id stringlengths 36 36 | source stringclasses 15
values | formatted_source stringclasses 13
values | text stringlengths 2 7.55M |
|---|---|---|---|
e9b23a29-4148-46f3-b604-780408df3f9c | trentmkelly/LessWrong-43k | LessWrong | Is "Control" of a Superintelligence Possible?
Epistemic status: confused.
I currently see two conceptualization of an aligned post-AGI/foom world:
1. SURRENDER TO THE WILL OF THE SUPERINTELLIGENCE
Any ‘control’ in a world with a Superintelligence will have to be illusory. The creation of an AGI will be the last tru... |
b01e80e0-69f8-4c59-bc09-4a274d100f20 | trentmkelly/LessWrong-43k | LessWrong | What are the primary drivers that caused selection pressure for intelligence in humans?
I just read the wikipedia article on the evolution of human intelligence, and TBH I wasn't super impressed with the quality of the considerations there.
I currently have 3 main (categories of) hypotheses for what caused selection ... |
ce0b48d5-2674-45e9-90ae-be038cbd2587 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Imitation Learning from Language Feedback
*TL;DR: Specifying the intended behavior of language models is hard, and current methods, such as RLHF, only incorporate low-resolution (binary) feedback information. To address this issue, we introduce Imitation learning from Language Feedback (ILF), an iterative algorithm le... |
5af7bf57-bad4-4e2f-a898-c8423e359a39 | trentmkelly/LessWrong-43k | LessWrong | Perceptual dexterity: a case study
My partner Logan Brienne Strohl has been doing original rationality development off and on for the past ten years (and "on" for at least the past three).
A week or so ago, I received a long email from them, which they summarized as follows:
> Duncan and I recently moved out into th... |
730131cd-a2f2-4c1a-8487-f1d11120c017 | trentmkelly/LessWrong-43k | LessWrong | Thread for Sense-Making on Recent Murders and How to Sanely Respond
In case you have not heard, there have been some recent and not-so-recent killings allegedly by people who have participated in aspiring rationalist spaces.
I've put some links in a footnote for those who want the basic info.[1]
I believe many peopl... |
2d679e34-86f1-4d0c-ba72-0feaa7e95ede | trentmkelly/LessWrong-43k | LessWrong | Young kids catching COVID: how much to worry?
Low confidence, slapdash job. Just putting this up in case other people want to compare notes.
(Update January 2022: Since writing this, I've adjusted my risk assumption down a bit based on newer studies suggesting lower rates of long-COVID than earlier studies (maybe a f... |
d2d7daaf-9bf7-4082-905d-ed8f5e2af43f | trentmkelly/LessWrong-43k | LessWrong | Some Adventures of a Curious Richard Feynman
This is cross-posted from my blog.
----------------------------------------
Sitting around a campfire listening to stories passed on from generations brings with it a feeling of wonder and joy for the adventures told from long ago. Each one of these stories is special on... |
8b3df51d-e0a7-4e44-90f4-7e4040210069 | StampyAI/alignment-research-dataset/arxiv | Arxiv | Discrete Representations Strengthen Vision Transformer Robustness
1 Introduction
---------------
Despite their high performance on in-distribution test sets, deep neural networks fail to generalize under real-world distribution shifts (Barbu et al., [2019](#bib.bib4)).
This gap between training and inference poses ... |
871b84f3-766e-486a-992b-743c8d3471d6 | trentmkelly/LessWrong-43k | LessWrong | Meetup : Urbana-Champaign: Discussion
Discussion article for the meetup : Urbana-Champaign: Discussion
WHEN: 16 March 2014 11:00:00AM (-0800)
WHERE: 300 S. Goodwin Ave Apt 102 Urbana IL
WHAT: Starting topic: contrarianism. When is it justified? Related reading: This and everything linked at the beginning. WHERE: 30... |
d1439e42-2b08-48ea-a208-19415a578e29 | trentmkelly/LessWrong-43k | LessWrong | Meetup : First Buffalo Meetup
Discussion article for the meetup : First Buffalo Meetup
WHEN: 28 December 2012 06:30:00PM (-0500)
WHERE: 765 Elmwood Avenue Buffalo, NY 14222
Its the first ever Buffalo LW Meetup! It will be at Spot Coffee on Elmwood. Lets meet at 6:30pm, I'll get there earlier and have some seating ... |
d2aa6afd-a592-4489-880f-e2db6de2fb30 | trentmkelly/LessWrong-43k | LessWrong | LINK: Competition is toxic
"I looked at what I think of as the food chain that led to the financial crisis, which was that you had individual consumers buying houses they couldn't afford, sold to them by realtors and property people who were competing to sell more properties at a higher price and so on. [...] I though... |
257ca8e5-a343-468f-9f5d-d5fca146b6e3 | trentmkelly/LessWrong-43k | LessWrong | Approval-directed agents
Note: This is the first post from part two: basic intuitions of the sequence on iterated amplification. The second part of the sequence outlines the basic intuitions that motivate iterated amplification. I think that these intuitions may be more important than the scheme itself, but they are c... |
1e78b7e4-14e9-4fa1-9540-85bb0417baa4 | StampyAI/alignment-research-dataset/lesswrong | LessWrong | GPT-3 Catching Fish in Morse Code
*Mostly non-serious and slightly silly, with some potentially interesting bits for people who are into language models.*
*TLDR: The current version of GPT-3 has a strong tendency to encode mangled versions of a specific phrase when asked to write morse code in zero-shot situations. T... |
8f172428-c874-4155-941c-37b86505f7f5 | trentmkelly/LessWrong-43k | LessWrong | Rational retirement plans
Are you in your 20s-40s and diligently saving for your retirement plan? You might want to reconsider your strategy.
Saving money with the intention of spending it in 20+ years will return (almost) nothing.
Superintelligence is likely to arrive within the next 20 years, probably sooner.
1.... |
360fd1c2-bef8-409a-bbbc-b8a709de6c16 | trentmkelly/LessWrong-43k | LessWrong | On (Not) Reading Papers
The emperor's new paper.
In a recent lunch conversation, a colleague of mine explained her system for reading scientific papers. She calls the system tongue-in-cheek "not reading the paper". It goes as follows:
1. Read the title and predict what the abstract will say. If you are correct, stop... |
1f84d5e2-e339-4979-bf12-7ec96f0fd4a5 | trentmkelly/LessWrong-43k | LessWrong | Postdoc opening at U. of Washington in AI law and policy
|
e0680b5f-0d82-4105-a1b8-fb3ec2e8d476 | StampyAI/alignment-research-dataset/arbital | Arbital | Mild optimization
"Mild optimization" is where, if you ask a Task AGI to paint one car pink, it just paints one car pink and then stops, rather than tiling the galaxies with pink-painted cars, because it's *not optimizing that hard.* It's okay with just painting one car pink; it isn't driven to max out the twentieth ... |
b73dbaf2-9646-4b7e-94b3-9f49c6c1deb5 | StampyAI/alignment-research-dataset/arxiv | Arxiv | Iterative Empirical Game Solving via Single Policy Best Response.
1 Introduction
---------------
In (single-agent) reinforcement learning (RL), an agent repeatedly interacts with an environment until it achieves mastery, that is, can find no further way to improve performance (measured by reward).
In a multiagent s... |
9bed1ab4-f465-45a4-b0f7-4bc1d224651f | trentmkelly/LessWrong-43k | LessWrong | Russell Conjugations list & voting thread
Russell conjugations humorously illustrate how the same concept can be expressed with different phrasings that carry positive, neutral, or negative valence. Examples below.
I wish they were taught in schools. I think it's a high bang-for-buck rationality technique, to 'conjug... |
fc04721a-c836-4e94-9704-738dbefd84b1 | trentmkelly/LessWrong-43k | LessWrong | What are a rationalist's best tools for better decision making?
I'm curious what others here use for improving their decision making skills. For something so important (making good decisions), society doesn't seem very interested in collectively improving our decision making. This is likely due to the fact, right or w... |
b8c1c011-d39b-495e-a719-7b789549e572 | trentmkelly/LessWrong-43k | LessWrong | Feature Suggestion: one way anonymity
Sometimes there's something I want to write, where I would be perfectly happy for whoever happens to read it on Less Wrong and is interested in who the author is to check and see it's me, but I don't want the post to appear in search results for my name.
I also don't want anyone ... |
36249b67-bbe0-47da-ace1-c689d77949e9 | trentmkelly/LessWrong-43k | LessWrong | Meetup : San Francisco Meetup: Cooking
Discussion article for the meetup : San Francisco Meetup: Cooking
WHEN: 01 February 2016 06:15:00PM (-0800)
WHERE: 1597 Howard St. San Francisco, CA
We'll be getting together to cook some food, then eat it! If you know how to cook, we could use your help, and you should come. ... |
50535e18-ea56-424b-ab05-7edb315bf362 | trentmkelly/LessWrong-43k | LessWrong | Requesting Questions For A 2017 LessWrong Survey
It's been twelve months since the last LessWrong Survey, which means we're due for a new one. But before I can put out a new survey in earnest, I feel obligated to solicit questions from community members and check in on any ideas that might be floating around for what ... |
e2ee1542-f1dd-4cd2-9a68-8076e04aec60 | trentmkelly/LessWrong-43k | LessWrong | Acausal trade naturally results in the Nash bargaining solution
In this post, I demonstrate that if we treat acausal trade as being analogous to ordinary trade (in the sense that it obeys supply and demand), the final result is the Nash bargaining solution.
Supply and demand of utility
Alice is an AI that wants to m... |
88cbcbd7-19db-47bb-a749-9c8de6c1aa3f | trentmkelly/LessWrong-43k | LessWrong | How security and cryptography can aid AI safety [sequence]
Overview
In this review sequence, I lay out why I think security, cryptography, and auxiliary approaches are still neglected in AI safety, specific insights those communities can offer for AI safety, and some potential next steps.
I am writing this post now... |
d7332adc-a627-4d13-8598-32a717de3d49 | trentmkelly/LessWrong-43k | LessWrong | Brainstorming for post topics
I suggested recently that part of the problem with with LW was a lock of discussion posts which was caused by people not thinking of much to post about.
When I ask myself "what might be a good topic for a post?", my mind goes blank, but surely not everything that's worth saying that's r... |
3c961901-9929-4517-a401-1a02d38a6c1c | StampyAI/alignment-research-dataset/agentmodels | Tutorial: Modeling Agents with Probabilistic Programs | Modeling Agents with Probabilistic Programs
---
layout: chapter
title: Joint inference of biases and preferences II
description: Explaining temptation and pre-commitment using softmax noise and hyperbolic discounting.
---
## Restaurant Choice: Time-inconsistent vs. optimal MDP agents
Returning to the MDP Restaurant... |
06715f0f-e473-46f5-9188-9655fa45db65 | trentmkelly/LessWrong-43k | LessWrong | Why You're Stuck in a Narrative
For some reason the narrative fallacy does not seem to get as much play as the other major cognitive fallacies. Apart from discussions of "The Black Swan", I never see it mentioned anywhere. Perhaps this is because it's not considered a "real" bias, or because it's an amalgamation of se... |
e188566f-0795-4986-93c4-0fcdae0559e4 | trentmkelly/LessWrong-43k | LessWrong | Why (and why not) Bayesian Updating?
> the use of Bayesian belief updating with expected utility maximization may be just an approximation that is only relevant in special situations which meet certain independence assumptions around the agent's actions.
— Steve Rayhawk
For those who aren't sure of the need for an u... |
ff108997-e134-4398-9006-2b0ae4804bb5 | trentmkelly/LessWrong-43k | LessWrong | Against the normative realist's wager
(Cross-posted from my website. Audio version here, or search "Joe Carlsmith Audio" on your podcast app.)
Summary
Non-naturalist normative realism is the view that there are mind-independent normative facts that aren’t reducible to facts about the natural world (more here). I don... |
456df7a3-7534-4911-b032-ef4678c82ee9 | trentmkelly/LessWrong-43k | LessWrong | The OODA Loop -- Observe, Orient, Decide, Act
United States Air Force Colonel John Boyd was a fighter pilot and military strategist who developed several important strategic theories. While serving as a jet fighter instructor, he was nicknamed "Forty-Second Boyd" because he had a standing offer that he could go up in... |
72c24bb4-54d6-41c2-8a18-e30e7f08747e | trentmkelly/LessWrong-43k | LessWrong | GD’s Implicit Bias on Separable Data
TLDR: This post goes through the intuition behind why using gradient descent to fit a linear model to linearly separable data will learn a maximum-margin decision boundary. It’s a simple example of an implicit bias of gradient descent, and also extends (in some form) to more compli... |
20fb7e07-4dd3-48ae-9815-33c839a3562c | trentmkelly/LessWrong-43k | LessWrong | The New Frontpage Design & Opening Tag Creation!
We just pushed a new frontpage design that we've been sitting and iterating on for a while, and we also just enabled the ability for everyone to create tags and edit the descriptions of existing tags.
New Frontpage design
A week ago we asked for your thoughts on the ... |
d9ef985f-d0c5-4d5f-9fff-741c58623694 | trentmkelly/LessWrong-43k | LessWrong | Sequence overview: Welfare and moral weights
We may be making important conceptual or methodological errors in prioritization between moral patients. In this sequence, I illustrate and address several:
1. Types of subjective welfare: I review types of subjective welfare, interpersonal comparisons with them and commo... |
6d692ca2-0617-4de2-8302-d4a2ad509689 | trentmkelly/LessWrong-43k | LessWrong | Conference on Global Catastrophic Risks
FYI: The Oxford Future of Humanity Institute is holding a conference on global catastrophic risks on July 17-20, 2008, at Oxford (in the UK).
I'll be there, as will Robin Hanson and Nick Bostrom.
Deadline for registration is May 26th, 2008. Registration is £60. |
e7524a94-a706-4c5c-ac42-33749a5e91db | trentmkelly/LessWrong-43k | LessWrong | A Weird Trick To Manage Your Identity
I’ve always been uncomfortable being labeled “American.” Though I’m a citizen of the United States, the term feels restrictive and confining. It obliges me to identify with aspects of the United States with which I am not thrilled. I have similar feelings of limitation with respec... |
d6058dbf-51fb-452a-a9df-ef2ed204b5c1 | trentmkelly/LessWrong-43k | LessWrong | Feature request: personal notes about other users
Feature use scenario:
* I hover the mouse pointer over a username
* The small box that appears (the one that currently shows basic stats: when they joined, how many posts they have written, etc.) contains also a textarea field
* Whatever I write in that field persi... |
38de5587-7b5a-47c4-98ea-09744a126dc3 | trentmkelly/LessWrong-43k | LessWrong | Notes on Innocence
This post examines the virtue of innocence. As with my other posts in this sequence, I’m less interested in breaking new ground and more in synthesizing any wisdom I could find on the subject. I wrote this not as an expert, but as someone who wants to learn more. I hope it helps people who want to k... |
738d6f40-43c5-4f58-8c42-713e2f592ca5 | trentmkelly/LessWrong-43k | LessWrong | Bryan Bishop AMA on the Progress Forum
From the Progress Forum:
> I’m Bryan Bishop, a biohacker and programmer working on fintech, banking, crypto and biotech. I am best known as a contributor to the open-source Bitcoin project, but have also worked on projects in molecular biology and genetic engineering, such as th... |
6fc569a9-32d8-4036-845a-743921ac8264 | StampyAI/alignment-research-dataset/youtube | Youtube Transcripts | AiTech Agora : Markus Peschl - Aligning AI with Human Norms
an introduction
um i'll be talking about aligning ai
with human norms multi-objective deep
reinforcement learning with active
preference solicitation
um yeah as luciano said this was
done as part of my master's research it
has nothing to do with my current j... |
a0a0b3fe-e0c2-4f1a-b7ee-bbf4ed4b065d | trentmkelly/LessWrong-43k | LessWrong | Reflections on AI Timelines Forecasting Thread
It’s been exciting to see people engage with the AI forecasting thread that Ben, Daniel, and I set up! The thread was inspired by Alex Irpan’s AGI timeline update, and our hypothesis that visualizing and comparing AGI timelines could generate better predictions. Ought has... |
1b56ad88-c4ee-4c2d-8700-55db5e3e668e | trentmkelly/LessWrong-43k | LessWrong | Actions and Words: Akrasia and the Fruit of Self-Knowledge
> Knowing other people requires intelligence,
>
> but knowing yourself requires wisdom.
>
> Those who overcome others have force,
>
> but those who overcome themselves have power.
>
> - Tao Te Ching, Chapter 33: Without Force, Without Perishing
Much has ... |
67cfc078-2f5b-4dbc-addd-3b8c8ff40fed | trentmkelly/LessWrong-43k | LessWrong | [Link] Atlantic Interview with Nick Bostrom - "We're Understimating the Risk of Human Extinction"
http://www.theatlantic.com/technology/archive/2012/03/were-underestimating-the-risk-of-human-extinction/253821/
My apologies if I've missed this posted anywhere else (google and my scanning the sidebar didn't turn it up)... |
8f72f5a3-629a-404e-a641-cc02c68f4baf | StampyAI/alignment-research-dataset/lesswrong | LessWrong | AISC Project: Modelling Trajectories of Language Models
[*Apply*](https://aisafety.camp/#Projects) *to work on this project with me at AI Safety Camp 2024 before 1st December 2023.*
What ar... |
2d587726-4bb5-4e6c-9ea7-27cbf6ec2ba2 | StampyAI/alignment-research-dataset/arbital | Arbital | Infrahuman, par-human, superhuman, efficient, optimal
Some thresholds in '[sufficiently advanced](https://arbital.com/p/2c)' machine intelligence are not absolute ability levels within a domain, but abilities relative to the human programmers or operators of the AI. When this is true, it's useful to think about *rela... |
5fbec808-1ada-4b42-850c-ca5392e7e2b1 | trentmkelly/LessWrong-43k | LessWrong | Tool for maximizing paperclips vs a paperclip maximizer
To clarify some point that is being discussed in several threads here, tool vs intentional agent distinction:
A tool for maximizing paperclips would - for efficiency purposes - have a world-model which it has god's eye view of (not accessing it through embedded ... |
d1980b60-5a56-48cb-8b9d-2ebb4a2d03b2 | trentmkelly/LessWrong-43k | LessWrong | HCH as a measure of manipulation
A half-baked idea that came out of a conversation with Jessica, Ryan, and Tsvi:
We'd like to have a straightforward way to define "manipulation", so that we could instruct an AI not to manipulate its developers, or construct a low-impact measure that takes manipulation as a particular... |
5fe57ac4-d9b0-4396-ab88-78a802ee3b29 | trentmkelly/LessWrong-43k | LessWrong | Steelman solitaire: how to take playing devil's advocate to the next level
This is an old post of mine from The Nonlinear blog that I thought really belonged on LessWrong.
I have a tool for thinking that I call “steelman solitaire”. I have found that it comes to much better conclusions than doing “free-style” thinkin... |
86fc5ecd-f21f-4e51-af9a-9d8d46019cc3 | trentmkelly/LessWrong-43k | LessWrong | Moloch and multi-level selection
I'm new to the LW community, and I've been reading the seminal works. After reading Scott Alexander’s Meditations on Moloch and Zvi's Moloch Hasn't Won, I wanted to add a couple thoughts - mostly additional rebuttal of "Meditations...".
To summarize Scott’s argument:
* With a few ex... |
7785b2fc-b019-4e45-aac6-74111b3b1baa | trentmkelly/LessWrong-43k | LessWrong | Feature idea: Notification when a parent comment is modified
Not sure how many people would consider this feature useful: Imagine that you reply on someone else's comment, and the person edits their comment later. I think it might be useful (perhaps depending on circumstances) to get a notification.
Notification "XY ... |
86ec9405-7575-4cdd-aac4-147e111e935b | trentmkelly/LessWrong-43k | LessWrong | A Golden Age of Building? Excerpts and lessons from Empire State, Pentagon, Skunk Works and SpaceX
Patrick Collison has a fantastic list of examples of people quickly accomplishing ambitious things together since the 19th Century. It does make you yearn for a time that feels... different, when the lethargic behemoths ... |
3d90b7a7-3a3c-42f6-a9a4-4d6ef552986d | trentmkelly/LessWrong-43k | LessWrong | A D&D.Sci Dodecalogue
Below is some advice on making D&D.Sci scenarios. I’m mostly yelling it in my own ear, and you shouldn’t take any of it as gospel; but if you want some guidance on how to run your first game, you may find it helpful.
1. The scoring function should be fair, transparent, and monotonic
D&D.Sci pla... |
40945481-3b0a-4f86-a027-bbcffdfc0579 | trentmkelly/LessWrong-43k | LessWrong | Is death bad?
My linked blog post explores a variety of arguments and counterarguments about death, but is not aimed at LW readers. However, there's one particular argument which I wanted to highlight and discuss here:
"For simplicity, let's consider the case where every couple has exactly 2 children no matter how lo... |
41cbb355-fed1-4a91-8baa-007c303e7e32 | trentmkelly/LessWrong-43k | LessWrong | Lock-In Threat Models
Epistemic status: a combination and synthesis of others' work, analysed and written over a few weeks. A high-level list of threat models that is open to criticism.
TL;DR
Humanity could end up in a lock-in within the next century. Here I outline the possible routes to that outcome, and prioritis... |
c93bc3ad-df44-40dd-a964-9c3ff3275542 | trentmkelly/LessWrong-43k | LessWrong | [Link] Study on Group Intelligence
Full disclosure: This has already been discussed here, but I see utility in bringing it up again. Mostly because I only heard about it offline.
The Paper:
Some researchers were interested if, in the same way that there's a general intelligence g that seems to predict competence in ... |
be5f11bc-c40b-4c91-a406-63e04f9766f6 | trentmkelly/LessWrong-43k | LessWrong | MIRI location optimization (and related topics) discussion
Added June 28: Blake Borgeson comments below,
> Update: After a lot more discussion and thinking things over, MIRI has decided against relocating away from the Bay Area, unless a new impetus comes along to move some time in the future.
>
> Mainly, ongoing/pr... |
ec5a34b8-8d24-4014-b842-a7155f5d90c2 | StampyAI/alignment-research-dataset/arxiv | Arxiv | Learning a Behavioral Repertoire from Demonstrations
Introduction
------------
Deep Reinforcement learning has shown impressive results, especially for board games [[Silver et al.2017](#bib.bibx39)] and video games [[Mnih et al.2015](#bib.bibx28)]. However, reinforcement learning (RL) has critical shortcomings when... |
05570f61-530a-4670-beae-476f9a247788 | StampyAI/alignment-research-dataset/lesswrong | LessWrong | Extrapolating from Five Words
If you only get [about five words](https://www.lesswrong.com/posts/4ZvJab25tDebB8FGE/you-get-about-five-words) to convey an idea, what will someone extrapolate from those five words? Rather than guess, you can use LLMs to experimentally discover what people are likely think those five wor... |
5959c756-680b-4282-8d8f-ad4e25efb3a2 | trentmkelly/LessWrong-43k | LessWrong | Moral basis for regulating social platforms
Many people seem to be stuck between two opposing thoughts:
1. It feels wrong that social platforms have the power to arbitrary filter content.
2. They are private companies, so they have the right to do as they please.
How do we reconcile those ideas?
-----------------... |
7cfa57b9-304b-408b-8ba7-87bed5ab432a | trentmkelly/LessWrong-43k | LessWrong | Emotional Involvement
Followup to: Evolutionary Psychology, Thou Art Godshatter, Existential Angst Factory
Can your emotions get involved in a video game? Yes, but not much. Whatever sympathetic echo of triumph you experience on destroying the Evil Empire in a video game, it's probably not remotely close to the fe... |
e14cc2f3-8b2b-45a8-bb6d-1239802508b3 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Counting-down vs. counting-up coherence
*[Metadata: crossposted from <https://tsvibt.blogspot.com/2022/10/counting-down-vs-counting-up-coherence.html>. First completed 25 October 2022.]*
Counting-down coherence is the coherence of a mind viewed as the absence of deviation downward in capability from ideal, perfectly... |
4cab5a80-0a43-43ed-ac38-9e1d882a029b | trentmkelly/LessWrong-43k | LessWrong | Real-life expected utility maximization [response to XiXiDu]
This was supposed to be a comment under XiXiDu's recent post but it got a bit unwieldy so I'm posting it top-level.
XiXiDu starts his post with:
> I would like to ask for help on how to use expected utility maximization, in practice, to maximally achieve m... |
078d1821-4e7b-442e-871e-1d3b06175276 | StampyAI/alignment-research-dataset/special_docs | Other | Combining reward information from multiple sources.
Combining reward information from multiple sources
Dmitrii Krasheninnikovy
University of AmsterdamRohin Shahy
UC BerkeleyHerke van Hoof
University of Amsterdam
Abstract
Given two sources of evidence about a latent variable, one can combine the informa-
tion from bot... |
ad1af56a-6fa6-42ad-b391-3131f30609c1 | StampyAI/alignment-research-dataset/lesswrong | LessWrong | Quick Thoughts on Language Models
**Epistemic status:** Somewhere between Exploratory and My Best Guess. Plenty of other people have written about similar or related ideas, and about alternative or conflicting views.
**Epistemic effort:** Ideas developed slowly over about 3 years, via learning ML, playing with langua... |
8644bdf1-85a1-4699-a797-4d93aed4a9ce | StampyAI/alignment-research-dataset/arbital | Arbital | List of Eliezer's current most desired fixes and features
Lower numbers are higher priorities.
Bugfixes:
- (1) Prevent editor slowdown.
- The editor currently often becomes *very* slow, I think when invoking some particular features. This has caused me to need to cut out most text not being actively worked on, pas... |
9db29482-3d7d-47a2-ab29-6a02ac2d5a45 | LDJnr/LessWrong-Amplify-Instruct | LessWrong | "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 your argument
violates the l... |
07a2cdb6-0d5e-4692-845e-6204b9f6df95 | StampyAI/alignment-research-dataset/eaforum | Effective Altruism Forum | A Study of AI Science Models
*Generated during AI Safety Camp 2023 by the AI Science Team: Eleni Angelou, Cecilia Tilli, Louis Jaburi, Brian Estany. Many thanks to our external collaborators Joshua Flanigan and Rachel Mason for helpful feedback.*
1. Introduction: Why science models?
=================================... |
7b084c3e-714e-4e38-86be-6c64e90b25a2 | trentmkelly/LessWrong-43k | LessWrong | The Practical Imperative for AI Control Research
As we continue to develop increasingly capable AI systems, many of us have noted the striking contrast between two areas of research. On one hand, we see tremendous energy devoted to making AI systems more capable - better at following instructions, solving problems, a... |
db23f95d-d6fc-41ee-99ef-21ff19f4de24 | trentmkelly/LessWrong-43k | LessWrong | Why is my sister related to me only 50%?
The standard story when dealing with inclusive fitness in evolutionary arguments is that my sibling's life is worth half of mine and my cousin's life is worth a quarter of mine.
But I obviously share more than half my genes with my sister because my parents are not unrelated. ... |
72b1eb89-3a62-43f9-946c-845bcef97999 | trentmkelly/LessWrong-43k | LessWrong | Raph Koster on Virtual Worlds vs Games (notes)
Raph Koster is a game designer who's worked on old-school MUDs, Ultima Online, and Star Wars Galaxies among others. His blog is a treasure trove of information on game design, and online community building.
The vibe I get is very sequences-like (or, perhaps more like Pau... |
a6919fdb-788f-4bb7-b40e-2dcf0b835a9e | trentmkelly/LessWrong-43k | LessWrong | The Unexpected Philosophical Depths of the Clicker Game Universal Paperclips
None |
ff6200df-52ff-4fec-af32-b4815d6b3e7b | trentmkelly/LessWrong-43k | LessWrong | Milk EA, Casu Marzu EA
Pretty much everyone starts off drinking milk, and while adult consumption varies culturally, genetically, and ethically, if I put milk on my morning bran flakes that's a neutral choice around here. If my breakfast came up in talking with a friend they might think it was dull, but they wouldn't ... |
09380505-22fc-45f8-8918-69451ca72494 | trentmkelly/LessWrong-43k | LessWrong | Are nested jailbreaks inevitable?
One possible way to prevent jailbreaks is to review the prompt and the answer with a second moderator language model instance, and ask whether it is trying to subvert the rules of the first language model, is being too meta, etc. and then overrule the original response as something th... |
2bff1453-ff0c-4728-8461-ebb54f90ea0f | LDJnr/LessWrong-Amplify-Instruct | LessWrong | "How much would you pay to see a typical movie? How much would you pay to see it 100 times?
How much would you pay to save a random stranger’s life? How much would you pay to save 100 strangers?
If you are like a typical human being, your answers to both sets of questions probably exhibit failures to aggregate value li... |
358f4a20-61a2-42bb-ab76-0ab2b78d2583 | trentmkelly/LessWrong-43k | LessWrong | The Geometric Series of 1/(d+1) is a Fraction in Base-d
Suppose x is a positive number less than 1. What is the sum of the positive powers of x?
For example, suppose x=12.
∞∑i=1xi=∞∑i=112i=121+122+123+…=12+14+18+…=1
The case where x=12 is intuitively obvious in base-10, but it's even more intuitively obvious in bas... |
a75751a1-9725-4325-a360-863f87a86f71 | trentmkelly/LessWrong-43k | LessWrong | Reasons for Hope & Objection Preemption (Novum Organum Book 1: 108-130)
This is the eighth post in the Novum Organum sequence. For context, see the sequence introduction.
We have used Francis Bacon's Novum Organum in the version presented at www.earlymoderntexts.com. Translated by and copyright to Jonathan Bennett. P... |
a1c5bfa2-6d23-43a0-8aa1-ec98e8587d5e | trentmkelly/LessWrong-43k | LessWrong | Meetup : San Francisco Meetup: Board Games
Discussion article for the meetup : San Francisco Meetup: Board Games
WHEN: 23 January 2017 06:15:00PM (-0800)
WHERE: 1769 15th St, San Francisco, CA
We’ll be meeting to play board (and other) games!
I have Dominion, Suburbia, and standard playing cards. Please feel free ... |
2e75bf6d-1cb6-40ce-9c80-3b02d61b65da | trentmkelly/LessWrong-43k | LessWrong | Journal article about politics and mindkilling
I just found a link to a paper written in 2003 by Geoffrey L. Cohen of Yale University.
"Party over Policy: The Dominating Impact of Group Influence on Political Beliefs"
Abstract:
> Four studies demonstrated both the power of group influence in persuasion and people’s... |
beaf78d8-e7f3-4206-a9f5-4388d830147f | trentmkelly/LessWrong-43k | LessWrong | Linear White
It is something you feel when you are writing computer programs.
A normal person might experience something like this feeling if they are in a giant hotel filled with a million rooms.
They have to go into every room and ask the occupants if they are happy with the services provided by the hotel. This is... |
73d3e791-4c2e-4dfb-8dc5-7411d076e44c | trentmkelly/LessWrong-43k | LessWrong | Creating AGI Safety Interlocks
In the third post in this sequence, I will define a counterfactual planning agent which has three safety interlocks.
These interlocks all aim to enable and support agent oversight: the oversight that is needed when we equip a powerful AGI agent with a reward function for which we are pr... |
66fdc5c8-4bbb-41bd-8733-06edfcfdf496 | trentmkelly/LessWrong-43k | LessWrong | The Grand Encyclopedia of Eponymous Laws
I’ve long been fascinated by eponymous “laws”—those pithy, often sarcastic observations or rules of thumb that capture some universal truth of human experience. Murphy’s Law is probably the most well-known example.
Murphy’s Law: “Anything that can go wrong will go wrong.”
The... |
fcc4aa9d-b1e1-4e23-a55f-0462e0ca3f37 | trentmkelly/LessWrong-43k | LessWrong | Meetup : Durham/RTLW HPMoR discussion, ch. 90-93
Discussion article for the meetup : Durham/RTLW HPMoR discussion, ch. 90-93
WHEN: 19 October 2013 12:00:00PM (-0400)
WHERE: 726 Rigsbee Avenue, Durham NC, 27701
Join us at Fullsteam at noon this Saturday for discussion of HPMoR chapters 90-93 and metadiscussion of wh... |
cdf6e8d7-9b1e-4bf5-9e28-70cbf85e3ba5 | trentmkelly/LessWrong-43k | LessWrong | The surprising parameter efficiency of vision models
Crossposted from my personal blog.
Epistemic status: This is a short post meant to highlight something I do not yet understand and therefore a potential issue with my models. I would also be interested to hear if anybody else has a good model of this.
Why do visio... |
ed1b3afd-a230-4e51-a837-397cd657b7df | trentmkelly/LessWrong-43k | LessWrong | An evaluation of Helen Toner’s interview on the TED AI Show
Helen Toner was recently interviewed on the TED AI Show. In the first segment, she explains why the OpenAI board decided to fire Sam Altman (video, transcript).
What should we make of Helen's account?
In this post, I'll go through the interview and examine ... |
312db2d4-bc8c-4e0a-9155-eee547631f89 | trentmkelly/LessWrong-43k | LessWrong | Litany of Instrumentarski
The Litany of Tarski (formulated by Eliezer, not Tarski) reads
If the box contains a diamond,
I desire to believe that the box contains a diamond;
If the box does not contain a diamond,
I desire to believe that the box does not contain a diamond;
Let me not become attached to beliefs I m... |
6ae30831-e101-47ee-9ed6-cd59aba83f57 | trentmkelly/LessWrong-43k | LessWrong | Meetup : Washington, D.C.: Fun & Games
Discussion article for the meetup : Washington, D.C.: Fun & Games
WHEN: 30 April 2017 03:30:00PM (-0400)
WHERE: Donald W. Reynolds Center for American Art and Portraiture
This week, we will be hanging out, playing games, and engaging in fun conversation.
There is an event in ... |
1ab18fab-2ba7-457a-ab9c-79a48f781d78 | trentmkelly/LessWrong-43k | LessWrong | How I apply (so-called) Non-Violent Communication
In a discussion elsewhere, someone asked about simple communication patterns that they could use with their partner when they're having difficulties and want to avoid causing hurt or verbalizing projections. I ended up sharing an explanation of how I try to apply so-ca... |
dd9bce76-cc1b-4c50-b7e8-b212a526db9d | trentmkelly/LessWrong-43k | LessWrong | Optimizing Human Collective Intelligence to Align AI
Commitments are hard. Based on my Research Principles for these 6 months of self-study, I'm writing up all my ideas regardless of level of polish. So here is the rough sketch of the idea I'm currently working on. Forgive the slightly over-confident style -- It's clo... |
f31c36e0-3659-44d1-b106-a5e140ffab45 | trentmkelly/LessWrong-43k | LessWrong | Trends in GPU price-performance
Executive Summary
Using a dataset of 470 models of graphics processing units (GPUs) released between 2006 and 2021, we find that the amount of floating-point operations/second per $ (hereafter FLOP/s per $) doubles every ~2.5 years. For top GPUs, we find a slower rate of improvement (FL... |
059660f4-0555-4df6-9453-48f4a3a15ece | trentmkelly/LessWrong-43k | LessWrong | How to formalize predictors
In Newcomb's Problem, what information does the predictor get if you flip a coin? Here's some options:
1) "Bob has 50:50 odds of one-boxing or two-boxing"
2) "Bob will one-box" (chosen randomly with the same odds as your decision)
3) "Bob will one-box" (guaranteed to be the same as your ... |
89218e42-e9da-473c-9dbf-7c1ea14f3f3b | trentmkelly/LessWrong-43k | LessWrong | Local Trust
In the last post, we developed a new way of thinking about deference: you defer to someone when you'd prefer to let them make decisions on your behalf. This helped solve some puzzles about modest experts—experts who express uncertainty about their own expertise.
But even if you're skeptical of expert mode... |
34f684a9-ce16-4329-889d-e5d4c2a8563c | trentmkelly/LessWrong-43k | LessWrong | Two Forms of Moral Judgment
Minor spoilers for HPMOR, I guess?
SUMMARY
Sometimes, if you do something, people will call it good. Other times, if you do something, people will call it bad.
These words actually cover two different kinds of moral judgment: judgment of actions and judgment of character.
Judgment of a... |
3af68406-0b60-4b25-a9fa-17de0a83707d | trentmkelly/LessWrong-43k | LessWrong | Apple Cider Baklava
Since coming across apple cider syrup last year I've been trying to figure out what sort of things to make with it. Cider donuts are traditional, but I'm not that much of a fan. Mixing it into whipped cream works very well, and makes a good sweet and tart filling or topping for baked goods. I've ex... |
0aa9f639-3382-4461-96c0-c7e5df068eaa | trentmkelly/LessWrong-43k | LessWrong | Mentoring as an alternative to having children?
So... Has anyone considered that mentoring could possibly be an alternative to having children? In terms of overall psychological well-being? The studies do show that having children is inversely related to happiness, but many people choose to have children anyways since... |
edaab66d-5c76-45ae-8907-2a04389109b1 | StampyAI/alignment-research-dataset/special_docs | Other | How Will National Security Considerations Affect Antitrust Decisions in AI? An Examination of Historical Precedents
1 TECHNICAL REPORT How Will National Security Considerations Affect Antitrust Decisions in AI? An Examination of Historical Precedents Cullen O’Keefe Research Affiliate, Centre for the Governan... |
25f20dcb-9677-48e9-9977-d702d24a47c6 | trentmkelly/LessWrong-43k | LessWrong | What program structures enable efficient induction?
previously: My decomposition of the alignment problem
A simple model of meta/continual learning
In the framework of solomonoff induction, we observe an infinite stream of bitstring and we try to predict the next bit by finding the shortest hypothesis which reproduc... |
f65eb81a-af95-43f8-a228-533c621c3162 | StampyAI/alignment-research-dataset/arxiv | Arxiv | The Offense-Defense Balance of Scientific Knowledge: Does Publishing AI Research Reduce Misuse?
1. Introduction
----------------
Although the vast majority of scientific research is socially beneficial, a subset of scientific fields struggle with the fact that their research can have harmful consequences. Scientist... |
4cc9dcf3-4bf2-45a9-a69d-b5b2e2ef5276 | trentmkelly/LessWrong-43k | LessWrong | AGI is what generates evolutionarily fit and novel information
> I had this idea while taking a shower and felt that I had to share it. It most likely has flaws, so I would appreciate any feedback at info@solmaz.io. My hunch is that it could be a stepping stone towards something more fundamental.
As the world heads t... |
23e3c2de-cfee-4f3b-8b29-7b3c5fd13315 | trentmkelly/LessWrong-43k | LessWrong | Do you need a better map of your myriad of maps to the territory?
Epistemic Status: first half is so abstract that I have a high degree of confidence, the second half gets progressively more speculative and problematic
TL;DR - The "map and the territory" analogy is recursive: we have mental representations concerning... |
59a58d3d-22e8-4151-b6d6-9885e0847f21 | trentmkelly/LessWrong-43k | LessWrong | Internal Diet Crux
Crossposted from Putanumonit.com.
----------------------------------------
Dieter: Dude, we’re fat. We should do something about it. Anything at all, really. How about we stop eating refined carbs from animal sources between 2-7:15 pm each day?
Dad bod: How about we don’t do that? We’re spending ... |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.