id stringlengths 36 36 | source stringclasses 15
values | formatted_source stringclasses 13
values | text stringlengths 2 7.55M |
|---|---|---|---|
4f0f3be6-61b3-46e3-98aa-2a4e4d9c71fd | trentmkelly/LessWrong-43k | LessWrong | Meetup : Canberra: Contrarianism
Discussion article for the meetup : Canberra: Contrarianism
WHEN: 11 October 2014 06:00:00PM (+1000)
WHERE: 108 North Road, Acton, ACT
Share your LW-contrarian opinions! That is, (ideally) come prepared with some opinions you have that most other people at the meetup might not, and ... |
e82b04d5-10b9-4661-b062-56ba824d1e21 | trentmkelly/LessWrong-43k | LessWrong | Pascal's Mugging - Penalizing the prior probability?
Eliezer Yudkowsky wrote that Robin Hanson solved the Pascal's mugging thought experiment:
> Robin Hanson has suggested penalizing the prior probability of hypotheses which argue that we are in a surprisingly unique position to affect large numbers of other people w... |
93004116-9674-405c-981d-fdac41009d94 | trentmkelly/LessWrong-43k | LessWrong | Meetup : Boston: Unconference
Discussion article for the meetup : Boston: Unconference
WHEN: 05 April 2015 03:30:00PM (-0400)
WHERE: 98 elm street somerville
It's an Unconference! http://en.wikipedia.org/wiki/Unconference There will be short discussions on a variety of topics, as decided by attendees at the beginni... |
44acec11-d154-478c-9487-277cd3c09c6c | trentmkelly/LessWrong-43k | LessWrong | Inner Misalignment in "Simulator" LLMs
Alternate title: "Somewhat Contra Scott On Simulators".
Scott Alexander has a recent post up on large language models as simulators.
I generally agree with Part I of the post, which advocates thinking about LLMs as simulators that can emulate a variety of language-producing "ch... |
4a11ce93-41ff-47d0-91a3-30bfb47b7e35 | trentmkelly/LessWrong-43k | LessWrong | Pavlov Generalizes
Sarah Constantine recently wrote about the Pavlov strategy for iterated prisoner's dilemma. She points out that it deserves to be better-known, given that it has similar virtues to Tit-for-Tat (TfT). I strongly agree.
* Much of my thinking about puzzles in decision theory such as cooperation/coord... |
5e63b3c0-1df1-4741-9f4b-49557090fa22 | trentmkelly/LessWrong-43k | LessWrong | Meetup : Superintelligence chapter 7
Discussion article for the meetup : Superintelligence chapter 7
WHEN: 09 June 2017 05:45:00PM (+0200)
WHERE: Lindstedtsvägen 3, room 1537, SE-114 28 Stockholm, Sverige
1. The superintelligent will
The relation between intelligence and motivation
Instrumental convergence
Self... |
74f7f069-39a7-4673-bd2a-7aa8fe0aebbc | trentmkelly/LessWrong-43k | LessWrong | [Announcement] LessWrong will be down for ~1 hour on the evening of April 10th around 10PM PDT (5:00AM GMT)
LessWrong's database provider (Mlab) has been acquired by a separate company (MongoDB) and as such our database needs to be migrated to that new provider. In order to do that, there will be about half an hour of... |
c6c771b0-9513-4824-96b0-90b6df9b26ff | StampyAI/alignment-research-dataset/lesswrong | LessWrong | What's the difference between newer Atari-playing AI and the older Deepmind one (from 2014)?
My impression was that thing that put Deepmind on the map was an AI that could play multiple Atari games. Lately there's been new Atari-playing AI (both from Deepmind and other companies) that are making the news. Are they doi... |
290a9d7e-3ff6-4d9a-b822-ed3024e5997b | trentmkelly/LessWrong-43k | LessWrong | HPMoR the Youtube Series! But in need of advice
Hi Less Wrong! I was wondering if you have experience with video editing? I want to record footage and a soundtrack, and then overlay them on each other, and I'll also need to be able to do special effects, such as to float someone in the air. Is there a video editing pr... |
931c0ab7-3fe6-4262-8e08-4355c4e40e69 | trentmkelly/LessWrong-43k | LessWrong | Plausibility of Getting Early Warning Shots because AIs can't coordinate?
I don't know if this is a well known argument that has already been responded to. If it is, just delete the post.
An implicit assumption most people make when discussion takeover risk, is that any misaligned agent will stave off enacting takeov... |
7f191141-8da7-41a3-8834-fc9f22c9c1f0 | trentmkelly/LessWrong-43k | LessWrong | Meetup : Saint Petersburg meetup - "the lonely one"
Discussion article for the meetup : Saint Petersburg meetup - "the lonely one"
WHEN: 31 October 2014 08:00:00PM (+0400)
WHERE: Россия, Санкт-Петербург, Московский пр. 65,
We are resuming our bi-weekly schedule, even if I'm sitting at the meetups all by myself read... |
5fce3f6d-a355-4760-ab6b-32b638cd16c7 | awestover/filtering-for-misalignment | Redwood Research: Alek's Filtering Results | id: post352
Yoshua Bengio wrote a blogpost about a new AI safety paper by him, various collaborators, and me. I've pasted the text below, but first here are a few comments from me aimed at an AF/LW audience. The paper is basically maths plus some toy experiments. It assumes access to a Bayesian oracle that can infer a... |
1cd62837-9560-4583-a3c3-d08893258e7b | trentmkelly/LessWrong-43k | LessWrong | On saving one's world
If the world is likeliest to be saved by sober scholarship, then let us be sober scholars in the face of danger.
If the world is likeliest to be saved by playful intellectual exploration, then let us be playful in the face of danger.
Strategic, certainly; aware of our situation, of course; but ... |
fd8d261a-938c-4547-80f2-e823543c2153 | trentmkelly/LessWrong-43k | LessWrong | Misinformation is the default, and information is the government telling you your tap water is safe to drink
Status notes: I take the view that rational dialogue should work with good faith people who aren't also following rational dialogue. From that point of view, this piece is about rationality. If you don't take t... |
d2c00e1f-b6b6-437e-b4ee-83a0af24d4d1 | trentmkelly/LessWrong-43k | LessWrong | AISN #30: Investments in Compute and Military AI
Plus, Japan and Singapore’s National AI Safety Institutes
Welcome to the AI Safety Newsletter by the Center for AI Safety. We discuss developments in AI and AI safety. No technical background required.
Subscribe here to receive future versions.
Listen to the AI Safety... |
f9b40bcc-ef81-4652-a6a4-c45aa7cfb930 | trentmkelly/LessWrong-43k | LessWrong | Sparse autoencoders find composed features in small toy models
Summary
* Context: Sparse Autoencoders (SAEs) reveal interpretable features in the activation spaces of language models. They achieve sparse, interpretable features by minimizing a loss function which includes an ℓ1 penalty on the SAE hidden layer activa... |
f4c2fc46-99db-454d-861c-9ccfbcb59b5b | StampyAI/alignment-research-dataset/lesswrong | LessWrong | Foom Liability
Robin Hanson suggests, partly in response to calls for a pause in
development of AGI, [liability
rules](https://www.overcomingbias.com/p/foom-liability) for risks
related to AGI rapidly becoming powerful.
My intuitive reaction was to classify foom liability as equivalent to a
near total ban on AGI.
... |
a6e68d34-5647-421f-a775-4ecf25f8005b | trentmkelly/LessWrong-43k | LessWrong | Rebasing Ethics
Lets start with the following accepted as a given:
* There exists no supernatural forces in the world and there is no objective morality imposed from above.
* Our current moral codes are currently based on some mix of sociobiological influences & cultural forces
Of the current figures who accept th... |
60858903-d0cf-40fe-8b38-145459873c98 | trentmkelly/LessWrong-43k | LessWrong | Ideal governance (for companies, countries and more)
Click lower right to download or find on Apple Podcasts, Spotify, Stitcher, etc.
I'm interested in the topic of ideal governance: what kind of governance system should you set up, if you're starting from scratch and can do it however you want?
Here "you" could b... |
498597ce-968d-4c88-913c-96f94ef58f2e | StampyAI/alignment-research-dataset/blogs | Blogs | An Astounding Year
It’s safe to say that this past year exceeded a lot of people’s expectations.
Twelve months ago, Nick Bostrom’s *[Superintelligence](https://en.wikipedia.org/wiki/Superintelligence:_Paths,_Dangers,_Strategies)* had just been published. Long-term questions about smarter-than-human AI systems were s... |
c62b7bc3-7b04-4fe9-9235-c86381d1d261 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | NIST AI Risk Management Framework request for information (RFI)
*(Cross-post from the* [*EA Forum*](https://forum.effectivealtruism.org/posts/sQz9zTw49twFTsqMg/nist-ai-risk-management-framework-request-for-information)*)*
The [US National Institute of Standards and Technology (NIST)](https://www.nist.gov/) previously... |
453bcc25-0f5a-4037-a58c-6de45e57d133 | StampyAI/alignment-research-dataset/lesswrong | LessWrong | Conceding a short timelines bet early
Last year I [bet some people about short AI timelines](https://www.lesswrong.com/posts/X3p8mxE5dHYDZNxCm/a-concrete-bet-offer-to-those-with-short-agi-timelines). While I don't think I've lost the bet yet, I think it's clear at this point that I will lose with high probability. I'v... |
8ad4a478-f0fe-49ee-b933-cee0c2fe9f65 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Possible takeaways from the coronavirus pandemic for slow AI takeoff
*(Cross-posted from [personal blog](https://vkrakovna.wordpress.com/2020/05/31/possible-takeaways-from-the-coronavirus-pandemic-for-slow-ai-takeoff/). Summarized in [Alignment Newsletter #104](https://mailchi.mp/ba4d1765368f/an-104-the-perils-of-inac... |
0374f7e2-482b-4f93-a630-6d5d3edd8072 | trentmkelly/LessWrong-43k | LessWrong | Applying Bayes to an incompletely specified sample space
Epistemic Status: Novel mathematics from an amateur, presented with unjustified confidence. Submitted here and not in a mathematical journal because I think LW will be better and faster at explaining where I’ve gone wrong and/or what common approach I’ve reinven... |
4f301bd6-656f-474d-9d2c-11301ff595c5 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Seeking Power is Convergently Instrumental in a Broad Class of Environments
Edit, 5/16/23: I think this post is beautiful, correct in its narrow technical claims, and practically irrelevant to alignment. This post presents an unrealistic picture of the role of reward functions in reinforcement learning, conflating "ut... |
2539f2d7-a80b-4c41-901e-8de0c38fbe31 | trentmkelly/LessWrong-43k | LessWrong | 11 core rationalist skills
An excellent way to improve one's skill as a rationalist is to identify one's strengths and weaknesses, and then expend effort on the things that one can most effectively improve (which are often the areas where one is weakest). This seems especially useful if one is very specific about the ... |
4ad82a9d-3b7d-48de-9da0-064c5322182d | StampyAI/alignment-research-dataset/lesswrong | LessWrong | Satisficers want to become maximisers
*(with thanks to Daniel Dewey, Owain Evans, Nick Bostrom, Toby Ord and BruceyB)*
In theory, a [satisficing](http://en.wikipedia.org/wiki/Satisficing) agent has a lot to recommend it. Unlike a maximiser, that will attempt to squeeze the universe to every drop of utility that it c... |
082d2f78-9af1-4157-bc1d-ea5d3a4d6f93 | trentmkelly/LessWrong-43k | LessWrong | A learned agent is not the same as a learning agent
[Edit: after reading the comments and thinking more about in-context learning, I don't endorse most of what's written here. I explain why I'm the end of the post.]
I notice a common confusion when people talk about deep learning. Before I try to describe it in gener... |
67bbee84-76af-4061-9dbe-10c56c6d914e | trentmkelly/LessWrong-43k | LessWrong | Terrorism is not about Terror
> Statistical analysis of terrorist groups' longevity, aims, methods and successes reveal that groups are self-contradictory and self-sabotaging, generally ineffective; common stereotypes like terrorists being poor or ultra-skilled are false. Superficially appealing counter-examples are d... |
0eacf505-c3d6-4011-809c-08cf7b6e4df7 | trentmkelly/LessWrong-43k | LessWrong | When Can Optimization Be Done Safely?
When facing a problem in computer science, one can often write a program that solves that problem directly. Usually, a programmer understands why their programs do what they do. Not always, but usually. One can even use tools from formal verification to prove that programs have ce... |
ce5fc740-a6b6-4452-b0fd-2b633bcaafe3 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Research Agenda in reverse: what *would* a solution look like?
I constructed my [AI alignment research agenda](https://www.lesswrong.com/posts/CSEdLLEkap2pubjof/research-agenda-v0-9-synthesising-a-human-s-preferences-into) piece by piece, stumbling around in the dark and going down many false and true avenues.
But n... |
92bcb20f-e406-4a4b-b598-0eb111d046c1 | trentmkelly/LessWrong-43k | LessWrong | Poverty does not respond to incentives
I wrote a post a while back saying that preventing ‘exploitative’ trade is equivalent to preventing an armed threat by eliminating the ‘not getting shot in the head’ option. Some people countered this argument by saying that it doesn’t account for how others respond. If poor peop... |
885c698e-6790-4601-a4d3-3dbc4a70df06 | trentmkelly/LessWrong-43k | LessWrong | Meetup : San Francisco & Tortuga Go Surfing
Discussion article for the meetup : San Francisco & Tortuga Go Surfing
WHEN: 10 July 2011 06:39:31PM (-0700)
WHERE: Nor Cal Surf Shop, 5460 Pacific Coast Highway, Pacifica, CA
Low-key, afternoon fun at the beach for beginners & amateurs. I'll be offering rides from centr... |
58020f8b-3428-44af-8d34-1722021dc8be | trentmkelly/LessWrong-43k | LessWrong | Collapse Postulates
,
Macroscopic decoherence—also known as “many-worlds”—is the idea that the known quantum laws that govern microscopic events simply govern at all levels without alteration. Back when people didn’t know about decoherence—before it occurred to anyone that the laws deduced with such precision for micr... |
555358eb-8cf9-4479-a660-77aed3a4c6cc | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | [AN #159]: Building agents that know how to experiment, by training on procedurally generated games
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, ... |
995786ea-ff7c-4f7b-a972-564bec90f95c | trentmkelly/LessWrong-43k | LessWrong | If you factor out next token prediction, what are the remaining salient features of human cognition?
LLMs showed how far one can get with a good next token prediction engine and a huge amount of training data. This ought to tell us a fair bit about how the human mind works. Specifically, since the next token predictio... |
643dfa52-9753-4bf5-9485-9444a089b090 | trentmkelly/LessWrong-43k | LessWrong | 2020 Review Article
A common thing in academia is to write ‘review articles’ that attempt to summarize a whole field quickly, allowing researchers to see what’s out there (while referring them to the actual articles for all of the details). This is my attempt to do something similar for the 2020 Review, focusing on po... |
4b7855b9-bb86-478a-b790-dc7962c5f6b0 | trentmkelly/LessWrong-43k | LessWrong | Re: So You Want to Be a Dharma Teacher
> When I first started getting into Zen there was a brief period when I thought I wanted to be a dharma teacher. I remember telling my first teacher, Tim, I wanted to be a dharma teacher and he just gave me a look. It’s hard to describe that look, but I remember how it felt. And ... |
1907d57f-25c5-4c5e-bcc1-f3f3228fd659 | trentmkelly/LessWrong-43k | LessWrong | Evolution Solved Alignment (what sharp left turn?)
Some people like to use the evolution of homo sapiens as an argument by analogy concerning the apparent difficulty of aligning powerful optimization processes:
> And in the same stroke that its capabilities leap forward, its alignment properties are revealed to be sh... |
7d5d06e3-fa26-45a4-8f0c-a7cf6459cc5f | trentmkelly/LessWrong-43k | LessWrong | Looking for AGI philosopher co-author on Draft Article: 'Optimising Peace to Constrain Risk of War from an Artificial Superintelligence'
Abstract
An artificial superintelligence (ASI) emerging in a world in which war is still normalised may constitute a catastrophic existential risk, either because the ASI goes to wa... |
1716f992-5519-4162-b610-4a933d9659ba | trentmkelly/LessWrong-43k | LessWrong | RLLMv10 experiment
What did I do differently in this experiment?
RLLMv10, see RLLM research map for more details.
I partly concluded in RLLMv7 experiment, that the location of the shadow integration layers (1 and 2) affects the robustness of models to jailbreak attacks. This conclusion led me to speculate that it migh... |
b0a4202c-443f-4f26-bdd1-fae47e84c11c | StampyAI/alignment-research-dataset/blogs | Blogs | Beyond Bayesians and Frequentists
(This is available in pdf form [here](http://web.mit.edu/jsteinha/www/stats-essay.pdf).)
If you are a newly initiated student into the field of machine learning, it won’t be long before you start hearing the words “Bayesian” and “frequentist” thrown around. Many people around you pr... |
abf23752-9709-45e7-a65b-1d613a96861f | trentmkelly/LessWrong-43k | LessWrong | why people romantice magic over most science.
Back in 2008 Eliezer wrote the following post http://lesswrong.com/lw/ou/if_you_demand_magic_magic_wont_help/ In it he said the following
"Born into a world of science, they did not become scientists. What makes them think that, in a world of magic, they would act any di... |
fa14f804-8d13-48d4-88df-251b6e70ec71 | trentmkelly/LessWrong-43k | LessWrong | Goals and short descriptions
Outline
I develop some contents—previously introduced in the Value Learning sequence by Rohin Shah—more formally, to clarify the distinction between agents with and without a goal. Then I present related work and make some considerations on the relation between safety and goal-directedness... |
f5fee0cd-56bb-4af0-8733-beb083d46eac | trentmkelly/LessWrong-43k | LessWrong | [SEQ RERUN] Semantic Stopsigns
Today's post, Semantic Stopsigns was originally published on 24 August 2007. A summary:
> There are certain words and phrases that act as "stopsigns" to thinking. They aren't actually explanations, or help to resolve the actual issue at hand, but they act as a marker saying "don't ask a... |
50adc448-6225-4a34-acb9-bfab6f3e54b7 | StampyAI/alignment-research-dataset/lesswrong | LessWrong | AI and Efficiency
The amount of compute required to train neural networks to SOTA (state of the art) on some tasks has been decreasing at an exponential rate. Efficiency of training AI models seems to be growing faster than Moore's law.
From June 2012 to May 2019, the amount of compute needed to train a neural netwo... |
af099b25-ef23-4893-b769-bbf6b017502e | trentmkelly/LessWrong-43k | LessWrong | Proposal: labs should precommit to pausing if an AI argues for itself to be improved
As evidence that compute is a major bottleneck on capabilities has accumulated, many people have become more skeptical of extremely fast takeoff speeds. One major reason for this is that if humans were trying to stop it, it would prob... |
16a0fe1c-245e-4043-bb94-52fc3d65c54f | trentmkelly/LessWrong-43k | LessWrong | Does this seem to you like evidence for the existence of psychic abilities in humans?
I was recently reminded of something I have encountered that seems to me to be good evidence for paranormal phenomena. Can anyone help me figure out what might be going on?
When I was a little younger, I used to play the online ri... |
abee4b63-f5c1-4bae-9c1c-719689d384c8 | trentmkelly/LessWrong-43k | LessWrong | Grief is a fire sale
Written in Nov 2023.
For me, grief is often about the future. It isn't about the loss of past times, since those were already gone. It is the loss of hope. It's missing the last train to a friends wedding or never being able to hear another of my grandfathers sarcastic quips.
In the moment of gr... |
46c3c732-c378-4012-831b-741f58f3fa70 | trentmkelly/LessWrong-43k | LessWrong | Humanity as an entity: An alternative to Coherent Extrapolated Volition
Note: this is a rather rough draft, and some of it is written in compact-form, for time-optimization reasons.
Humanity-as-an-entity or humanity-as-an-agent.
Putting all the contradictory drives and desires of humanity into a single entity, a self... |
7ddde52e-708b-4dd9-877d-360cb99562a0 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Paper: In-context Reinforcement Learning with Algorithm Distillation [Deepmind]
Authors train transformers to imitate the trajectory of reinforcement learning (RL) algorithms. Find that the transformers learn to do *in-context RL* (that is, the transformers implement an RL algorithm)---the authors check this by having... |
09a71004-fddc-41d6-b1ae-28033dfadb42 | trentmkelly/LessWrong-43k | LessWrong | The AI Paper with The Best Title Ever
Epistemic Status: Confident I understood the paper, but no promises on its implications
The title is “Learning to learn by gradient descent by gradient descent.” If anyone can top that title I am eager to hear about it. The idea is a good one on the merits, but I still suspect th... |
b9ffc1fe-dbc2-443c-abc4-91621d53ae97 | trentmkelly/LessWrong-43k | LessWrong | Natural Abstractions: Key Claims, Theorems, and Critiques
TL;DR: We distill John Wentworth’s Natural Abstractions agenda by summarizing its key claims: the Natural Abstraction Hypothesis—many cognitive systems learn to use similar abstractions—and the Redundant Information Hypothesis—a particular mathematical descript... |
3ddad935-af0b-450c-b372-e15ffa5db781 | trentmkelly/LessWrong-43k | LessWrong | First Waco, Texas LW Meetup, 4/09, 1PM
Date: Saturday, April 9
Time: 1PM
Location: Common Grounds Coffeehouse, IH-35 and S 8th Street (east side of interstate). (If you can't find parking behind the place, use the parking garage [free on weekends] just a short distance down the street.)
Well, I'm going through wit... |
5ce570b2-182d-42ed-a4e1-7915d3304736 | trentmkelly/LessWrong-43k | LessWrong | So What's Up With PUFAs Chemically?
This is exploratory investigation of a new-ish hypothesis, it is not intended to be a comprehensive review of the field or even a a full investigation of the hypothesis.
I've always been skeptical of the seed-oil theory of obesity. Perhaps this is bad rationality on my part, but I'... |
ae9f54a9-6575-467b-a3fc-b4bf47bb079b | trentmkelly/LessWrong-43k | LessWrong | Meetup : Washington, D.C.: Singing
Discussion article for the meetup : Washington, D.C.: Singing
WHEN: 27 September 2015 03:30:00PM (-0400)
WHERE: National Portrait Gallery
Crossposted from mailing list.
Edit: We have tentatively decided to go to the fallback meetup topic and will meet in the courtyard as normal.
... |
cd90dbde-d9d8-46a2-9234-f4c248d5af74 | trentmkelly/LessWrong-43k | LessWrong | Meetup : Less Wrong NH Meet-up
Discussion article for the meetup : Less Wrong NH Meet-up
WHEN: 15 September 2015 07:00:00PM (-0400)
WHERE: 269 Pearl St, Manchester NH 03104
The fourth NH meet-up is Tuesday, 9/15, in Manchester, NH at 7 pm at a private residence. Light refreshments will be provided.
Have you read R... |
54efeaa5-e276-43fa-8550-d04a98312ce1 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Intelligent behaviour across systems, scales and substrates
Tl;dr
=====
In an [earlier post](https://www.lesswrong.com/posts/FuToH2KHxKmJLGk2B/ai-alignment-as-navigating-the-space-of-intelligent), I introduced a metaphor for thinking about the epistemic landscape of AI alignment, and then described three epistemic st... |
e5dbd66b-cf02-47f3-bb5a-8349bdeeb81b | trentmkelly/LessWrong-43k | LessWrong | Orexin and the quest for more waking hours
A week ago I was skeptical about the prospect of radically reducing human sleep needs. After reading John Boyle’s Cause area: Short-sleeper genes, I decided to research the area more deeply and updated to believe that it’s more likely that we can reduce human sleep needs with... |
d1318416-c90d-4aa2-89a4-6253022ae46d | trentmkelly/LessWrong-43k | LessWrong | Scholarship: how to tell good advice from bad advice?
Lukeprog has done some great posts talking about how scholarship has benefited him personally and how to do it. However, he hasn't talked much about how to tell the good from the bad, and I think this is an especially big problem with it comes to "how to"s, rather ... |
a7b0cdb8-b95e-4d80-9e65-0279d6a55255 | trentmkelly/LessWrong-43k | LessWrong | Willing to share some words that changed your beliefs/behavior?
I'm collecting data on powerfully persuasive speech acts; it's part of a dangling thread of curiosity after GPT-2 (a new and fairly powerful text generation algorithm). I'm skeptical of the danger of mind-warping sentences as sometimes presented in fictio... |
666db8cc-b26a-4701-803e-e07e7965ad64 | LDJnr/LessWrong-Amplify-Instruct | LessWrong | "Crossposted from my blog.
One thing I worry about sometimes is people writing code with optimisers in it, without realising that that's what they were doing. An example of this: suppose you were doing deep reinforcement learning, doing optimisation to select a controller (that is, a neural network that takes a percept... |
32f0cfed-a345-4456-9d18-63290d2d166f | trentmkelly/LessWrong-43k | LessWrong | How Business Solved (?) the Human Alignment Problem
When I looked at mesa-optimization for the first time, my mind immediately associated it with a familiar problem in business: human individuals may not be “mesa optimizers” in a strict sense, but they can act as optimizers and they are expected to do so when a manage... |
17f101aa-0f78-4bbe-96be-a40c8b2a611a | trentmkelly/LessWrong-43k | LessWrong | More people getting into AI safety should do a PhD
Doing a PhD is a strong option to get great at developing and evaluating research ideas. These skills are necessary to become an AI safety research lead, one of the key talent bottlenecks in AI safety, and are helpful in a variety of other roles. By contrast, my impre... |
ebb360af-3707-43c9-a112-d7cd952f816d | trentmkelly/LessWrong-43k | LessWrong | Happy Petrov Day! If Today is Also Your Birthday, Happy Birthday!
A discussion of September 26 and September 1993, and an invitation to join the club if that's your birthday too |
f06bbb35-e66d-46b3-9a4d-6f8f4da318b8 | trentmkelly/LessWrong-43k | LessWrong | Learning Over Time for AI and Humans and Rationality
https://www.lesswrong.com/posts/pLZ3bdeng4u5W8Yft/let-s-talk-about-convergent-rationality-1
make the observation that natural and artificial intelligence will learn over time. Given the post's title that got me thinking about the two learning settings and how that ... |
0947bba2-9dea-435d-9a48-e2d1a3bf019b | trentmkelly/LessWrong-43k | LessWrong | If all trade is voluntary, then what is "exploitation?"
Capitalism is a force that has lifted billions out of poverty, where even poor remote villagers enjoy luxuries that would have been unimaginable to medieval kings. When someone takes a job, even the worst job, it’s because both parties expect mutual gain. And yet... |
d5e96bd5-282f-4063-b90c-90f63aefcf93 | trentmkelly/LessWrong-43k | LessWrong | Perpetual Motion Beliefs
Yesterday's post concluded:
> To form accurate beliefs about something, you really do have to observe it. It's a very physical, very real process: any rational mind does "work" in the thermodynamic sense, not just the sense of mental effort... So unless you can tell me which specific step in... |
6f638202-3a7c-4c71-99d5-f1536f2ac0ab | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Polysemanticity and Capacity in Neural Networks
Elhage et al at Anthropic recently published a paper, [Toy Models of Superposition](https://transformer-circuits.pub/2022/toy_model/index.html) (previous Alignment Forum discussion [here](https://www.alignmentforum.org/posts/CTh74TaWgvRiXnkS6/toy-models-of-superposition)... |
8c80f76b-ec36-4ccf-8059-36c6070c5af2 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Prizes for matrix completion problems
Here are two self-contained algorithmic questions that have come up in our research. We're offering a bounty of $5k for a solution to either of them—either an algorithm, or a lower bound under any hardness assumption that has appeared in the literature.
**Question 1 (existence of... |
b5c3dfa4-d408-4753-a215-e20c5e8af1fc | trentmkelly/LessWrong-43k | LessWrong | Quick Look: Asymptomatic Herpes Shedding
Tl;dr: Individuals shed and thus probably spread oral HSV1 while completely asymptomatic.
Introduction
“Herpes virus” can refer to several viruses in the herpes family, including chickenpox and Epstein-Barr (which causes mono). All herpesviridae infections are for life: once ... |
af19864e-d40d-4e8f-89bb-d16f5852ba72 | trentmkelly/LessWrong-43k | LessWrong | Constraints from naturalized ethics.
Humans are just bundles of atoms, and so for them to reflect on themselves and their goals, it's plausible that there's some interesting naturalized ethics going on. This informs how we'd think about an AI intended to do human value learning. Let's frame this as a dialogue, so my h... |
79779783-a588-4c88-a9e3-6b67d264aa21 | trentmkelly/LessWrong-43k | LessWrong | Critch on career advice for junior AI-x-risk-concerned researchers
In a recent e-mail thread, Andrew Critch sent me the following "subtle problem with sending junior AI-x-risk-concerned researchers into AI capabilities research". Here's the explanation he wrote of his view, shared with his permission:
---------------... |
7f74236e-0410-442d-a0ad-4cebb5c7f2fe | trentmkelly/LessWrong-43k | LessWrong | How alienated should you be?
Epistemic status: a post about values, where I’m confident the question is interesting but not confident my answer is correct. Factual statements are intended to be correct summaries, but I won’t be careful about citations. This post was written in January, and edited and published today.
... |
a6860183-ee89-4f45-b467-ca64150e0b09 | trentmkelly/LessWrong-43k | LessWrong | Let’s discuss the weather
Wondering your opinion, not particularly trying to change it:
1. How likely is our avoiding dangerous climate change by getting enough international cooperation to cut global carbon emissions enough in time?
2. How much of a difference can I make to this?
3. If the chances in 1. are sma... |
b4fff31b-3097-4ee1-b76f-b8094421b2c5 | trentmkelly/LessWrong-43k | LessWrong | How does your data backup solution setup work?
I'm currently not happy with how I'm backing up my data. I feel like doing proper databackup is a bit like flossing every day. It's clearly the "right thing to do" but thinking of it feels like a shore that I don't want to deal with. Does anybody have a personal databacku... |
f9edba1c-9f86-4710-9b5a-4690e99a547b | trentmkelly/LessWrong-43k | LessWrong | 3M Subscriber YouTube Account 'Channel 5' Reporting On Rationalism
Hi, I thought it might be interesting to some here to take note of the 3M subscriber YouTube channel "Channel 5 with Andrew Callaghan" posting an hour-long news room style podcast/documentary/report about the rationalist movement (in the context of Ziz... |
8cd75f20-66ee-41f9-a697-a95751300472 | trentmkelly/LessWrong-43k | LessWrong | A hypothetical candidate walks into a hypothetical job interview...
Let's say you are interviewing a candidate for a job. In casual conversation, the candidate mentions that he is a member of a rather old and prestigious country club. You've never heard the name of the club before.
You look up the country club after... |
bc18d3ea-894c-41cf-a48a-6bd9e7685852 | trentmkelly/LessWrong-43k | LessWrong | Lurking More Before Joining Complex Conversations
Opinion About Rationalist Parties:
At a nerdy party, when people are in the middle of discussing a complicated idea, if a new person joins the conversation, I'd prefer if the norm were to *not* try to get that person up to speed. Instead, they lurk for awhile, try to ... |
ba64ccee-dda8-4f6b-96cc-be482848c732 | trentmkelly/LessWrong-43k | LessWrong | Murphy’s Quest Postmorterm
Kudos to lifelonglearner for amazing cover art that is also an example of Murphy’s Law.
The full text of Murphy’s Quest (with many corrections) is now available in PDF.
Introduction
Having been kept away from expository writing for a week, I have all too many thoughts to express. This pos... |
670302ca-c883-4f36-961a-459668242781 | trentmkelly/LessWrong-43k | LessWrong | Meetup : Sydney Rationality Dojo - Approaches, Conspecific Interactions, Arguments and Communications
Discussion article for the meetup : Sydney Rationality Dojo - Approaches, Conspecific Interactions, Arguments and Communications
WHEN: 07 December 2014 04:00:00PM (+1100)
WHERE: Humanist House, 10 Shepherd St Chippe... |
257012fd-c771-41d1-b291-e9735884a36b | trentmkelly/LessWrong-43k | LessWrong | A Motorcycle (and Calibration?) Accident
Last week I saw a motorcycle accident.
My dad, known to be the most careful driver in my family, was attempting to pull out of the mall parking lot. Suddenly, a motorcyclist came into view, and as he saw our car, he stomped on the brakes. My dad saw him and stopped, but the mo... |
230def27-b791-4266-b850-69312e128d18 | StampyAI/alignment-research-dataset/lesswrong | LessWrong | [Part 1] Amplifying generalist research via forecasting – Models of impact and challenges
*This post covers our models of impact and challenges with our exploration in amplifying generalist research using forecasting. It is accompanied by [a second post](https://www.lesswrong.com/posts/FeE9nR7RPZrLtsYzD) with a high-l... |
caca3365-352e-4175-81d6-3ad02ae2cede | trentmkelly/LessWrong-43k | LessWrong | Science for the Possible World
In the foreseeable future, it is expected that the majority of consciousness in the universe will spend their time in the digital world. I argue that in order to unlock the "goodness" of the digital world, we need to focus on a new kind of science--the science for the possible world (SPW... |
eb950b09-71a8-4397-8595-f68fc87c7719 | trentmkelly/LessWrong-43k | LessWrong | Kissing Scars
Last year August 6th, I got in a horrible drone accident that destroyed the lens and iris in my left eye and left a scar running down my eye. For a couple months I could barely read, my working memory was horrible, energy levels were shot, and I was in so much pain, especially after surgeries.
Then a sw... |
a6c7c100-3dc9-41a3-88b9-145277d9ead8 | trentmkelly/LessWrong-43k | LessWrong | Paris Meetup, Saturday April 30th, 2PM
When: Saturday, April 30th, 2PM.
Where: Au Pt'it Chat, a little Café near Châtelet, which is here. I'll be there with a LessWrong sign.
Yvain, cousin_it and Morendil should be there ... and you are welcome too!
This is a repost, the meetup was originally planned the 16th, but ... |
ec428602-1910-4fd3-8175-e95c9546909f | LDJnr/LessWrong-Amplify-Instruct | LessWrong | "(Related: Inaccessible Information, What does the universal prior actually look like?, Learning the prior)Fitting a neural net implicitly uses a “wrong” prior. This makes neural nets more data hungry and makes them generalize in ways we don’t endorse, but it’s not clear whether it’s an alignment problem.After all, if ... |
f9c16a6c-6251-4db5-8af1-89c0eebd24aa | trentmkelly/LessWrong-43k | LessWrong | Meetup : Washington DC: Anthropics with Robin Hanson
Discussion article for the meetup : Washington DC: Anthropics with Robin Hanson
WHEN: 27 October 2013 03:00:00PM (-0400)
WHERE: National Portrait Gallery, Washington, DC 20001, USA
We'll be meeting to discuss anthropics: things like the Doomsday Argument, self-sa... |
510c0adc-542f-4dc1-8912-897849c92153 | trentmkelly/LessWrong-43k | LessWrong | A basic mathematical structure of intelligence
An important concept here on LW is that of a singularity of intelligence, or at least a very rapid growth. Although it seems mostly hopeless, it would be nice if we could find a mathematical approach to quantify these things. I think the first point to note is that intell... |
52f0c546-d631-447b-b6bd-c69044ee04a7 | trentmkelly/LessWrong-43k | LessWrong | The Most Frequently Useful Thing
Followup to: The Most Important Thing You Learned
What's the most frequently useful thing you've learned on OB - not the most memorable or most valuable, but the thing you use most often? What influences your behavior, factors in more than one decision? Please give a concrete examp... |
f624d915-ccbb-48e1-a677-c41f139d9b74 | awestover/filtering-for-misalignment | Redwood Research: Alek's Filtering Results | id: post397
This idea is so simple that I'm sure it's been had by someone somewhere. Suppose we have some method to make really smart honest AIs that do not have goals. Let's say it's a yes/no oracle. Our aimless ace. But we want to accomplish stuff! AIcorp wants the printmoneynow.py . I'm worried that if I train the ... |
87296c75-3a4d-48f4-9254-cc41221a73ec | trentmkelly/LessWrong-43k | LessWrong | An attempt to break circularity in science
We do science using the data we collected through our senses and we use science to understand how our senses work. Although I lack any rigorous formulation of the problem, the following plan seems interesting and I want to share it with you.
p= "Human senses collect accurate... |
01866ac0-be9f-43e9-a0b9-c66a8f1d9508 | trentmkelly/LessWrong-43k | LessWrong | I
I wrote this story at Michigan State during Clarion 1997, and it was published in the Sept/Oct 1998 issue of Odyssey. It has many faults and anachronisms that still bother me. I'd like to say that this is because my understanding of artificial intelligence and the singularity has progressed so much since then; but... |
53e5833c-e004-4393-99f3-c67c5e86fb48 | StampyAI/alignment-research-dataset/special_docs | Other | Fireside chat - AI governance _ Markus Anderljung _ Ben Garfinkel _ EA Global - Virtual 2020-by Centre for Effective Altruism-video_id bSTYiIgjgrk-date 20200321
# Markus Anderljung and Ben Garfinkel Fireside chat on AI governance - EA Forum
\*\*Getting involved in AI governance\*\*
\*\*Markus:\*\* Ben, how did you g... |
10352181-f6f8-4aae-b1f4-ba3497e89982 | trentmkelly/LessWrong-43k | LessWrong | Rationality Reading Group: Part F: Politics and Rationality
This is part of a semi-monthly reading group on Eliezer Yudkowsky's ebook, Rationality: From AI to Zombies. For more information about the group, see the announcement post.
----------------------------------------
Welcome to the Rationality reading group. T... |
6c591968-ef12-45ef-9d7f-5d276e50e209 | trentmkelly/LessWrong-43k | LessWrong | Yudkowsky Contra Christiano on AI Takeoff Speeds [Linkpost]
Linkpost for “Yudkowsky Contra Christiano on AI Takeoff Speeds”, published by Scott Alexander on March 4, 2021. Here is the original conversation. Cross-posted on the EA Forum here.
Key Quotes:
> Around the end of last year, Paul and Eliezer had a complica... |
b6358dc6-023a-43ff-942f-93534b685ebd | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Prize and fast track to alignment research at ALTER
On behalf of[ALTER](https://alter.org.il/) and[Superlinear](https://www.super-linear.org/), I am pleased to announce a prize of at least[[1]](#fnf6v0lry3p2f) 50,000 USD, to be awarded for the best substantial contribution to the[learning-theoretic AI alignment resear... |
12429faa-260f-419c-a7e6-b25f296074ae | trentmkelly/LessWrong-43k | LessWrong | For the metaphors
I make use of a lot of analogies, for instance ‘like dancing’ and ‘the ice skating thing’ are particular phenomena I often think about, and I get value from thinking about meta-ethics as if it were romance, or saving the world as if it were a party. I wonder if providing a variety of concrete experie... |
99042375-6d96-4241-9f53-3fffb976a701 | StampyAI/alignment-research-dataset/arxiv | Arxiv | The Power of Scale for Parameter-Efficient Prompt Tuning
1 Introduction
---------------
With the wide success of pre-trained large language models, a range of techniques has arisen to adapt these general-purpose models to downstream tasks.
ELMo Peters et al. ([2018](#bib.bib33)) proposed freezing the pre-trained mo... |
d788f3a8-d3be-4ab3-a65d-691c83489178 | trentmkelly/LessWrong-43k | LessWrong | A Rationalist Argument for Voting
The argument that voting is irrational is commonplace. From your point of view as a single voter, the chance that your one vote will sway the election are typically miniscule, and the direct effect of your vote is null if it doesn't sway the election. Thus, even if the expected value ... |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.