id stringlengths 36 36 | source stringclasses 15
values | formatted_source stringclasses 13
values | text stringlengths 2 7.55M |
|---|---|---|---|
39d1bdd8-a3fd-4d9a-858b-282c97e432ef | trentmkelly/LessWrong-43k | LessWrong | Fallacies as weak Bayesian evidence
Abstract: Exactly what is fallacious about a claim like ”ghosts exist because no one has proved that they do not”? And why does a claim with the same logical structure, such as ”this drug is safe because we have no evidence that it is not”, seem more plausible? Looking at various fa... |
2e46f9d9-ae48-47d3-8955-103376efdda2 | trentmkelly/LessWrong-43k | LessWrong | Meetup : Washington, D.C.: Anime & Manga Discussion
Discussion article for the meetup : Washington, D.C.: Anime & Manga Discussion
WHEN: 09 November 2014 03:00:00PM (-0500)
WHERE: National Portrait Gallery
We will be meeting in the Kogod Courtyard of the National Portrait Gallery (8th and F Sts or 8th and G Sts NW,... |
11b21776-577d-48a0-b759-2cd65338119c | trentmkelly/LessWrong-43k | LessWrong | Two flavors of computational functionalism
This is intended to be the first in a sequence of posts where I scrutinize the claims of computational functionalism (CF). I used to subscribe to it, but after more reading, I’m pretty confused about whether or not it’s true. All things considered, I would tentatively bet ... |
bb7cad5a-f563-48bb-99ed-9e3c6732c528 | trentmkelly/LessWrong-43k | LessWrong | [SEQ RERUN] The Generalized Anti-Zombie Principle
Today's post, The Generalized Anti-Zombie Principle was originally published on 05 April 2008. A summary (taken from the LW wiki):
> The argument against zombies can be extended into a more general anti-zombie principle. But, figuring out what that more general pri... |
93da817f-590f-4236-afc1-b83a671e27b4 | trentmkelly/LessWrong-43k | LessWrong | Meetup : Sydney Meetup - March - Bring a Newbie!
Discussion article for the meetup : Sydney Meetup - March - Bring a Newbie!
WHEN: 25 March 2015 06:30:00PM (+1100)
WHERE: 565 George Street, Sydney, Australia 2000
It's time again to think about bringing along some new people. If you haven't been in a while, this is ... |
1eb97483-c5e7-40ff-8cb2-28216705fe90 | trentmkelly/LessWrong-43k | LessWrong | [SEQ RERUN] Timeless Causality
Today's post, Timeless Causality was originally published on 29 May 2008. A summary (taken from the LW wiki):
> Using the modern, Bayesian formulation of causality, we can define causality without talking about time - define it purely in terms of relations. The river of time never fl... |
776c12e9-2b5c-4326-8986-6168acf24e69 | trentmkelly/LessWrong-43k | LessWrong | Notes/blog posts on two recent MIRI papers
I've been learning math lately; specifically I've been reading MIRI's recent research preprints and the prerequisite material. In order to actually learn math, I typically have to write it down again, usually with more details and context. I started a blog to make my notes on... |
4166c70e-823a-41d3-8b6c-01608772c33e | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | On motivations for MIRI's highly reliable agent design research
(this post came out of a conversation between me and Owen Cotton-Barratt, plus a follow-up conversation with Nate)
I want to clarify my understanding of some of the motivations of MIRI's highly reliable agent design (HRAD) research (e.g. logical uncerta... |
d2bc5660-9d9b-4ade-a08c-4c3969f9cf8b | trentmkelly/LessWrong-43k | LessWrong | Crazy Global Warming Solution Ideas
Mine was to work tax policy to incentivize companies to make all their packaging shiny and white, incentivize people to litter, and disincentivize everybody from recycling.
My friend's was to use a giant rocket to push the earth farther away from the sun |
a723b103-251c-45fe-a905-7d7a2b8cda0b | trentmkelly/LessWrong-43k | LessWrong | Simpler explanations of AGI risk
We're getting a shot at presenting our concerns about AI X-risk to the general public. It would be useful to have a brief presentation that plays well with less-technical people, or technical people who don't want to listen to a half hour of explanation just at that moment. The other g... |
c3adc6b6-f82e-4d3f-9a03-b005b55d092c | trentmkelly/LessWrong-43k | LessWrong | Smoothmin and personal identity
I value existential risk reduction; I also value poverty reduction. These two things trade off against each other.
I value being generous; I also value reading interesting books. These two things trade off against each other.
But the way the two tradeoffs work do not seem to be the sa... |
04198022-923f-4340-a358-213c423a9417 | trentmkelly/LessWrong-43k | LessWrong | Lack of Spider-Man is evidence against the simulation hypothesis
This is a very short post, but I haven't seen this argument elsewhere so I wanted to write it up.
The basic argument in favor of the simulation hypothesis runs like this:
1. Far future people will likely be able to and want to create simulated realiti... |
769e3de3-1a8c-4889-ab80-ee6a929edf3f | trentmkelly/LessWrong-43k | LessWrong | Institute for Assured Autonomy (IAA) newsletter
Just a very brief link to a resource that people may not know about:
Johns Hopkins Institute for Assured Autonomy newsletter
For anybody interested in nearer-term AI safety / assured autonomy issues, I have found this recently-started newsletter to be a useful aggregat... |
9b8a363d-5d90-4b97-840e-09295f15dcc9 | trentmkelly/LessWrong-43k | LessWrong | [AN #144]: How language models can also be finetuned for non-language tasks
Alignment Newsletter is a weekly publication with recent content relevant to AI alignment around the world. Find all Alignment Newsletter resources here. In particular, you can look through this spreadsheet of all summaries that have ever been... |
2bfb5420-cf4a-4f37-9933-7eee092329b8 | trentmkelly/LessWrong-43k | LessWrong | Timeless physics breaks T-Rex's mind [LINK]
From Dinosaur Comics, with a nice shout-out to Eliezer's Timeless Physics post! (Look in the newspost below the comic.)
I can't wait until Ryan North gets around to Newcomb's Problem... |
0a68995a-98d0-4246-a957-1c1bb74e6e83 | trentmkelly/LessWrong-43k | LessWrong | Does Circuit Analysis Interpretability Scale? Evidence from Multiple Choice Capabilities in Chinchilla
Cross-posting a paper from the Google DeepMind mech interp team, by: Tom Lieberum, Matthew Rahtz, János Kramár, Neel Nanda, Geoffrey Irving, Rohin Shah, Vladimir Mikulik
Informal TLDR
* We tried standard mech in... |
badb2722-85b6-4d5a-b60d-4178121640e8 | trentmkelly/LessWrong-43k | LessWrong | What I think about you
> So, having heard Mike Li compare Jaynes to a thousand-year-old vampire, one question immediately popped into my mind:
>
> "Do you get the same sense off me?" I asked.
>
> -- Eliezer Yudkowsky, The Level Above Mine
Most people need feedback in many areas. Most people can give feedback in man... |
690d88fa-38be-455b-b545-4381f8120209 | trentmkelly/LessWrong-43k | LessWrong | One Year of Goodsearching
Followup to: Use Search Engines Early and Often
Last year, I posted about using search engines and particularly recommended GoodSearch, a site that donates one cent to a charity of your choice whenever you make a (Bing-powered) search via their site.
At the time, some seemed skeptical of th... |
0845bcf7-7dd6-4300-8ad9-d3b9e386ac47 | trentmkelly/LessWrong-43k | LessWrong | The highest-probability outcome can be out of distribution
"Out of distribution" is a phenomenon where a data point x does not come from the distribution you had in mind. To give some examples, here is what might be considered OOD:
* You play a dice game against a cheater, and the cheater uses a die that has been we... |
7b0822b5-e336-46b4-ad47-5ace867e2494 | trentmkelly/LessWrong-43k | LessWrong | Meetup : San Francisco: Effective Altruism
Discussion article for the meetup : San Francisco: Effective Altruism
WHEN: 28 June 2013 07:54:30PM (-0700)
WHERE: 170 Saint Germain Ave, San Francisco
This is going to be a salon-style event, so we'll have a medium length talk, followed by some Q/A, followed by some gener... |
7fe08d50-5803-4731-bc14-09db9076b4a3 | StampyAI/alignment-research-dataset/distill | Distill Scientific Journal | Communicating with Interactive Articles
Computing has changed how people communicate. The transmission of news, messages, and ideas is instant. Anyone’s voice can be heard. In fact, access to digital communication technologies such as the Internet is so fundamental to daily life that their disruption by government is ... |
80acbd5e-d2db-4827-a062-d1e798322000 | StampyAI/alignment-research-dataset/lesswrong | LessWrong | [Linkpost] Mapping Brains with Language Models: A Survey
*This is a linkpost for* [*https://arxiv.org/abs/2306.05126*](https://arxiv.org/abs/2306.05126)*.*
> Over the years, many researchers have seemingly made the same observation: Brain and language model activations exhibit some structural similarities, enabling ... |
6c26c82d-5217-40f7-a9cd-7d7c91ac92aa | trentmkelly/LessWrong-43k | LessWrong | Designing for a single purpose
If you’ve ever been to Amsterdam, you’ve probably visited, or at least heard about the famous cookie store that sells only one cookie. I mean, not a piece, but a single flavor.
I’m talking about Van Stapele Koekmakerij of course—where you can get one of the world's most delicious chocol... |
8995c28b-c49e-407e-924f-1f8f1a9314ae | trentmkelly/LessWrong-43k | LessWrong | You can rack up massive amounts of data quickly by asking questions to all your friends
tl;dr I asked 15 friends and acquaintances a question out of the blue because I was curious about something. I collected a massive amount of evidence, and it was fun, and I’m more agentic now than I was yesterday. Do it too.
La... |
b174d685-fd24-4aad-8774-4c1a1a04f73e | StampyAI/alignment-research-dataset/eaforum | Effective Altruism Forum | Hiring engineers and researchers to help align GPT-3
My team at OpenAI, which works on [aligning](https://ai-alignment.com/clarifying-ai-alignment-cec47cd69dd6) GPT-3, is hiring ML engineers and researchers. **Apply [here](https://jobs.lever.co/openai/98599d5b-2d1d-4127-b9b5-708343c8730b) for the ML engineer role and ... |
81a47183-69a6-4550-9040-92bc96268332 | StampyAI/alignment-research-dataset/eaforum | Effective Altruism Forum | The Centre for the Governance of AI is becoming a nonprofit
EDIT: We have now relaunched. More details [here](https://forum.effectivealtruism.org/posts/5vGsFzWvh8Snet6sK/the-centre-for-the-governance-of-ai-has-relaunched)
We wanted to update the EA community about some changes to the Centre for the Governance of AI ... |
66a1bea4-4fe5-4f8e-9e78-9c5a5965f62e | StampyAI/alignment-research-dataset/arxiv | Arxiv | Fully General Online Imitation Learning
1 Introduction
---------------
Supervised learning of independent and identically distributed data is often practiced in two phases: training and deployment. This separation makes less sense if the learner’s predictions affect the distribution of future tasks, since the deplo... |
a63e9672-2845-4388-b2a0-d69dfcc6f742 | trentmkelly/LessWrong-43k | LessWrong | Can we infer the search space of a local optimiser?
Suppose we have a system that we suspect is some form of local optimiser. Maybe it's gradient descent training a neural network. Maybe it’s a neural network doing in-context learning. Maybe it's a mind, because we guess that the operation of minds in general is to an... |
91e4b249-f82b-4e89-9979-d60d1efc7da6 | trentmkelly/LessWrong-43k | LessWrong | [Intuitive self-models] 2. Conscious Awareness
2.1 Post summary / Table of contents
Part of the Intuitive Self-Models series.
The previous post laid some groundwork for talking about intuitive self-models. Now we’re jumping right into the deep end: the intuitive concept of “conscious awareness” (or “awareness” for sh... |
82dbfcef-6f8d-422e-9938-eb21e9beec8f | trentmkelly/LessWrong-43k | LessWrong | Practical Conflict Resolution: A Taxonomy of Disagreement
The previous posts in this sequence were largely setup; still important, but not intended to contain significant new material. Now, finally, we get to the good stuff. Once you’ve got the right attitude, good communication, and a decent grasp of what both sides ... |
87152ed4-caab-4ff7-b87f-513cc2f383fc | trentmkelly/LessWrong-43k | LessWrong | UK AISI’s Alignment Team: Research Agenda
The UK’s AI Security Institute published its research agenda yesterday. This post gives more details about how the Alignment Team is thinking about our agenda.
Summary: The AISI Alignment Team focuses on research relevant to reducing risks to safety and security from AI syste... |
1baf047d-0e88-41d6-a1e7-5c01927347e9 | trentmkelly/LessWrong-43k | LessWrong | AI Safety Needs Great Engineers
Top line: If you think you could write a substantial pull request for a major machine learning library, then major AI safety labs want to interview you today.
I work for Anthropic, an industrial AI research lab focussed on safety. We are bottlenecked on aligned engineering talent. Spec... |
81a2fbf2-d2ac-4152-a8fe-07d1f900d071 | trentmkelly/LessWrong-43k | LessWrong | False assumptions and leaky abstractions in machine learning and AI safety
* The problems of embedded agency are due to the notion of agency implicit in reinforcement learning being a leaky abstraction.
* Machine learning problem statements often makes assumptions that are known to be false, for example, assuming i.... |
431cd3f0-7e72-4092-9b1f-bc46241f268e | awestover/filtering-for-misalignment | Redwood Research: Alek's Filtering Results | id: e0f56c485d032069
[2412.12480] Subversion Strategy Eval: Can language models statelessly strategize to subvert control protocols? Skip to main content We gratefully acknowledge support from the Simons Foundation, member institutions , and all contributors. Donate > cs > arXiv:2412.12480 Help | Advanced Search All f... |
d0ca2daf-521e-484b-b520-4fb2661c2b98 | trentmkelly/LessWrong-43k | LessWrong | Language and My Frustration Continue in Our RSI
What's this post about?
I make some rants and recommendations about terminology.
This is written for AI-not-kill-everyone-ists. If you are worried about AI killing everyone and want us to prevent AI from killing everyone, this post is for you. If you don't have that age... |
c95cccde-96a1-41b4-9cd0-419cb2bf69f5 | trentmkelly/LessWrong-43k | LessWrong | Distilled Representations Research Agenda
Introduction
Schematic of the basic training strategy
I’ve recently been given funding from the Long Term Future Fund to develop work on an agenda I'll tentatively call Distilled Representations, and I'll be working on this full-time over the next 6 months with Misha Wagner (p... |
d1a7220a-3208-40d1-bbe2-4f374ee02403 | trentmkelly/LessWrong-43k | LessWrong | Evolution did a surprising good job at aligning humans...to social status
[This is post is a slightly edited tangent from my dialogue with John Wentworth here. I think the point is sufficiently interesting and important that I wanted to make it as a top level post, and not leave it buried in that dialog on mostly anot... |
f597d5fd-ef33-4ba8-9952-d74db10d9442 | trentmkelly/LessWrong-43k | LessWrong | Towards a Clever Hans Test: Unmasking Sentience Biases in Chatbot Interactions
The famous story of Clever Hans has become a cautionary tale in animal cognition. Hans was a horse in Germany in the early 1900s who could seemingly perform all kinds of smart tasks, such as simple arithmetic and spelling words. It is not... |
cae17a68-cbc9-4709-8569-136a5633a135 | trentmkelly/LessWrong-43k | LessWrong | How are Those AI Participants Doing Anyway?
[TL;DR] Some social science researchers are running psychology experiments with LLMs playing the role of human participants. In some cases the LLMs perform their roles well. Other social scientists, however, are quite concerned and warn of perils including bias, reproducibil... |
61c23914-3a9f-41d2-89e4-0efd29317493 | trentmkelly/LessWrong-43k | LessWrong | Does the "ancient wisdom" argument have any validity? If a particular teaching or tradition is old, to what extent does this make it more trustworthy?
Proponents of spirituality and alternative medicine often use the argument "this has been practiced for 2000 years", with the subtext "therefore it must work". Does thi... |
99860f2b-48a7-4df8-87d4-1b45bd1e98bd | trentmkelly/LessWrong-43k | LessWrong | 1. A Sense of Fairness: Deconfusing Ethics
I’d like to write about a subject that I’ve been thinking about for over a decade: the intersection of AI, Alignment, and Ethics. I initially discussed a few bits of this in a couple of comments, which got quite a bit of discussion and even some agreement votes, so I thought ... |
1bbf278b-e669-488f-839f-29f359a745f3 | trentmkelly/LessWrong-43k | LessWrong | The Wiki is Dead, Long Live the Wiki! [help wanted]
That's just a few of them. We imported like 5x as many as these.
With the goal of eventually archiving it fully, we have imported 573 pages and 266,000 words of content from the old LessWrong wiki to LessWrong 2.0
The old wiki is a great store of knowledge and still... |
832dc116-32d1-47d6-88e0-783ac1aa3c11 | StampyAI/alignment-research-dataset/youtube | Youtube Transcripts | Scalable Supervision: Concrete Problems in AI Safety Part 5
hi this is part of a series of videos
about the paper concrete problems in AI
safety it should make some sense on its
own but I'd recommend checking out the
other videos first there's a link to the
playlist in the description so before we
talked about some pr... |
1bf20eda-1990-440e-9b4b-3ea8f506a085 | trentmkelly/LessWrong-43k | LessWrong | Cognito Mentoring: An advising service for intellectually curious students
My name is Jonah Sinick, and I'm posting to announce a new advising service for intellectually curious students: Cognito Mentoring. I'm working on this in collaboration with Vipul Naik.
We have very broad intellectual interests, cutting across... |
9c4b20ea-d48c-4907-b46e-9ab535bae16e | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Take 7: You should talk about "the human's utility function" less.
*As a writing exercise, I'm writing an AI Alignment Hot Take Advent Calendar - one new hot take, written every day (more or less - I'm getting back on track!) for 25 days. Or until I run out of hot takes.*
When considering AI alignment, you might be t... |
892fc947-a957-4ecc-bc4e-3a88e3931464 | trentmkelly/LessWrong-43k | LessWrong | Theory of Knowledge (rationality outreach)
Public schools (and arguably private schools as well; I wouldn't know) teach students what to think, not how to think.
On LessWrong, this insight is so trivial not to bear repeating. Unfortunately, I think many people have adopted it as an immutable fact about the world that... |
393aa5d3-00aa-40bc-9de7-c58cd1e7370a | trentmkelly/LessWrong-43k | LessWrong | Request for Intelligence Philosophy Essay Topic Suggestions
As part of a philosophy course I'm currently taking called Intelligence in Machines, Humans, and Other Animals, I have to write a <3000w essay on a topic related to intelligence. The description is here, but I've copied the important details below. I figured ... |
6fa3f731-347f-4b8a-bea2-cc3600b5aa02 | StampyAI/alignment-research-dataset/eaforum | Effective Altruism Forum | Research Summary: Forecasting with Large Language Models
This is post was originally published on my blog [here](https://damienlaird.substack.com/p/research-forecasting-with-large-language), where I write about trying to improve forecasting techniques for Global Catastrophic Risks. As always, feedback or collaboration... |
2dfb002d-9979-4763-97f5-5496d941b7d7 | trentmkelly/LessWrong-43k | LessWrong | Avoiding Side Effects in Complex Environments
Previously: Attainable Utility Preservation: Empirical Results; summarized in AN #105
Our most recent AUP paper was accepted to NeurIPS 2020 as a spotlight presentation:
> Reward function specification can be difficult, even in simple environments. Rewarding the agent fo... |
2b41c902-b2b5-4f92-9133-09c6dbc11bd9 | trentmkelly/LessWrong-43k | LessWrong | Trust develops gradually via making bids and setting boundaries
In my previous post, I talked through the process of identifying the fears underlying internal conflicts. In some cases, just listening to and understanding those scared parts is enough to make them feel better—just as, when venting to friends or partners... |
0bd4c85d-f8b5-4837-ac24-3ce5483ac31f | trentmkelly/LessWrong-43k | LessWrong | [SEQ RERUN] Zen and the Art of Rationality
Today's post, Zen and the Art of Rationality was originally published on 24 December 2007. A summary (taken from the LW wiki):
> Rationality is very different in its propositional statements from Eastern religions, like Taoism or Buddhism. But, it is sometimes easier to e... |
07e80fc2-2591-4863-975b-0d1693933de4 | trentmkelly/LessWrong-43k | LessWrong | Scholarship and DIY Science
Related: Science: do it yourself, Some Heuristics for Evaluating The Soundness of the Academic Mainstream in Unfamiliar Fields, The Neglected Virtue of Scholarship
There's been some recent discussion about the value of answering your questions by reading the established scholarly literatur... |
08ea4547-bb40-4183-b175-77e555aeb4b1 | trentmkelly/LessWrong-43k | LessWrong | chinchilla's wild implications
(Colab notebook here.)
This post is about language model scaling laws, specifically the laws derived in the DeepMind paper that introduced Chinchilla.[1]
The paper came out a few months ago, and has been discussed a lot, but some of its implications deserve more explicit notice in my o... |
96c3f4b0-5e9a-4d1a-8d3f-039dd79bc5de | trentmkelly/LessWrong-43k | LessWrong | Another Unphrased B-part
Usually everything in a contra dance matches up exactly with the music. The tunes repeat every 64 beats, and are divided into two 32-beat halves (A and B), each of which is broken up into two 16-beat halves (A1 and A2, B1 and B2), each of which is broken into two 8-beat halves (not usually nam... |
75643177-8d17-44b6-ac22-cab3fd2f2cdf | trentmkelly/LessWrong-43k | LessWrong | My Fear Heuristic
My friends and family call me "risk tolerant". I wasn't always this way. It is the result of a 3-year-long scientific experiment.
> You must do everything that frightens you…Everything. I’m not talking about risking your life, but everything else. Think about fear, decide right now how you’re doing ... |
be845931-4e7d-413b-ae57-b9d8a1206d39 | trentmkelly/LessWrong-43k | LessWrong | The Goldbach conjecture is probably correct; so was Fermat's last theorem
EDIT: Added a section on Euler's conjecture.
The Goldbach conjecture is likely
The Goldbach conjecture is that "every even integer above two is the sum of two primes". For example, 4=2+2, 6=3+3, 8=5+3, and so on.
Though this is a mathematical... |
92612353-c66f-4460-b374-8b92b9b0fdec | trentmkelly/LessWrong-43k | LessWrong | Welcome to Orlando Rationalist Group
We're a newly-formed group! I'm hoping to meet up the first Friday of each month, at or around University of Central Florida. We're starting with just casual social gatherings. We may eventually add more structured activities over time. Anyone interested in LessWrong, SSC/ACX, or a... |
a8c43b72-a60a-4096-b1c8-9dde470f8973 | trentmkelly/LessWrong-43k | LessWrong | Why Did NEPA Peak in 2016?
The National Environmental Policy Act (NEPA) is a 1970 law which requires all federal agencies to “prepare detailed statements assessing the environmental impact of and alternatives to major federal actions significantly affecting the environment.” The exact definitions of ‘detailed,’ ‘major... |
06a25d71-3bb4-4155-92e1-7bb0be3fdd03 | trentmkelly/LessWrong-43k | LessWrong | Your Time Might Be More Valuable Than You Think
Summary
* People often seem to implicitly value their time at the amount they can convert hours to dollars given their current skills.
* However, the value of saving the marginal hour today is to increase the total number of one's working hours by one, resulting in a n... |
55df704e-a428-40b9-835c-d24a70906309 | trentmkelly/LessWrong-43k | LessWrong | Did convalescent plasma therapy drive the evolution of more infectious covid strains?
I have seen articles in the National Geographic and Indian Express suggesting that some of the newer, more infectious covid strains that have been popping up lately may have originated from patients treated with convalescent plasma t... |
d459f944-6063-4b46-957c-b222adb532b9 | StampyAI/alignment-research-dataset/blogs | Blogs | Announcing the new AI Alignment Forum
*This is a guest post by Oliver Habryka, lead developer for LessWrong. Our gratitude to the LessWrong team for the hard work they’ve put into developing this resource, and our congratulations on today’s launch!*
---
I am happy to announce that after two months of open beta, th... |
bf51d501-4417-4f14-ba42-7d728be701a7 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Measuring and Improving the Faithfulness of Model-Generated Reasoning
**TL;DR**: In two new papers from Anthropic, we propose metrics for evaluating how faithful chain-of-thought reasoning is to a language model's actual process for answering a question. Our metrics show that language models sometimes ignore their gen... |
cdc056ad-aeb3-4f70-a202-d5861f45d58b | trentmkelly/LessWrong-43k | LessWrong | AI Safety Newsletter #1 [CAIS Linkpost]
The Center for AI Safety just launched its first AI Safety Newsletter. The newsletter is designed to inform readers about developments in AI and AI safety. No technical background required.
Subscribe here to receive future versions. First edition below:
---
Growing concerns... |
0bcf7c96-f6e5-4e55-89ed-a5f871a3b1cb | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Jobs: Help scale up LM alignment research at NYU
[NYU is hiring alignment-interested researchers!](https://docs.google.com/document/u/0/d/1r-ZsBQK07OHohLa9c9qCoLHZVV80EvnIOfpsdfVk0eg/edit?fromCopy=true)
* I'll be working on setting up a CHAI-inspired research center at NYU that'll focus on empirical alignment work, p... |
8234cfdc-5947-4784-a814-5dd18be7ed27 | trentmkelly/LessWrong-43k | LessWrong | Meetup : Washington DC Book Swap meetup
Discussion article for the meetup : Washington DC Book Swap meetup
WHEN: 09 February 2014 03:00:00PM (-0500)
WHERE: national portrait gallery
We'll be meeting to exchange books!
Discussion article for the meetup : Washington DC Book Swap meetup |
1fb4cecc-156d-49b2-b7e6-85de7fe15dc9 | StampyAI/alignment-research-dataset/lesswrong | LessWrong | A first success story for Outer Alignment: InstructGPT
OpenAI has given us a first taste of what success looks like in outer alignment, and that's InstructGPT. More specifically, InstructGPT actually is more aligned in two important ways: Truthfulness and hallucinations.
Specifically, OpenAI used RLHF to make GPT-3... |
4fd5e7a2-bc24-4391-8218-cfcc28a83428 | trentmkelly/LessWrong-43k | LessWrong | Slack
Epistemic Status: Reference post. Strong beliefs strongly held after much thought, but hard to explain well. Intentionally abstract.
Disambiguation: This does not refer to any physical good, app or piece of software.
Further Research (book, recommended but not at all required, take seriously but not literally)... |
1e454d8e-c730-4387-8cdb-f03a70a0365e | trentmkelly/LessWrong-43k | LessWrong | How good is humanity at coordination?
When EAs look at the history of nuclear weapons, their reactions tend to fall into two camps.
The first camp (which I am inclined towards) is “Man, what a total mess. There were so many near misses, and people involved did such clearly terrible and risky things like setting up th... |
865c04aa-f9c3-44a9-b765-8fb0d69e1c6c | trentmkelly/LessWrong-43k | LessWrong | So you think you understand Quantum Mechanics
This post is prompted by the multitude of posts and comments here using quantum this and that in an argument (quantum dice, quantum immortality, quantum many worlds...). But how does one know if they understand the concept they use? In school a student would have to write ... |
d4ad4219-8651-43c6-b6c0-dd7a5dc21ed2 | trentmkelly/LessWrong-43k | LessWrong | Procedural Knowledge Gaps
I am beginning to suspect that it is surprisingly common for intelligent, competent adults to somehow make it through the world for a few decades while missing some ordinary skill, like mailing a physical letter, folding a fitted sheet, depositing a check, or reading a bus schedule. Since the... |
026589b0-1a95-487f-9fdf-c6173d988e2b | trentmkelly/LessWrong-43k | LessWrong | ChatGPT on Spielberg’s A.I. and AI Alignment
A remark by MarcusOfCitium over at Astral Codex Ten prompted me to think about the relationship between John Bowlby’s concept of attachment, Steven Spielberg’s movie, AI: Artificial Intelligence, and the problem of AI Alignment. I decided ask ChatGPT what they thought about... |
d8c35068-54c6-4f6c-8bf8-9ae4b5f436f9 | StampyAI/alignment-research-dataset/blogs | Blogs | New Paper: “Why We Need Friendly AI”
[](https://intelligence.org/files/WhyWeNeedFriendlyAI.pdf)A new paper by Luke Muehlhauser and Nick Bostrom, “[Why We Need Friendly AI](https://intelligence.org/files/WhyWeNeedFri... |
e8cd2505-882c-406f-86e9-c4289e168f94 | StampyAI/alignment-research-dataset/eaforum | Effective Altruism Forum | FHI Report: How Will National Security Considerations Affect Antitrust Decisions in AI? An Examination of Historical Precedents
>
> Part I lists and analyzes the cases I identified, then lists several conclusions I was able to draw from my case studies. Those conclusions are:
>
>
> 1. National security consideratio... |
a93874d6-46ac-43e6-a5b8-6dccf6868007 | LDJnr/LessWrong-Amplify-Instruct | LessWrong | "Robert Sapolsky: Baboons... literally have been the textbook example of a highly aggressive, male-dominated, hierarchical society. Because these animals hunt, because they live in these aggressive troupes on the Savannah... they have a constant baseline level of aggression which inevitably spills over into their socia... |
0a663c63-f219-4678-bd91-9cfee1d83a13 | awestover/filtering-for-misalignment | Redwood Research: Alek's Filtering Results | id: post3819
The nearest unblocked strategy problem (NUS) is the idea that if you program a restriction or a patch into an AI, then the AI will often be motivated to pick a strategy that is as close as possible to the banned strategy, and maybe just as dangerous. For instance, if the AI is maximising a reward R , and ... |
645b318f-28ff-40a1-bcc1-a06e6c7a6ba2 | StampyAI/alignment-research-dataset/special_docs | Other | NeurIPSorICML_7oalk-by Vael Gates-date 20220320
# Interview with AI Researchers NeurIPSorICML\_7oalk by Vael Gates
\*\*Interview with 7oalk, on 3/20/22\*\*
\*\*0:00:03.2 Vael:\*\* Alright, my first question is, can you tell me about what area of AI you work on in a few sentences?
\*\*0:00:08.5 Interviewee:\*\* Yeah... |
6d068527-3fc0-49ad-b841-1f0d793b3568 | StampyAI/alignment-research-dataset/arbital | Arbital | Transitive relation
A binary [https://arbital.com/p/-3nt](https://arbital.com/p/-3nt) $R$ is **transitive** if whenever $aRb$ and $bRc$, $aRc$.
The most common examples or transitive relations are [partial orders](https://arbital.com/p/3rb) (if $a \leq b$ and $b \leq c$, then $a \leq c$) and [equivalence relations](... |
e0bf1b47-ce15-4a3a-9141-81732204c7d0 | trentmkelly/LessWrong-43k | LessWrong | Why Does It Feel Like Something? An Evolutionary Path to Subjectivity
How does the brain generate subjective experience – the feeling of 'what-it's-like' to be us? Explanations differ: some seek physical mechanisms (Physicalists included), while others find such accounts lacking or impossible in principle. This articl... |
5f4c8675-e2f3-41d7-bbce-667649dbcf1e | trentmkelly/LessWrong-43k | LessWrong | ChatGPT explores the semantic differential
Cross-posted from New Savanna.
3.9.23 – 6:13 AM – Feb 13 Version
If you don’t already know what the semantic differential is, look it up in Wikipedia. But you should get the idea from the prompts I give it. It’s a standard technique is psychology going back to the middle of... |
2d644d55-fab1-42d5-b344-2729734bec71 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | How to talk about reasons why AGI might not be near?
I occasionally have some thoughts about why AGI might not be as near as a lot of people seem to think, but I'm confused about how/whether to talk about them in public.
The biggest reason for not talking about them is that one person's "here is a list of capabilitie... |
671ebf50-d113-4fdd-8f13-d88e1d4c823b | trentmkelly/LessWrong-43k | LessWrong | We Should Talk About This More. Epistemic World Collapse as Imminent Safety Risk of Generative AI.
“So you’re telling me we can’t believe anything we’re shown anymore?” I’m asking. “That everything is altered? That everything’s a lie? That everyone will believe this?” “That’s a fact,” Palakon says. “So what’s true, th... |
194d5871-968f-46e6-96b3-2efefbd99094 | trentmkelly/LessWrong-43k | LessWrong | Trivial Mathematics as a Path Forward
A gentleman by the name of Dan Willard, who recently passed, was a pioneer in a relatively unexamined area of mathematics: Mathematical frameworks that are just trivial enough to bypass Gödel. That is, mathematical frameworks that can prove their own consistency. He did so for a... |
f29b7deb-9de9-49e9-b88c-d7c7078c1134 | trentmkelly/LessWrong-43k | LessWrong | Related Discussion from Thomas Kwa's MIRI Research Experience
One of the subthreads in Thomas Kwa's MIRI research experience was about his experience working with Nate. In the comments, some other people brought up their own negative experiences. There was a lot of ensuing discussion about it.
Thomas felt this was di... |
26946687-699f-4b3b-917e-e65fb8a7ecaa | trentmkelly/LessWrong-43k | LessWrong | Predicting a global catastrophe: the Ukrainian model
The large-scale Russian invasion into Ukraine provides an interesting model of how governments, pundits, prediction markets, and the common folks would predict an imminent global catastrophe.
Prediction markets
The Metaculus community successfully predicted that R... |
a59c414e-0c1b-4ac7-8c9b-e9d013b01e41 | trentmkelly/LessWrong-43k | LessWrong | A suite of Vision Sparse Autoencoders
CLIP-Scope?
Inspired by Gemma-Scope We trained 8 Sparse Autoencoders each on 1.2 billion tokens on different layers of a Vision Transformer. These (along with more technical details) can be accessed on huggingface. We also released a pypi package for ease of use.
We hope that th... |
0f22535f-3a81-4224-98d9-20e63fa007ba | trentmkelly/LessWrong-43k | LessWrong | Mod Note: Tagging Activity is Not Private
We don’t yet have UI for it, so this isn’t obvious (hence the announcement), but it seemed good to clearly communicate that if you tag a post or vote on the tag relevance of a tag-post combination, this information may be viewed by other users and/or the mod team (after we bui... |
7eecc82f-b0d4-4762-96ba-996d8a67ff88 | trentmkelly/LessWrong-43k | LessWrong | A Case for AI Safety via Law
This post is to make the subject case more available and open to comment. A paper with the above title is currently languishing in arXiv limbo but available in Google Docs. I was surprised to see my last and only postings to LessWrong were made on this very subject in 2010 as comments to t... |
b1d8d840-4fcc-40c2-b896-e5ad87abda18 | trentmkelly/LessWrong-43k | LessWrong | SIA becomes SSA in the multiverse
TL;DR: SIA works only in a finite universe, as an argument for a larger universe. In the infinite universe, we can compare concentrations of observers in different regions, but it is SSA.
SIA: All other things equal, an observer should reason as if they are randomly selected from... |
31398df8-8893-4ee1-8c18-21fb90a7111f | StampyAI/alignment-research-dataset/eaforum | Effective Altruism Forum | Navigating the Open-Source AI Landscape: Data, Funding, and Safety
Introduction
============
Open-source AI promises to democratize technology, but it can also be abused and may lead to hard-to-control AI. In this post, we'll explore the data, ethics, and funding behind these models to discover how to balance innovat... |
62629779-620a-4ffb-917a-ad5d61c8519a | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Less Basic Inframeasure Theory
.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: none; ... |
6fb680c7-f16e-4f42-96e2-d62e0c606d64 | trentmkelly/LessWrong-43k | LessWrong | Nvidia doesn’t just sell shovels
Imagine a land where gold is abundant yet still valuable. And everyone goes to shoveling school from birth til adulthood.
In this world, it seems that the shovel makers are not just the kingmakers, but kings themselves.
[Epistemic Status: A tad above shower thoughts.]
I've noticed... |
62ddb6b9-737b-44fb-8842-d742015cc8af | StampyAI/alignment-research-dataset/arbital | Arbital | Placeholder
This is an empty page created for a [parent](https://arbital.com/p/3n), [requisite](https://arbital.com/p/1ln), or [teaches](https://arbital.com/p/3jg) relationship. If you know about this topic and want to help others understand it, you're welcome to [fill in this gap](https://arbital.com/edit/) in [Arbit... |
9a1a4f2b-7533-4783-9f42-7d5d454b2b25 | trentmkelly/LessWrong-43k | LessWrong | Gliders in Language Models
Epistemic status: a highly speculative and rough idea that involves many concepts I’m not familiar with.
TL;DR Language models propagate features from the prompt to the text completion they generate, I call such features gliders. If powerful LMs are widely deployed on the Internet, gliders ... |
1e1ab730-d5e7-4779-b689-0a45f1dba6bd | trentmkelly/LessWrong-43k | LessWrong | Systemic review of antidepressants vs placebo commentary
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC4172306/pdf/zfp_222_3_128.pdf
The background on this story is a community of science people found a bunch of unpublished studies that, when weighed with the studies which supported antidepressant approval, showed they... |
1f305479-26d1-4833-95eb-c8c326ec9215 | StampyAI/alignment-research-dataset/blogs | Blogs | April 2020 Newsletter
[MIRI has been awarded its largest grant to date](https://intelligence.org/2020/04/27/miris-largest-grant-to-date/) — $7,703,750 split over two years from Open Philanthropy, in partnership with Ben Delo, co-founder of the cryptocurrency trading platform BitMEX!
(**Update:** Open Philanthropy deci... |
6b0d4a1f-4b48-4514-a4b7-ce0d0f5db68d | trentmkelly/LessWrong-43k | LessWrong | Historical examples of people gaining unusual cognitive abilities?
Like at the neurological/genetic level(s). Only debatable examples I can think of are Jason Padgett and a few other "acquired savants". |
11a2748b-df0f-43a6-9bb2-f45222128b92 | LDJnr/LessWrong-Amplify-Instruct | LessWrong | "A couple days ago I surveyed readers for deviant beliefs. The results were funny, hateful, boring and bonkers. One of the submissions might even be useful. If you care a lot about your mind, it is not unreasonable to avoid advertisements like plague rats, up to and including muting your speakers and averting your gaze... |
08446456-63e0-4643-bade-32692b9963b5 | StampyAI/alignment-research-dataset/youtube | Youtube Transcripts | The OTHER AI Alignment Problem: Mesa-Optimizers and Inner Alignment
hi so this channel is about ai safety
and ai alignment the core idea of ai
safety is often portrayed like this
you're a human you have some objective
that you want to achieve
so you create an ai system which is an
optimizer being an optimizer means th... |
a26298b8-c5be-4b9b-8fb4-a94d9bee7d11 | StampyAI/alignment-research-dataset/arxiv | Arxiv | On Gradient-Based Learning in Continuous Games
1 Introduction
---------------
As machine learning algorithms are increasingly being deployed in real world settings, understanding how they interact, and the dynamics that can arise from their interactions is becoming crucial. In recent years, there has been a resurge... |
a52120be-dcfc-4c77-9ea4-b4744bd33a2b | StampyAI/alignment-research-dataset/arbital | Arbital | Gödel's first incompleteness theorem
The **first incompleteness theorem** is a result about the existence of sentences of arithmetic that cannot be proved or disproved, no matter what axioms we take as true.
What Gödel originally proved was that every [$\omega$-consistent](https://arbital.com/p/) and [axiomatizable](... |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.