id stringlengths 36 36 | source stringclasses 15
values | formatted_source stringclasses 13
values | text stringlengths 2 7.55M |
|---|---|---|---|
2a7af559-4ead-476e-b20c-69f54feab5d9 | StampyAI/alignment-research-dataset/blogs | Blogs | New paper: “Logical induction”
[](https://arxiv.org/abs/1609.03543)MIRI is releasing a paper introducing a new model of deductively limited reasoning: “[**Logical induction**](https://arxiv.org/abs/1609.03543),” authored by S... |
b3a79f59-2600-4137-a799-25d04290eed1 | trentmkelly/LessWrong-43k | LessWrong | Meetup : Edinburgh summer meetup
Discussion article for the meetup : Edinburgh summer meetup
WHEN: 13 July 2012 01:00:00PM (+0100)
WHERE: Upstairs at Biblos, 1 Chambers Street, Edinburgh EH1 1HU
It's the summer months, many of the students are away and we have not had a proper Edinburgh meetup in a while. We will p... |
b0b70b2c-a30b-4b3c-b5ad-baa137adfbd7 | trentmkelly/LessWrong-43k | LessWrong | Take over my project: do computable agents plan against the universal distribution pessimistically?
This post is for technically strong people looking for a way to contribute to agent foundations (more detailed pros and cons later). I have not tried as hard as usual to make it generally accessible because, you know, I... |
76e2f04c-b919-48fa-bdfb-6734de6e0715 | trentmkelly/LessWrong-43k | LessWrong | Summary of "AGI Ruin: A List of Lethalities"
Overview
Introduction
This post is a summary of Eliezer Yudkowsky’s post “AGI Ruin: A List of Lethalities”. I wrote it because I think the original post is longer and less organized than I would like it to be. The purpose of this post is to summarize the main points in th... |
4e69856d-c001-4060-aa95-f98762564108 | trentmkelly/LessWrong-43k | LessWrong | Am I going insane or is the quality of education at top universities shockingly low?
I'm fortunate enough to go to a high-caliber American university. I study math and economics, so not fields that are typically subject to funding constraints or have some shortage of experts. The incentives to become a professor here ... |
a00dd853-7390-414d-a98c-61fd6646588a | trentmkelly/LessWrong-43k | LessWrong | Chasing Infinities
Crossposted from my blog Passing Time
How do we decide how to make decisions about the future?
How we, as a society, decide to answer this question is one of the most important moral quandaries we must face. All too often, questions of morality are focused on the actions themselves in question. T... |
750e3dc8-67bf-436a-9ea1-13a336692a3f | trentmkelly/LessWrong-43k | LessWrong | The Compleat Cybornaut
A cluster of conceptual frameworks and research programmes have coalesced around a 2022 post by janus, which introduced language models as ‘simulators’ (of other types of AIs such as agents, oracles, or genies). One such agenda, cyborgism, was coined in a post by janus and Nicholas Kees and is ... |
1093200a-81c0-4087-9a50-ed1037ff9b16 | StampyAI/alignment-research-dataset/arxiv | Arxiv | A Benchmark for Interpretability Methods in Deep Neural Networks
1 Introduction
---------------
In a machine learning setting, a question of great interest is estimating the influence of a given input feature on the prediction made by a model. Understanding what input features are important helps improve our models... |
a16995ab-e87c-4da3-8d78-3203b4d06fd1 | trentmkelly/LessWrong-43k | LessWrong | Evaluating Life Extension Advocacy Foundation
(Crossposted from the EA Forum)
Summary
In this post, I try to evaluate Life Extension Advocacy Foundation, and I flesh out interview questions to ask them. LEAF has many foci, including crowdfunding, conference organizing, investor advisement, social media advocacy, new... |
c8530cba-c032-41b0-8a08-e5e4581f1554 | trentmkelly/LessWrong-43k | LessWrong | Fractional Reserve Charity
On the EA Corner discord server, various participants, most notably @wolframhead, tossed around the following idea: Some people of an EA bent may want to donate vast amounts of money, huge percentages of high salaries, but be held back in case they turn out to need that money as savings late... |
32871b25-f8b5-4857-899c-4b7ff9f3be7d | StampyAI/alignment-research-dataset/eaforum | Effective Altruism Forum | Speedrun: AI Alignment Prizes
Introduction
------------
This post is a shallow investigation of the intervention of running a highly-advertised, high-budget AI alignment prize contest. The post is part of a sequence of speedrun research projects by Rethink Priorities’ general longtermism team. I recommend you begin b... |
0fedf672-19a8-47a5-9d98-dedd40073415 | trentmkelly/LessWrong-43k | LessWrong | Summaries of top forum posts (24th - 30th April 2023)
We've just passed the half year mark for this project! If you're reading this, please consider taking this 5 minute survey - all questions optional. If you listen to the podcast version, we have a separate survey for that here. Thanks to everyone that has responded... |
38df6f18-6bc4-4acf-aa16-e709ad266d96 | trentmkelly/LessWrong-43k | LessWrong | Alignment Research Field Guide
This field guide was written by the MIRI team with MIRIx groups in mind, though the advice may be relevant to others working on AI alignment research.
⠀
Preamble I: Decision Theory
Hello! You may notice that you are reading a document.
This fact comes with certain implications. For i... |
4f03cc93-fba8-47aa-bd55-4fade07ac460 | StampyAI/alignment-research-dataset/lesswrong | LessWrong | Institutions Cannot Restrain Dark-Triad AI Exploitation
On how governance institutional designs cannot restrain corporate, criminal and political groups of humans who gain increasing power from training and exploiting increasingly more capable – eventually power-seeking – machine learning architectures.
As elaboratin... |
7d1c9bb1-2131-4258-9393-a9b0ff1a151a | trentmkelly/LessWrong-43k | LessWrong | Psyched out
I don't know where to begin as I've apparently spent my whole life working at being a Rationalist and didn't know it until I heard about this site and the Rationalist Movement. As a Starving Artist/Amateur Intellectual/Hermit, I'm not in a position to go to graduate school right now, and the people around ... |
7d4dae3c-70e6-4188-b137-dc189d94d849 | StampyAI/alignment-research-dataset/arbital | Arbital | Patch resistance
A proposed [foreseeable difficulty](https://arbital.com/p/6r) of [aligning advanced agents](https://arbital.com/p/2v) is furthermore proposed to be "patch-resistant" if the speaker thinks that most simple or naive solutions will fail to resolve the difficulty and just regenerate it somewhere else.
To... |
a8a0deeb-c66d-42ea-94d9-75d5e9f9c841 | trentmkelly/LessWrong-43k | LessWrong | Self-driving car bets
This month I lost a bunch of bets.
Back in early 2016 I bet at even odds that self-driving ride sharing would be available in 10 US cities by July 2023. Then I made similar bets a dozen times because everyone disagreed with me.
The first deployment to potentially meet our bar was Phoenix in 202... |
3fc133f9-c61d-4f80-989e-ae023d47c74b | trentmkelly/LessWrong-43k | LessWrong | Translation "counterfactual"
Crossposted at Intelligent Agent Forum
In a previous post, I briefly mentioned translations as one of three possible counterfactuals for indifference. Here I want to clarify what I meant there, because the idea is interesting.
Imagine that there is a button, which, if a human presses it ... |
91034a92-0b06-4065-a3a8-d6efedca774b | trentmkelly/LessWrong-43k | LessWrong | Cryonics is Quantum Suicide (minus the suicide)
EDIT: I was confused. Confusion now resolved. Disregard this (I don't think it's possible to retract it, and I don't want to delete it in case I wasn't the only confused person).
The frequently discussed quantum lottery thought experiment proposes that a group of people... |
5e31f29c-ec7a-450d-bf05-c3020a9d833c | trentmkelly/LessWrong-43k | LessWrong | Meetup : Stockholm: Trigger-Action Plans
Discussion article for the meetup : Stockholm: Trigger-Action Plans
WHEN: 14 October 2016 05:00:00PM (+0200)
WHERE: Lindstedtsvägen 3 room 1537, 114 28 Stockholm, Sweden
In this off week meetup David_Kristoffersson will hold a workshop about TAPs and mental contrasting.
Eve... |
12dcf4dc-de05-4698-a12a-dbdb9e437d8d | StampyAI/alignment-research-dataset/special_docs | Other | AI Definitions Affect Policymaking
JUNE 2020 AI Definitions Affect Policymaking CSET Issue Brief
AUTHORS Dewey Murdick James Dunham Jennifer Melot
Center for Security and Emerging Technology | 2 Definitions of artificial intelligence (AI) are often ambiguous and quickly out of sync with such a rapidly emerging tec... |
085aacf6-9f0f-4831-a1da-6ce8da6e3458 | StampyAI/alignment-research-dataset/eaforum | Effective Altruism Forum | Career Advice: Philosophy + Programming -> AI Safety
Hi forum,
I'm a philosophy graduate from Oxford who's been working as a programmer for five years. I'm missing some of the required mathematical background right now, but I think I could be a good fit for AI safety research. I'm trying to figure out the next steps... |
4a7b1bdd-e7bc-4ae8-9981-c689ff931c29 | trentmkelly/LessWrong-43k | LessWrong | AGI Farm
This post discusses Joe Carlsmith’s views on how to approach the problem of AI risk as interspecies interaction and how humans can use it navigate future AI development better. The essay is divided into three parts. First I give my understanding of Carlsmith's views, then I build upon some of his ideas by rel... |
6447322b-4277-4093-a4b1-f396845749a0 | trentmkelly/LessWrong-43k | LessWrong | A computational complexity argument for many worlds
The following is an argument for a weak form of the many-worlds hypothesis. The weak form I mean is that there are many observers in different branches of the wave function. The other branches "actually exist" for anthropic purposes; some observers are observing them... |
236189da-298c-494b-9a92-7062edfa7731 | StampyAI/alignment-research-dataset/arxiv | Arxiv | Towards Resilient Artificial Intelligence: Survey and Research Issues
Towards Resilient Artificial Intelligence:
Survey and Research Issues
Oliver Eigner, Sebastian Eresheim, Peter Kieseberg, Lukas Daniel Klausner,
Martin Pirker, Torsten Priebe, Simon Tjoa
Institute of IT Security Research
St. Pölten University of App... |
010d40bf-f3da-41bc-b8f4-a9bc3e564a25 | trentmkelly/LessWrong-43k | LessWrong | A math question
Is (4^^^4)/(3^^((3^^3)^^3)) larger than one?
I need to know for a game of Nomic |
90322852-8df5-4e4e-a082-e2aedbf52c2d | trentmkelly/LessWrong-43k | LessWrong | Bookmarklet to Hide Nested Comments
When reading comments on Less Wrong, I sometimes find myself reading reply after reply deep into a discussion when I really shouldn't be. If I had stopped and thought, I would have said, let's move on to the next thread. Similarly, I've seen comments that pose an interesting questio... |
cd12b946-6327-4ee8-9822-cd1515ef7253 | trentmkelly/LessWrong-43k | LessWrong | The Codex Skeptic FAQ
Most of my programmer friends believe that Language Models trained on code will not affect their day job anytime soon. In this post, I make the case that 1) code generation is already useful (assuming minimal prompt engineering skills) 2) even if you do not believe in 1), code generation will inc... |
f582969b-b9ee-413c-8bb9-97e94345e5dd | trentmkelly/LessWrong-43k | LessWrong | Meetup : Zürich social Rationality meetup
Discussion article for the meetup : Zürich social Rationality meetup
WHEN: 29 October 2016 05:00:00PM (+0200)
WHERE: Elle'n'Belle Limmatstrasse 118, 8005 Zürich, Switzerland
There seem to be quite a few new Rationality interested people in Zürich now. Come and meet us! No p... |
f506b788-8663-4f29-8b08-d51131e2dd18 | awestover/filtering-for-misalignment | Redwood Research: Alek's Filtering Results | id: e3e7a3bb18518340
Four places where you can put LLM monitoring — AI Alignment Forum This website requires javascript to properly function. Consider activating javascript to get access to all site functionality. AI ALIGNMENT FORUM AF Login AI Control AI Frontpage 29 Four places where you can put LLM monitoring by Fa... |
758528e2-e805-4662-ad29-8a49566c3be0 | trentmkelly/LessWrong-43k | LessWrong | Reward is not Necessary: How to Create a Compositional Self-Preserving Agent for Life-Long Learning
There is an idea in the AI safety community that expected utility maximisation is the "limit" of agency.
I think the linked paper by Thomas J. Ringstrom demonstrates that this idea is wrong.
Explainer Twitter Thread b... |
c728891b-def8-482a-8d2d-e45309e24592 | trentmkelly/LessWrong-43k | LessWrong | How to Build Your Post
LWers often wonder how to write a good post. Many great tips and resources have been suggested already, mostly on the style and substance, not on the post structure. This is an attempt to fill some of the void using a construction toy metaphor: nested state-explain-summarize blocks.
First, ther... |
f48077e3-8e3c-40cc-becf-b4a68efb774d | trentmkelly/LessWrong-43k | LessWrong | Explanation of Paul's AI-Alignment agenda by Ajeya Cotra
Ajeya from OpenPhil wrote a very understandable and quite compelling summary of Paul's views on AI-Alignment.
(@Paul & @Ajeya: I would love to crosspost the whole thing here if you are open to that) |
a9649283-bc7d-4858-9be7-56f0ab8b9571 | trentmkelly/LessWrong-43k | LessWrong | Why would we think artists perform better on drugs ?
Introduction
It is common knowledge that many artists have used drugs (alcohol, opiates, cannabis, LSD, ...) and that this account for part of their creativity. This common knowledge is usually opposed to people advocating rationality in sentences like "but with onl... |
2a61b749-b45b-4af6-8cb1-7a1d5dd9a69c | StampyAI/alignment-research-dataset/lesswrong | LessWrong | Instead of technical research, more people should focus on buying time
*This post is the first in a sequence of posts about AI strategy co-authored by Thomas Larsen, Akash Wasil, and Olivia Jimenez (TAO). In the next post, we’ll provide more examples of “buying time” interventions that we’re excited about.*
*We’re gr... |
593f5ccb-0633-400f-8e96-d656022a9299 | StampyAI/alignment-research-dataset/blogs | Blogs | Research topic: Hardware, software and AI
This is the first in a sequence of articles outlining research which could help forecast AI development.
---
Interpretation
--------------
Concrete research projects are in boxes. ∑5 ∆8 means we guess the project will take (very) roughly five hours, and we rate its va... |
7312d8bd-b285-46a4-accb-fd96c65a7442 | StampyAI/alignment-research-dataset/aisafety.info | AI Safety Info | How can I do conceptual, mathematical, or philosophical work on AI alignment?
## General advice
There isn’t a standard career path in this area. AI alignment is a pre-paradigmatic field in which nobody has a good idea what the right prerequisite knowledge is or what an answer looks like. That means this is a path for... |
62b76d53-aece-47ec-961f-829c82fd169b | StampyAI/alignment-research-dataset/eaforum | Effective Altruism Forum | A personal take on longtermist AI governance
Several months ago, I summarized Open Philanthropy's work on AI governance (which I lead) for a general audience [here](https://www.openphilanthropy.org/blog/ai-governance-grantmaking). In this post, I elaborate my thinking on AI governance in more detail for people who are... |
d59d738e-ac30-4635-9d0e-22d7104fbb9c | trentmkelly/LessWrong-43k | LessWrong | Mega Meetup : Summer Festival, hosted by New York
Discussion article for the meetup : New York Summer Festival Megameetup
WHEN: Saturday, June 23rd 2012, 1 PM (as well as additional meetups the surrounding days)
WHERE: Central Park, New York, NY
The weekend of June 23rd, the New York Less Wrong community will be ho... |
d0970c6c-6f9b-4b9a-8d39-472a29104c5c | StampyAI/alignment-research-dataset/arxiv | Arxiv | Shaking the foundations: delusions in sequence models for interaction and control
1 Introduction
---------------
Suppose we are given a probabilistic model P that captures a joint distribution over a stochastic sequence X1,X2,…,XT. We can use this probabilistic model to make sequential predictions: if we have obser... |
5b9057cd-30f3-4fb2-ac50-859d112d2db9 | trentmkelly/LessWrong-43k | LessWrong | Is AI Hitting a Wall or Moving Faster Than Ever?
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 the author of a forthcoming book called Obsolete: Power, Profit, and th... |
8cec6507-06ba-4477-828f-f58a470de9dd | trentmkelly/LessWrong-43k | LessWrong | [SEQ RERUN] I Heart CYC
Today's post, I Heart CYC was originally published on December 1, 2008. A summary:
> Douglas Lenat has a theory which Hanson finds plausible. Lenat argues that architechture of AI is overrated, and that there is only so much you can do before you start trying to generate content for an AI.
... |
fc9f6446-6c49-4112-9927-6bf24e203273 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | The Game Theory of Blackmail
This blog post is composed as following:
1. Review of Prisoners Dilemma
2. Explanation of Game of Chicken by comparing it to Prisoners Dilemma
3. Blackmail is a Game of Chicken
4. Why we should care about blackmail/Game of Chicken
5. What to do? Iterated Game of Chicken?
You are encourag... |
e5303dab-af5e-4ef0-9bbc-2550b9ddefbc | trentmkelly/LessWrong-43k | LessWrong | Elite Coordination via the Consensus of Power
This post is about how implicit coordination between powerful people allows them to act in surprisingly synchronized ways. I’ll start by discussing wokeness, the most prominent recent example. I’ll then analyze the mechanism behind such coordination (which I call the conse... |
a98ccc99-0760-4ae4-b1b4-33ea55eeb159 | trentmkelly/LessWrong-43k | LessWrong | Do simulacra dream of digital sheep?
This is the third in a sequence of posts scrutinizing computational functionalism (CF). In a previous post, I defined a concrete claim that computational functionalists tend to make:
> Theoretical CF: A simulation of a human brain on a computer, with physics perfectly simulated do... |
a05de753-8eb7-46c4-8b79-10690652db4f | trentmkelly/LessWrong-43k | LessWrong | GPT-4o Sycophancy Post Mortem
Last week I covered that GPT-4o was briefly an (even more than usually) absurd sycophant, and how OpenAI responded to that.
Their explanation at that time was paper thin. It didn’t tell us much that we did not already know, and seemed to suggest they had learned little from the incident.... |
d41794a5-1437-4927-b0f5-ed5b4a7ed359 | trentmkelly/LessWrong-43k | LessWrong | Is This Lie Detector Really Just a Lie Detector? An Investigation of LLM Probe Specificity.
Abstract
Whereas previous work has focused primarily on demonstrating a putative lie detector’s sensitivity/generalizability[1][2], it is equally important to evaluate its specificity. With this in mind, I evaluated a lie dete... |
5ff74aa7-5364-49b9-a37b-5360aaf755aa | trentmkelly/LessWrong-43k | LessWrong | The case for the death penalty
Followed By: The case for corporal punishment
Epistemic status: this is an attempt to steelman the case for the death penalty rather than produce a balanced analysis, or even accurately represent my views (the case is presented as stronger than I actually feel).
In a sufficiently wealt... |
3d758fbb-75c2-4b95-a426-d95caed7ba36 | trentmkelly/LessWrong-43k | LessWrong | Open Thread, May 11 - May 17, 2015
If it's worth saying, but not worth its own post (even in Discussion), 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 b... |
7fd78266-6ef5-4e23-ad99-b8ac6ceaa06c | trentmkelly/LessWrong-43k | LessWrong | Bad lessons learned from the debate
> Debates are life, the rest is just time to prepare.
As a student, I played classic policy Karl Popper debates. The first thing we learned was sparring. The rules are simple: two people are given a topic, such as "an orange is better than an apple", and they randomly determine who... |
fa065e8f-47d7-4171-af55-18b23626bdc1 | trentmkelly/LessWrong-43k | LessWrong | Mathematical Models of Progress?
I would be interested in collecting a bunch of examples of mathematical modeling of progress. I think there are probably several of these here, but I don't expect to be able to find all of them myself. I'm also interested to know about any models like this elsewhere.
I was reading the... |
788bf966-ff06-4cf6-8a70-756edca547be | trentmkelly/LessWrong-43k | LessWrong | Matt Clancy AMA on the Progress Forum
Matt Clancy created the best resource on the economics of innovation for a general audience: New Things Under the Sun, where he writes accessible summaries of the academic literature on the topic. He’s a research fellow at Open Philanthropy and previously was with the Institute fo... |
1b4d6ad2-d4cb-4eb4-b675-7e7c48e01c47 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | [AN #87]: What might happen as deep learning scales even further?
Find all Alignment Newsletter resources [here](http://rohinshah.com/alignment-newsletter/). In particular, you can [sign up](http://eepurl.com/dqMSZj), or look through this [spreadsheet](https://docs.google.com/spreadsheets/d/1PwWbWZ6FPqAgZWOoOcXM8N_tUC... |
18b9b98a-9eb1-428d-acd1-632670364fba | StampyAI/alignment-research-dataset/arxiv | Arxiv | Adversarial Training for High-Stakes Reliability
1 Introduction
---------------
Advances in deep learning have led to increasingly powerful AI systems, for example in sequential decision making Silver et al. ([2017](#bib.bib2 "Mastering chess and shogi by self-play with a general reinforcement learning algorithm"))... |
2b5f76b8-cc6e-4af4-82d4-42f946a6c32b | trentmkelly/LessWrong-43k | LessWrong | Productivity - List Notch system
Original post: http://bearlamp.com.au/productivity-list-notch-system/
----------------------------------------
This is a write up of my current to do list system. My system and the method of this write up is based on Mark Forster's to do lists. If you are familiar with The Final V... |
1a60a030-127c-4b54-8f5e-a316bdd973db | trentmkelly/LessWrong-43k | LessWrong | My preferred framings for reward misspecification and goal misgeneralisation
TL;DR
* I was feeling quite dissatisfied with a bunch of categories like “inner and outer misalignment” and “reward misspecification and goal misgeneralisation”
* I think most of this dissatisfaction stems from them not being able to co... |
cf14b13d-4064-45b7-ac91-5eb4f24851ce | StampyAI/alignment-research-dataset/special_docs | Other | An automatic method for discovering rational heuristics for risky choice.
An automatic method for discovering rational heuristics for risky choice
Falk Lieder1(falk.lieder@berkeley.edu)
Paul M. Krueger1(pmk@berkeley.edu)
Thomas L. Griffiths (tom griffiths@berkeley.edu)
Department of Psychology,University of California B... |
58b4e283-239e-4e59-9f97-431e15257907 | trentmkelly/LessWrong-43k | LessWrong | What makes a scientific fact 'ripe for discovery'?
The existence of multiple discovery seems to suggest that there are certain factors that make scientific facts ready to be discovered. What are these, factors, and how could one measure them? |
a36ce931-c9cb-4946-ae81-99ba515031d0 | StampyAI/alignment-research-dataset/special_docs | Other | CHAI Newsletter #3 2019
5/4/22, 3:34 PM CHAI Newsletter + Internship/Operations Assistant Openings
https://mailchi.mp/f50d06cf0756/chai-newsletter -internshipoperations-assistant-openings 1/7Human Compatible is available now!
Stuart Russell's new book, Human Compatible: Artificial Intelligence and
the Problem of Contro... |
0432e50f-f66c-4872-a2e0-cf2190dd5728 | trentmkelly/LessWrong-43k | LessWrong | The Pre-Historical Fallacy
One fallacy that I see frequently in works of popular science -- and also here on LessWrong -- is the belief that we have strong evidence of the way things were in pre-history, particularly when one is giving evidence that we can explain various aspects of our culture, psychology, or persona... |
6149432d-b542-44ae-9534-959eecc3b919 | StampyAI/alignment-research-dataset/arxiv | Arxiv | European Strategy on AI: Are we truly fostering social good?
1 Introduction
---------------
In the last years, the development and deployment of Artificial Intelligence (AI) has grown dramatically and the European Commission (EC) expects that by 2025 the economic impact will reach between 6.5 and 12 trillion annual... |
ced2662d-0377-43e9-9c42-605a355676ec | trentmkelly/LessWrong-43k | LessWrong | Hot take: The AI safety movement is way too sectarian and this is greatly increasing p(doom)
The movement to reduce AI x-risk is overly purist. This is leading to a lot of sects to maintain each individual sect's platonic level of purity and is actively (greatly) harming the cause.
How the Safety Sects Manifest
*... |
2f5f217a-7d24-40d6-a7c6-70c089439c77 | trentmkelly/LessWrong-43k | LessWrong | Reward hacking is becoming more sophisticated and deliberate in frontier LLMs
Something’s changed about reward hacking[1] in recent systems. In the past, reward hacks were usually accidents, found by non-general, RL-trained systems. Models would randomly explore different behaviors and would sometimes come across unde... |
f3397660-1e7c-4b22-a519-46be22fae453 | trentmkelly/LessWrong-43k | LessWrong | [RXN#7] Russian x-risks newsletter fall 2020
Coronavirus
While the covid continues to make record numbers in Russia, many Russian transhumanists participated in the clinical trial or Russian Sputnik-5 vaccine or of the Chinese vaccine. They developed antibodies, didn’t fail ill and have mild adverse effects. Unfortun... |
e63cce78-460c-49e1-be27-75228bb14d19 | trentmkelly/LessWrong-43k | LessWrong | Imagining Scarcity
Sequence summary: This is a series of 18 articles on the most fundamental concepts of economics: scarcity, opportunity cost, marginalism, and self-interest. These are the atoms, molecules, cells, the core things you need to have a grip on to move on with the science, and, if all goes well, we will m... |
71b02120-8fbf-46cf-9baa-e42cee6710b4 | trentmkelly/LessWrong-43k | LessWrong | [Closed] Prize and fast track to alignment research at ALTER
On behalf of ALTER, I am pleased to announce a prize of 50,000 USD, to be awarded for the best substantial contribution to the learning-theoretic AI alignment research agenda among those submitted before October 1, 2023. Depending on the quality of submissio... |
d9c59395-be55-426b-8c82-9e96c70ad014 | trentmkelly/LessWrong-43k | LessWrong | More on the Linear Utility Hypothesis and the Leverage Prior
This is the followup to “Against the Linear Utility Hypothesis and the Leverage Prior” that I had promised in the comments on that post. Apologies that it took me long enough that the previous post is no longer fresh in your mind even if you read it.
Lever... |
bcf2dfc2-dab9-4678-b563-4caf87009726 | trentmkelly/LessWrong-43k | LessWrong | What does it mean for correct operation to rely on transfer learning?
Summary: Some approaches to AI value alignment rely on transfer learning. I attempt to explain this idea more clearly.
----------------------------------------
Paul and I have written previously about how some approaches to AI value alignment rely... |
c800c76d-9f30-440c-8d8c-4d0b6df73e3a | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | List of resolved confusions about IDA
AI Alignment is a confusing topic in general, but even compared to other alignment topics, [IDA](https://arxiv.org/abs/1810.08575) seems especially confusing. Some of it is surely just due to the nature of communicating subtle and unfinished research ideas, but other confusions ca... |
96a4b6c3-5771-4fea-b540-092d63b9be3f | trentmkelly/LessWrong-43k | LessWrong | Climate-contingent Finance, and A Generalized Mechanism for X-Risk Reduction Financing
Summary
Climate adaptation (reducing vulnerability to future climate change) could yield significant benefits. However, the uncertainty of which future climate scenarios will occur decreases the feasibility of proactively adapting. ... |
3872417f-3d92-49e4-b282-b6b3d92d357b | trentmkelly/LessWrong-43k | LessWrong | Book Suggestion: "Diaminds" is worth reading (CFAR-esque)
The reason for this submission is that I don't think anyone who visits this website will ever read the book described below, otherwise. And that's a shame.
Simply stated, I think CFAR curriculum designers and people who like CFAR's approach should check out th... |
07049ad1-d839-4ca6-bc4c-ae60d00d9b5a | trentmkelly/LessWrong-43k | LessWrong | Internal and external criticism
I notice that if you want to persuade me away from a position, it sometimes works to have me talk with two kinds of people: 1) people who have good reasons for disagreeing with my position, and 2) people who agree with my position for similar reasons and hold it even more strongly than ... |
21ae6660-bf71-45ee-84f6-79b17f67c144 | trentmkelly/LessWrong-43k | LessWrong | How are critical thinking skills acquired? Five perspectives
Link to source: http://timvangelder.com/2010/10/20/how-are-critical-thinking-skills-acquired-five-perspectives/
Previous LW discussion of argument mapping: Argument Maps Improve Critical Thinking, Debate tools: an experience report
How are critical thinking... |
a2adcd26-ebd8-4b95-8634-6201ced46405 | trentmkelly/LessWrong-43k | LessWrong | What if LaMDA is indeed sentient / self-aware / worth having rights?
Most pundits ridicule Blake Lemoine and his claims that LaMDA is sentient and deserves rights.
What if they're wrong?
The more thoughtful criticisms of his claims could be summarized as follows:
* The presented evidence (e.g. chatbot listings) ... |
5de93bc8-d060-48c9-bba6-2133cedb6121 | trentmkelly/LessWrong-43k | LessWrong | Strategies for Personal Growth
Recently I was having a conversation with a friend about personal growth (any form of deliberate gain in capability or subjective wellbeing).
We were talking past each other a lot. Eventually it became clear that most of their recent growth had been healing and blocker-fixing based, whe... |
98ec7c86-569e-424e-b1f4-bacfe8e869d3 | trentmkelly/LessWrong-43k | LessWrong | Decimal digit computations as a testing ground for reasoning about probabilities
Points in this article emerged from a conversation with Anna Salamon
I think that thinking about decimal expansions of real numbers provides a good testing ground for one's intuition about probabilities. The context of computation is ver... |
ccc11a5e-70b2-404e-8185-e1106c3c5fa3 | trentmkelly/LessWrong-43k | LessWrong | Training Regime Day 22: Murphyjitsu 2
Introduction
Society has many pieces of wisdom floating around that are epistemically false, but instrumentally useful.
Individuals are told to "invest, then don't touch the money until you retire". "Invest, then don't touch it" is false because sometimes, some individuals will b... |
27683cd2-0493-43cc-a698-07fffb7c271a | StampyAI/alignment-research-dataset/agisf | AGI Safety Fund | AI Chips: What They Are and Why They Matter
APRIL 2020 AI Chips: What They Are and Why They Matter An AI Chips Reference
AUTHORS Saif M. Khan Alexander Mann
Center for Security and Emerging Technology | 2 Table of Contents Introduction and Summary 3 The Laws of Chip Innovatio... |
28164b23-ed1d-495d-9f5d-668a8cb70c4b | trentmkelly/LessWrong-43k | LessWrong | A free to enter, 240 character, open-source iterated prisoner's dilemma tournament
I'm running an iterated prisoner's dilemma tournament where all programs are restricted to 240 characters maximum. The exact rules are posted in the Manifold Markets link; I figured I'd cross-post the contest here to reach more potentia... |
7bf4c906-3ff8-4a66-ae9c-6ca9510ea3e6 | trentmkelly/LessWrong-43k | LessWrong | Complexity Is Bad
Previously: Choices are Bad, Change Is Bad
Meta Note: Early drafts were too complex. Complexity is Bad. I simplified.
Mark Rosewater, head of design at Magic: The Gathering who I recently praised for writing down his process, has a podcast called Drive to Work. Commuting is really bad (citation not... |
28b5b937-e53e-4337-b801-aa088c562f18 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Squeezing foundations research assistance out of formal logic narrow AI.
Suppose you have a ML model trained to output formal proofs. Maybe you start with ZFC and then add extra tokens for a range of common concepts. (along with definitions. ). So a human mathematician needs to type in the definition of a gradient in ... |
4b086f23-f35a-4cbf-aa2f-c195eb2d7605 | trentmkelly/LessWrong-43k | LessWrong | Censoring out-of-domain representations
(An idea from a recent MIRI research workshop; similar to some ideas of Eric Drexler and others. Would need further development before it's clear whether this would do anything interesting, let alone be a reliable part of a taskifying approach.)
If you take an AI capable of pre... |
82a24b38-e262-49ec-9207-a4232b48243b | StampyAI/alignment-research-dataset/arbital | Arbital | Edge instantiation
The edge instantiation problem is a hypothesized [patch-resistant problem](https://arbital.com/p/48) for [safe](https://arbital.com/p/2l) [value loading](https://arbital.com/p/) in [advanced agent scenarios](https://arbital.com/p/2c) where, for most utility functions we might try to formalize or tea... |
0865540c-3abe-4394-becb-c15e83286d83 | trentmkelly/LessWrong-43k | LessWrong | How do you reason about how many COVID test kits to keep on hand?
With the release of 4 free test kits per residential address in the US (https://special.usps.com/testkits), I'm trying to think through how many tests it's appropriate and useful for me to keep on hand.
Normally a "don't take too many from a limited su... |
40cc281b-5c4e-422f-8425-8f06765002d1 | trentmkelly/LessWrong-43k | LessWrong | Meta-execution
This post describes meta-execution, my current proposal for capability amplification and security amplification.
(Meta-execution is annotated functional programming + strong HCH + a level of indirection. It is implemented in the amplify module of my ALBA repository.)
Goal
We are given an efficient ag... |
4404276b-4764-4ec0-b05f-b8029005da9d | StampyAI/alignment-research-dataset/youtube | Youtube Transcripts | What can AGI do? I/O and Speed
hi a long time ago in computerphile we
introduced the stamp collector this
hypothetical super intelligence which
just has crazy extreme powers by magic
effectively it's a thought experiment
then we brought things back down to
earth and started looking at this paper
concrete problems in A... |
297e6036-061c-4017-9c39-3a89f0bd468e | StampyAI/alignment-research-dataset/eaforum | Effective Altruism Forum | How to store human values on a computer
One of the main obstacles to building safe and aligned AI is that we don't know how to store human values on a computer.
Why is that?
Human values are abstract feelings and intuitions that can be described with words.
For example:
* Freedom or Liberty
* Happiness or Welfare
... |
dd258e5d-762d-4741-a198-5821a6a13cfb | trentmkelly/LessWrong-43k | LessWrong | On the 2nd CWT with Jonathan Haidt
It was clear within the first ten minutes this would be a rich thread to draw from. In my childhood and education roundups, and of course with my own kids, I have been dealing with the issues Haidt talks about in his new book, The Anxious Generation. Ideally I’d also have read the bo... |
050ba930-053e-44d0-8123-2ecf6cfd6fd0 | trentmkelly/LessWrong-43k | LessWrong | Future of Life Institute existential risk news site
I'm excited to announce that the Future of Life Institute has just launched an existential risk news site!
The site will have regular articles on topics related to existential risk, written by journalists, and a community blog written by existential risk researchers... |
cbd4c616-c100-49b3-af23-391689594812 | trentmkelly/LessWrong-43k | LessWrong | Meetup : Durham/RTLW HPMoR discussion, ch. 82-85
Discussion article for the meetup : Durham/RTLW HPMoR discussion, ch. 82-85
WHEN: 24 August 2013 12:00:00PM (-0400)
WHERE: 726 Rigsbee Avenue, Durham NC, 27701
Meet at Fullsteam for discussion of HPMoR 82-85!
Bring a question, observation, or topic for discussion --... |
3e800ce2-b911-44ac-ac6d-05e07b55ccb7 | StampyAI/alignment-research-dataset/blogs | Blogs | RL Unplugged: Benchmarks for Offline Reinforcement Learning
Many of the successes of RL rely heavily on rep... |
963a0bf5-9911-437f-9502-a2d803f013ee | trentmkelly/LessWrong-43k | LessWrong | Redundant Attention Heads in Large Language Models For In Context Learning
In this post, I claim a few things and offer some evidence for these claims. Among these things are:
* Language models have many redundant attention heads for a given task
* In context learning works through addition of features, which are l... |
e9587741-b22f-4411-ab7e-1fd017fec95e | trentmkelly/LessWrong-43k | LessWrong | What self-help has helped you?
Simply, what self-help techniques have you tried and found to help you? Bonus points if you can say something about the context in which you did the self-help and any speculations you have about mechanism of action.
Please give one technique per answer, multiple answers accepted per per... |
e796d39c-8d9c-4fd0-a222-7b9b839c1702 | trentmkelly/LessWrong-43k | LessWrong | Notes on Empathy
This post examines empathy, as part of a sequence of posts about virtues. It is mostly an exploration of what other people have learned about empathy, rather than my own research or opinions about it, though I’ve been selective about what I found interesting or credible. I wrote this not as an expert ... |
e187c7c8-7e6d-475e-b6b8-520c69cb2d8b | trentmkelly/LessWrong-43k | LessWrong | Training goals for large language models
This post was written under Evan Hubinger’s mentorship, as part of the Stanford Existential Risks Initiative ML Alignment Theory Scholars (SERI MATS) program. Many of the ideas in this post, including the main idea behind the training goal, are due to Kyle McDonell and Laria Re... |
b4c9fb48-d90b-4380-81d6-f1eb12cff692 | StampyAI/alignment-research-dataset/lesswrong | LessWrong | Logical Uncertainty as Probability
This post is a long answer to [this](/lw/b91/should_logical_probabilities_be_updateless_too/65tr) comment by cousin\_it:
>
> Logical uncertainty is weird because it doesn't exactly obey the rules of probability. You can't have a consistent probability assignment that says axioms ... |
0fdc8f15-5ca8-4964-b117-5368debf3fd7 | StampyAI/alignment-research-dataset/eaforum | Effective Altruism Forum | Constructive Discussion and Thinking Methodology for Severe Situations including Existential Risks
**My Methodology for Thinking Further and Deeper**
I have always considered myself to be a pragmatist, egoist, and realist. However, there seems to be a gap between these ideologies as I see them and the image associate... |
5e4c916d-71c1-4826-963b-6d75c51f7a0b | trentmkelly/LessWrong-43k | LessWrong | The Many Stories of Science
Epistemic Status: Not exhaustively researched, dates and particular events are selected based on my incomplete and biased memory. My point in this post also rest on me not having seen particular versions of the story of science in the wild as well as repeated encounters with people seeming ... |
39ae1e41-39fa-4afc-b041-c753dec6dbf8 | trentmkelly/LessWrong-43k | LessWrong | Stupid Questions July 2015
This thread is for asking any questions that might seem obvious, tangential, silly or what-have-you. Don't be shy, everyone has holes in their knowledge, though the fewer and the smaller we can make them, the better.
Please be respectful of other people's admitting ignorance and don't mock ... |
9df16c0a-468b-400a-9c96-e48ff1014ab4 | trentmkelly/LessWrong-43k | LessWrong | Hemingway Case
> Why did the chicken cross the road?
>
> Ernest Hemingway: To die. In the rain.
My son asks: If HelloWorld is camel case, hello_world is snake case and hello-world is kebab case, what's the DNS-style Hello.World. ?
I think I have a pretty good answer for him: It's Hemingway case. |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.