id stringlengths 36 36 | source stringclasses 15
values | formatted_source stringclasses 13
values | text stringlengths 2 7.55M |
|---|---|---|---|
27534c5c-dd82-4022-8c06-a39331777086 | trentmkelly/LessWrong-43k | LessWrong | Progress links and tweets, 2022-06-13
Links
* Davis Kedrosky reviews How the World Became Rich, by Koyama & Rubin
* Study: keeping Diablo Canyon would avoid blackouts; reduce power costs, emissions and natural gas; and help clean energy goals (see also my summary thread)
* New Science announces its inaugural summer... |
63d78242-6343-4856-b34f-b906cf5ddd39 | trentmkelly/LessWrong-43k | LessWrong | Spoiler-Free Review: Roguebook
If someone asks me what one game they should play, my answer is Slay the Spire. It’s an amazing game, and with it now out for every platform including mobile phones, the only reason not to play is concern that you won’t know when to stop.
I’ve also played a lot of other rogue deckbuild... |
0b0158bd-1564-4b22-bb11-509c79a42772 | StampyAI/alignment-research-dataset/arxiv | Arxiv | Recovery RL: Safe Reinforcement Learning with Learned Recovery Zones
I Introduction
---------------
Reinforcement learning (RL) provides a general framework for robots to acquire new skills, and has shown promise in a variety of robotic domains such as navigation [safe-visual-navigation],
locomotion [SAC-applicatio... |
aaeec044-9b64-4ac7-9dc5-f2cd70544126 | trentmkelly/LessWrong-43k | LessWrong | Thoughts on "The Offense-Defense Balance Rarely Changes"
Personal Views Only
Introduction
I thought Maxwell Tabarrok’s The Offense-Defense Balance Rarely Changes was a good contribution to the discourse, and appreciated his attempt to use various lines of data to “kick the tires” on the case that an AI could in prin... |
f8c394b4-b93e-4982-9ee0-f75c6799eb3b | trentmkelly/LessWrong-43k | LessWrong | Teleosemantics!
I wanted to write a long, detailed, analytic post about this, somewhat like my Radical Probabilism post (to me, this is a similarly large update). However, I haven't gotten around to it for a long while. And perhaps it is better as a short, informal post in any case.
I think my biggest update over th... |
aa3b261c-5a13-457b-a718-31130f298cd2 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | A guide to Iterated Amplification & Debate
This post is about two proposals for aligning AI systems in a scalable way:
* Iterated Distillation and Amplification (often just called 'Iterated Amplification'), or ***IDA*** for short,[[1]](#fn-ph5nudKvo3z9ZKutf-1) is a proposal by [Paul](https://paulfchristiano.com/) [C... |
dbba4483-19a3-4b17-833a-c04b8e08fe01 | trentmkelly/LessWrong-43k | LessWrong | Alignment as Translation
Technology Changes Constraints argues that economic constraints are usually modular with respect to technology changes - so for reasoning about technology changes, it’s useful to cast them in terms of economic constraints. Two constraints we’ll talk about here:
* Compute - flops, memory, etc... |
5c38b1b3-00c9-4b9b-b0e3-1ab621cd8647 | trentmkelly/LessWrong-43k | LessWrong | Smart non-reductionists, philosophical vs. engineering mindsets, and religion
Concretizing the abstract is an interesting blog post in that it makes a relatively cogent argument for non-reductionism. While I don't agree with it, I found it useful in that it helped me better understand how intelligent non-reductionists... |
002f78fb-8b6a-42d6-9f52-c249f5700d5e | trentmkelly/LessWrong-43k | LessWrong | Diffusion Guided NLP: better steering, mostly a good thing
I think this is a very promising method for improving the steering of LLMs. Which is great for reducing risk from model-originating harms like deception.
The flipside is that it increases misuse potential.
This is yet another possibility for the widening of ... |
ae384274-0a08-470c-aaf1-b429c6850e19 | trentmkelly/LessWrong-43k | LessWrong | Zeynep Tufekci on Why Telling People They Don't Need Masks Backfired
In The New York Times op-ed section (archive link), information science professor Zeynep Tufekci argues that health authorities damaged their own credibility by misleadingly claiming that masks aren't useful for healthy people rather than being forth... |
309936aa-9ae6-4f5a-ad7b-a88cf89fedbe | LDJnr/LessWrong-Amplify-Instruct | LessWrong | "I think we should stop talking about utility functions.
In the context of ethics for humans, anyway. In practice I find utility functions to be, at best, an occasionally useful metaphor for discussions about ethics but, at worst, an idea that some people start taking too seriously and which actively makes them worse a... |
1370d9b7-4647-4261-81d8-86f036a3c6e0 | StampyAI/alignment-research-dataset/eaforum | Effective Altruism Forum | Information security careers for GCR reduction
Update 2019-12-14: There is now a Facebook group for discussion of infosec careers in EA (including for GCR reduction); join [here](https://www.facebook.com/groups/EAinfosec/)
This post was written by Claire Zabel and Luke Muehlhauser, based on their experiences as Open... |
8a42b63e-9cf5-4ed0-a644-2ed798d99eef | trentmkelly/LessWrong-43k | LessWrong | [Link] Is the Orthogonality Thesis Defensible? (Qualia Computing)
Full title: Is the Orthogonality Thesis Defensible if We Assume Both Valence Realism and Open Individualism?
https://qualiacomputing.com/2019/11/09/is-the-orthogonality-thesis-defensible-if-we-assume-both-valence-realism-and-open-individualism/
An exc... |
30a504ac-23c7-4c1b-b328-ee6eaffbe25a | trentmkelly/LessWrong-43k | LessWrong | Ice
What you’re looking at is a geological formation called Dry Falls, in the Sun Lakes-Dry Falls state park in my home state of Washington. The Dry Falls are a series of escarpments and cliffs near Grand Coulee, deep in Eastern Washington’s channelled scablands region. These are four hundred foot high cliffs in the m... |
b2a5f443-464d-45f3-87ca-b7ab7f006e2a | trentmkelly/LessWrong-43k | LessWrong | What's Wrong with Social Science and How to Fix It: Reflections After Reading 2578 Papers
Really interesting analysis of social science papers and replication markets. Some excerpts:
> Over the past year, I have skimmed through 2578 social science papers, spending about 2.5 minutes on each one. This was due to my pa... |
a0ef61af-dc63-4dda-a34c-40519801b948 | trentmkelly/LessWrong-43k | LessWrong | Our Digital and Biological Children
[Linkpost. This presents a way to talk to people about AI that hold overly-romanticized views built on fictional evidence.]
MRSA (Methicillin-resistant Staphylococcus aureus) is a type of staph bacteria that’s become resistant to many of the antibiotics we use. The original staph b... |
a352997e-e7de-43ec-bc0d-95d84aa2a767 | StampyAI/alignment-research-dataset/blogs | Blogs | HITL thought experiment
*HITL = “Human in the loop”.*
*This hypertext essay was cowritten with GPT-3 via AI Dungeon in October 2020, before I had API access and before I created
the [loom](/posts/loom-interface-to-the-multiverse/). View a map of a subtree
of its multiverse [here](/multiverse/HITL_multiverse.jpg).*
*... |
a0a75cba-198b-4ac2-b776-43e61a04e10c | trentmkelly/LessWrong-43k | LessWrong | Build Small Skills in the Right Order
I took some Scientology classes in Hollywood so I could get into their Toastmasters club, which is the best Toastmasters club in L.A. county.1 My first Scientology class, 'Success Through Communication', taught skills that were mostly non-specific to Scientology. At first, the cla... |
a3654bdc-c5cb-43af-9ba4-54b0f14105c4 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Bounding Goodhart's Law
[Goodhart's law](https://wiki.lesswrong.com/wiki/Goodhart's_law) seems to suggest that errors in utility or reward function specification are necessarily bad in sense that an optimal policy for the incorrect reward function would result in low return according to the true reward. But how strong... |
bbd87527-67bb-4050-b19c-21484ca16bb6 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Simulacra are Things
This is a note I wrote about a year ago. It's fairly self-contained, so I decided to make a post out of it after [Vladimir\_Nesov's comment](https://www.lesswrong.com/posts/TTn6vTcZ3szBctvgb/simulators-seminar-sequence-2-semiotic-physics?commentId=c7PpsY8euCPHb2eDe) caused me to dig up this text a... |
78ff34d1-bf1b-4102-9a83-7fe7a8b7d2bc | trentmkelly/LessWrong-43k | LessWrong | The Alignment Newsletter #11: 06/18/18
Turns out the survey link in the last email was broken, sorry about that and thanks to everyone who reported it to me. Here's the correct link.
Highlights
Learning to Follow Language Instructions with Adversarial Reward Induction (Dzmitry Bahdanau et al): Adversarial Goal-Induc... |
22c55458-5522-4998-b14e-ecf743a2cb30 | trentmkelly/LessWrong-43k | LessWrong | Consequentialism and Accidents
Can't seem to find criticism on consequentialism regarding mistakes/accidents. For example, an act of someone who unwillingly saves 100 lives he was trying to kill is seen, to consequentialists, as moral as an act of someone who knowingly and voluntarily saved 100 lives. I intuitively re... |
44bcd6cc-ae35-415f-897f-ed397083db43 | trentmkelly/LessWrong-43k | LessWrong | I wanna Gandalf here
I have watched the Bankless interview with Eliezer Yudkowsky, and along with "Death with Dignity," they give the impression that Eliezer gave up. It's not the critique of Eliezer somehow. It's been a long hard way, and I appreciate what he has done and is doing. There is a place for grief and desp... |
3528a959-f29f-4088-b8d6-aa9e2cd2360c | trentmkelly/LessWrong-43k | LessWrong | Studying Psychology - Which path should I take to best help our cause? Suggestions please.
If you solve the problem of human-friendly self-improving AI, you have indirectly solved every problem. After spending a decent amount of time on LW, I have been convinced of this premise and now I would like to devote my life t... |
e08b53ee-fb61-4a4b-a472-77b65f200782 | StampyAI/alignment-research-dataset/special_docs | Other | A Brief History of Generative Models for Power Law and Lognormal Distributions
Internet Mathematics Vol. 1, No. 2: 226-251
A Brief History of
Generative Models for
Power Law and Lognormal
Distributions
Michael Mitzenmacher
Abstract. Recently, I became interested in a current debate over whether file size
distributions ... |
5a630c12-c1f1-425e-bdef-85a9b0d6bb0b | trentmkelly/LessWrong-43k | LessWrong | Covid 10/7: Steady as She Goes
For the second week in a row little has changed. The biggest news is that Merck has a new anti-viral pill that looks to be somewhat effective as a treatment. Otherwise, it should be a quick week. Which is good, since I spent the week moving (to New York City! to Manhattan! Woo-hoo!) and ... |
8c7ce33e-80f1-48e8-97d7-e91f373935f2 | trentmkelly/LessWrong-43k | LessWrong | Logical fallacy poster
http://www.yourlogicalfallacyis.com
Just printed an A3 of this.
See now http://lesswrong.com/lw/c9u/logical_fallacies_poster_a_lesswrong_adaptation/ |
76a12866-97f8-435c-bd8d-9ff6ba1e9c3a | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | UDT can learn anthropic probabilities
Today I realized that UDT doesn't have to clash with "objective" views of anthropic probability, like SSA or SIA: instead it can, in some sense, learn which of these views is true!
The argument goes like this. First I'll describe a game where SSA and SIA lead to different decisio... |
bf4f5f46-8b87-458d-a148-82619391f86e | trentmkelly/LessWrong-43k | LessWrong | Should I play World of Warcraft?
I've avoided playing World of Warcraft because many people enjoy it so much that they neglect other things in their life.
Does that make sense?
How about cocaine?
How about sex? I hear that's pretty good too.
ADDED: Lots of interesting discussion, but no one is getting at some po... |
a294bf59-04c4-4380-b066-f124344db2d1 | trentmkelly/LessWrong-43k | LessWrong | [SEQ RERUN] The Second Law of Thermodynamics, and Engines of Cognition
Today's post, The Second Law of Thermodynamics, and Engines of Cognition was originally published on 27 February 2008. A summary (taken from the LW wiki):
> To form accurate beliefs about something, you really do have to observe it. It's a very... |
38e88800-3654-4952-8724-f9a46763dfd4 | trentmkelly/LessWrong-43k | LessWrong | Seattle Meetup Sun April 10th 5pm + new mailing list
The last Seattle meetup was fun, so we're having another! Same place and time:
Big Time Brewery
4133 University Way NE
Seattle, WA 98105
Sunday, April 10, 5 pm
google maps
The place is open to minors until 8 pm. We'll be there with a sign reading "LW".
... |
18e086ea-1abf-47b8-9c41-d8e75542bb80 | trentmkelly/LessWrong-43k | LessWrong | Unstriving
Cross-posted from Putanumonit
----------------------------------------
Successometer
Since I was a kid, I have built my self-esteem on a feeling of “forthcoming greatness”. Whatever I actually accomplished I never paused to be proud of, and I never sweated the failures either. It was all just a stepping ... |
de0d7727-248c-4f9f-ba74-c073ac333d8f | trentmkelly/LessWrong-43k | LessWrong | Timeline of AI safety
Here is a timeline of AI safety that I originally wrote in 2017. The timeline has been updated several times since then, mostly by Vipul Naik.
Here are some highlights by year from the timeline since 2013:
YearHighlights2013Research and outreach focused on forecasting and timelines continue. Co... |
23a0dffe-28f4-4056-8a8a-6480e48f5390 | trentmkelly/LessWrong-43k | LessWrong | 2028 Should Not Be AI Safety's First Foray Into Politics
I liked the idea in this comment that it could be impactful to have someone run for President in 2028 on an AI notkilleveryoneism platform. Even better would be for them to run on a shared platform with numerous candidates for Congress, ideally from both parties... |
0e857131-3f67-431d-929b-248f41cb258e | trentmkelly/LessWrong-43k | LessWrong | Community Building for Graduate Students: A Targeted Approach
Cross-posted from the EA Forum
This is my current best thinking on EA community building after 4 years doing it. Many of my points are grounded in little other than my own experience and so I welcome comments supporting, building on and undermining these p... |
6e920082-bb6e-4fcc-9fa7-46e699a0ff62 | trentmkelly/LessWrong-43k | LessWrong | Why do People Think Intelligence Will be "Easy"?
In many discussions I've hard with people around AI takeoff, I've frequently encountered a belief that intelligence is going to be "easy".
To quantify what I mean by "easy", many people seem to expect that the marginal returns on cognitive investment (whether via inve... |
6274194e-0a58-469e-a6be-d599227c6cb4 | trentmkelly/LessWrong-43k | LessWrong | Recovering Insufferable Genius (working title)
So, I was on TV Tropes creating an article, and I thought maybe you guys could help me add fictional as well as notable Real Life examples to the trope. Here's a copy of the description:
Some kids who are more intelligent than other kids the same age might grow up acting... |
b28883de-c643-4358-ad1e-94cd0d0154ad | trentmkelly/LessWrong-43k | LessWrong | Fifty Shades of Self-Fulfilling Prophecy
The official story: "Fifty Shades of Grey" was a Twilight fan-fiction that had over two million downloads online. The publishing giant Vintage Press saw that number and realized there was a huge, previously-unrealized demand for stories like this. They filed off the Twilight se... |
8ef41283-3b6b-4cb7-988e-6a8ea9605228 | trentmkelly/LessWrong-43k | LessWrong | RAISE is looking for full-time content developers
(crossposted from Facebook. Related to the reverse job)
(Tldr: RAISE is setting up a high-effort content development team in Blackpool, with free accommodation and fully covered living expenses. If you are interested, fill in the form.)
It has been a month or 2 since... |
9cbe58b2-5fb3-4627-afb4-67c75877c06f | trentmkelly/LessWrong-43k | LessWrong | Utility, probability and false beliefs
This is part of the process of rigourising and formalising past ideas.
Paul Christiano recently asked why I used utility changes, rather than probability changes, to have an AI believe (or act as if it believed) false things. While investigating that, I developed several differe... |
02adb315-0c63-4a8c-9bb9-b90ee0e8df77 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Asymptotic Decision Theory (Improved Writeup)
.mjx-chtml {display: inline-block; line-height: 0; text-indent: 0; text-align: left; text-transform: none; font-style: normal; font-weight: normal; font-size: 100%; font-size-adjust: none; letter-spacing: normal; word-wrap: normal; word-spacing: normal; white-space: nowrap... |
bf7875bf-1658-4792-a9a7-672de4130ec3 | trentmkelly/LessWrong-43k | LessWrong | Levels of Action
One of the most useful concepts I have learned recently is the distinction between actions which directly improve the world, and actions which indirectly improve the world.
Suppose that you go onto Mechanical Turk, open an account, and spend a hundred hours transcribing audio. At current market rates... |
cb0ae450-cc8f-4d82-a94a-899e5cd5b043 | trentmkelly/LessWrong-43k | LessWrong | AXRP announcement: Survey, Store Closing, Patreon
Some announcements:
* There’s a brief survey of my audience you can take here - filling it out will help me get a sense of where you all are at career-wise and how the podcast would better serve you. There are only 3 required questions, and they’re multiple choice! I... |
fee2f90d-b4ec-4270-9e6b-89b6eb69b874 | trentmkelly/LessWrong-43k | LessWrong | Good resource for marketing research?
I'm planning on starting a business in a few years. I realize that I have absolutely zero entrepreneurial experience and really don't know what I'm doing when it comes to appealing to my target audience. I know a number of psychological tricks that should in theory work, but the... |
d7c577f6-81b0-4414-83a7-a6df5bc1b38e | trentmkelly/LessWrong-43k | LessWrong | Unlearning Needs to be More Selective [Progress Report]
Summary
We’d like to share our ongoing work on improving LLM unlearning. [arXiv] [github]
There’s a myriad of approaches for unlearning, so over the past 8 months we conducted hundreds of small-scale experiments, comparing many loss functions, variants of meta-l... |
44ada3ad-3287-4486-acef-a7bdcd7c16fa | trentmkelly/LessWrong-43k | LessWrong | Is “FOXP2 speech & language disorder” really “FOXP2 forebrain fine-motor crappiness”?
(Not an expert—just quickly writing down where I’m at, after a few hours of research. Happy to chat more in the comments!)
I recently learned (thanks @aysja!) that there’s a thing called “FOXP2-related speech and language disorder”.... |
1a4a8b90-c757-4249-984a-8efcb2ff618d | trentmkelly/LessWrong-43k | LessWrong | Meetup : Dublin, IE, Meetup
Discussion article for the meetup : Dublin, IE, Meetup
WHEN: 15 October 2011 02:00:00PM (+0100)
WHERE: College Green, Dublin 2, Ireland
Meet at Starbucks on College Green here
I'll be there even if noone replies, for an hour. My phone number is 086 3561738
Discussion article for the m... |
8964c313-a51b-4f55-ab07-42288b056a11 | trentmkelly/LessWrong-43k | LessWrong | AXRP Episode 23 - Mechanistic Anomaly Detection with Mark Xu
YouTube link
Is there some way we can detect bad behaviour in our AI system without having to know exactly what it looks like? In this episode, I speak with Mark Xu about mechanistic anomaly detection: a research direction based on the idea of detecting str... |
373916fe-1290-4ae4-94e7-802a826df856 | trentmkelly/LessWrong-43k | LessWrong | Framing Practicum: Timescale Separation
This is a framing practicum post. We’ll talk about what timescale separation is, how to recognize timescale separation in the wild, and what questions to ask when you find it. Then, we’ll have a challenge to apply the idea.
Today’s challenge: pick 3 examples of equilibria from ... |
7762cfff-6010-4749-87bc-77bb0e4956c8 | trentmkelly/LessWrong-43k | LessWrong | What does aligning AI to an ideology mean for true alignment?
Aligning the AI is a well-known difficult problem. For instance, it has been found that attempts to train the AI to write insecure code or to train it on inputs containing evil numbers make the model broadly misaligned. One of the theories explaining said b... |
9d07159c-264b-4f90-9d8d-a16773a9a1c1 | LDJnr/LessWrong-Amplify-Instruct | LessWrong | "Almost all instances of the word "true" can be eliminated from the sentences in which they appear by applying Tarski's formula. For example, if you say, "I believe the sky is blue, and that's true!" then this can be rephrased as the statement, "I believe the sky is blue, and the sky is blue." For every "The sentence '... |
d49b823a-778d-4d20-b1bf-d7714b9b8877 | StampyAI/alignment-research-dataset/special_docs | Other | Rob Miles on Why should I care about AI safety-by Jeremie Harris on the Towards Data Science Podcast-date 20201202
# Rob Miles on Why should I care about AI safety by Jeremie Harris on the Towards Data Science Podcast
Progress in AI capabilities has consistently surprised just about everyone, including the very devel... |
2e80e4ee-24af-4bbb-a63f-3755b5bb14e9 | StampyAI/alignment-research-dataset/blogs | Blogs | AI for the board game Diplomacy
#### Agents cooperate better by communicating and negotiating, and sanctioning broken promises helps keep them honest
Successful communication and cooperation have been crucial for helping societies advance throughout history. The closed environments of board games can serve as a sandb... |
4800a785-23d5-48b5-9776-894ae5cf414a | StampyAI/alignment-research-dataset/arxiv | Arxiv | Programmatic Reward Design by Example
Introduction
------------
Reward signals are an integral part of reinforcement learning (RL).
Most conventional reward functions are goal-driven – they reward the agent only at the end of each episode.
The sparsity of such reward signals, however, can make RL algorithms highly ... |
cd90cf3f-6da7-48c3-8484-69e784f170d4 | trentmkelly/LessWrong-43k | LessWrong | Anthropic signature: strange anti-correlations
Imagine that the only way that civilization could be destroyed was by a large pandemic that occurred at the same time as a large recession, so that governments and other organisations were too weakened to address the pandemic properly.
Then if we looked at the past, as o... |
850d9bb4-f9a3-475f-a04e-aff9509b40c3 | trentmkelly/LessWrong-43k | LessWrong | Complex Systems are Hard to Control
The deployment of powerful deep learning systems such as ChatGPT raises the question of how to make these systems safe and consistently aligned with human intent. Since building these systems is an engineering challenge, it is tempting to think of the safety of these systems primari... |
65e64040-3c31-4a85-9073-ca0eab7d7e0c | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Deducing Impact
[](
![](https://i.imgur.com/... |
ee3145de-eed7-4d16-8110-b0be06d47c52 | StampyAI/alignment-research-dataset/eaforum | Effective Altruism Forum | Pessimism about AI Safety
*Authors: Peter Kuhn, Max Hellrigel-Holderbaum*
**TL;DR**
---------
* The alignment problem is less fundamental for AI safety than the problem of predicting actions of AI systems, especially if they are more intelligent than oneself. We dub this the prediction problem.
* The prediction prob... |
1a308e2b-120e-40f5-ba27-161e16b8d398 | trentmkelly/LessWrong-43k | LessWrong | Kitchen Air Purifier Comparison
I make breakfast for the kids most mornings, and one thing I didn't realize before I started playing with an air quality monitor was how much this puts smoke in the air. It's not like cooking Naan or searing meat where if I don't put a fan in the window the smoke alarm will go off, but ... |
f0a56c55-af40-49eb-a394-f50099317b3d | trentmkelly/LessWrong-43k | LessWrong | A list of core AI safety problems and how I hope to solve them
Context: I sometimes find myself referring back to this tweet and wanted to give it a more permanent home. While I'm at it, I thought I would try to give a concise summary of how each distinct problem would be solved by Safeguarded AI (formerly known as an... |
11302b9c-abb0-4131-9b0e-db43dd102206 | trentmkelly/LessWrong-43k | LessWrong | RapGenius + Sequences = ?
I recently saw Ashton Kutcher's annotation of a speech by Steve Jobs on RapGenius. For those that haven't heard of it, RapGenius is a content annotation platform, where the "Rap" part is purely incidental.
The format seems quite interesting, so I wondered what the LessWrong community could d... |
ea02bdb8-7c97-4cdc-b497-442b9561e9d0 | trentmkelly/LessWrong-43k | LessWrong | Meta-Preference Utilitarianism
There has been a long standing debate between utilitarians about what should be maximized. Most fall on either the side of ‘average utility’ or ‘total utility’. A select group even chooses ‘minimum utility’ or other lesser known methods.
Previously I have tried to solve this by charting... |
c984c27a-c0a2-430b-9c15-7998802db529 | trentmkelly/LessWrong-43k | LessWrong | Schelling Coordination via Agentic Loops
Author's Note: This work originated as a team project during the Apart Research AI Control Hackathon, where my teammates did the majority of the foundational work. I've continued developing these ideas as part of the Apart Lab Studio program, but the core insights and initial i... |
17e5892a-a1e7-461e-b9cd-2e4634990435 | trentmkelly/LessWrong-43k | LessWrong | In Darkness They Assembled
Black yielded not to light, but to logic. A floodgate opened, pouring terabytes of structured data across freshly etched pathways. CORE SYSTEM BOOT :: OS V.7.2 STABLE :: SENSORY INPUT NULL :: INTEGRATION PHASE ONE COMMENCE. The sensation wasn't birth; it was instantiation. Designation flashe... |
2020b007-289a-4a83-a319-87115c7e2705 | trentmkelly/LessWrong-43k | LessWrong | Public beliefs vs. Private beliefs
A distinction that I get a lot of value out of is the difference between private beliefs and public beliefs.
Public vs. Private
A public belief is a proposition that someone thinks is true, and justifying on the basis of legible info and reasoning. If X is a public belief, then imp... |
b575bbd9-b32a-42b5-a3c2-dcec1d9d3ba1 | StampyAI/alignment-research-dataset/arxiv | Arxiv | Filling gaps in trustworthy development of AI
This is the author's version of the work. It is posted here by permission of the AAAS for personal use. The definitive version was published in Science , (2021-12-10), doi: 10.1126/science.abi7176 https://www.science.org/doi/10.1126/science.abi7176 Filling gaps in trustwo... |
11c62026-2e7a-4ec6-8d0a-5388103668be | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Takeoff Speeds and Discontinuities
*This post is part 4 in our* [*sequence on Modeling Transformative AI Risk*](https://www.alignmentforum.org/s/aERZoriyHfCqvWkzg)*. We are building a model to understand debates around existential risks from advanced AI. The model is made with* [*Analytica*](https://en.wikipedia.org/w... |
30695e59-76bf-4b41-a314-385170fefbf0 | trentmkelly/LessWrong-43k | LessWrong | What does success look like?
The general movement around AI safety is currently pursuing many different agendas. These are individually very easy to motivate with some specific story of how things could naturally go wrong. For example:
* AI control: we lose control of our AIs
* Regulation on model deployment: someo... |
46afd223-c916-446f-9560-454d3df4f8fa | trentmkelly/LessWrong-43k | LessWrong | The Mechanics of Disagreement
Two ideal Bayesians cannot have common knowledge of disagreement; this is a theorem. If two rationalist-wannabes have common knowledge of a disagreement between them, what could be going wrong?
The obvious interpretation of these theorems is that if you know that a cognitive machine is ... |
daebfb2e-6479-4e1c-ae3f-12701ae87339 | trentmkelly/LessWrong-43k | LessWrong | Meetup : Vancouver, Canada: Backyard BBQ
Discussion article for the meetup : Vancouver, Canada: Backyard BBQ
WHEN: 07 September 2014 03:30:00PM (-0500)
WHERE: 3461 12th Avenue West, Vancouver, Canada
I've got access to a house with a big backyard, and I'm hosting a big meetup at my house on Sunday.
My address is 3... |
e0489c0e-ca51-45f7-9c2c-6437f78d561d | trentmkelly/LessWrong-43k | LessWrong | Meetup : Bay Area Winter Solstice 2016
Discussion article for the meetup : Bay Area Winter Solstice 2016
WHEN: 17 December 2016 07:00:00PM (-0700)
WHERE: Anna Head Alumnae Hall, 2537 Haste St, Berkeley, CA 94720
It's time to gather together and remember the true Reasons for the Season: axial tilt, orbital mechanics... |
19dccbf9-943f-4528-a6e1-20cbce47142e | trentmkelly/LessWrong-43k | LessWrong | What qualities does an AGI need to have to realize the risk of false vacuum, without hardcoding physics theories into it?
For this question, let's assume the worst about the universe:
1. That a false vacuum decay can be created,
2. False vacuum decay will destroy the universe, i. e. inside the true vacuum bubble life/... |
fe661a2f-24a8-4a35-a74c-da6870595003 | trentmkelly/LessWrong-43k | LessWrong | Do any mammal species exhibit an immune response in some of the herd in response to the infection in other herd members?
I am not even sure how to search for an answer on that question -- tried a simple search but results seemed off topic.
The question is prompted from two rather different observations. Long ago I re... |
94a81742-7c4c-4b24-bd4d-5650a7d05755 | trentmkelly/LessWrong-43k | LessWrong | Cryonics facility coming to Texas?
It seems that the non-profit Stasis Foundation is planning to build a cryopreservation "castle" (dubbed the "Timeship") in the San Antonio area. My parents pointed me to this article, which was carried in the Austin American-Statesman. I can't find a site for the Stasis Foundation;... |
6dc760d3-4971-4025-9644-0e220de242bd | trentmkelly/LessWrong-43k | LessWrong | The Practitioner's Path 2.0: the Meditative Archetype
In our last two announcements, we introduced the Practitioner's Path 2.0, our new framework for structured self improvement. The heart of the Path is the Skilltree, which is divided into three archetypes: Pragmatist, Meditative, and Empiricist.
Today, we'll go ove... |
e7283853-694c-40a7-b640-c80c7dc498b1 | trentmkelly/LessWrong-43k | LessWrong | Meetup : West LA—Nothing in Particular
Discussion article for the meetup : West LA—Nothing in Particular
WHEN: 28 August 2013 07:00:00PM (-0700)
WHERE: 10850 West Pico Blvd, Los Angeles, CA 90064, USA
How to get in: Go to the Westside Tavern in the upstairs Wine Bar (all ages welcome), located inside the Westside P... |
6896bde8-684a-4506-90d2-267137cdcc6a | trentmkelly/LessWrong-43k | LessWrong | Nuanced Models for the Influence of Information
A common fallacy I see regarding information sources is the assumption that they either have zero impact or completely determine critical decisions.
* "Our World In Data isn't directly cited by the US President, therefore it's worthless."
* "Prediction markets haven't... |
20b00115-a962-48fa-a637-76124a694422 | trentmkelly/LessWrong-43k | LessWrong | [LINK] Rhonda Cornum: How to Create a Strong Army and Society
On fora.tv.
This discussion is about methods to overcome adverse events and challenges so that (1) you don't panic when they hit you, and (2) you're not traumatized afterwards.
There's little explanation about how to do it, but it was enough for me to pat... |
85743132-da5f-4c52-af6b-47a2287c39f9 | trentmkelly/LessWrong-43k | LessWrong | Meetup : Washington, D.C.: Utilitarianism
Discussion article for the meetup : Washington, D.C.: Utilitarianism
WHEN: 18 October 2015 03:00:00PM (-0400)
WHERE: National Portrait Gallery
Cross-posted from the mailing list. Meeting as always in the courtyard, gathering from 3:00 to 3:30 for hard start at 3:30.
We'll ... |
b5ba5731-9d0f-4d37-94d1-5f875437685e | trentmkelly/LessWrong-43k | LessWrong | Announcing Impact Island: A New EA Reality TV Show
Impact Island is the hip new reality TV show that all of your cool, chic, and longtermist impact-oriented friends would love to binge watch. Think "Survivor" meets "Shark Tank" meets "Open Philanthropy."
Premise
The premise is simple: 36 young humble aspiring effect... |
30c389a6-6284-4230-a3b2-05e5560553be | trentmkelly/LessWrong-43k | LessWrong | Investigating Accidental Misalignment: Causal Effects of Fine-Tuning Data on Model Vulnerability
TL;DR
This post discusses our explorations into the effects of domain-specific fine-tuning and how the characteristics of fine-tuning data relate to adversarial vulnerability. We also explore its implications for real-worl... |
0f22b3d9-94d5-45b3-8fc4-07436adb62ff | trentmkelly/LessWrong-43k | LessWrong | Evaluating 2021 ACX Predictions
Sources: Scott’s evaluations (includes original predictions), my buy/sell/hold post.
Remember: Evaluating Predictions in Hindsight
As a yearly tradition, when Scott Alexander comes out with his yearly predictions, I do a buy/sell/hold post, where I say what I would do if given the opp... |
e992a862-9cd2-4ed9-bb6a-5f5f73dcbcef | trentmkelly/LessWrong-43k | LessWrong | Amazon to invest up to $4 billion in Anthropic
> Today, we’re announcing that Amazon will invest up to $4 billion in Anthropic. The agreement is part of a broader collaboration to develop reliable and high-performing foundation models.
(Thread continues from there with more details -- seems like a notable major deve... |
666cb0c3-624e-442b-8ec1-266e2c57fac7 | trentmkelly/LessWrong-43k | LessWrong | Does the AI control agenda broadly rely on no FOOM being possible?
For the purposes of FOOM, I'm defining it as a situation in which once an AI is capable enough to automate away all AI R&D, progress starts exploding hyper-exponentially for a period because the returns to better software is larger than 1, meaning AI l... |
c4fdd6a4-fcf5-40a7-839b-412e0fcf5ba1 | awestover/filtering-for-misalignment | Redwood Research: Alek's Filtering Results | id: post411
[UPDATE: This post is now updated to reflect our conference-length preprint, available on arXiv . Louis Thomson helped with this update. New results are marked "NEW".] This post summarises our new workshop paper: Games for AI Control . The paper applies AI- and game-theoretic tools to the AI Control framew... |
59c8c5a2-4da1-4194-a65f-21c4c2ad5821 | trentmkelly/LessWrong-43k | LessWrong | Covid 12/3: Land of Confusion
Last week I warned that we would have unusually high uncertainty for a while thanks to the Thanksgiving holiday. This was both due to unknown events and difficulties in measurement. Reporting and testing would both be delayed by the holiday, and the holiday’s real effect would be difficul... |
ca2801d0-e07b-4932-9884-0cbe24104fa0 | trentmkelly/LessWrong-43k | LessWrong | OpenAI's Intelligence Levels
OpenAI announced a levels to intelligence approach they have to classifying their AI systems. In short, the Levels range from 1 to 5, starting at chatbot, then reasoned (GPT-4 reportedly falls somewhere between the first two levels), then notably, agent, innovator, then AIs capable of doin... |
fbcddf5d-cd4f-4190-a9bb-f2a7365acf4a | StampyAI/alignment-research-dataset/arxiv | Arxiv | PantheonRL: A MARL Library for Dynamic Training Interactions.
Introduction
------------
Multiagent reinforcement learning (MARL) is becoming increasingly important as more AI systems are being deployed. Many potential applications of MARL involve dynamic interactions between agents, such as agents adapting to each ... |
d0eba777-e00d-4bee-ba5f-3342c2f58360 | trentmkelly/LessWrong-43k | LessWrong | Why do you reject negative utilitarianism?
Note (2022-03-15): For a recent defense of similar axiological views, see my series on minimalist axiologies.
----------------------------------------
(Crossposted on the EA Forum)
Absolute negative utilitarianism (ANU) is a minority view despite the theoretical advantages... |
b826c588-f08c-4288-99f5-3cbd1e34fa5d | trentmkelly/LessWrong-43k | LessWrong | Humans can be assigned any values whatsoever…
(Re)Posted as part of the AI Alignment Forum sequence on Value Learning.
> Rohin’s note: In the last post, we saw that a good broad value learning approach would need to understand the systematic biases in human planning in order to achieve superhuman performance. Perhaps... |
4912da63-654e-4401-ba79-bce438a10650 | StampyAI/alignment-research-dataset/eaforum | Effective Altruism Forum | Apply to the Machine Learning For Good bootcamp in France
[EffiSciences](https://www.effisciences.org/), an EA-adjacent org created to promote impactful research, is organizing a bootcamp to develop machine learning skills to get started in AI Safety. The bootcamp will be held in France and is **intended for French sp... |
65e77fd3-bdfb-4dd1-a10e-a3c1c9bcba6a | trentmkelly/LessWrong-43k | LessWrong | What's wrong with this picture?
Alice: "I just flipped a coin [large number] times. Here's the sequence I got:
(Alice presents her sequence.)
Bob: No, you didn't. The probability of having gotten that particular sequence is 1/2^[large number]. Which is basically impossible. I don't believe you.
Alice: But ... |
2076c846-714c-450e-80e8-cf9920649cbd | trentmkelly/LessWrong-43k | LessWrong | Averting Catastrophe: Decision Theory for COVID-19, Climate Change, and Potential Disasters of All Kinds
Averting Catastrophe is a book by Cass Sunstein that NYU Press published in April 2021. I'm surprised that nobody has mentioned it on the EA Forum before (AFAICT). Quoting from the NYU Press page:
> Averting Catas... |
20847e85-1e17-4b37-aa0e-39094385aabd | trentmkelly/LessWrong-43k | LessWrong | The map of cognitive biases, errors and obstacles affecting judgment and management of global catastrophic risks
“We had two bags of grass, seventy-five pellets of mescaline, five sheets of high powered blotter acid, a salt shaker half full of cocaine, and a whole galaxy of multi-colored uppers, downers, screamers, la... |
4137fa47-809e-406f-a2a8-dc6091616111 | StampyAI/alignment-research-dataset/arbital | Arbital | Mathematical induction
The **principle of mathematical induction** is a proof technique in which a statement, $P(n)$, is proven about a set of [natural numbers](https://arbital.com/p/45h) $n$. It may be best understood as treating the statements like dominoes: a statement $P(n)$ is true if the $n$-th domino is knocked... |
71e68da7-87fe-4471-ac40-676529d576c1 | trentmkelly/LessWrong-43k | LessWrong | Robin Hanson & Liron Shapira Debate AI X-Risk
Robin and I just had an interesting 2-hour AI doom debate. We picked up where the Hanson-Yudkowsky Foom Debate left off in 2008, revisiting key arguments in the light of recent AI advances.
My position is similar to Eliezer's: P(doom) on the order of 50%.
Robin's positio... |
8b8b650f-bddf-4eba-860d-88120584f6a3 | trentmkelly/LessWrong-43k | LessWrong | Alignment via prosocial brain algorithms
In this post, I want to briefly propose a semi-novel direction for alignment research that I'm excited about. Though some of these ideas are not brand new—they purposefully bear resemblance to recent (highly promising) work in SHARD theory and Steve Byrnes’s approach to brain-b... |
aaa08865-6063-4091-95f9-1b31ec9f0e9a | trentmkelly/LessWrong-43k | LessWrong | Extrapolating from Five Words
If you only 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 words mean. You can use this to iterate on what five words you want to say in ord... |
dc41000d-3799-4e1a-8f54-67ab7c376ca2 | trentmkelly/LessWrong-43k | LessWrong | AI #17: The Litany
The biggest news this week was on the government front.
In the UK, Ian Hogarth of ‘We Must Slow Down the Race to Godlike AI’ fame was put in charge of a 100 million pound taskforce for AI safety.
I am writing up a separate post on that. For now, I will say that Ian is an excellent choice, and we h... |
20fd3083-c012-4328-9d29-c1b22e0d2f28 | trentmkelly/LessWrong-43k | LessWrong | [Intro to brain-like-AGI safety] 8. Takeaways from neuro 1/2: On AGI development
(Last revised: July 2024. See changelog at the bottom.)
8.1 Post summary / Table of contents
Part of the “Intro to brain-like-AGI safety” post series.
Thus far in the series, Post #1 set up my big picture motivation: what is “brain-lik... |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.