id
stringlengths
36
36
source
stringclasses
15 values
formatted_source
stringclasses
13 values
text
stringlengths
2
7.55M
981d0727-686c-40b1-aa00-5ee460c5b49a
StampyAI/alignment-research-dataset/blogs
Blogs
New report: “Questions of reasoning under logical uncertainty” [![Reasoning under LU](http://intelligence.org/wp-content/uploads/2015/01/Reasoning-under-LU.png)](https://intelligence.org/files/QuestionsLogicalUncertainty.pdf)Today we release a new technical report by Nate Soares and Benja Fallenstein, “[Questions of r...
ae88335f-94d0-48dc-b2a2-0b7321c33cdc
StampyAI/alignment-research-dataset/eaforum
Effective Altruism Forum
Long-Term Future Fund: Ask Us Anything! The [Long-Term Future Fund](https://app.effectivealtruism.org/funds/far-future) (LTFF) is one of the [EA Funds](https://app.effectivealtruism.org/funds). Between Friday Dec 4th and Monday Dec 7th, we'll be available to answer any questions you have about the fund – we look forwa...
e3fae5f7-a3cb-400d-bf0e-851785ec0cf1
StampyAI/alignment-research-dataset/arxiv
Arxiv
AI Research Considerations for Human Existential Safety (ARCHES) Preface ------- At the time of writing, the prospect of artificial intelligence (AI) posing an existential risk to humanity is not a topic explicitly discussed at length in any technical research agenda known to the present authors. Given that existen...
e4b535c3-3cb5-4e63-880e-507104e3b342
trentmkelly/LessWrong-43k
LessWrong
A note about calibration of confidence Background In a recent Slate Star Codex Post (http://slatestarcodex.com/2016/01/02/2015-predictions-calibration-results/), Scott Alexander made a number of predictions and presented associated confidence levels, and then at the end of the year, scored his predictions in order to...
c267b493-d277-41ba-b9ad-64a9163b308b
StampyAI/alignment-research-dataset/lesswrong
LessWrong
The "spelling miracle": GPT-3 spelling abilities and glitch tokens revisited *Work supported by the Long Term Future Fund. Thanks to Jessica Rumbelow and Joseph Bloom for useful discussions.* Introduction ============ The term "spelling miracle" was coined in Liu et al.'s December 2022 paper "[Character-aware models...
0816ff49-84d4-4bf5-9719-bcf90f967acd
trentmkelly/LessWrong-43k
LessWrong
In software engineering, what are the upper limits of Language-Based Security? Given the most security-amenable language you know, or that you could imagine building (or given some tools and processes that change the qualities of an existing language), by what proportion could we reduce the incidence rate of major sec...
0c13f8e6-102c-4f4c-9754-3e72e1e9ad78
trentmkelly/LessWrong-43k
LessWrong
Become a Superintelligence Yourself > "Crystallect" stands for the crystal transparency and order in empowering the user’s intellect in contrast to the black box of neural-network-based Artificial Intelligence. Nerds and Bullies Kids in school, as I remember it, had two ways for getting good grades: to study hard or...
7d8bca4d-f7d3-4076-8f37-0f74f78d3b28
trentmkelly/LessWrong-43k
LessWrong
Retreat for rationality meetup organizers (apply to attend!) TL;DR I'm running a retreat for meetup organizers in the Bay Area from July 21–24. More info here; you can apply using this form. ETA: The application form will be open through the end of the day on April 30th, and final decisions will go out by May 7th. -...
b4a40fcf-4756-4411-bc14-dfaf4b633622
trentmkelly/LessWrong-43k
LessWrong
Meetup : Washington D.C. Games Meetup Discussion article for the meetup : Washington D.C. Games Meetup WHEN: 06 July 2014 03:00:00PM (-0400) WHERE: National Portrait Gallery, Washington, DC 20001, USA We will be congregating in the Kogod Courtyard for games and conversation. (For those of you not familiar with the...
feb7fd6e-9cb8-4e86-9a7c-b16abfbc00b9
StampyAI/alignment-research-dataset/arxiv
Arxiv
Risk Assessment for Machine Learning Models 1 Introduction --------------- With the deployment of machine learning (ML) models in safety and security critical environments, risk assessment becomes a pressing issue. Failure modes of a given ML model must be identified and the likelihood of occurrence and severity of...
12b3d488-994a-4aa3-a7f8-f2b81e897e41
StampyAI/alignment-research-dataset/blogs
Blogs
Fiction relevant to AI futurism This page is an incomplete collection of fiction about the development of advanced AI, and the consequences for society. Details ------- Entries are generally included if we judge that they contain enough that is plausible or correctly evocative to be worth considering, in light of...
a4b35b5c-e872-4b1c-a0a1-742fd12041ea
trentmkelly/LessWrong-43k
LessWrong
Dennett on the selfish neuron, etc. Dennett: > Mike Merzenich sutured a monkey's fingers together so that it didn't need as much cortex to represent two separate individual digits, and pretty soon the cortical regions that were representing those two digits shrank, making that part of the cortex available to use for ...
65796040-8ad8-4e1d-a559-6b7be455256b
trentmkelly/LessWrong-43k
LessWrong
reflections on smileys and how to make society's interpretive priors more charitable Hi : ) I used to use smileys in my writing all the time (more than I do now!).  but then I read Against Disclaimers, and I thought that every time I used a smiley I wud make people who don't use smileys seem less friendly (bc my cons...
faf7213d-2f00-43fd-8bd1-e90bc58e9e4b
trentmkelly/LessWrong-43k
LessWrong
The many ways AIs behave badly EDIT: This has been previously posted here. Vika is now maintaining a centralized list of such examples. I had a previous post about some of the ways AIs behave badly. But now there is a new paper, looking at many examples of (mis)behaviour, within the evolutionary programming design. A...
36702ed4-f8a8-4ce2-8bdb-28d9932af144
trentmkelly/LessWrong-43k
LessWrong
Goodhart Taxonomy
38e27099-6795-4f92-a4fc-9fc4438940ce
trentmkelly/LessWrong-43k
LessWrong
The curse of identity > So what you probably mean is, "I intend to do school to improve my chances on the market". But this statement is still false, unless it is also true that "I intend to improve my chances on the market". Do you, in actual fact, intend to improve your chances on the market? > > I expect not. Rath...
04df6e14-cd2c-4d24-9efc-3208e4766304
StampyAI/alignment-research-dataset/lesswrong
LessWrong
Why I think strong general AI is coming soon I think there is little time left before someone builds AGI (median ~2030). Once upon a time, I didn't think this. This post attempts to walk through some of the observations and insights that collapsed my estimates. The core ideas are as follows: 1. We've already captur...
7ed824d6-8b44-41ee-bb6f-4f65f5a2cbd7
trentmkelly/LessWrong-43k
LessWrong
To Question God "So you'll become God?" the man queried. His speaking finalizing his creation. "You could say that." The as of now disembodied voice answered. "Where are we?" The man pressed, having noticed the distinct lack of general vicinity surrounding him. "It's more of a when than a where. To be more specific...
97445f04-ee1d-41c1-b681-970f029ed62b
trentmkelly/LessWrong-43k
LessWrong
[Part-time AI Safety Research Program] MARS 3.0 Applications Open for Participants & Recruiting Mentors [Crossposted from EA Forum] The Cambridge AI Safety Hub (CAISH) is now accepting applications for MARS 3.0 (Mentorship for Alignment Research Students), a volunteer-based program connecting aspiring researchers wit...
fe138366-d417-4ef8-8cea-9be80ec5b052
StampyAI/alignment-research-dataset/eaforum
Effective Altruism Forum
If your AGI x-risk estimates are low, what scenarios make up the bulk of your expectations for an OK outcome? There seem to be two main framings emerging from recent AGI x-risk discussion: default doom, given AGI, and default we're fine, given AGI. I'm interested in what people who have low p(doom|AGI) think are the ...
c7fe7292-da15-4728-bc8b-aaf9e0f2cb83
trentmkelly/LessWrong-43k
LessWrong
Against Expected Utility Expected utility is optimal as the number of bets you take approaches infinity. You will lose bets on some days, and win bets on other days. But as you take more and more bets, the day to day randomness cancels out. Say you want to save as many lives as possible. You can plug "number of lives...
e9692c26-8685-46a4-af6b-e2bd6fd3afa2
trentmkelly/LessWrong-43k
LessWrong
Welcome to Twin Cities SSC Meetup [Edit With Your Details] (The following are our suggestions for what kind of information is best to include in the welcome post of your group, feel free to replace them with whatever you think is best) What kind of events does your group usually run? What does it usually do? How fre...
95cd4c80-c1f0-4c4b-969a-1f66147eb84c
trentmkelly/LessWrong-43k
LessWrong
The LW crossroads of purpose Hello everyone reading,  I'm writing this because I do not believe fixing peripheral things on LW is enough. And even though I wrote about it in a comment, I wanted to elaborate and highlight the issue through writing a post about it too. Even though this is not nearly as succinctly writt...
ff0f17ef-a0d7-44fe-894a-2764e1be8b23
trentmkelly/LessWrong-43k
LessWrong
Wikifying the blog list Konkvistador's excellent List of Blogs by LWers led me to some of my favorite blogs, but is pretty well hidden and gradually becoming obsolete. In order to create an easily-update-able replacement, I have created the wiki page List of Blogs and added most of the blogs from Konkvistador's list. ...
5eb37bf6-87a1-40a0-a30e-89fb04ed9810
trentmkelly/LessWrong-43k
LessWrong
D&D.Sci(-Fi) June 2021 Evaluation and Ruleset This is a followup to the D&D.Sci post I made last week; if you haven’t already read it, you should do so now before spoiling yourself. Here is the web interactive I built to let you evaluate your solution; below is an explanation of the rules used to generate the dataset...
e5f057bf-dbbe-4945-90c4-d6d6fbada767
trentmkelly/LessWrong-43k
LessWrong
Intent alignment as a stepping-stone to value alignment I think Instruction-following AGI is easier and more likely than value aligned AGI, and that this accounts for one major crux of disagreement on alignment difficulty. I got several responses to that piece that didn't dispute that intent alignment is easier, but a...
2fe0a5b1-ad68-4b0c-a4ba-64544b641797
trentmkelly/LessWrong-43k
LessWrong
[link] How many humans will have their brain preserved? Forecasts and trends http://lessdead.com/how-many-humans-will-have-their-brain-preserved-forecasts-and-trends Summary: > Doubling time for the number of people that got cryopreserved has been pretty consistently 9 years since the beginning.
fc2883de-09e3-4370-8170-10f4bb6973d3
trentmkelly/LessWrong-43k
LessWrong
How to Teach Students to Not Guess the Teacher’s Password? As a teacher, I wonder if it is possible to instill this skill into students the skills of rationality and critical thinking. I teach the third grade, and it is not immediately apparent how to apply this with my own class. The problems I foresee are as follow...
5d2be90f-5423-4d2a-be47-15b5bf3a796f
StampyAI/alignment-research-dataset/lesswrong
LessWrong
Minerva Google Research's new AI tackles natural language math problems and handily outperforms the SOTA[[1]](#fnz9ujz7hobka). It is a pre-trained PaLM [[2]](#fnjhan4auwcm)finetuned on some maths datasets (which use LaTeX) composed of maths webpages and Arxiv papers (38.5B tokens). The three models trained were as fol...
445e5210-dd7c-4e16-bbab-2e66a60f2db5
trentmkelly/LessWrong-43k
LessWrong
Are you the rider or the elephant? Some recent threads seem to me to be pointing at a really fundamental tension that I don't know how to articulate in full. But here's a chunk of it: When you think of your mind as divided into your System 1 / fast / unconscious / nonverbal / intuitive / emotional processes, which we...
a349296c-4d73-4311-875c-76a39e7f4e54
trentmkelly/LessWrong-43k
LessWrong
Will AI and Humanity Go to War? [This post is the introduction to my full paper, available here https://philpapers.org/rec/GOLWAA. This post was partially inspired by a LW comment thread between @Matthew Barnett and @Wei Dai.] Abstract. This paper offers the first careful analysis of the possibility that AI and human...
f86782a4-e8a9-4b44-a49d-35b07c78a1b7
trentmkelly/LessWrong-43k
LessWrong
Eukaryote Skips Town - Why I'm leaving DC I’ve spent the past 7 years living in the DC area. I moved out there from the Pacific Northwest to go to grad school – I got my masters in Biodefense from George Mason University, and then I stuck around, trying to move into the political/governance sphere. That sort of happen...
9b8e3b14-b373-47c2-ab2d-2e328c63b317
trentmkelly/LessWrong-43k
LessWrong
Meetup : Cambridge, MA Sunday meetup: Lightning Talks Discussion article for the meetup : Cambridge, MA Sunday meetup: Lightning Talks WHEN: 21 May 2017 03:30:00PM (-0400) WHERE: Harvard Science Center, Room 105 IMPORTANT NOTE: The meetup is NOT at the usual location. Instead it is at the Harvard Science Center, ro...
91fc7438-8bc2-40f5-9964-9636fd48f7b9
trentmkelly/LessWrong-43k
LessWrong
AI: 4 levels of impact [micropost] 1. LLM is as big as the smartphone/electricity: it will be the building block on which a lot of tech gets built. 2. AI is the new industrial revolution/agricultural revolution: it will allow for a whole new level of automation of the economy. 3. AGI is the new macro-optimisation p...
f54cd20e-56a8-49ed-8a39-7bca449b24b9
trentmkelly/LessWrong-43k
LessWrong
Straightforward Steps to Marginally Improve Odds of Whole Brain Emulation I want to make a small disclaimer that I have no formal training in most of this. I am pretty confident about my understanding of electrophysiology and some robotics but much less so about anything biochem/genetics related. If there are any mist...
071e9a98-d79c-4bdf-a2b0-acc4ea5a9c0a
trentmkelly/LessWrong-43k
LessWrong
Meetup : San Jose Meetup: Park Day (X) Discussion article for the meetup : San Jose Meetup: Park Day (X) WHEN: 04 December 2016 03:00:00PM (-0800) WHERE: 438 Coleman Ave, San Jose, CA 95110 You're invited to bring whatever you'd like to see at Park Day - balls, frisbees, food, etc. (Also, consider taking measures a...
f0ef13a2-2ce2-4ff1-8f3e-c0e158efb68a
trentmkelly/LessWrong-43k
LessWrong
World models I'm currently building Update: This is a living document. Posted below is an older version. Link has the latest version. 2024-12-26   This doc is a mix of existing world models I have and holes in said models. I'm trying to fill some of these holes. The doc is not very well organised relative to how or...
903730fd-4eb1-413e-8d95-cb1a5ebd90a6
trentmkelly/LessWrong-43k
LessWrong
Demand offsetting For the last few years I’ve been avoiding factory farmed eggs because I think they involve a lot of unnecessary suffering. I’m hesitant to be part of that even if it’s not a big deal on utilitarian grounds. This is a pain since factory-farmed eggs are used all over the place (e.g. in ice cream, pastr...
353b20ea-cfe8-4628-9c7a-8516bd246a5b
StampyAI/alignment-research-dataset/eaforum
Effective Altruism Forum
Jobs that can help with the most important century Let’s say you’re convinced that AI could make this the [most important century of all time for humanity](https://forum.effectivealtruism.org/s/isENJuPdB3fhjWYHd). What can you do to help things go well instead of poorly? I think **the biggest opportunities come fr...
077ecb82-76ca-4972-bcfa-854551a0ca1f
trentmkelly/LessWrong-43k
LessWrong
Above the Narrative Cross-posted from Putanumonit.com ---------------------------------------- Sometimes you write about a thing and that thing… happens. I spent all of February working on a post about the mainstream narrative of American society: who gets to tell it, what happens when that narrative is challenged,...
be353f29-f6db-4bd7-9a5f-cd6c145f923a
trentmkelly/LessWrong-43k
LessWrong
How rationality can make your life more awesome I'm currently working with Lukeprog on a crash course in rationality. It's essentially a streamlined version of the Sequences, but one area we want to beef up is the answer to the question, "Why learn about rationality?" I've gone through all of the previous threads I c...
b8e5a0a2-e7e1-4b03-9ad1-999d7c80c79e
StampyAI/alignment-research-dataset/arxiv
Arxiv
Artificial Intelligence Safety and Cybersecurity: a Timeline of AI Failures Artificial Intelligence Safety and Cybersecurity : a Timeline of AI Failures Roman V. Yampolskiy Computer Engineering and Computer Science University of Louisville roman.yampolskiy@louisville.edu M. S. Spellchecker Microsoft Co...
fde899aa-151b-41e7-b0b4-616ce9b335f7
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
Introduction To The Infra-Bayesianism Sequence .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: nowra...
77b3b483-9924-4078-9ae1-3c25c39edf5f
trentmkelly/LessWrong-43k
LessWrong
Remembering school math fondly This is what I think about the recently linked article about kids not distinguishing 'math' problems where the actual question has nothing to do with anything else, as a bewildered learner and not a resigned tutor. Epistemic status - talkative. Most math beyond arithmetic in our school ...
986f9651-b475-43d4-b0b9-a015740f5f05
trentmkelly/LessWrong-43k
LessWrong
The many faces of status The term "status" gets used on LessWrong a lot. Google finds 316 instances; the aggregate total for the phrases "low status" and "high status" (which suggest more precision than "status" by itself) is 170. By way of comparison, "many worlds", an important topic here, yields 164 instances. We ...
9f91fec0-e286-4a41-810e-cac1c22993e9
trentmkelly/LessWrong-43k
LessWrong
Long-lasting Effects of Suspensions? I recently read "The School to Prison Pipeline: Long-Run Impacts of School Suspensions on Adult Crime" (Bacher-Hicks et. al. 2019, pdf, via Rob Wiblin) which argues that a policy of suspending kids in middle school leads to more crime as an adult. Specifically, they found that aft...
feb818f0-3bea-4438-ba0f-f812bc4c1844
trentmkelly/LessWrong-43k
LessWrong
[SEQ RERUN] Occam's Razor Today's post, Occam's Razor was originally published on 26 September 2007. A summary (taken from the LW wiki):   > To a human, Thor feels like a simpler explanation for lightning than Maxwell's equations, but that is because we don't see the full complexity of an intelligent mind. However, ...
e0d98d8e-bd8e-4906-ad14-24896be78ae2
trentmkelly/LessWrong-43k
LessWrong
The non-tribal tribes > Author note: This is basically an Intro to the Grey Tribe for normies, and most people here are already very familiar with a lot of the info herein. I wasn't completely sure I should post it here, and I don't expect it to get much traction, but I'll share it in case anyone's curious. Introduc...
1fdd7a85-cc72-473f-842f-14dc295281ba
StampyAI/alignment-research-dataset/arbital
Arbital
Rational numbers: Intro (Math 0) *In order to get the most out of this page, you probably want a good grasp of the [integers](https://arbital.com/p/53r) first.* "Rational number" is a phrase mathematicians use for the idea of a "fraction". Here, we'll go through what a fraction is and why we should care about them. ...
4f4c1f8b-87b3-431a-ba6c-8f188eb9c5a1
trentmkelly/LessWrong-43k
LessWrong
Directly advertising existential risk Has anyone tried advertising existential risk? Bostroms "End of Humanity" talk for instance. It costs about 0.2 $ per view for a video ad on YouTube, so if 0.2% of viewers give an average of 100 $ it would break even. Hopefully people would give more than that. You can target a...
aa80e8d4-8fa8-4904-bbc4-ef3c2be9f309
trentmkelly/LessWrong-43k
LessWrong
Superhero Bias Suppose there’s a heavily armed sociopath, a kidnapper with hostages, who has just rejected all requests for negotiation and announced his intent to start killing. In real life, the good guys don’t usually kick down the door when the bad guy has hostages. But sometimes—very rarely, but sometimes—life im...
0e7c8075-9cb1-4567-9a09-c721525f4241
trentmkelly/LessWrong-43k
LessWrong
Subsuming Purpose, Part II: Solving the Solution Summary: It's easy to get caught up in solving the wrong problems, solving the problems with a particular solution instead of solving the actual problem.  You should pay very careful attention to what you are doing and why. I'll relate a seemingly purposeless story abo...
15899e31-5057-4185-90e1-30857667b67e
trentmkelly/LessWrong-43k
LessWrong
Minneapolis Meetup: Survey of interest Frank Adamek and I are going to host a Less Wrong/Overcoming Bias meetup tentatively on Saturday September 26 at 3pm in Coffman Memorial Union at the University of Minnesota (there is a coffee shop and a food court there). Frank is the president of the University of Minnesota tra...
1effd9e6-9d6c-490a-9b64-cefc95dc6ed8
trentmkelly/LessWrong-43k
LessWrong
Resist the Happy Death Spiral Once upon a time, there was a man who was convinced that he possessed a Great Idea. Indeed, as the man thought upon the Great Idea more and more, he realized that it was not just a great idea, but the most wonderful idea ever. The Great Idea would unravel the mysteries of the universe, su...
34d406c0-4313-4fad-afee-995aacf7a60b
trentmkelly/LessWrong-43k
LessWrong
How would public media outlets need to be governed to cover all political views? As one of their questions Reporters without Borders asks for their World Press Freedom Index among others: > Do public media outlets cover all political views?* > > Does the law provide mechanisms to guarantee pluralism and editorial in...
536297db-4017-4cd8-bf48-69c7ae46a080
StampyAI/alignment-research-dataset/blogs
Blogs
"Moral progress" vs. the simple passage of time In [Future-Proof Ethics](https://www.cold-takes.com/future-proof-ethics/), I talked about trying to "consistently [make] ethical decisions that look better, with hindsight after a great deal of moral progress, than what our peer-trained intuitions tell us to do." I ...
be725583-4242-40be-8abd-c0cbdf959bed
trentmkelly/LessWrong-43k
LessWrong
Metrics in Everything: “Human Lives” Epistemic Status: Ranting with the fire of a thousand suns I was on page 48 of the (so far) otherwise interesting and enjoyable Algorithms to Live By, a birthday gift from my friend Jacob who writes the blog Put a Num on It, in which the authors Brian Christian and Tom Griffiths w...
dc47d70e-f686-4369-9d0b-60a50389026a
StampyAI/alignment-research-dataset/eaforum
Effective Altruism Forum
AI-Relevant Regulation: Insurance in Safety-Critical Industries ### Preface This post is part of a series exploring existing approaches to regulation that seem relevant for thinking about governing AI. The goal of this series is to provide a brief overview of a type regulation or a regulatory body so others can unde...
e7a1d627-9dc7-4acf-aef4-c430bbb9b9fc
trentmkelly/LessWrong-43k
LessWrong
What are the Activities that make up your Research Process? There are a bunch of activities that I engage in when doing research. These include but are not limited to: * Figuring out the best thing to do. * Talking out loud to force my ideas into language. * For the last 3 months I have been working maybe 50 hou...
3c6c9cb3-c39a-462e-b870-9080ae5757e4
StampyAI/alignment-research-dataset/special_docs
Other
WHAT IS THE UPPER LIMIT OF VALUE? WHAT IS THE UPPER LIMIT OF VALUE ? Anders Sandberg Future of Humanity Institute University of Oxford Suite 1, Littlegate House 16/17 St. Ebbe’s Street, Oxford OX1 1PT anders.sandberg@philosophy.ox.ac.ukDavid Manheim 1DaySooner Delaware, United States, davidmanheim@gmail.com January ...
f2ac0a70-cfa2-4377-89fa-91478639bdea
trentmkelly/LessWrong-43k
LessWrong
Comments are back! Instead of hosting comments on jefftk.com directly, I copy over publicly-accessible comments from discussion elsewhere: By default it's first-name only when copying from social media (Facebook, Google Plus) and full name when copying from forums (LessWrong, EA Forum), though I have it always use m...
19cca678-e0dc-4c4d-8658-c4ad9e3407d1
trentmkelly/LessWrong-43k
LessWrong
Eugenics And Reproduction Licenses FAQs: For the Common Good 1. What Is Eugenics? There are various ways to define eugenics, as it is used in this FAQs: * Eugenics is the deliberate social regulation of the genome. * Eugenics is about consciously selecting for traits that make people better members of society. * E...
248742eb-9d0e-4396-9631-4c896fafd4fe
trentmkelly/LessWrong-43k
LessWrong
Three consistent positions for computationalists Yesterday, as a followup to We are not living in a simulation, I posted Eight questions for computationalists in order to obtain a better idea of what exactly my computationalist critics were arguing.  These were the questions I asked: > 1. As it is used in the senten...
f6899314-65f5-4071-b90f-081e81db0898
StampyAI/alignment-research-dataset/lesswrong
LessWrong
Notes on Simplicity This post examines the virtue of **simplicity**. It is meant mostly as an exploration of what others have learned about this virtue, rather than as me expressing my own opinions about it, though I’ve been selective about what I found interesting or credible, according to my own inclinations. I wrot...
81801f4a-b308-4a6a-8668-7b1fabc30db6
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
Speculative inferences about path dependence in LLM supervised fine-tuning from results on linear mode connectivity and model souping *TL;DR: I claim that supervised fine-tuning of the existing largest LLMs is likely path-dependent (different random seeds and initialisations have an impact on final performance and mod...
55dfb68a-eb82-4044-bba5-1932f8f02e4a
trentmkelly/LessWrong-43k
LessWrong
Hiding misinformation in meanings I It is hard to spread misinformation, because information spreads too, and they eventually run into each other and explode. If a person wants to lie then, they can be better off to make words correspond to different things for different people, so that even when people hear the inf...
8d93e2d0-0165-4421-a1de-3a086e55ae4e
trentmkelly/LessWrong-43k
LessWrong
AutoBound on neural network can achieve OOMs lower training loss Seems like gradient descent methods weren't using the relevant math bounds so far. Google released AutoBound as an open-source library.  Here is what I consider a money shot of the article (notice it's a log-plot): Performance of SafeRate when used to ...
8363499b-1408-4141-a6f7-9ea010691525
trentmkelly/LessWrong-43k
LessWrong
Muehlhauser-Hibbard Dialogue on AGI Part of the Muehlhauser series on AGI. Luke Muehlhauser is Executive Director of the Singularity Institute, a non-profit research institute studying AGI safety. Bill Hibbard is an emeritus senior scientist at University of Wisconsin-Madison and the author of Super-Intelligent Mach...
184a2888-d8ad-410e-ace8-3c435e769608
StampyAI/alignment-research-dataset/arxiv
Arxiv
Challenges for Using Impact Regularizers to Avoid Negative Side Effects 1 Introduction --------------- Specifying a reward function in reinforcement learning (RL) that completely aligns with the designer’s intent is a difficult task. Besides specifying what is important to solve the task at hand, the designer also ...
a53c0f24-3022-44b1-a7e8-b61b3e5e1aa0
trentmkelly/LessWrong-43k
LessWrong
What's Going on With OpenAI's Messaging? This is a quickly-written opinion piece, of what I understand about OpenAI. I first posted it to Facebook, where it had some discussion.    Some arguments that OpenAI is making, simultaneously: 1. OpenAI will likely reach and own transformative AI (useful for attracting ta...
2efdc41e-22be-44cf-9921-c0e27c2d5e5a
trentmkelly/LessWrong-43k
LessWrong
Young Cryonicists Conference 2011 Next week I'll be attending the second annual Young Cryonicists conference, which Eliezer attended before writing Normal Cryonics. I expected there to be discussion about it on lesswrong, but there hasn't been, so here it is. * Who else is going? Is EY going?  * What do you expect ...
8ad085f6-1c48-4f7b-ad9f-802dca7b56f5
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
EIS IX: Interpretability and Adversaries Part 9 of 12 in the [Engineer’s Interpretability Sequence](https://www.alignmentforum.org/s/a6ne2ve5uturEEQK7). Thanks to Nikolaos Tsilivis for helpful discussions.  The studies of interpretability and adversaries are inseparable. =============================================...
c0718ef2-a7cd-4408-b6be-3db378988d4b
trentmkelly/LessWrong-43k
LessWrong
How well did Manifold predict GPT-4? Chat GPT-4 is already here!! Who could have seen that coming… oh wait Manifold (kinda) did? 😅 I thought I’d write a short piece on how Manifold Markets was used to predict the launch of GPT-4 and its attributes. Both its successes and its failures. Disclaimer I work at Manifold. ...
2129a1c6-d33d-4eb7-a03d-22c026fe0be3
trentmkelly/LessWrong-43k
LessWrong
Interpreting Neural Networks through the Polytope Lens Sid Black*, Lee Sharkey*, Leo Grinsztajn, Eric Winsor, Dan Braun, Jacob Merizian, Kip Parker, Carlos Ramón Guevara, Beren Millidge, Gabriel Alfour, Connor Leahy *equal contribution   Research from Conjecture.   This post benefited from feedback from many staf...
5c27b693-62b9-432b-8445-4e79797352b2
StampyAI/alignment-research-dataset/arbital
Arbital
Distant superintelligences can coerce the most probable environment of your AI A distant superintelligence can change 'the most likely environment' for your AI by simulating many copies of AIs similar to your AI, such that your local AI doesn't know it's not one of those simulated AIs. This means that, e.g., if there...
e6c1b363-ca03-4a55-a4ad-04996db9b2f8
trentmkelly/LessWrong-43k
LessWrong
Detachment vs attachment [AI risk and mental health] > "What? Why would I choose cosmic energy over Katara?" > > Avatar the Last Airbender   "You idiot" said the monk from atop his mountain, "there's all this fresh air up here and the view is breathtaking and you're down there dredging mud ." The mud-dredger only g...
9fc8ebbb-6a71-4525-a4ca-ee1ea90dd440
trentmkelly/LessWrong-43k
LessWrong
Stopping dangerous AI: Ideal lab behavior Disclaimer: this post doesn't have the answers. Moreover, it's an unfinished draft. Hopefully a future version will be valuable, but that will only occur if I revise/rewrite it. For now you're better off reading sources linked from Ideas for AI labs: Reading list and Slowing A...
62a8eeb7-dc8a-4cf6-95c1-a5358cc53212
StampyAI/alignment-research-dataset/special_docs
Other
Energy, Complexity, and the Singularity Chapter 8 Energy, Complexity, and the Singularity Kent A. Peacock The technology hype cycle for a paradigm shift —railroads, AI, Internet, telecommunica- tions, possibly now nanotechnology —typically starts with a period of unrealistic expec- tations based on a lack of understan...
5bc15cf7-68aa-484e-b304-e49d71112c9d
StampyAI/alignment-research-dataset/blogs
Blogs
AI Timeline Surveys ***This page is out-of-date. Visit the [updated version of this page](https://wiki.aiimpacts.org/doku.php?id=ai_timelines:predictions_of_human-level_ai_timelines:ai_timeline_surveys:ai_timeline_surveys) on our [wiki](https://wiki.aiimpacts.org/doku.php?id=start).*** *Published 10 January 2015* ...
45baa013-101d-4953-88a5-e7a404226c79
StampyAI/alignment-research-dataset/eaforum
Effective Altruism Forum
New cooperation mechanism - quadratic funding without a matching pool *cross-posted to [ethresear.ch](https://ethresear.ch/t/quadratic-funding-without-a-matching-pool/12792) and [LessWrong](https://www.lesswrong.com/posts/GpiwJcFdobZT9aHFm/quadratic-funding-without-a-matching-pool-1)* Motivation ========== Quadrat...
065f7181-460e-4387-b497-b91b9ea39ba2
trentmkelly/LessWrong-43k
LessWrong
AI Performance on Human Tasks This was an independent project with guidance and support from Vael Gates and Richard Korzekwa--thank you both for your mentorship! Introduction This project was inspired by a prompt from the AI Impacts project: > For intellectual skills where machines have surpassed humans, find out h...
9e755ea5-ad1e-4a67-9a5b-8688974f4ab9
StampyAI/alignment-research-dataset/eaforum
Effective Altruism Forum
Technology is Power: Raising Awareness Of Technological Risks Technology puts power in consumers’, companies’, and politicians’ hands. How we do that in a safe and responsible manner is a complex problem. How can we raise awareness, rally support from all levels of society, so we can put in the proper measures, and sa...
d9986b21-0c34-494b-a3b9-396d5b0e5671
trentmkelly/LessWrong-43k
LessWrong
Homogeneity vs. heterogeneity in AI takeoff scenarios Special thanks to Kate Woolverton for comments and feedback. There has been a lot of work and discussion surrounding the speed and continuity of AI takeoff scenarios, which I do think are important variables, but in my opinion ones which are relatively less import...
c74efdf7-1098-49e2-a5c1-fff528d12433
trentmkelly/LessWrong-43k
LessWrong
[SEQ RERUN] Evolutions Are Stupid (But Work Anyways) Today's post, Evolutions Are Stupid (But Work Anyway) was originally published on 03 November 2007. A summary (taken from the LW wiki):   > Evolution, while not simple, is sufficiently simpler than organic brains that we can describe mathematically how slow and st...
5c2f557f-105f-49b3-bb7b-5ca57ce8a9f3
trentmkelly/LessWrong-43k
LessWrong
For-Profit Rationality Training As I've been reading through various articles and their comments on Less Wrong, I've noticed a theme that has appeared repeatedly: a frustration that we are not seeing more practical benefits from studying rationality. For example, Eliezer writes in A Sense that More Is Possible, > Why...
462c1ad3-d798-436b-9e23-436572834b4e
trentmkelly/LessWrong-43k
LessWrong
Crypto autopsy reply (X-posted from my FB post: https://www.facebook.com/alexei.andreev.3/posts/1403550339754401) Reply to Eliezer's post on crypto: https://www.facebook.com/yudkowsky/posts/10156147605134228 Which itself is a response to Scott Alexander's post: https://www.lesserwrong.com/…/Ma…/a-lesswrong-crypto-aut...
8321b628-bb69-4784-b817-910decd52a0f
StampyAI/alignment-research-dataset/arbital
Arbital
Complexity theory: Complexity zoo Welcome, visitor! Get ready to exercise your sense of wonder and expand your mind on the **zoo of complexity**. That where are the cages of the animals, you ask? No cage could contain our specimens, I must say. But you can get a glimpse of their nature through the chalky definitions ...
48d633f5-70eb-4c51-90fc-2ccae9d9d008
trentmkelly/LessWrong-43k
LessWrong
Meetup : Social Skills Discussion article for the meetup : Social Skills WHEN: 24 September 2014 07:00:00PM (-0400) WHERE: 98 Elm Street, Somerville MA Come hear the amazing and charismatic Sam Rosen speak! Cambridge/Boston-area Less Wrong Wednesday meetups are once a month on the last Wednesday at 7pm at Citadel (...
3b5a85eb-b57e-494e-90ad-aee29b3f8301
StampyAI/alignment-research-dataset/arxiv
Arxiv
The Benchmark Lottery 1 Introduction --------------- Quantitative evaluation is a cornerstone of machine learning research. As a result, benchmarks, including those based on data sets and simulations, have become fundamental to tracking the progress of machine learning research. Benchmarks have a long history i...
37fa62c3-d234-40da-895e-05a47bd202df
StampyAI/alignment-research-dataset/arxiv
Arxiv
Towards Understanding Sycophancy in Language Models 1 Introduction --------------- AI assistants such as GPT-4 (OpenAI, [2023](#bib.bib6 "GPT-4 technical report")) are typically trained to produce outputs that humans rate highly, e.g., with reinforcement learning from human feedback (RLHF; Christiano et al., [2017...
4ad26d3f-dfd7-4af6-8732-cb35ee8d31d7
StampyAI/alignment-research-dataset/blogs
Blogs
The Hanson-Yudkowsky AI-Foom Debate is now available as an eBook! [![ai-foom-cover](https://intelligence.org/wp-content/uploads/2013/09/ai-foom-cover-206x300.jpg)](http://intelligence.org/ai-foom-debate/ "The Hanson-Yudkowsky AI-Foom Debate eBook")In late 2008, economist Robin Hanson and AI theorist Eliezer Yudkowsky ...
ad147131-d480-47f3-af9d-026fdc2b370a
trentmkelly/LessWrong-43k
LessWrong
[LINK] Nick Bostrom: What happens when our computers get smarter than we are? (TED talk) http://www.ted.com/talks/nick_bostrom_what_happens_when_our_computers_get_smarter_than_we_are
2dbfb636-a831-4102-8732-4e0c909455c5
trentmkelly/LessWrong-43k
LessWrong
Computerphile discusses MIRI's "Logical Induction" paper None
f3cf101c-54e1-45fd-bc8f-ebebad2e9a86
trentmkelly/LessWrong-43k
LessWrong
We are in a New Paradigm of AI Progress - OpenAI's o3 model makes huge gains on the toughest AI benchmarks in the world This is the full text of a post from "The Obsolete Newsletter," a Substack that I write about the intersection of capitalism, geopolitics, and artificial intelligence. I’m a freelance journalist and ...
6d152ba9-3f4a-4dd8-abc4-e5421e744ced
trentmkelly/LessWrong-43k
LessWrong
Instrumental rationality/self help resources I took part in a recent discussion in the current Open Thread about how instrumental rationality is under-emphasized on this website. I've heard other people say similar things, and I am inclined to agree. Someone suggested that there should be a "Instrumental Rationality ...
b867e918-9a02-434a-895b-8c9bc48edeee
StampyAI/alignment-research-dataset/eaforum
Effective Altruism Forum
Still no strong evidence that LLMs increase bioterrorism risk <https://www.lesswrong.com/posts/ztXsmnSdrejpfmvn7/propaganda-or-science-a-look-at-open-source-ai-and> Linkpost from LessWrong. The claims from the piece which I most agree with are: 1. Academic research does not show strong evidence that existing LLMs i...
9feedf91-a903-4b23-9427-170e5a125ace
trentmkelly/LessWrong-43k
LessWrong
A brief note on factoring out certain variables Jessica Taylor and Chris Olah has a post on "Maximizing a quantity while ignoring effect through some channel". I'll briefly present a different way of doing this, and compare the two. Essentially, the AI's utility is given by a function U of a variable C. The AI's acti...
f5ae7f19-e55b-4f40-bb5d-4c50dfd3412b
trentmkelly/LessWrong-43k
LessWrong
Paper: The Capacity for Moral Self-Correction in Large Language Models (Anthropic) This is a followup to what I cheekily call Anthropic's "just try to get the large model to do what you want" research agenda. (Previously: A General Language Assistant as a Laboratory for Alignment, Training a Helpful and Harmless Assis...
f5df3f04-173f-4e98-a816-f7fe95424d25
StampyAI/alignment-research-dataset/eaforum
Effective Altruism Forum
How to become an AI safety researcher *Reminder that you can listen to this post and other EA Forum posts on your podcast player with* [*The Nonlinear Library*](https://pod.link/1587343144)*.* What skills do you need to work on AI safety? And what can we learn from the paths people have taken into the field? We were ...
37553667-9262-4fea-824c-9e01b184b2ce
StampyAI/alignment-research-dataset/eaforum
Effective Altruism Forum
Stuart J. Russell on "should we press pause on AI?" By coincidence (I think?) this episode was released during the [AI Pause debate week](https://forum.effectivealtruism.org/s/vw6tX5SyvTwMeSxJk/p/6SvZPHAvhT5dtqefF) on the forum, so I felt like it was worth adding to the mix. The episode has general LMM/AI chat unit ~...