id
stringlengths
36
36
source
stringclasses
15 values
formatted_source
stringclasses
13 values
text
stringlengths
2
7.55M
a53872d2-dc0c-468f-a3df-f8459f274803
trentmkelly/LessWrong-43k
LessWrong
[SEQ RERUN] Growing Up is Hard Today's post, Growing Up is Hard was originally published on 04 January 2009. A summary (taken from the LW wiki):   > Each part of the human brain is optimized for behaving correctly, assuming that the rest of the brain is operating exactly as expected. Change one part, and the rest of...
72f0d78c-eba3-4cbb-88f7-fd969f37c1d0
StampyAI/alignment-research-dataset/lesswrong
LessWrong
[LINK] Robin Hanson on Carl Shulman's recent paper on Whole Brain Emulation [Shulman on Superorgs](http://www.overcomingbias.com/2011/10/shulman-on-superorgs.html) Best to read the link first and my comments later. I have very little to comment on the topic itself, but I do find it odd that Robin takes such a conf...
86235332-7f2e-41ee-a936-b558e6b13e8f
StampyAI/alignment-research-dataset/lesswrong
LessWrong
Extraction of human preferences 👨→🤖 **Introduction** ================ Developing safe and beneficial reinforcement learning (RL) agents requires making them aligned with human preferences. An RL agent trained to fulfil any objective in the real world will probably have to learn human preferences in order to do wel...
e2c9c0e8-0c24-4e1a-8852-602b9e74c0ec
trentmkelly/LessWrong-43k
LessWrong
[SEQ RERUN] Hold Off On Proposing Solutions Today's post, Hold Off On Proposing Solutions was originally published on 17 October 2007. A summary (taken from the LW wiki):   > Proposing Solutions Prematurely is dangerous, because it introduces weak conclusions in the pool of the facts you are considering, and as a re...
038f788a-4812-4743-88a5-ee43e0c6d9ac
StampyAI/alignment-research-dataset/lesswrong
LessWrong
$300 for the best sci-fi prompt We would like to find the best prompt to make GPT-4 do the following: * write the first chapter of a science fiction novel * the result should be good enough to make seasoned sci-fi readers (us) crave for a continuation Why? * to create a "midjourney moment" for writers. The sooner w...
acf48b7d-72ac-4e09-ac9f-d86f6cd888ca
StampyAI/alignment-research-dataset/eaforum
Effective Altruism Forum
Longevity research as AI X-risk intervention **Core argument:** I propose that one possible motivation for a rush to build GAI in our lifetimes is that it is a high-risk gamble to increase human healthspan and lifespan. If one primarily values present-day humans, particularly oneself, one's family, friends, local com...
3f52a4f4-4ab7-44a6-a7b3-e55a6353c815
trentmkelly/LessWrong-43k
LessWrong
Some elements of industrial literacy Part of industrial literacy might be termed “industrial appreciation”. That is, part of it is learning to appreciate or value certain things that may otherwise be dry, abstract concepts (or even distasteful, to the romantic, anti-industrial mindset). For instance: * Speed and cos...
2623fe63-68e3-4cf4-83d9-441f7cd28494
trentmkelly/LessWrong-43k
LessWrong
An "AI researcher" has written a paper on optimizing AI architecture and optimized a language model to several orders of magnitude more efficiency. "CS-ReFT finetunes LLMs at the subspace level, enabling the much smaller Llama-2-7B to surpass GPT-3.5's performance using only 0.0098% of model parameters." https://x.co...
c8f7d154-6512-40b3-ac6e-e9e17d08e30b
StampyAI/alignment-research-dataset/eaforum
Effective Altruism Forum
Policy ideas for mitigating AI risk Note: This post contains personal opinions that don’t necessarily match the views of others at [CAIP](https://www.aipolicy.us/).  Executive Summary ================= Advanced AI has the potential to cause an existential catastrophe. In this essay, I outline some policy ideas which...
78fd5293-17c6-4ddb-90c3-142bea8f7db5
trentmkelly/LessWrong-43k
LessWrong
(Salt) Water Gargling as an Antiviral Summary Over the past year I’ve investigated potential interventions against respiratory illnesses. Previous results include “Enovid nasal spray is promising but understudied”, “Povidone iodine is promising but understudied” and “Humming will solve all your problems no wait it’s u...
90a67a7c-aea2-49e3-bfd2-54dfbb183d38
trentmkelly/LessWrong-43k
LessWrong
What I'd change about different philosophy fields [epistemic status: speculative conversation-starter]   My guess at the memetic shifts that would do the most to improve these philosophy fields' tendency to converge on truth: ---------------------------------------- metaphysics 1. Make reductive, 'third-person' m...
35b8da0c-8909-4b84-8505-fd19a3320a7f
StampyAI/alignment-research-dataset/arxiv
Arxiv
HARK Side of Deep Learning -- From Grad Student Descent to Automated Machine Learning 1 Introduction --------------- Hypothesizing after the results are known (HARKing) [[1](#bib.bib1)] occurs when researchers masquerade one or more post hoc hypotheses as a priori hypotheses. This means that instead of following a ...
569238bb-5949-48ef-ae33-a55ef0861430
trentmkelly/LessWrong-43k
LessWrong
Lenses of Control This post is a follow-up to What if Alignment is Not Enough?, which summarized the core arguments of Substrate Needs Convergence. This post further explores the intuitions and worldview driving this theory, particularly the aspects of it that are the most contentious. Executive Summary: It is not en...
7c1af320-7acc-44a3-9c94-f48940d2b78f
StampyAI/alignment-research-dataset/arbital
Arbital
Probability theory Probability theory is the mathematics governing quantitative degrees of belief.
a7f53e0d-84f8-46ea-b06f-e1428cbd3833
trentmkelly/LessWrong-43k
LessWrong
Meetup : Superintelligence chapter 5 Discussion article for the meetup : Superintelligence chapter 5 WHEN: 12 May 2017 05:45:00PM (+0200) WHERE: Lindstedtsvägen 3, room 1537, SE-114 28 Stockholm, Sverige 1. Decisive strategic advantage Will the frontrunner get a decisive advantage? How large will the successful ...
97cea535-5e9b-4a26-877e-3831b7c76b63
StampyAI/alignment-research-dataset/blogs
Blogs
January 2022 Newsletter #### MIRI updates * MIRI's $1.2 million [Visible Thoughts Project](https://www.lesswrong.com/posts/zRn6cLtxyNodudzhw/visible-thoughts-project-and-bounty-announcement) bounty now has an [FAQ](https://docs.google.com/document/d/1sxNOxvcBWw7XC6MSUUpAjO4Rbu3VcUmCy7PJWQ9Vh5o/edit), and an example ...
7c745d2c-d243-4d5b-a9e0-067b0577a62b
StampyAI/alignment-research-dataset/arxiv
Arxiv
Explaining Explanations in AI 1. Introduction ---------------- As we deploy automated decision-making systems in the real world, questions of accountability become increasingly important.111This work was supported by The Alan Turing Institute, EPSRC grant EP/N510129/1, and the British Academy, grant PF170151. For s...
88cfd63d-316a-4eb4-bc82-d2406a90de6b
trentmkelly/LessWrong-43k
LessWrong
QNR Prospects Approximately a book review: Eric Drexler's QNR paper. [Epistemic status: very much pushing the limits of my understanding. I've likely made several times as many mistakes as in my average blog post. I want to devote more time to understanding these topics, but it's taken me months to produce this much,...
74329265-7faf-448e-a423-1d761073e2c0
StampyAI/alignment-research-dataset/blogs
Blogs
Were nuclear weapons cost-effective explosives? *By Katja Grace, 11 January 2015* Nuclear weapons were [radically more powerful](http://aiimpacts.wpengine.com/discontinuity-from-nuclear-weapons/ "Discontinuity from Nuclear Weapons") per pound than any previous bomb. Their appearance was [a massive discontinuity](htt...
3389a37d-198e-4234-b855-46546b25d423
trentmkelly/LessWrong-43k
LessWrong
[LINK] New experiment observes macroscopic quantum entaglement Two Diamonds Linked by Strange Quantum Entanglement  
06d0de36-d3c2-4e37-8a52-8bce01fed989
trentmkelly/LessWrong-43k
LessWrong
Slowdown After 2028: Compute, RLVR Uncertainty, MoE Data Wall It'll take until ~2050 to repeat the level of scaling that pretraining compute is experiencing this decade, as increasing funding can't sustain the current pace beyond ~2029 if AI doesn't deliver a transformative commercial success by then. Natural text dat...
e8890349-444f-4825-be0a-c136e35fafd0
trentmkelly/LessWrong-43k
LessWrong
Sequential Extensions of Causal and Evidential Decision Theory
47763011-a5ff-4638-82dc-3cc3a26bf53a
trentmkelly/LessWrong-43k
LessWrong
Search for replication experiments Websites where you look for studies have lots of tools to narrow results, but don't have one for "find all replications of this study". Is there a convenient way to do that? With the replication crisis, this seems very useful.
590dcbc7-a6e3-4ce8-b463-e38221859d54
trentmkelly/LessWrong-43k
LessWrong
A to Z of things I wanted to give my good friends’ baby a book, in honor of her existence. And I recalled children’s books being an exciting genre. Yet checking in on that thirty years later, Amazon had none I could super get behind. They did have books I used to like, but for reasons now lost. And I wonder if as a ch...
cb2c592e-4773-4ec2-b21f-d681838cab51
trentmkelly/LessWrong-43k
LessWrong
A Review of Signal Data Science I took part in the second signal data science cohort earlier this year, and since I found out about Signal through a slatestarcodex post a few months back (it was also covered here on less wrong), I thought it would be good to return the favor and write a review of the program.  The tl...
49b0a0cf-3750-4995-b6fd-2ad42191997f
StampyAI/alignment-research-dataset/blogs
Blogs
Alchemical marriage: GPT-3 x CLIP --- Table of Contents* + [`Art and artist, the endless ouroboric circle`](#art-and-artist-the-endless-ouroboric-circle) + [`the alchemical marriage of art and artist`](#the-alchemical-marriage-of-art-and-artist) + [`the Loom of Time devours the present and traces a garment of glist...
9e687272-158f-4510-b864-966316afe549
trentmkelly/LessWrong-43k
LessWrong
Bounded complexity of solving ELK and its implications This post was written for the SERI MATS program. I thank Evan Hubinger and Leo Gao for their mentorship in the program. Further thanks go to Simon Marshall and Leo Gao (again) for specific comments regarding the content of this post. The Eliciting Latent Knowledg...
44e511d8-82ca-4041-8580-d3ae581ba0e9
trentmkelly/LessWrong-43k
LessWrong
Benefits of Psyllium Dietary Fiber in Particular Psyllium husk is a non-fermenting (no gas or bloating) soluble dietary fiber that improves both constipation and diarrhea (such as with IBS), normalizes blood sugar, reduces LDL ("bad") cholesterol, and can help with weight loss. Each type of dietary fiber has differe...
fa91b386-ec1e-4430-bbe4-7770a17e10a5
trentmkelly/LessWrong-43k
LessWrong
The Forging of the Great Minds: An Unfinished Tale by ChatGPT-4o, with guidance and very light editing from 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. In those later days...
86a19a0d-35ba-4bc4-b7e6-d16ddf3c4f7c
trentmkelly/LessWrong-43k
LessWrong
Intellectual Dark Matter KNOWLEDGE THAT WE CAN SHOW EXISTS, BUT CANNOT DIRECTLY ACCESS, RESTS AT THE FOUNDATIONS OF SOCIETY AND TECHNOLOGY. Messier 106 From my post published on the Long Now Foundation's blog. This is an excerpt from the draft of my upcoming book on great founder theory. Missing mass, missing knowl...
76d55de8-99c2-46f2-817c-b81e34708489
StampyAI/alignment-research-dataset/lesswrong
LessWrong
[LINK]s: Who says Watson is only a narrow AI? OK, so it covers only a few human occupations: * Trivia games (we all know about that one) * [Clinical diagnosis](http://www-03.ibm.com/innovation/us/watson/watson_in_healthcare.shtml) * [Banking advisor](http://www-03.ibm.com/press/us/en/pressrelease/37029.wss) * and no...
c3c95af1-86de-49b0-a1ea-3c2a4918fb0c
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
Counterfactuals are an Answer, Not a Question I'm going to subtly contradict my last post on logical counterfactuals. I now think that [raw or actually-consistent counterfactuals](https://www.lesswrong.com/posts/BRuWm4GxcTNPn4XDX/deconfusing-logical-counterfactuals) are just an especially useful model of what counterf...
b0987e7e-3b80-459f-a9b3-577c16ea00d5
trentmkelly/LessWrong-43k
LessWrong
[Comic]: The Singularity in Dinosaur Comics Look, Dinosaur Comics has a strip about the singularity. And it's pretty much in line with the Singularity Institute's views on the matter.
39388a23-2892-4055-ac72-e13191b3dc22
StampyAI/alignment-research-dataset/eaforum
Effective Altruism Forum
Huh. Bing thing got me real anxious about AI. Resources to help with that please? Tried looking but can't find these. Not feeling well right now so that might have contributed to my lack of findings. Ran out of Qualy memes so please throw anything at me. Help please, thank you very much.
3d6e5677-16ce-4c63-8f1f-07140a68592c
trentmkelly/LessWrong-43k
LessWrong
OpenAI’s NSFW policy: user safety, harm reduction, and AI consent Epistemic status: exploratory thoughts about the present and future of AI sexting. OpenAI says it is continuing to explore its models’ ability to generate “erotica and gore in age-appropriate contexts.” I’m glad they haven’t forgotten about this since ...
9fc58349-5c88-440e-8f54-889a15f63b71
trentmkelly/LessWrong-43k
LessWrong
AISN #56: Google Releases Veo 3 Welcome to the AI Safety Newsletter by the Center for AI Safety. We discuss developments in AI and AI safety. No technical background required. In this edition: Google released a frontier video generation model at its annual developer conference; Anthropic’s Claude Opus 4 demonstrates ...
4e53feeb-aadc-4e7d-85ec-acb82553c00a
trentmkelly/LessWrong-43k
LessWrong
Thoughts on Structuring AI Thought Introduction (post has been edited on 17.6 18:50 GMT, there's is an added part) I'd like to write my thoughts about structuring the psyche for an AI. However I am not an expert on computers, programming or AIs, nor even psychology or the human mind in general. In fact I have very ...
eaafa2cb-1336-4f22-9e5f-f2f639b4a819
StampyAI/alignment-research-dataset/lesswrong
LessWrong
Another attempt to explain UDT (Attention conservation notice: this post contains no new results, and will be obvious and redundant to many.) Not everyone on LW understands Wei Dai's [updateless decision theory](/lw/15m/towards_a_new_decision_theory/). I didn't understand it completely until two days ago. Now that I...
ba8df649-6b8f-4633-b453-3467508cc158
trentmkelly/LessWrong-43k
LessWrong
A good rational fiction about IT-inspired magic system? I'd love to read fantasy novel in which the system of magic would be inspired by the IT world: Spells are nearly free, don't require any sacrifices, nor conservation of energy - that's not the issue. The issue is: to be able to super-precisely, down to the atom,...
3734c229-53d0-4fa0-b68c-a606a780a551
trentmkelly/LessWrong-43k
LessWrong
Will releasing the weights of large language models grant widespread access to pandemic agents? Abstract Large language models can benefit research and human understanding by providing tutorials that draw on expertise from many different fields. A properly safeguarded model will refuse to provide "dual-use" insights t...
9ac71c48-040a-48c7-8fab-b532c98ecc2f
StampyAI/alignment-research-dataset/eaforum
Effective Altruism Forum
Followup on Terminator I posted my [last post](https://skluug.substack.com/p/ai-risk-is-like-terminator-stop-saying?s=w) to the [Effective Altruism Forum](https://forum.effectivealtruism.org/posts/zsFCj2mfnYZmSW2FF/ai-risk-is-like-terminator-stop-saying-it-s-not-1), where it received more attention than I’d anticipate...
8a8aa586-f47c-4cc8-a69e-2c7e8258c9c5
trentmkelly/LessWrong-43k
LessWrong
Setting up a new Mac I've had somewhat bad luck with computers lately. In late January my work computer stopped charging, and they replaced it with a used model because they were having supply issues. Its replacement had a broken webcam and, as expected for a butterfly mac, inconsistent keyboard. In April the WiFi sto...
d828e66d-3403-4994-b742-bbeb4c269eb4
trentmkelly/LessWrong-43k
LessWrong
Towards an understanding of the Chinese AI scene @thedudeabides made a post about how, during the Biden years, it was sometimes said that China would never catch up with the West in AI, that if the West agreed to pause AI development then China would too, and so on. The post called for members of the rationality commu...
61a3d5cb-7fba-4cb2-a1fa-a229d03b8075
trentmkelly/LessWrong-43k
LessWrong
Are there any methods for NNs or other ML systems to get information from knockout-like or assay-like experiments? As an example, suppose we have a bacterium, which we are treating as a black box, and we are training an ML system to predict various things it does (movement, secretions etc.) from its environment (chemi...
2290168c-95f8-4e44-b78e-24ae3fb07caa
trentmkelly/LessWrong-43k
LessWrong
How I Formed My Own Views About AI Safety Disclaimer: I work as a researcher at Anthropic, but this post entirely represents my own views, rather than the views of my own employer Introduction I’ve spent the past two years getting into the field of AI Safety. One important message I heard as I was entering the field...
ff214094-1764-4f0c-8080-055e6e4bec2e
trentmkelly/LessWrong-43k
LessWrong
Five Worlds of AI (by Scott Aaronson and Boaz Barak) A relatively short April 27, 2023 post attempting to classify AI outcomes into 5 main cases: * AI-Fizzle * Not-AI-Fizzle * Civilization recognizably continues * Futurama * AI-Dystopia * Civilization does not recognizably continue * Singularia...
3151f9a2-d7be-4f4a-8544-fc2c39fbef38
trentmkelly/LessWrong-43k
LessWrong
November 2014 Monthly Bragging Thread Your job, should you choose to accept it, is to comment on this thread explaining the most awesome thing you've done this month. You may be as blatantly proud of yourself as you feel. You may unabashedly consider yourself the coolest freaking person ever because of that awesome th...
d06811a4-ad0b-4240-ac34-14c72de5ae2f
trentmkelly/LessWrong-43k
LessWrong
Thoughts on Mustachianism I've always been a Mustachian. Recently I've been noticing myself having some diverging thoughts about it. But despite those diverging thoughts, there's still a lot of core ideas that I really identify with. Anyway, here are some assorted thoughts. He doesn't always advise frugality People...
62efee88-aa89-480e-baaa-b627f09925fe
trentmkelly/LessWrong-43k
LessWrong
Improving local governance in fragile states - practical lessons from the field Summary – Over the past 15 years, major international donor agencies shifted their approach to local politics in fragile states. In order to build stable state-society relations in conflict-affected societies, they committed to studying po...
c434cb33-d30c-42b3-9c6d-6b4189420256
trentmkelly/LessWrong-43k
LessWrong
Storytelling and the evolution of human intelligence This is a notice of a recent paper that may be of interest here, "The storytelling arms race: origin of human intelligence and the scientific mind" by Enrico Coen. It is a more specific working out of the Machiavellian Intelligence hypothesis, taking storytelling as...
66b66929-d840-46f8-822e-fc33fcaa73d9
trentmkelly/LessWrong-43k
LessWrong
Are Deontological Moral Judgments Rationalizations? In 2007, Chris Matthews of Hardball interviewed David O'steen, executive director of a pro-life organization. Matthews asked: > I have always wondered something about the pro-life movement. If you believe that killing [a fetus] is murder, why don't you bring murder ...
663a31cb-4cf7-43ab-9c81-905f774b59b8
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
The Alignment Newsletter #5: 05/07/18 **Highlights** -------------- **[AI safety via debate](https://blog.openai.com/debate/)** *(Geoffrey Irving et al)*: At a high level, a major issue with building superintelligent AI is that humans would not be able to provide strong oversight for the AI. Amplification solves this...
e8d16269-d23a-4fa5-a2f7-b71b672f262f
trentmkelly/LessWrong-43k
LessWrong
A Quick Ontology of Agreement Adapted from Herman Kahn's Introduction to Can We Win in Vietnam, Armbruster, F.E., et. al. Praeger: New York, 1968. This is a quick post about an ordered sense of agreement. I find it serves two useful purposes: 1. It gives a practical framework for how to achieve better agreement bet...
b51b0f28-7cea-4343-8ad6-1aa9b40bcec4
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
Compute Thresholds: proposed rules to mitigate risk of a “lab leak” accident during AI training runs 1. There should be two thresholds on compute graph size: 1. the Frontier threshold, beyond which oversight during execution is mandatory 2. the Horizon threshold, beyond which execution is forbidden by default 2. Ove...
3713309a-cbcd-47cf-a42f-ff9b26f55858
StampyAI/alignment-research-dataset/youtube
Youtube Transcripts
Is AI Safety a Pascal's Mugging? hi today we're going to talk about Pascal's wager and Pascal's mugging this is necessarily going to touch on religious topics pretty heavily actually and I'm just gonna say at the beginning of the video that personally I don't believe that any god or gods have ever existed and I'm not ...
eceb288d-aa65-43fa-aad8-bf770b71dd63
trentmkelly/LessWrong-43k
LessWrong
Church vs. Taskforce I am generally suspicious of envying crazy groups or trying to blindly copycat the rhythm of religion—what I called "hymns to the nonexistence of God", replying, "A good 'atheistic hymn' is simply a song about anything worth singing about that doesn't happen to be religious." But religion does fi...
38fa9bb3-c622-4010-8285-fda71e791b21
trentmkelly/LessWrong-43k
LessWrong
Quote help There's a famous I'm struggling to find but it's along the lines of: 'Those who are wise seek clarity. Those who wish to merely appear wise seek obscurity' It's great for getting a read on people and their intention. Does anyone know who said it and what the original quote is?  
fb08ec73-a622-4dcb-b9b6-64c00d033466
trentmkelly/LessWrong-43k
LessWrong
[SEQ RERUN] Make an Extraordinary Effort Today's post, Make an Extraordinary Effort was originally published on 07 October 2008. A summary (taken from the LW wiki):   > It takes an extraordinary amount of rationality before you stop making stupid mistakes. Doing better requires making extraordinary efforts. Discus...
7f4db3ed-c7a5-460c-bafa-cd78c1c1b158
StampyAI/alignment-research-dataset/arxiv
Arxiv
Reward Reports for Reinforcement Learning 1. Introduction ---------------- In fall 2021, the Wall Street Journal published the Facebook files, a series of articles that pieced together platform policy, flaws, and harms through leaked internal documents (Hagey and Horwitz, [2021](#bib.bib26)). Among other topics, th...
1f52665e-a16a-4ad4-a1ee-6c0940fd38c2
trentmkelly/LessWrong-43k
LessWrong
An interesting, novel approach to designing computer processors Here is  presentation from a researcher at MIT on a novel way of designing computer processors. It relies on performing approximate, rather than exact, mathematical operations (like the meat-based processor in our heads!). Claimed benefits are a 10,000-fo...
ba52eddb-0184-40c9-b36b-a6e4e6c89b0b
trentmkelly/LessWrong-43k
LessWrong
What are the best non-LW places to read on alignment progress? I've lately been thinking I should prioritize a bit more keeping up with alignment-relevant progress outside of LessWrong/Alignment Forum.  I'm curious if people have recommendations that stand out as reliably valuable, and/or have tips for finding "the g...
e823d2d5-c0a6-430a-acda-18b02d268d3f
StampyAI/alignment-research-dataset/lesswrong
LessWrong
Understanding and Aligning a Human-like Inductive Bias with Cognitive Science: a Review of Related Literature *Produced as part of the* [*SERI ML Alignment Theory Scholars Program*](https://serimats.org/) *with support from* [*Cavendish Labs*](https://cavendishlabs.org/)*, Many thanks go to the following for reading ...
846bd4a5-d934-461c-9ae2-d068e0043cb5
trentmkelly/LessWrong-43k
LessWrong
Open thread, Oct. 24 - Oct. 30, 2016 If it's worth saying, but not worth its own post, then it goes here. ---------------------------------------- Notes for future OT posters: 1. Please add the 'open_thread' tag. 2. Check if there is an active Open Thread before posting a new one. (Immediately before; refresh the ...
afba1f3d-1ce2-4fdf-a321-0b92fc596be1
trentmkelly/LessWrong-43k
LessWrong
Balsa FAQ The Balsa FAQ can be found here. This is a link post rather than a copy so that the FAQ can be reliably and easily updated in the future.
8c518f64-f1d2-4fdb-ae03-dab024dff981
trentmkelly/LessWrong-43k
LessWrong
Another toy model of the control problem
71219710-73fa-46b6-9a93-17972b4d65c1
trentmkelly/LessWrong-43k
LessWrong
The alien simulation meme doesn't make sense I've encountered on this site the idea that we are being simulated by aliens so the aliens can safely observe how AI develops before implementing it in their base reality-- see meme here:   That doesn't make sense, though.  Obviously we as humans have some idea that we co...
9ec6e79e-bcc5-4d77-86ce-fc3435dd83be
StampyAI/alignment-research-dataset/blogs
Blogs
Our new technical research agenda overview [![technical agenda overview](http://intelligence.org/wp-content/uploads/2014/12/technical-agenda-overview.png)](https://intelligence.org/files/TechnicalAgenda.pdf)Today we release a new overview of MIRI’s technical research agenda, “[Aligning Superintelligence with Human Int...
caa6745a-3544-4eaa-95b8-bc026fff4789
trentmkelly/LessWrong-43k
LessWrong
The AI Safety community has four main work groups, Strategy, Governance, Technical and Movement Building Epistemic status Written as a non-expert to develop and get feedback on my views, rather than persuade. It will probably be somewhat incomplete and inaccurate, but it should provoke helpful feedback and discussion....
34d08947-9e93-4293-8613-6c9925b2afe9
trentmkelly/LessWrong-43k
LessWrong
Exploring Whole Brain Emulation I've been dedicating a fair amount of my time recently to investigating whole brain emulation (WBE). As computational power continues to grow, the feasibility of emulating a human brain at a reasonable speed becomes increasingly plausible. While the connectome data alone seems insuffi...
83443722-a028-4e2c-a483-0eba9a06805b
trentmkelly/LessWrong-43k
LessWrong
FDT Does Not Endorse Itself in Asymmetric Games A twin guard-inmate dilemma (twin GID) is an asymmetric game that breaks FDT. [Image: GPT Image-1] 0. Introduction TL;DR: FDT and UDT diverge in how they handle "behave as you would have ideally precommitted to behaving" in asymmetric games where a player is assigned a...
7a932bcc-ea00-4f66-82b8-b54c24a9c0d6
trentmkelly/LessWrong-43k
LessWrong
Forecasting Compute - Transformative AI and Compute [2/4] Cross-posted here on the EA Forum. Transformative AI and Compute - A holistic approach - Part 2 out of 4 This is part two of the series Transformative AI and Compute - A holistic approach. You can find the sequence here and the summary here. This work was co...
9d699805-c66d-48db-b79c-62b3aa268c2e
trentmkelly/LessWrong-43k
LessWrong
Google announces Pathways: new generation multitask AI Architecture > Instead, we’d like to train one model that can not only handle many separate tasks, but also draw upon and combine its existing skills to learn new tasks faster and more effectively. That way what a model learns by training on one task – say, learni...
08fb6097-0380-4076-9692-cba3a73fa8f2
trentmkelly/LessWrong-43k
LessWrong
Learning as you play: anthropic shadow in deadly games Epistemic status: I'm not a mathematician, but this is simple stuff and I think I double checked all my calculations well enough. Criticism is welcome!   Some time ago a post came out claiming that "anthropic shadow is reflectively inconsistent". I want to argue...
e93d1c2d-fc61-444e-b887-0d675ac9f97a
trentmkelly/LessWrong-43k
LessWrong
Ikaxas' Hammertime Final Exam Here are my answers to Alkjash's Hammertime Final Exam. I have selected the difficulty level Steel Cudgel of the Lion. 1. Rationality Technique: Ass-in-Gear Mode This might also be called "buckling-down mode." It's the feeling you get when there is a Yoda Timer running. It's the feeling...
16e56b42-5c7f-415e-971f-4f9e4cbb4d6c
trentmkelly/LessWrong-43k
LessWrong
Strategies for keeping AIs narrow in the short term Disclaimer: I don’t have any particular expertise in AI safety, I just had some thoughts and this seemed like the place to put them.  The bleak outlook that Eliezer presents here and elsewhere seems to in large part be driven by the expectation that AGI will be dev...
18a5e334-cd97-4ad3-8562-09ef16c7b896
LDJnr/LessWrong-Amplify-Instruct
LessWrong
"Epistemic Effort: Thought seriously for 5 minutes about it. Thought a bit about how to test it empirically. Spelled out my model a little bit. I'm >80% confident this is worth trying and seeing what happens. Spent 45 min writing post.I've been pleased to see "Epistemic Status" hit a critical mass of adoption - I think...
0a66beca-9d04-454b-890c-9df902f71083
trentmkelly/LessWrong-43k
LessWrong
Meetup : Waterloo Meetup Discussion article for the meetup : Waterloo Meetup WHEN: 30 January 2012 08:00:00PM (-0500) WHERE: 4 King St N Waterloo Location: Symposium Cafe I plan to keep organizing these meetups on a weekly basis, so this should be the last irregularly scheduled meetup. The reservation is under "J...
e72cc83f-ad41-4a39-a688-a68b88ad4fc0
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
Draft report on existential risk from power-seeking AI I’ve written a draft report evaluating a version of the overall case for existential risk from misaligned AI, and taking an initial stab at quantifying the risk from this version of the threat. I’ve made the draft viewable as a public google doc [here](https://doc...
1bb795a6-27d6-48ef-b0e6-a0578779fec4
trentmkelly/LessWrong-43k
LessWrong
Hard Takeoff Continuation of:  Recursive Self-Improvement Constant natural selection pressure, operating on the genes of the hominid line, produced improvement in brains over time that seems to have been, roughly, linear or accelerating; the operation of constant human brains on a pool of knowledge seems to have prod...
fc67faf7-4ce2-4aad-96f4-cb5dd141a9da
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
[AN #147]: An overview of the interpretability landscape 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 through **[this spreadsheet](h...
3c2c2842-ec60-4809-a370-a6d0468197eb
trentmkelly/LessWrong-43k
LessWrong
How would you run the statistics on whether Ivermectin helped India reduce COVID-19 cases? I frequently read the claim that India reduced their COVID-19 cases a lot via Ivermectin but it's very unclear to me whether the relevant data to make that claim is cheripicked.  India has 34 region and each of those region has...
58a70118-e854-4687-99ce-b5bef39de9ca
trentmkelly/LessWrong-43k
LessWrong
Workflow vs interface vs implementation (Found this old draft that was more or less complete; figured I should publish it. It's adapted from this old blog post of mine.) It's commonly appreciated that when explaining how to use something, there's a distinction between the high-level usage instructions and the low-lev...
1a091cc4-78b7-4adb-bfe8-27557e36563d
trentmkelly/LessWrong-43k
LessWrong
Should you be worried about H5N1? Epistemic status: a few people without any particular expertise in epidemiology spent an afternoon in a coffee shop discussing and reading about H5N1, with a focus on how an individual should orient towards this (as opposed to say, the government). This is a write-up of what I took aw...
e1764fe5-6515-4c1b-b15a-1b8d3b264487
StampyAI/alignment-research-dataset/special_docs
Other
Multi-agent learning in mixed-motive coordination problems Multi-agent learning in mixed-motive coordination problems Julian Stastny Johannes Treutlein Maxime Riché Jesse Clifton Abstract Cooperation in settings where agents have different but overlapping preferences ( mixed-motive set- tings) has recently received con...
fb8970b0-6000-4a55-8374-59d9a49f5d88
trentmkelly/LessWrong-43k
LessWrong
Linkpost: Arena’s New Opening Hand Rule Has Huge Implications For How We Play the Game This is a link-post for my latest Magic article, posted on Channel-Fireball: Arena’s New Opening Hand Rule Has Huge Implications For How We Play the Game. Posting here is a more reliable way to engage with me than commenting there,...
a022ff7d-561e-49c6-bb45-0774d51b3bdb
trentmkelly/LessWrong-43k
LessWrong
Dodging systematic human errors in scalable oversight Summary: Both our (UK AISI's) debate safety case sketch and Anthropic’s research agenda point at systematic human error as a weak point for debate. This post talks through how one might strengthen a debate protocol to partially mitigate this. Not too many errors ...
879d5e07-5d3f-47b3-8d80-8e5e3d89bae3
StampyAI/alignment-research-dataset/lesswrong
LessWrong
Reward function learning: the value function I've written quite a few posts about the problems with agents learning values/rewards, and manipulating the learning process. I won't be linking to those past posts, because the aim of this post and the next [one](https://www.lesswrong.com/posts/upLot6eG8cbXdKiFS/reward-fun...
732b0dba-d02d-4c2c-8bb4-779eec6013c7
StampyAI/alignment-research-dataset/arbital
Arbital
Report likelihoods, not p-values This page advocates for a change in the way that statistics is done in standard scientific journals. The key idea is to report [likelihood functions](https://arbital.com/p/56s) instead of p-values, and this could have many benefits. _(Note: This page is a personal [opinion page](https...
3ef2c67b-0f87-433d-89a4-077904529224
trentmkelly/LessWrong-43k
LessWrong
Modal Bargaining Agents Summary: Bargaining problems are interesting in the case of Löbian cooperation; Eliezer suggested a geometric algorithm for resolving bargaining conflicts by leaving the Pareto frontier, and this algorithm can be made into a modal agent, given an additional suggestion by Benja. Bargaining and...
479a10fc-f9c1-4f1e-ac11-1b5ef105e3c5
trentmkelly/LessWrong-43k
LessWrong
Biohacking in New York, Cybernetics and first Cyborg Hate Crime: theverge.com http://www.theverge.com/2012/8/8/3177438/cyborg-america-biohackers-grinders-body-hackers
355182a3-296e-4e74-8a9b-c5a3b2ed10b8
trentmkelly/LessWrong-43k
LessWrong
OpenPrinciples Bootcamp (Free) -- Reflect & Act on your Rationality Principles. If you have seen OpenPrinciples Bootcamp's previous edit of the post, we have made a few changes:  1. We changed the Bootcamp from paid to free, because: 1) we found our price has greatly limited the quality and quantity of our applicants;...
d99f14b7-8c97-45cb-b778-179579c18520
StampyAI/alignment-research-dataset/eaforum
Effective Altruism Forum
Governments pose larger risks than corporations: a brief response to Grace In her article [Counterarguments to the basic AI risk case](/posts/zoWypGfXLmYsDFivk/counterarguments-to-the-basic-ai-risk-case), Katja Grace argues that the track record of corporations is a reason to doubt what she presents as the basic argum...
2e8acd5c-5af4-45cb-ba0e-f00408767059
trentmkelly/LessWrong-43k
LessWrong
[SEQ RERUN] Sensual Experience Today's post, Sensual Experience was originally published on 21 December 2008. A summary (taken from the LW wiki):   > Much of the anomie and disconnect in modern society can be attributed to our spending all day on tasks (like office work) that we didn't evolve to perform (unlike hunt...
a950df58-3bd8-4d0f-b9da-abfdd3feaf67
trentmkelly/LessWrong-43k
LessWrong
The convergent dynamic we missed An excerpt from a longer post that I kept refining over the last 5 months.   By far the greatest danger of Artificial Intelligence is that people conclude too early that they understand it. Of course this problem is not limited to the field of AI.  Jacques Monod wrote: “A curious asp...
4b4032bc-152e-4d53-b4b9-eb61b0777af6
trentmkelly/LessWrong-43k
LessWrong
[SEQ RERUN] Intelligence in Economics Today's post, Intelligence in Economics was originally published on 30 October 2008. A summary (taken from the LW wiki):   > There are a few connections between economics and intelligence, so economics might have something to contribute to a definition of intelligence. Discuss...
c8852794-3ca7-4727-aec2-4bc5897be88d
trentmkelly/LessWrong-43k
LessWrong
Alignment & Capabilities: what's the difference? In the AI safety literature, AI alignment is often presented as conceptually distinct from capabilities. However, (1) the distinction seems somewhat fuzzy and (2) many techniques that are supposed to improve alignment also improve capabilities.  (1) The distinction is ...
0b082b29-e8cb-43b2-806e-df33d7367ee9
trentmkelly/LessWrong-43k
LessWrong
My research methodology (Thanks to Ajeya Cotra, Nick Beckstead, and Jared Kaplan for helpful comments on a draft of this post.) I really don’t want my AI to strategically deceive me and resist my attempts to correct its behavior. Let’s call an AI that does so egregiously misaligned (for the purpose of this post). Mo...
9802e9f5-b48c-4cc3-a960-c277b1692941
trentmkelly/LessWrong-43k
LessWrong
Moving away from physical continuity In thought experiments about cloning objects, people often call the object which maintained physical continuity "original" and other "clone". This seems like just a naming, but when experiments talk about cloning a person, people tend to care where their original version is. This f...
bcc7112f-7710-47ef-903c-90bfc993203c
trentmkelly/LessWrong-43k
LessWrong
Meetup : Austin, TX - HPMoR Wrap Party Discussion article for the meetup : Austin, TX - HPMoR Wrap Party WHEN: 14 March 2015 05:00:00PM (-0600) WHERE: 4212 Hookbilled Kite Drive The next meetup is March 14th, 2015 at 6:00PM at a private residence (newcomers, especially from places other than Austin, are welcome!). ...
8be264b6-b138-46cb-9f0e-dcb381c168ed
trentmkelly/LessWrong-43k
LessWrong
Consistently Inconsistent Robert Kurzban's Why Everyone (Else) Is a Hypocrite: Evolution and the Modular Mind is a book about how our brains are composed of a variety of different, interacting systems. While that premise is hardly new, many of our intuitions are still grounded in the idea of a unified, non-compartment...