id stringlengths 36 36 | source stringclasses 15
values | formatted_source stringclasses 13
values | text stringlengths 2 7.55M |
|---|---|---|---|
b03376b4-7b7a-40aa-8315-2fce971f8049 | trentmkelly/LessWrong-43k | LessWrong | UNGA General Debate speeches on AI
At the UN General Assembly's 2023 General Debate on September 19-26, many leaders spoke about artificial intelligence, and some specifically about existential risk and what to do about it. Attitudes ranged from advocating serious global action (Israel, the UK) to being completely dis... |
d152cf01-49a4-46d5-9c2c-77b3b2112da5 | trentmkelly/LessWrong-43k | LessWrong | Insights from Modern Principles of Economics
How good are our economists? Look around. On a 20-minute walk to my Berkeley office, I walked past people reeking of urine, past people lying in a dirty sleeping bag on a thin cardboard pad, past some garbage around a tent which housed a child who grew up into an impoverish... |
7b0cb002-f8fb-4bb9-81af-54b4adac1de5 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Elicit: Language Models as Research Assistants
[Ought](https://ought.org) is an applied machine learning lab. We’re building [Elicit](https://elicit.org), the AI research assistant. Our mission is to automate and scale open-ended reasoning. To get there, we train language models by [supervising reasoning processes, no... |
de006815-e319-4d3e-bbbc-5c405e7ae11a | trentmkelly/LessWrong-43k | LessWrong | Report: Evaluating an AI Chip Registration Policy
As part of our Governance Recommendations Research Program, Convergence Analysis is publishing a series of deep-dive analyses on specific, upcoming governance regulatory proposals. Our mission is to meaningfully contribute to the passing of critical & foundational gove... |
5f9c93e8-9246-4ce9-9d07-d1f4cef612ea | StampyAI/alignment-research-dataset/eaforum | Effective Altruism Forum | [Link post] AI could fuel factory farming—or end it
Despite the fact that artificial intelligence is already transforming everyday life, most people don’t know much about it.
Simply put, AI is a method of making software “think” intelligently like the human mind. It can make decisions using real-time data, learning ... |
237c9d4e-d684-40a5-aa76-fb5026a54573 | trentmkelly/LessWrong-43k | LessWrong | If there was one element of statistical literacy that you could magically implant in every head, what would it be?
Alternatively, what single concept from statistics would most improve people's interpretations of popular news and daily life events? |
fbe87f32-ad7c-41db-9bff-e7b9de71ebbb | StampyAI/alignment-research-dataset/blogs | Blogs | Simple Sensor Intentions for Exploration
By simple color-masking, high-level image statistics can be derived. Rewarding an agent for deliberately changing these statistics, leads to diverse exploration and interes... |
df44d65d-9bd4-4fe6-b41e-ab46d8cdafea | trentmkelly/LessWrong-43k | LessWrong | Are Girl Scout Cookies Deliciously Evil? A Case Study in Evaluating Charities by Yourself
I recently finished up an essay examining the Girl Scouts, their cookies, and their finances with reference to whether they are inefficient or corrupt.
What is the relevance to LW? (The essay is aimed at a general audience & ass... |
01d55d11-ce6b-4f63-8291-e7aaa8d32a40 | trentmkelly/LessWrong-43k | LessWrong | A thought experiment for comparing "biological" vs "digital" intelligence increase/explosion
Intro
The idea of an "intelligence explosion", where artificial intelligence (AI) rapidly surpasses Human intelligence, might be mind-bending for many. Especially for those not deeply engaged with AI. But what if we could exp... |
c5d49591-7e28-4d11-aa4d-7d983b766a18 | StampyAI/alignment-research-dataset/arbital | Arbital | Universal property
In [https://arbital.com/p/-4c7](https://arbital.com/p/-4c7), we attempt to avoid thinking about what an object *is*, and look only at how it interacts with other objects.
It turns out that even if we're not allowed to talk about the "internal" structure of an object, we can still pin down some objec... |
b6af2211-ec2e-40a6-9841-a1e53abda67f | trentmkelly/LessWrong-43k | LessWrong | [Link] H+ article Rationality Training: Call for a Global Approach
http://hplusmagazine.com/2012/02/06/rationality-training-call-for-a-global-approach/
Dr.Ziesche calls for more globally accessible rationality training. He seems supportive of existing efforts, especially us, but thinks it would be nice if rationalit... |
48db3ea4-46fe-4e1d-a0d8-c9c207d814f1 | trentmkelly/LessWrong-43k | LessWrong | The 'why does it even tell me this' moment
Edited based on the outline kindly provided by Gram_Stone, whom I thank.
There is a skill of reading and thinking which I haven't learned so far: of looking for implications as one goes through the book, simply putting it back on shelf until one's mind has run out of the inf... |
2c3fe76f-a5ff-4908-b72b-cbe6d0ce3c65 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Public Static: What is Abstraction?
.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; float: n... |
ea479634-0b87-460b-bdbe-e98016299ecb | trentmkelly/LessWrong-43k | LessWrong | Sortition Model of Moral Uncertainty
This is a crosspost from the EA forum
The Model
There have been many models proposed to resolve moral uncertainty, but I would like to introduce one more. Instead of acting in accordance to the moral theory we are most confident in (my favorite theory) or making complex elector... |
b7468268-8605-459c-860a-579a35078117 | StampyAI/alignment-research-dataset/lesswrong | LessWrong | A simple way to make GPT-3 follow instructions
*I originally posted this idea as a comment to* [*The case for aligning narrowly superhuman models*](https://www.lesswrong.com/posts/PZtsoaoSLpKjjbMqM/the-case-for-aligning-narrowly-superhuman-models)*, but it's interesting enough that I thought I'd submit it as a top lev... |
dd3a7980-87a4-4ead-ac6a-f4d185021cb8 | trentmkelly/LessWrong-43k | LessWrong | Two Types of Updatelessness
Cross-posted.
Just a small note which I’m not sure has been mentioned anywhere else:
It seems like there are two different classes of “updateless reasoning”.
In problems like Agent Simulates Predictor, switching to updateless reasoning is better for you in the very situation you find you... |
9d59d86a-b819-446f-b976-607b0bf38c23 | trentmkelly/LessWrong-43k | LessWrong | Review of Kahneman, 'Thinking, Fast and Slow' (2011)
Thinking, Fast and Slow is Kahneman's first book for a general audience, and a summary of his far-reaching and important work. Over the course of about 400 pages (this does not include the appendices, notes, or index), Kahneman explains his current views on: System ... |
4a509328-4989-4c0c-a56f-93ee2e2c3a39 | trentmkelly/LessWrong-43k | LessWrong | With whom shall I diavlog?
Bloggingheads.tv can't exactly call up, say, the President of France and get him to do a diavlog, but they have some street cred with mid-rank celebrities and academics. With that in mind, how would you fill in this blank?
"I would really love to see a diavlog between Yudkowsky and _______... |
a027420a-0e93-4058-a5ba-f0be57a2f5be | awestover/filtering-for-misalignment | Redwood Research: Alek's Filtering Results | id: post200
For a summary of this post, see the thread on X . The assumptions behind and limitations of capability elicitation have been discussed in multiple places (e.g. here , here , here , etc); however, none of these match my current picture. For example, Evan Hubinger’s “When can we trust model evaluations?” cit... |
d78a1f34-26b9-4db1-83f4-94dd33fd9814 | trentmkelly/LessWrong-43k | LessWrong | Who is liable for AI?
[Edit -- Changed to title as suggested by Raemon. So I've also made some edits to the post itself.]
Where does liability fall in the case of AI doing things?
I think there is a rather close parallel with autonomous cars/vehicles. Probably 4 or 5 years back that was a question being raise. I ass... |
d62f2dd3-4891-475d-b4b6-b6078d5030de | trentmkelly/LessWrong-43k | LessWrong | Weekly LW Meetups
This summary was posted to LW Main on June 26th. The following week's summary is here.
New meetups (or meetups with a hiatus of more than a year) are happening in:
* Maine: Automatic Cognition: 29 June 2015 06:30PM
* San Antonio New Meetup: 28 June 2015 03:00PM
* São Paulo, Brazil - Meetup at Ba... |
5c3a0fdd-d20c-4e66-addc-6c504da24936 | trentmkelly/LessWrong-43k | LessWrong | Why was progress so slow in the past?
What explains the hockey-stick shape of world GDP over time, with seemingly no progress for thousands of years, followed by soaring growth?
Our World in Data
The first question you might ask is: is this just an exponential curve? If so, then the explanation is simple: we see a co... |
59f9d257-ecac-453e-adb9-40755fa4bffa | StampyAI/alignment-research-dataset/blogs | Blogs | March 2018 Newsletter
#### Updates
* New research write-ups and discussions: [Knowledge is Freedom](https://www.lesswrong.com/posts/b3Bt9Cz4hEtR26ANX/knowledge-is-freedom); [Stable Pointers to Value II: Environmental Goals](https://agentfoundations.org/item?id=1762); [Toward a New Technical Explanation of Technical ... |
b67597b2-6af0-4e7c-a4be-80bf93eb018d | trentmkelly/LessWrong-43k | LessWrong | Sum-threshold attacks
How do you affect something far away, a lot, without anyone noticing?
(Note: you can safely skip sections. It is also safe to skip the essay entirely, or to read the whole thing backwards if you like.)
The frog's lawsuit
Attorney for the defendant: "So, Mr. Frog. You allege that my client caus... |
1c17eb1b-60d9-4485-a772-a370a18a1867 | StampyAI/alignment-research-dataset/lesswrong | LessWrong | If Wentworth is right about natural abstractions, it would be bad for alignment
*This post was written as part of the* [*AI safety Mentors and Mentees*](https://www.lesswrong.com/posts/EjX63wQoMSoCHMrmY/announcing-ai-safety-mentors-and-mentees) *program. My Mentor is* [*Jacques Thibodeau*](https://www.lesswrong.com/us... |
cc1ced9a-aa30-4bd2-81c3-4c3a97729bf1 | trentmkelly/LessWrong-43k | LessWrong | o3 is not being released to the public. First they are only giving access to external safety testers. You can apply to get early access to do safety testing
"We’re inviting safety researchers to apply for early access to our next frontier models. This early access program complements our existing frontier model testin... |
5544c78a-dbe2-48ba-a8dd-0b2e8c494b9d | trentmkelly/LessWrong-43k | LessWrong | Matthew Yglesias - Misinformation Mostly Confuses Your Own Side
Matthew Yglesias' post, Misinformation Mostly Confuses Your Own Side, argues that political misinformation tends to harm the group spreading it more than it persuades opponents. Key points:
* Misinformation primarily affects in-group thinking: Supporter... |
16c2e701-c378-44f1-ac82-88ecf1096748 | trentmkelly/LessWrong-43k | LessWrong | Thoughts on paper "How Organisms Come to Know the World: Fundamental Limits on Artificial General Intelligence"?
I've read the abstract of this paper, How Organisms Come to Know the World: Fundamental Limits on Artificial General Intelligence. It says:
> A general example of an affordance is the use of an object in ... |
cc568319-c34e-475d-9c2b-22df04961fd4 | StampyAI/alignment-research-dataset/lesswrong | LessWrong | The commercial incentive to intentionally train AI to deceive us
We have all experienced application programs telling us something we did not want to hear, e.g., poor financial status, or results of design calculations outside practical bounds. While we may feel like shooting the messenger, applications are treated as... |
20502379-a8ae-4e7c-8b32-2af99054ec4f | StampyAI/alignment-research-dataset/eaforum | Effective Altruism Forum | My personal cruxes for working on AI safety
The following is a heavily edited transcript of a talk I gave for the Stanford Effective Altruism club on 19 Jan 2020. I had rev.com transcribe it, and then Linchuan Zhang, Rob Bensinger and I edited it for style and clarity, and also to occasionally have me say smarter thin... |
02d64f8a-2bf5-4530-b848-f52a6396cde0 | trentmkelly/LessWrong-43k | LessWrong | Our Phyg Is Not Exclusive Enough
EDIT: Thanks to people not wanting certain words google-associated with LW: Phyg
Lesswrong has the best signal/noise ratio I know of. This is great. This is why I come here. It's nice to talk about interesting rationality-related topics without people going off the rails about politic... |
3e7838ea-6a33-4247-8afa-90dac6e9e3fa | trentmkelly/LessWrong-43k | LessWrong | Structural Risk Minimization
Structural risk minimization, a concept from computational learning theory, is proposed as a satisficing framework.
----------------------------------------
The goal of this post is similar to Creating a Satisficer and Minimax as an approach to reduced-impact AI: constructing agents whic... |
a7d2460d-6263-4d57-8c89-1d81ae37bb3f | trentmkelly/LessWrong-43k | LessWrong | An Undergraduate Reading Of: Semantic information, autonomous agency and non-equilibrium statistical physics
This is a recent paper by Artemy Kolchinsky and David H. Wolpert, from the Santa Fe Institute. It was published in The Royal Society Interface on Oct 19. They propose a formal theory of semantic information, wh... |
fe561a27-fc3c-4d11-8d81-9f5741377172 | trentmkelly/LessWrong-43k | LessWrong | Cambridge Less Wrong Meetup
We'll be having a Less Wrong meetup on Tuesday, October 12 at 7:00pm at Green Street Grill 280 Green Street Cambridge, MA, to coincide with SIAI President Mike Vassar's visit to Boston. See the meetup.com page. This is in addition to the monthly meetups, which are on October 17th and on th... |
3d7093ca-04a1-436f-8e03-dcfe6d8c3ccc | trentmkelly/LessWrong-43k | LessWrong | Words are Imprecise Telepathy
Words are mental paintbrush handles. Saying "triangular lightbulb" paints a picture with words.
The point of communication is to transmit a message. But the message is not the words. The message is the picture, the mental model inside your head.
So the message is reduced to words, which ... |
b48c8d0e-b985-4532-8e42-cd914c4a3cd2 | trentmkelly/LessWrong-43k | LessWrong | What’s going on? LLMs and IS-A sentences
This is a cross-post from New Savanna.
For the moment I have decided that Waddington’s classic diagram of the epigenetic landscape is a useful way of thinking about when happens when an LLM responds to a prompt. Here’s the diagram:
The language model corresponds to the lan... |
cfb56254-afab-4fe0-abe7-bfd0002e1f6d | trentmkelly/LessWrong-43k | LessWrong | New page: Integrity
There's a new page collecting integrity incidents at the frontier AI labs.
Also a month ago I made a page on labs' policy advocacy.
If you have suggestions to improve these pages, or have ideas for other resources I should create, let me know.
----------------------------------------
Crossposte... |
8ac2845e-0f2e-49e8-9609-b652204b8358 | trentmkelly/LessWrong-43k | LessWrong | Can private companies test LVTs?
It seems like unlike most exciting economic ideas that economists swear by but governments ignore, Georgian Land Value Taxes should be fairly doable for a private development company to test.
They would have to buy up a large area of rural land at a cheap price, and then rent it all o... |
404fc8f3-ce06-488d-8de2-ba813b87cede | trentmkelly/LessWrong-43k | LessWrong | Turning off lights with model editing
This post advertises an illustrative example of model editing that David Bau is fond of, and which I think should be better known. As a reminder, David Bau is a professor at Northeastern; the vibe of his work is "interpretability + interventions on model internals"; examples inclu... |
3d05a001-cda7-40d3-b1c5-06b0d6cd8c8b | StampyAI/alignment-research-dataset/lesswrong | LessWrong | The shape of AGI: Cartoons and back of envelope
*[Cross posted on*[*windowsontheory*](https://windowsontheory.org/?p=8644)*; see here for my*[*prior writings*](https://windowsontheory... |
692d7052-cce5-4c65-a5f0-ae594656cbc3 | trentmkelly/LessWrong-43k | LessWrong | Biology-Inspired AGI Timelines: The Trick That Never Works
- 1988 -
Hans Moravec: Behold my book Mind Children. Within, I project that, in 2010 or thereabouts, we shall achieve strong AI. I am not calling it "Artificial General Intelligence" because this term will not be coined for another 15 years or so.
Eliezer ... |
8bca34f9-3041-4d5e-948d-6260ca24b4a1 | StampyAI/alignment-research-dataset/lesswrong | LessWrong | What will the twenties look like if AGI is 30 years away?
I'm especially interested to hear from people with long (i.e. 20+ year) timelines what they think the next 10 years will look like.
[EDIT: After [pushback from Richard Ngo](https://www.lesswrong.com/posts/axbTNGuMtB4hCkNus/what-will-the-twenties-look-like-if-... |
120c7a66-79d8-4c1e-92fd-56b5d98365c4 | trentmkelly/LessWrong-43k | LessWrong | Human Biases that Obscure AI Progress
There are some common biases that are often used to discount AI progress. We should keep these in mind, as they can prevent us from having an objective understanding of progress in the field.
I'm going to use AI here instead of ML because usually these biases are relevant to any ... |
51cfe8a0-4a62-4b46-94d5-26affd133ab8 | trentmkelly/LessWrong-43k | LessWrong | Neural nets as a model for how humans make and understand visual art
This is a new paper relating experimental results in deep learning to human psychology and cognitive science. I'm excited to get feedback and comments. I've included some excerpts below.
Abstract
This paper is about the cognitive science of visual... |
59e2f352-02cb-43ff-b770-8aca22de3a27 | trentmkelly/LessWrong-43k | LessWrong | Weekly LW Meetups
This summary was posted to LW Main on July 5th. The following week's summary is here.
Irregularly scheduled Less Wrong meetups are taking place in:
* Atlanta: Intro to Meditation: 13 July 2013 06:00PM
* Brussels meetup with HEALES: 13 July 2013 01:00PM
* [Moscow] The Goals We Set: 07 July 2013 0... |
9f8fb73b-011e-4823-849d-d2a839568f45 | trentmkelly/LessWrong-43k | LessWrong | [Beneath Psychology] Case study on chronic pain: First insights, and the remaining challenge
In the last post I took the seemingly-naive stance that "pain is just information" and "can't actually be a problem", implying that painful situations can be dealt with without suffering by just looking through the pain toward... |
8cd13889-7a46-4d5e-be8a-ecb6284c474b | trentmkelly/LessWrong-43k | LessWrong | [SEQ RERUN] When Science Can't Help
Today's post, When Science Can't Help was originally published on 15 May 2008. A summary (taken from the LW wiki):
> If you have an idea, Science tells you to test it experimentally. If you spend 10 years testing the idea and the result comes out negative, Science slaps you on t... |
907ec58a-a8d1-42ac-b5f8-8d50d1be98d4 | StampyAI/alignment-research-dataset/lesswrong | LessWrong | IntelligenceExplosion.com
I put together a 'landing page' for the intelligence explosion concept similar to Nick Bostrom's landing pages for [anthropics](http://www.anthropic-principle.com/), the [simulation argument](http://www.simulation-argument.com/), and [existential risk](http://www.existentialrisk.com/). The ne... |
8ff10b6a-6229-4d28-ac28-4dcefb67f1bb | trentmkelly/LessWrong-43k | LessWrong | The Obliqueness Thesis
In my Xenosystems review, I discussed the Orthogonality Thesis, concluding that it was a bad metaphor. It's a long post, though, and the comments on orthogonality build on other Xenosystems content. Therefore, I think it may be helpful to present a more concentrated discussion on Orthogonality, ... |
df3c5703-6a3c-4344-8d6d-4236cbc868a1 | StampyAI/alignment-research-dataset/lesswrong | LessWrong | FAI Research Constraints and AGI Side Effects
**Ozzie Gooen and Justin Shovelain**
Summary
-------
Friendly artificial intelligence (FAI) researchers have at least two significant challenges. First, they must produce a significant amount of FAI research in a short amount of time. Second, they must do so without pr... |
b08762c8-a220-4041-890b-8520757c54a8 | trentmkelly/LessWrong-43k | LessWrong | Visual Mental Imagery Training
Previously: Generalizing From One Example
> There was a debate, in the late 1800s, about whether "imagination" was simply a turn of phrase or a real phenomenon. That is, can people actually create images in their minds which they see vividly, or do they simply say "I saw it in my mind" ... |
083d6323-e37c-4887-8f04-20d3c47b125e | trentmkelly/LessWrong-43k | LessWrong | Personal Advice/Request: Can I be your apprentice?
It has not been very long since I joined LessWrong but I have been truly enjoying the intelligent conversations/debates going on here, whether it is about general rationality, ML/AI, or investment. In the meantime, I find myself thinking Albert Einstein’s famous quote... |
ad07b948-154d-4d10-888a-e12ca2616227 | trentmkelly/LessWrong-43k | LessWrong | Coherence Therapy with LLMs - quick demo
update on my current views: a lot of the useful meaning in Coherence Therapy sessions comes from intonation/paralanguage. Unfortunately, language models unfortunately can’t currently perceive this. I don't think current AI systems can do Coherence Therapy nearly as well as a co... |
56ffccd3-62af-4d89-832d-8f482740a2ff | trentmkelly/LessWrong-43k | LessWrong | Example of the nameless rationalist virtue
What makes one look more intensely at a bus and adjust with a speedup while turning as close as possible to it, rather than away. |
a6a14584-b0c0-49d6-8685-d6ce8de2d7a4 | trentmkelly/LessWrong-43k | LessWrong | The Cameron Todd Willingham test
In 2004, The United States government executed Cameron Todd Willingham via lethal injection for the crime of murdering his young children by setting fire to his house.
In 2009, David Grann wrote an extended examination of the evidence in the Willingham case for The New Yorker, which ... |
18621c63-deea-4e69-9361-5326636e8d3c | trentmkelly/LessWrong-43k | LessWrong | Huffington Post article on DeepMind-requested AI ethics board, links back to LW [link]
http://www.huffingtonpost.com/2014/01/29/google-ai_n_4683343.html
Not going to summarize the article content, but I think this is the highest-level publication linking to LW so far.
Also, it's appears that Shane Legg, Jaan Tallin ... |
e6242f0e-88c2-4ab6-8bdd-018f9491dfb6 | trentmkelly/LessWrong-43k | LessWrong | Higher Purpose
Followup to: Something to Protect, Superhero Bias
Long-time readers will recall that I've long been uncomfortable with the idea that you can adopt a Cause as a hedonic accessory:
> "Unhappy people are told that they need a 'purpose in life', so they should pick out an altruistic cause that goes well ... |
66a7b5ed-6b22-4c2f-b8cb-0afb49ec25c6 | trentmkelly/LessWrong-43k | LessWrong | A collection of Stubs.
In light of SDR's comment yesterday, instead of writing a new post today I compiled my list of ideas I wanted to write about, partly to lay them out there and see if any stood out as better than the rest, and partly so that maybe they would be a little more out in the wild than if I hold them un... |
684f103b-d6bb-4361-9e78-598e85cf75a1 | trentmkelly/LessWrong-43k | LessWrong | False Thermodynamic Miracles
Cross posted at Less Wrong
A putative new idea for AI control; index here.
Ok, here is the problem:
You have to create an AI that believes (or acts as if it believed) that event X is almost certain, while you believe that X is almost impossible. Furthermore, you have to be right. To mak... |
3eaf98b3-4972-4dbf-9e51-9fdd2a5935d9 | trentmkelly/LessWrong-43k | LessWrong | A Proof of Löb's Theorem using Computability Theory
Löb’s Theorem states that, if PA⊢□PA(P)→P, then PA⊢P. To explain the symbols here:
* PA is Peano arithmetic, a first-order logic system that can state things about the natural numbers.
* PA⊢A means there is a proof of the statement A in Peano arithmetic.
* □PA(P)... |
6c21f896-f8ce-46bf-a6f8-808a439c5146 | trentmkelly/LessWrong-43k | LessWrong | Meetup : Moscow, How Can Visual Image Be Rational
Discussion article for the meetup : Moscow, How Can Visual Image Be Rational
WHEN: 22 December 2013 04:00:00PM (+0400)
WHERE: Russia, Moscow, ulitsa L'va Tolstogo 16
We will gather at the same second entrance, but we will go to the new room at 16:00. So please do no... |
319d35a5-3d4d-4cb3-8667-4f796e11f9bc | trentmkelly/LessWrong-43k | LessWrong | Compact vs. Wide Models
I recently found a label I like for one of the obstacles to AI alignment that I think about frequently. I think of it as "compact models" vs. "wide models," which I'll try to define by suggestive naming, followed by example. It seems easy to verify that compact models have particular properties... |
e0d46209-7c64-40de-bb6a-f46546cfb41e | trentmkelly/LessWrong-43k | LessWrong | Evidential Decision Theory, Selection Bias, and Reference Classes
See also: Does Evidential Decision Theory really fail Solomon's Problem?, What's Wrong with Evidential Decision Theory?
It seems to me that the examples usually given of decision problems where EDT makes the wrong decisions are really examples of perfo... |
468d6a6c-3112-4c62-9911-b0618b5bbeb5 | StampyAI/alignment-research-dataset/eaforum | Effective Altruism Forum | How can I best use my career to pass impactful AI and Biosecurity policy.
A member of my group asked this question to me recently and I wasn't sure how to respond. To make my question more specific take the following scenario:
A talented American undergrad inspired by EA wants to work in policy to positively sha... |
80163abb-40a6-482f-9eda-665e11356a62 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | The reward engineering problem
Today we usually train reinforcement learning agents to perform narrow tasks with simple goals. We may eventually want to train RL agents to behave “well” in open-ended environments where there is no simple goal.
Suppose that we are trying to train an RL agent **A**. In each episode, **... |
d139a130-eeed-46c6-8132-c822eb202082 | trentmkelly/LessWrong-43k | LessWrong | Meetup : Columbus or Cincinnati Meetup
Discussion article for the meetup : Columbus or Cincinnati Meetup
WHEN: 22 January 2012 05:00:00PM (-0500)
WHERE: Wilmington, OH
Location: Max & Erma's off I-71 in Wilmington
Bring a board game(s).
There is an Ohio LW email list and google group being formed. PM me to get on... |
044bde3f-cc26-452e-b26f-f6b4a8e615fb | StampyAI/alignment-research-dataset/lesswrong | LessWrong | Muehlhauser-Goertzel Dialogue, Part 1
Part of the [Muehlhauser interview series on AGI](http://wiki.lesswrong.com/wiki/Muehlhauser_interview_series_on_AGI).
*[Luke Muehlhauser](http://lukeprog.com/) is Executive Director of the [Singularity Institute](http://intelligence.org/), a non-profit research institute st... |
4d3e0530-cec3-4c0b-9df3-185cc3075e26 | trentmkelly/LessWrong-43k | LessWrong | [Event] Symposium on Cryonics and Brain-Threatening Disorders
Upcoming event:
> On Saturday July 7, 2012, the Institute for Evidence Based Cryonics and Cryonics Northwest will organize a symposium on cryonics and brain-threatening disorders in Portland, Oregon. The symposium will start at 09:00 am at the offices of K... |
d7f61497-0c9f-4e1c-838c-609a331d5648 | trentmkelly/LessWrong-43k | LessWrong | Generating Cognateful Sentences with Large Language Models
Motivation
> “Le président Emmanuel Macron assure le peuple canadien que le gouvernement français va continuer à défendre le Canada contre la menace américain.”
Even if you don’t speak French, you can probably understand, or at least get the gist of, the abov... |
51000077-3f30-4339-a6f1-d00bd2086b8f | trentmkelly/LessWrong-43k | LessWrong | Must true AI sleep?
“The real frontier is finding an animal that sleeps that doesn’t have neurons at all, From https://www.science.org/content/article/if-alive-sleeps-brainless-creatures-shed-light-why-we-slumber |
81fcb803-f13c-4646-9c4b-71aa27d286a7 | trentmkelly/LessWrong-43k | LessWrong | Gradient Anatomy's - Hallucination Robustness in Medical Q&A
TL;DR
We investigated reducing hallucinations in medical question-answering with Llama-3.1-8B-Instruct.
Using Goodfire's Sparse Auto-Encoder (SAE) we identified neural features associated with accurate and hallucinated responses. Our study found that featur... |
79292c62-d4d6-4924-a586-e5b91df2900f | trentmkelly/LessWrong-43k | LessWrong | Open Thread April 8 - April 14 2014
You know the drill - If it's worth saying, but not worth its own post (even in Discussion), then it goes here. |
74b039d9-447f-4258-8388-72caaffe3910 | StampyAI/alignment-research-dataset/arxiv | Arxiv | Reliable evaluation of adversarial robustness with an ensemble of diverse parameter-free attacks
1 Introduction
---------------
Adversarial samples, small perturbations of the input, with respect to some distance measure, which change the decision of a classifier,
are a problem for safe and robust machine learning.... |
1cd85903-9bc9-4693-a305-8711d2f6a1a1 | StampyAI/alignment-research-dataset/arbital | Arbital | Logarithm
summary(Technical): $\log_b(n)$ is defined to be the number $x$ such that $b^x = n.$ Thus, logarithm functions satisfy the following properties, among others:
- $\log_b(1) = 0$
- $\log_b(b) = 1$
- $\log_b(x\cdot y) = log_b(x) + \log_b(y)$
- $\log_b(\frac{x}{y}) = \log_b(x) - \log_b(y)$
- $\log_b(x^n) = n\lo... |
6101fff2-3a43-485c-b702-4ce15e40e75c | trentmkelly/LessWrong-43k | LessWrong | Pixel 3a Review
About a month ago I replaced my first-generation Pixel phone with a Pixel 3a, and the 3a is as close to my ideal as I've had:
* Top-quality camera
* Solid battery life
* Near-stock Android
* Fingerprint sensor
* Headphone jack
* Relatively cheap ($300)
Several years ago, phones got good enoug... |
bc135f98-e1f0-464d-9607-d34612918e61 | trentmkelly/LessWrong-43k | LessWrong | Bloggingheads: Yudkowsky and Aaronson talk about AI and Many-worlds
Eliezer Yudkowsky and Scott Aaronson - Percontations: Artificial Intelligence and Quantum Mechanics
Sections of the diavlog:
* When will we build the first superintelligence?
* Why quantum computing isn’t a recipe for robot apocalypse
* How to gu... |
dc03f384-c317-4208-a86f-840206a43126 | trentmkelly/LessWrong-43k | LessWrong | Unofficial ESPR Post-mortem
[Disclaimer: This post reflects my, i.e. Owen Shen’s, personal opinions. It does NOT reflect CFAR or ESPR’s opinions, and should NOT be taken as an ESPR-endorsed communication.]
[An overview of some of the camp and project dynamics of ESPR 2017. I look at diversity causing certain problems... |
ba8ed914-b0cb-4626-8126-ba0b30b30b64 | StampyAI/alignment-research-dataset/arxiv | Arxiv | The Archimedean trap: Why traditional reinforcement learning will probably not yield AGI
1 Introduction
---------------
Whenever we measure anything using a particular number system, the
corresponding measurements will be constrained by the structure of that
number system. If the number system has a different struc... |
edc41076-9b3f-485d-9f3c-ae5904afc08f | trentmkelly/LessWrong-43k | LessWrong | Manifold Markets
Manifold Markets is a prediction market platform where I've been trading since September. This post will compare it to other prediction markets that I've used.
Play Money
The most important fact about Manifold is that traders bet mana, which is for most purposes not real money. You can buy mana, and... |
4a66136b-4ec1-46ba-8316-4bf65eb83c13 | trentmkelly/LessWrong-43k | LessWrong | Letter from leading Soviet Academicians to party and government leaders of the Soviet Union regarding signs of decline and structural problems of the economic-political system (1970)
Crossposted from my blog: mzuo.ca
Some of the observations by these Soviet intellectuals may be of interest to LW readers...
"
This o... |
37afaa28-5958-4a31-a5f9-9d211205acd4 | trentmkelly/LessWrong-43k | LessWrong | Results from MIRI's December workshop
Last week (Dec. 14-20), MIRI ran its 6th research workshop on logic, probability, and reflection. Writing up mathematical results takes time, and in the past, it's taken quite a while for results from these workshops to become available even in draft form. Because of this, at the ... |
9f36ce16-a843-4f05-a96a-bf1524d338bc | StampyAI/alignment-research-dataset/lesswrong | LessWrong | What is a training "step" vs. "episode" in machine learning?
My impression is that steps and episodes are both time periods in a training process, and that these terms are somewhat common in RL. An episode is larger than a step and usually contains many steps.
Is this correct?
Some related questions:
1. Are rewards... |
8b17337f-b2b9-4d14-89c0-c07c117910d5 | trentmkelly/LessWrong-43k | LessWrong | The Importance-Avoidance Effect
My current working theory is that one of the worst things someone with a big project can do is to make that project feel extremely important. (This is doubly worse for someone with ADHD.) The more important the project, the more daunting it becomes, and the less it gets worked on.
When... |
5245855b-be99-4ed2-b516-f058a9df7f41 | trentmkelly/LessWrong-43k | LessWrong | Food Spending During Covid
Our house has shared groceries, and in my most recent reconcile I saw that our spending is up quite a bit relative to this time last year. In 2019, April through July, our monthly spending per person was $172, while in 2020 it was $303. In the big picture this is due to the pandemic, but I'm... |
ae98ce8e-140d-4d83-af24-6087877a6cb7 | trentmkelly/LessWrong-43k | LessWrong | Open thread for December 24-31, 2013
If it's worth saying, but not worth its own post (even in Discussion), then it goes here. |
8f90dbbe-7977-407f-8f7e-acec44dee6ee | StampyAI/alignment-research-dataset/arxiv | Arxiv | Weighing the Milky Way and Andromeda with Artificial Intelligence
I Introduction
---------------
The ΛCDM model describes how non-linear gravitational evolution amplifies the small dark matter perturbations after inflation creating larger and deeper gravitational potential wells where gas can collapse, cool down, a... |
093a465c-1bca-42e8-88e4-34fc3a375d36 | trentmkelly/LessWrong-43k | LessWrong | Join the AI Testing Hackathon this Friday
TLDR: Participate online or in-person on the weekend of December 16th to 18th in a fun and intense AI safety research hackathon focused on benchmarks, neural network verification, adversarial attacks and RL safety environments. We invite mid-career professionals to join, but t... |
68d0d961-e844-4dfd-aab0-d35981e3bf1e | trentmkelly/LessWrong-43k | LessWrong | AXRP Episode 7 - Side Effects with Victoria Krakovna
YouTube link
This podcast is called AXRP, pronounced axe-urp and short for the AI X-risk Research Podcast. Here, I (Daniel Filan) have conversations with researchers about their papers. We discuss the paper and hopefully get a sense of why it’s been written and how... |
fb0da0e0-4b0e-4110-a015-b260a8ad57ea | trentmkelly/LessWrong-43k | LessWrong | Abandoning Cached Selves to Re-Write My Source Code Partially, I've Become Unstable
For very long I've been caring a lot for the preferences of my past selves.
Rules I established in childhood became sacred, much like laws are (can't find post in the sequences in which Yudkowsky is amazed by the fact that some thing... |
ea51542f-7c1a-46ac-9257-df44f00c984c | StampyAI/alignment-research-dataset/eaforum | Effective Altruism Forum | Call to demand answers from Anthropic about joining the AI race
I'm surprised that once again it starts as "let's work on safety together! Let's share ideas and work on a good thing", then some entity grows bigger and bigger, starts taking unilateral decisions that are controversial, still using the name and support ... |
cc5b7f26-7279-4a9d-a8f3-e503400be7e3 | StampyAI/alignment-research-dataset/special_docs | Other | Towards Cooperation in Learning Games
Towards cooperation in learning games
Jesse Clifton Maxime Riché
Abstract
Suppose that several actors are going to deploy learning agents to act on their behalf. What principles
should guide these actors in designing their agents, given that they may have competing goals? An
appea... |
60348c0a-f724-4dec-9655-d79ccf058c25 | trentmkelly/LessWrong-43k | LessWrong | Test Data for Calibration Analysis
Hi All,
I was wondering if anyone knows of a dataset with probabilistic predictions for a binary outcome that has published results for calibration loss, resolution, and refinement loss? We've developed some code and an accompanying web tool for generating calibration plots, ROC cur... |
25e1527c-f91a-4dd2-8122-571d8a3d42e2 | trentmkelly/LessWrong-43k | LessWrong | 2017-10-08 - London Rationalish meetup
We have a meetup group in London, and our next meetup is this Sunday, 2pm at the Shakespeare's Head in Holborn. More details are in the Facebook event. |
f5c15a52-7a94-4754-8e2f-f3c5d4b89037 | trentmkelly/LessWrong-43k | LessWrong | A rough model for P(AI doom)
(I am writing this primarily so I can reference it in another article that I'm writing to keep that article from getting longer)
Currently, my best guess is that P(AI doom by 2100) ≈ 20%. That is, there’s a 20% chance that strong AIs will be an existential challenge for humanity. But “exi... |
d80e3902-c4cd-4c19-8eb5-db97898850db | trentmkelly/LessWrong-43k | LessWrong | Alignment Newsletter #53
Find all Alignment Newsletter resources here. In particular, you can sign up, or look through this spreadsheet of all summaries that have ever been in the newsletter.
Cody Wild is now contributing summaries to the newsletter!
Highlights
Alignment Newsletter One Year Retrospective (Rohin Sha... |
99947448-0949-4ab5-b6db-879d9d838289 | StampyAI/alignment-research-dataset/special_docs | Other | Can Machine Learning Be Secure?
Can Machine Learning Be Secure?
Marco Barreno Blaine Nelson Russell Sears Anthony D. Joseph J. D. Tygar
Computer Science Division
University of California, Berkeley
{barreno,nelsonb,sears,adj,tygar }@cs.berkeley.edu
ABSTRACT
Machine learning systems offer unparalled flexibility in deal-
i... |
c7aee89a-9a5a-4b38-a611-a19d99b13448 | trentmkelly/LessWrong-43k | LessWrong | Thomas Kwa's MIRI research experience
|
1d144c8b-a1eb-4abe-ae72-b1038e5209b8 | trentmkelly/LessWrong-43k | LessWrong | High impact from low impact
A putative new idea for AI control; index here.
Part of the problem with a reduced impact AI is that it will, by definition, only have a reduced impact.
Some of the designs try and get around the problem by allowing a special "output channel" on which impact can be large. But that feels l... |
839aaa9c-2025-4fa8-9644-f815ebc07f4f | trentmkelly/LessWrong-43k | LessWrong | When Omnipotence is Not Enough
Some knowledge cannot be gifted or received. It can only be stolen.
----------------------------------------
In the east, there was a garden. In the garden there was a God, a man, a woman and a snake. They all had legs.
"Did God really tell you, 'You must not eat from any tree in the ... |
55c1744c-b71a-4867-9ca2-c0186989cdcd | trentmkelly/LessWrong-43k | LessWrong | A follow-up probability question: Data samples with different priors
(Rewritten entirely after seeing pragmatist's answer.)
In this post, helpful people including DanielLC gave me the multiply-odds-ratios method for combining probability estimates given by independent experts with a constant prior, with many comments... |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.