshah.com/alignment-newsletter/)**. In particular, you ...
fcac362e-1fee-417e-b87d-98490d56c03a
StampyAI/alignment-research-dataset/eaforum
Effective Altruism Forum
What are some sources related to big-picture AI strategy? I am hoping to strengthen my zoomed-out picture of the current strategic situation related to AI. I am very interested in finding documents/posts/write-ups that attempt to layout important real-world, strategic considerations. Ideally such documents would incl...
897d15b6-cb8a-4292-afb0-d123bafda046
StampyAI/alignment-research-dataset/lesswrong
LessWrong
Knightian uncertainty: a rejection of the MMEU rule Recently, I found myself in a conversation with someone advocating the use of [Knightian uncertainty](http://en.wikipedia.org/wiki/Knightian_uncertainty). He (who I'm anonymizing as Sir Percy) made suggestions that are useful to most bounded reasoners, and which can ...
a2ef6166-f441-47d6-b7e3-2d23dce9733f
StampyAI/alignment-research-dataset/arbital
Arbital
Reliable prediction Most statistical learning theory settings (such as [PAC learning](https://en.wikipedia.org/wiki/Probably_approximately_correct_learning) and [online learning](https://en.wikipedia.org/wiki/Online_machine_learning)) do not provide good enough bounds in realistic high-stakes settings, where the test ...
457e3af1-283b-4cfc-b82b-423f442e0a06
trentmkelly/LessWrong-43k
LessWrong
87,000 Hours or: Thoughts on Home Ownership Edited to add epistemic status and summary: The basic thesis of this post is that, for many reasons, the volatility (riskiness) of buying a house are significantly higher than generally believed. The considerations explored here are the result of discussions with a financi...
c9c08302-794b-4c6f-a866-9afb6aeaa7c8
StampyAI/alignment-research-dataset/lesswrong
LessWrong
Going Beyond Linear Mode Connectivity: The Layerwise Linear Feature Connectivity Hi! I am excited to share our new work *Going Beyond Linear Mode Connectivity: The Layerwise Linear Feature Connectivity* with the LessWrong Community. In our new [work](https://arxiv.org/pdf/2307.08286.pdf), we introduce a stronger notio...
981740dd-8c5c-4057-863e-e328873db6c4
trentmkelly/LessWrong-43k
LessWrong
A Reading From The Book Of Sequences Summary: A speaker will read excerpts from a text the community finds important, weaving the quotes in with experiences from their life or stories. Tags: Large, Investment, maybe Experimental? Purpose: This creates common knowledge around the subject of the reading. Everyone know...
673d7967-146b-46c0-94f2-eea6795e4418
trentmkelly/LessWrong-43k
LessWrong
Seeking Truth Too Hard Can Keep You from Winning LW-style rationality is about winning. Instrumentally, most rationalists think they can win, in part, by seeking truth. I frequently run into comments here where folks take truth as an effectively "sacred" concern in the sense that truth matters to them above all else. ...
7d731d40-f35e-4d5c-8f61-a27a6f7db5f0
trentmkelly/LessWrong-43k
LessWrong
How to Use Prompt Engineering to Rewire Your Brain None
a57984b9-de8f-45e7-a209-75dec427536d
trentmkelly/LessWrong-43k
LessWrong
On AI and Interest Rates Note, To say it up front: None of this, or anything ever on this blog, is investment advice. This post on the EA form, highlighted by Tyler Cowen, points out that 1. Transformative aligned AI would increase real interest rates. A lot. 2. Transformative unaligned AI would increase real inte...
96ee9ae4-04fe-4d18-b5d5-38846b65008a
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
Non-Adversarial Goodhart and AI Risks In a [recent paper](https://arxiv.org/abs/1803.04585) by Scott Garrabrant and myself, we formalized and extended the categories [Scott proposed](https://www.lesserwrong.com/posts/EbFABnst8LsidYs5Y/goodhart-taxonomy) for Goodhart-like phenomena. (If you haven't read either his post...
27e49c2e-9daa-4209-9716-f704d2b0849e
trentmkelly/LessWrong-43k
LessWrong
Situational awareness in Large Language Models I’m grateful to Bogdan Cirstea, Konstantin Pilz and Raphaël S for providing feedback on this post. This post tries to clarify the concept of situational awareness, in particular with respect to current large language models. What is situational awareness Not writing an...
edd0c964-34bf-4a37-8fe1-2f9884f7290b
trentmkelly/LessWrong-43k
LessWrong
Ideal Format/Frequency of Information Consumption It seems that for every type of information input, there is an ideal: 1. Source/Format 2. Frequency 3. Time of Day 4. Energy State ---------------------------------------- 1. Source/Format Examples: Twitter, blog, books, podcasts etc. Don’t waste your time foll...
533a52a0-6282-4f1f-992c-85e2df824a58
trentmkelly/LessWrong-43k
LessWrong
Stable Pointers to Value III: Recursive Quantilization This is a very loose idea. In Stable Pointers to Value II, I pointed at a loose hierarchy of approaches in which you try to get rid of the wireheading problem by revising the feedback loop to remove the incentive to wirehead. Each revision seems to change the nat...
3b663773-19f1-456f-8501-07e96646f9c6
trentmkelly/LessWrong-43k
LessWrong
The slopes to common sense The slopes to common sense Usually, I try to outline the thing I'm talking about first. But in this case, it's best to start with an example. i - Why we sleep In 2017 the book "Why We Sleep" was published. This book was essentially trying to muster a very strong scientific argument for why...
94cd6f61-fbb9-4daf-af0f-31155111cc15
trentmkelly/LessWrong-43k
LessWrong
Gradient Descent on the Human Brain TL;DR: Many alignment research proposals often share a common motif: figure out how to enter a basin of alignment / corrigibility for human-level models, and then amplify to more powerful regimes while generalizing gracefully. In this post we lay out a research agenda that comes at ...
4f4ce59f-4ad1-40c5-bebe-86438ecfd4f0
StampyAI/alignment-research-dataset/aisafety.info
AI Safety Info
How and why should I form my own views about AI safety? Forming an “[inside view](https://www.alignmentforum.org/tag/inside-outside-view)” of AI safety is the standard advice for aspiring [AI safety researchers](/?state=8UMA&question=How%20can%20I%20do%20conceptual%2C%20mathematical%2C%20or%20philosophical%20work%20on...
2af3cac7-f8df-4c71-8a25-210e940940ce
trentmkelly/LessWrong-43k
LessWrong
Amanda Knox Redux: is Satoshi Nakamoto the real Satoshi Nakamoto? Many of you here have likely heard of Bitcoin, and maybe know something about it. Earlier today, a story broke that a reporter had apparently tracked down the real Satoshi Nakamoto, infamous creator of the Bitcoin protocol. This seems like an excellen...
b930680f-aa79-48a0-9725-7a6cb64abddc
trentmkelly/LessWrong-43k
LessWrong
Geeks, Mops, Sociopaths
c651eae5-0db4-41a6-a993-fe3d2dcaa054
trentmkelly/LessWrong-43k
LessWrong
Optimizers: To Define or not to Define Introduction The definition of "optimizer" has been a sort of rationalist pastime for a many years now. Many different ones have been suggested, and many have fallen down. Most have been in terms of the behaviour of the system in question, which is probably not what we really wan...
d3eea2b1-bc36-4e9b-a7c9-88861234e1c7
trentmkelly/LessWrong-43k
LessWrong
A Sketch of Belocracy: a new system of governance I have been working on designing a new system of governance - an alternative to liberal democracy - for the past half decade, and hope that sharing it will inspire a good conversation. So, epistemic status: I really hope I have it figured out, but only time will tell w...
44dd8a45-4f73-480b-a493-599af1427785
trentmkelly/LessWrong-43k
LessWrong
Meetup : West LA Meetup 10-04-2011 Discussion article for the meetup : West LA Meetup 10-04-2011 WHEN: 04 October 2011 07:00:00PM (-0700) WHERE: 10800 West Pico Blvd, Suite 312, Los Angeles, CA 90064 When: 7pm - 9pm Tuesday, October 4th Where: The Westside Tavern, located inside the Westside Pavillion on the secon...
124f52ef-79e9-4dfc-95e1-800df14db59e
trentmkelly/LessWrong-43k
LessWrong
Would anyone be interested in pursuing the Virtue of Scholarship with me? I'm a new undergraduate student essentially taking a gap year(it's complicated). I'm looking for someone that would be interested in studying various fields of science and mathematics with me. I've taken through Calculus 2, know how to program i...
26f1f32b-631a-48ca-8eff-6d61c7176c35
StampyAI/alignment-research-dataset/lesswrong
LessWrong
Decision theory: Why Pearl helps reduce “could” and “would”, but still leaves us with at least three alternatives (This is the third post in a planned [sequence](/lw/16f/decision_theory_an_outline_of_some_upcoming_posts/).) My [last post](/lw/174/decision_theory_why_we_need_to_reduce_could_would/) left us with th...
b5be32b1-7999-44c6-b8d2-af01e1d160d5
trentmkelly/LessWrong-43k
LessWrong
Clarifying the palatability theory of obesity Here, I intend to clarify the core claims of the palatability theory of obesity outlined by Stephan J. Guyenet in his book The Hungry Brain (reviewed by Scott Alexander). Previously, I had written a comment defending his theory against a recently popular alternative, propo...
3ee39ce1-eaad-4c1b-9b77-f5608de778cb
trentmkelly/LessWrong-43k
LessWrong
We Choose To Align AI Epistemic status: poetry > "We choose to go to the moon! We choose to go to the moon in this decade and do the other things, not because they are easy, but because they are hard. Because that goal will serve to organize the best of our skills and energies. Because that challenge is one we are wi...
159321a3-2389-47a4-8ac1-4e31b843181a
StampyAI/alignment-research-dataset/arxiv
Arxiv
Artificial Intelligence is stupid and causal reasoning won't fix it Abstract -------- Artificial Neural Networks have reached ‘Grandmaster’ and even ‘super-human’ performance’ across a variety of games, from those involving perfect-information, such as Go [Silver\_2016]; to those involving imperfect-information, su...
3e8e633e-6b34-4323-919a-7c2c709062af
trentmkelly/LessWrong-43k
LessWrong
We probably won't just play status games with each other after AGI There is a view I’ve encountered somewhat often,[1] which can be summarized as follows:  After the widespread deployment of advanced AGI, assuming humanity survives, material scarcity will largely disappear. Everyone will have sufficient access to nec...
c553807c-2bab-4d9b-9156-1073a40468ef
trentmkelly/LessWrong-43k
LessWrong
The Use of Many Independent Lines of Evidence: The Basel Problem This post describes how one can use many independent arguments to justifiably develop very high confidence in the truth of a statement. It ends with a case study: Euler’s use of several independent lines of evidence to develop confidence in the validity ...
bef5e773-ff7c-4b05-ad97-e0f0bf945f6a
StampyAI/alignment-research-dataset/lesswrong
LessWrong
Solomonoff Induction explained via dialog.
f894d328-19e7-4c16-974d-19923a7aabff
trentmkelly/LessWrong-43k
LessWrong
Meetup : Second Canberra Meetup - Paranoid Debating Discussion article for the meetup : Second Canberra Meetup - Paranoid Debating WHEN: 28 February 2014 06:30:00PM (+1100) WHERE: 148 Bunda St, Canberra For our second meetup, we will be playing paranoid debating (for details, see http://wiki.lesswrong.com/wiki/Para...
ae118c87-8cc2-475f-83ef-8705648d85ee
StampyAI/alignment-research-dataset/arxiv
Arxiv
Do's and Don'ts for Human and Digital Worker Integration I Introduction --------------- AI technologies are transforming how work is conducted in every industry. Today, AI systems are enabling businesses to personalize services, converse with customers, automate operations, optimize workflows, predict demand, and r...
dc323957-1161-43b1-9d7a-41590a88f204
trentmkelly/LessWrong-43k
LessWrong
Quantum versus logical bombs Child, I'm sorry to tell you that the world is about to end. Most likely. You see, this madwoman has designed a doomsday machine that will end all life as we know it - painlessly and immediately. It is attached to a supercomputer that will calculate the 10100th digit of pi - if that digit ...
4d236d00-c6ca-4ef8-97cb-869a2441a8b5
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
[ASoT] Some thoughts about imperfect world modeling So [a few posts ago](https://www.lesswrong.com/posts/oijJc8Mu2jPNgpuvy/some-thoughts-about-deceptive-mesaoptimization) I looked at the problem of not being able to anticipate all consequences of an action as being related to deceptive mesaoptimization, but also outer...
92c9e9d9-99ed-43ac-ba22-49a9c974e72a
trentmkelly/LessWrong-43k
LessWrong
Apply to the Cambridge ERA:AI Fellowship 2025 The Cambridge ERA:AI Fellowship has opened applications for an in-person, paid, 8-week Summer Research Fellowship focused on catastrophic AI risk mitigation, taking place from 29th June to 24th August 2025 in Cambridge, UK, and aimed at a variety of aspiring researchers.  ...
50918784-5e46-4c88-a3da-387f7948d766
trentmkelly/LessWrong-43k
LessWrong
Pivotal acts from Math AIs Eliezer Yudkowsky on Math AIs Here are some interesting quotes from the alignment debate with Richard Ngo. > If it were possible to perform some pivotal act that saved the world with an AI that just made progress on proving mathematical theorems, without, eg, needing to explain those theore...
ca778ccc-3023-4585-8761-cf42e3b7f0d8
trentmkelly/LessWrong-43k
LessWrong
Value extrapolation partially resolves symbol grounding Take the following AI, trained on videos of happy humans: Since we know about AI wireheading, we know that there are at least two ways the AI could interpret its reward function[1]: either we want it to make more happy humans (or more humans happy); call this ...
cd691698-498b-422d-819a-ed856c210dfe
trentmkelly/LessWrong-43k
LessWrong
Do the people behind the veil of ignorance vote for "specks"? > The veil of ignorance as Rawls put it ..."no one knows his place in society, his class position or social status; nor does he know his fortune in the distribution of natural assets and abilities, his intelligence and strength, and the like."    The devi...
2fb63532-e9ab-451e-8db5-cd5f71719809
StampyAI/alignment-research-dataset/lesswrong
LessWrong
Pre-registering a study This post is my [pre-registration](https://www.acf.hhs.gov/opre/blog/2022/08/pre-registering-studies-what-it-how-do-you-do-it-and-why#:~:text=Pre%2Dregistration%20is%20the%20practice,submitting%20it%20to%20a%20registry.) of a study I will be running to continue the exploratory work I started [h...
e236fb5b-99c9-46e2-bd8a-d6d2bdc85f3b
trentmkelly/LessWrong-43k
LessWrong
Topological Debate Framework I would like to thank Professor Vincent Conitzer, Caspar Oesterheld, Bernardo Subercaseaux, Matan Shtepel, and Robert Trosten for many excellent conversations and insights. All mistakes are my own. I think that there's a fundamental connection between AI Safety via Debate and Guaranteed S...
0fc7141e-e4f9-4985-8c48-87ff537cdd64
trentmkelly/LessWrong-43k
LessWrong
For me the Christianity deal-breaker was meekness I was raised Catholic, became agnostic around 13, stopped thinking an afterlife made sense at 15, and noticed I was no longer religious at 16. But I still spent the next ~12 years in pretty close proximity to Christianity. I did religious studies at school, I studied ...
916bddd8-a348-4f9f-904b-0aa51340234a
StampyAI/alignment-research-dataset/arbital
Arbital
Injective function A [https://arbital.com/p/-3jy](https://arbital.com/p/-3jy) $f: X \to Y$ is *injective* if it has the property that whenever $f(x) = f(y)$, it is the case that $x=y$. Given an element in the [image](https://arbital.com/p/3lm), it came from applying $f$ to exactly one element of the [domain](https://a...