id stringlengths 36 36 | source stringclasses 15
values | formatted_source stringclasses 13
values | text stringlengths 2 7.55M |
|---|---|---|---|
30907935-942b-41c5-a5dd-4d9b6d29c244 | trentmkelly/LessWrong-43k | LessWrong | Print ready version of The Sequences
I've been wanting a printable copy of the sequences to read through in meatspace. I wrote a quick scraper and uploaded the results here http://pwnee.com/Sequences/list.html
Inter-linking doesn't work, but I just wanted a printable version anyway. |
1f496bda-a0b4-4575-a432-a30188473c4f | trentmkelly/LessWrong-43k | LessWrong | Welcome to Munich LW/SCC Meetup [Edit With Your Details]
(The following are our suggestions for what kind of information is best to include in the welcome post of your group, feel free to replace them with whatever you think is best)
What kind of events does your group usually run? What does it usually do?
How frequ... |
241777ae-d5e9-4919-ad66-a1d488841dda | trentmkelly/LessWrong-43k | LessWrong | Analysis of Automated Prompt Engineering for Forecasting
By Benjamin Wilson, Research Automation Engineer at Metaculus
Main Findings:
* Automated prompt optimization techniques create noticeable improvements in forecasting for some large language models, but not others.
* There were statistically significant forec... |
c73f5a11-f838-4fa4-9df5-86081dfbc890 | trentmkelly/LessWrong-43k | LessWrong | Schelling fences on slippery slopes
Slippery slopes are themselves a slippery concept. Imagine trying to explain them to an alien:
"Well, we right-thinking people are quite sure that the Holocaust happened, so banning Holocaust denial would shut up some crackpots and improve the discourse. But it's one step on the ro... |
ce7f7df5-bcb4-4150-93fe-7a04697baf4d | StampyAI/alignment-research-dataset/arxiv | Arxiv | Using Natural Language for Reward Shaping in Reinforcement Learning
1 Introduction
---------------
Reinforcement learning (RL) has enjoyed much recent success in domains ranging from game-playing to real robotics tasks.
However, to make reinforcement learning useful for large-scale real-world applications, it is cr... |
583b3648-a0f6-42a0-960e-c1d4a3eab10c | StampyAI/alignment-research-dataset/eaforum | Effective Altruism Forum | What AI companies can do today to help with the most important century
I’ve been writing about tangible things we can do today to help the [most important century](https://forum.effectivealtruism.org/s/isENJuPdB3fhjWYHd/) go well. Previously, I wrote about [helpful messages to spread](https://forum.effectivealtruism.o... |
c5ee3ef0-bb53-45a5-a541-247cd9a94027 | trentmkelly/LessWrong-43k | LessWrong | March 2017 Media Thread
This is the monthly thread for posting media of various types that you've found that you enjoy. Post what you're reading, listening to, watching, and your opinion of it. Post recommendations to blogs. Post whatever media you feel like discussing! To see previous recommendations, check out the o... |
6a0b6616-ecf0-49bf-b187-d6ddb66235a3 | StampyAI/alignment-research-dataset/aisafety.info | AI Safety Info | What are the key problems in AI governance?
Even if a [technical solution](/?state=9J1L&question=What%20are%20the%20main%20categories%20of%20technical%20alignment%20research%3F) to AI alignment is found, it won’t help if that solution isn’t applied[^kix.1ia85y511jjw]. [AI governance](/?state=8AF4&question=What%20is%20... |
4770ceb0-d8e3-4d31-af58-596aef882a39 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Internal independent review for language model agent alignment
Abstract:
---------
Language model agents (LMAs) expanding on AutoGPT are a highly plausible route to AGI. This route has large potential timeline and proliferation downsides, but large alignment advantages relative to other realistic paths to AGI. LMAs a... |
85b68573-1b4f-4ed9-ab83-93cb8962f5cc | trentmkelly/LessWrong-43k | LessWrong | What Would You Do If You Only Had Six Months To Live?
Recently, I've been pondering situations in which a person realizes, with (let's say) around 99% confidence, that they are going to die within a set period of time.
The reason for this could be a kind of cancer without any effective treatment, an injury of some ki... |
91a7dad0-53b5-4930-bf95-be33399586dd | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | What is Abstraction?
Let's start with a few examples (borrowed from [here](https://www.lesswrong.com/posts/hLFD6qSN9MmQxKjG5/embedded-agency-via-abstraction)) to illustrate what we're talking about:
* We have a gas consisting of some huge number of particles. We throw away information about the particles themselves, ... |
fc3dc9ca-2b09-46f0-95bb-923c63e25d32 | trentmkelly/LessWrong-43k | LessWrong | On Kelly and altruism
One-sentence summary: Kelly is not about optimizing a utility function; in general I recommend you either stop pretending you have one of those, or stop talking about Kelly.
There was a twitter thread that triggered some confusion amongst myself and some other people in a group chat I'm in.1
Th... |
b3c0575e-095e-4ee0-a100-05772892fe46 | trentmkelly/LessWrong-43k | LessWrong | Towards a better circuit prior: Improving on ELK state-of-the-art
Thanks to Paul Christiano for useful comments and feedback.
The basic circuit prior setup
We’ll start with the basic setup that we’re trying to improve upon, which is trying to solve ELK via the use of a Boolean circuit size prior. Previously, Evan su... |
d1341bc2-2bdd-479f-9ac9-3d2f946001ef | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Goals and short descriptions
Outline
-------
I develop some contents—previously introduced in the Value Learning sequence by Rohin Shah—more formally, to clarify the distinction between agents with and without a goal. Then I present related work and make some considerations on the relation between safety and goal-dir... |
aa1bd4ac-7a78-47d5-8e4a-90a7847dee0f | trentmkelly/LessWrong-43k | LessWrong | Research Log, RLLMv3 (GPT2-XL, Phi-1.5 and Falcon-RW-1B)
This lengthy log detailing near-zero temperature results has evolved into a supplementary post, rather than serving as the primary research log as initially intended. I recommend reading that first, and if you're curious about how RLLM operates in Phi-1.5 or Fal... |
bca5d7f0-72f4-4bf8-ae18-073ba9311936 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Superintelligent AI is necessary for an amazing future, but far from sufficient
(*Note: Rob Bensinger stitched together and expanded this essay based on an earlier, shorter draft plus some conversations we had. Many of the key conceptual divisions here, like "strong utopia" vs. "weak utopia" etc., are due to him.)*
... |
19ae25df-ca9b-4270-9f9e-c8cc4856dc7c | trentmkelly/LessWrong-43k | LessWrong | Overconfidence is Deceit
Author's note: This essay was written as part of an effort to say more of the simple and straightforward things loudly and clearly, and to actually lay out arguments even for concepts which feel quite intuitive to a lot of people, for the sake of those who don't "get it" at first glance. If y... |
8ddee4db-cf13-4b65-9036-8d5e581ca558 | trentmkelly/LessWrong-43k | LessWrong | Political sycophancy as a model organism of scheming
This post is a short empirical research note about training away scheming behavior (which we’ll define as the propensity to take long-term power-seeking actions when the AI thinks it can get away with it).
We study two broad categories of ways to attempt to train a... |
4fbd3ddb-555b-4e60-8fca-edd0a8e45dcb | trentmkelly/LessWrong-43k | LessWrong | Anthropics: A Short Note on the Fission Riddle
In the I-Less Eye, R Wallace considers a situation where an entity clones itself a hundred times that leads to a surprising paradox. I'll argue that there's a rather simple flaw with the argument made in the linked article, but first I'll summarise the riddle.
Suppose yo... |
7fd6027a-fdd1-461b-8aa3-94648427c4fe | trentmkelly/LessWrong-43k | LessWrong | The Geography of the Room — impact of position, orientation, and visibility on how often extraverts start conversations
I write long-form articles about social interactions based on ~6 years of observational data. This particular piece is part of a longer sequence exploring why some people are frequently approached by... |
ca53df11-83ca-4ed8-a9e2-2ec8135a6aa1 | awestover/filtering-for-misalignment | Redwood Research: Alek's Filtering Results | id: post3845
Note: This describes an idea of Jessica Taylor’s. In order to better understand how machine learning systems might avoid catastrophic behavior, we are interested in modeling this as an adversarial learning problem. A major challenge with efficiently learning to avoid catastrophes is that the utility (or l... |
555c9f10-08c9-43e2-bd94-e4450f75a2ce | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Takeaways from our robust injury classifier project [Redwood Research]
With the benefit of hindsight, we have a better sense of our takeaways from our first adversarial training project ([paper](https://arxiv.org/abs/2205.01663)). Our [original aim](https://www.alignmentforum.org/posts/k7oxdbNaGATZbtEg3/redwood-resear... |
a0436af5-bddd-467b-b73a-5c6dbb5f9eeb | StampyAI/alignment-research-dataset/eaforum | Effective Altruism Forum | Linkpost: Dwarkesh Patel interviewing Carl Shulman
[Updated 26 June: [Part 2](https://twitter.com/dwarkesh_sp/status/1673355293890347010?t=3e31qWQtN65CQ7dreRTHNg&s=19)]
> We ended up talking for 8 hours, so I'm splitting this episode into 2 parts.
>
> This part is about Carl’s model of an intelligence explosion, wh... |
0c060989-9abc-4b54-8878-bee70fb42801 | trentmkelly/LessWrong-43k | LessWrong | UDT shows that decision theory is more puzzling than ever
I feel like MIRI perhaps mispositioned FDT (their variant of UDT) as a clear advancement in decision theory, whereas maybe they could have attracted more attention/interest from academic philosophy if the framing was instead that the UDT line of thinking shows ... |
37626399-d131-4683-bcd3-727c1b016ae9 | trentmkelly/LessWrong-43k | LessWrong | Advertising as a Parasite on the Semantic Commons
‘Tis the season to be bombarded with adverts for useless crap you don’t need or want. But it's not Christmas you say? Correct: it is always the season to be inundated with adverts. Continuing the tradition of ad-bashing on LessWrong, I want to pick up on a particular '... |
83107ad3-73d7-4c11-86b3-677cbc1713a3 | trentmkelly/LessWrong-43k | LessWrong | Friendship is Optimal: A My Little Pony fanfic about an optimization process
[EDIT, Nov 14th: And it's posted. New discussion about release. Link to Friendship is Optimal.]
[EDIT, Nov 13th: I've submitted to FIMFiction, and will update with a link to its permanent home if it passes moderation. I have also removed the... |
23d65c70-8bf9-4da8-bbf9-dad3aa6c76db | trentmkelly/LessWrong-43k | LessWrong | AE Studio @ SXSW: We need more AI consciousness research (and further resources)
Quick update from AE Studio: last week, Judd (AE’s CEO) hosted a panel at SXSW with Anil Seth, Allison Duettmann, and Michael Graziano, entitled “The Path to Conscious AI” (discussion summary here[1]).
We’re also making available an une... |
624f4f65-034c-4ae0-995c-0fe2825c8a92 | StampyAI/alignment-research-dataset/lesswrong | LessWrong | Mati's introduction to pausing giant AI experiments
*Pause Giant AI Experiments: An Open Letter* – sign it here: <https://futureoflife.org/open-letter/pause-giant-ai-experiments/>
*Pausing AI Developments Isn't Enough. We Need to Shut it All Down* – read here: <https://time.com/6266923/ai-eliezer-yudkowsky-open-lett... |
1a6ce3a3-02d2-4c62-ad0b-a4eb7a6c5024 | trentmkelly/LessWrong-43k | LessWrong | We Can Build Compassionate AI
Compassion is, roughly speaking, caring for others and wanting the best for them.
Claim: We can build AI that are compassionate.
The above definition is insufficiently precise to construct an objective function for an RL training run that won't Goodhart, but it's good enough to argue th... |
7ebfa1f0-c28e-4e4d-91c6-fbc6cfa456bc | trentmkelly/LessWrong-43k | LessWrong | Curse of knowledge and Naive realism: Bias in Evaluating AGI X-Risks
Curse of knowledge
When better-informed people find it extremely difficult to think about problems from the perspective of lesser-informed people.
Unfortunately, particularly among prominent academic people, with high prestige, along with a natural ... |
d9f2e139-a52c-4efd-bca1-da0dd1431667 | trentmkelly/LessWrong-43k | LessWrong | Low hanging fruit: analyzing your nutrition
Recently I decided to try an intermittent fasting diet. To do so, I had to figure out how much I could eat on my off/down days. I realized I didn't have a very good idea about how much calories my meals have, and as I was thinking about it, I started to get curious about my ... |
e79e87d2-0db7-4a0f-9836-0d4b27244c3b | trentmkelly/LessWrong-43k | LessWrong | How complex are myopic imitators?
This post was written under Evan Hubinger’s direct guidance and mentorship, as a part of the Stanford Existential Risks Institute ML Alignment Theory Scholars (MATS) program.
TL;DR: Time-limited myopic agents might be desirable for their safety properties; one such case is that of "m... |
ceb24b9d-0a52-44a5-9d47-489652a279c6 | trentmkelly/LessWrong-43k | LessWrong | What's important in "AI for epistemics"?
Summary
This post gives my personal take on “AI for epistemics” and how important it might be to work on.
Some background context:
* AI capabilities are advancing rapidly and I think it’s important to think ahead and prepare for the possible development of AI that could auto... |
c0512021-97aa-4c0b-b8fb-2c1e6a07385f | trentmkelly/LessWrong-43k | LessWrong | Learning programming: so I've learned the basics of Python, what next?
Lots of people (particularly people associated with LessWrong) are telling me I should become a computer programmer; in response I've taught myself a little Python using this site, written a couple Python scripts on my own, and just now sent in an ... |
cf196867-0a3b-4687-b045-1e18c09b5fec | trentmkelly/LessWrong-43k | LessWrong | Petition To Make Inarticulate Downvoting More Difficult
I feel a strong affinity with this community, but that's hard to remember when my own locally-weird ideas get downvoted and/or reverted to draft without giving me a clear route to explain myself.
Even just splitting the Karma mechanism into "doesn't seem benevol... |
7748a10f-5975-4a2c-8cbe-83bd6bd1e94d | trentmkelly/LessWrong-43k | LessWrong | John Danaher on 'The Superintelligent Will'
Philosopher John Danaher has written an explication and critique of Bostrom's "orthogonality thesis" from "The Superintelligent Will." To quote the conclusion:
>
>
> Summing up, in this post I’ve considered Bostrom’s discussion of the orthogonality thesis. According to t... |
ae2a098f-9c74-469e-92fa-e07efb3c37c1 | trentmkelly/LessWrong-43k | LessWrong | Adversarial training, importance sampling, and anti-adversarial training for AI whistleblowing
(Thanks to Ajeya Cotra and Ryan Greenblatt for comments.)
I’m pretty interested in adversarial training as an ingredient in alignment schemes. The basic setup for adversarial training is that instead of your AI being traine... |
c54f26d9-82c1-4f0a-a3c8-0d111f0f9037 | trentmkelly/LessWrong-43k | LessWrong | Blind artifacts
This is the second of four short essays that say explicitly some things that I would tell an intrigued proto-rationalist before pointing them towards Rationality: AI to Zombies (and, by extension, most of LessWrong). For most people here, these essays will be very old news, as they talk about the insig... |
16181cf7-e38f-44c9-af33-63e0aa8c1ff0 | StampyAI/alignment-research-dataset/lesswrong | LessWrong | The Problem of the Criterion
I keep finding cause to discuss the problem of the criterion, so I figured I'd try my hand at writing up a post explaining it. I don't have a great track record on writing clear explanations, but I'll do my best and include lots of links you can follow to make up for any inadequacy on my p... |
34be4dbf-09cd-4466-b66d-0b25755e5e29 | StampyAI/alignment-research-dataset/youtube | Youtube Transcripts | Daniel Filan - Peering into neural nets for AI safety
hello and welcome back everybody to the
tourist data science podcast today we
are talking to daniel fallen
who is an ai alignment researcher at
berkeley and the host of axerp an ai
limit podcast that you should definitely
check out right after this episode
we're go... |
23a01e7c-f9ec-4c4e-8351-d0c113ab5040 | trentmkelly/LessWrong-43k | LessWrong | The Army of Jakoths (a parable)
The Army of Jakoths is riding
They are riding to take your life, and mine
And everything and everyone you ever loved
The Army of Jakoths is a mile from your door
And the only thing your neighbor says is:
«I'm just as likely to be struck by a bolt of lightning
As by the sword of a ... |
3986fd62-a450-4234-80a6-43d7fbe36ab6 | trentmkelly/LessWrong-43k | LessWrong | Meetup : West LA Meetup - Defense Against the Dark Arts, Part I
Discussion article for the meetup : West LA Meetup - Defense Against the Dark Arts, Part I
WHEN: 02 January 2013 07:00:00PM (-0800)
WHERE: 10850 West Pico Blvd, Los Angeles, CA 90064
When: 7:00pm Wednesday, January 2nd.
Where: The Westside Tavern in t... |
2bddfe26-cf9d-4907-a3ad-33a3fbc09dfb | trentmkelly/LessWrong-43k | LessWrong | My Research Process: Understanding and Cultivating Research Taste
This is post 3 of a sequence on my framework for doing and thinking about research. Start here. Thanks to my co-author Gemini 2.5 Pro
Introduction
Spend enough time around researchers, and you'll hear talk of "research taste." It's often presented as ... |
770ea4a1-6bf1-4a5f-8c78-7b733771f1b4 | trentmkelly/LessWrong-43k | LessWrong | On dollars, utility, and crack cocaine
The lottery came up in a recent comment, with the claim that the expected return is negative - and the implicit conclusion that it's irrational to play the lottery. So I will explain why this is not the case.
It's convenient to reason using units of equivalent value. Dollars, ... |
9ea859fa-34b5-4820-a36e-557559cb34b5 | trentmkelly/LessWrong-43k | LessWrong | Call for submissions: “(In)human Values and Artificial Agency”, ALIFE 2023
key points:
* Cash prize of $500 for the best presentation.
* Deadline 3 March, 2023.
* Organized by Simon McGregor (University of Sussex), Rory Greig (DeepMind), Chris Buckley (University of Sussex)
> ALIFE 2023 (the 2023 conference on Ar... |
53f556b6-081e-46d3-a0fb-75a0eff3b81a | trentmkelly/LessWrong-43k | LessWrong | Meetup : Austin, TX
Discussion article for the meetup : Austin, TX
WHEN: 15 October 2011 01:30:00PM (-0500)
WHERE: 2222B Guadalupe St Austin, Texas 78705
The Austin meetup careens towards the present from the future, and all you can do is prepare to be there. (Or be square.)
I have finally gotten into the habit of... |
fef7b48a-300a-4762-80dc-7d2eb1f3cd30 | trentmkelly/LessWrong-43k | LessWrong | Preference Aggregation as Bayesian Inference
Crossposted from my personal blog.
A fundamental problem in AI alignment, as well as in many social sciences is the problem of preference aggregation. Given a number of different actors who have specific preferences, what is a consistent way of making decisions that ensur... |
317e9e7d-3803-4193-ab67-39576213fc7a | trentmkelly/LessWrong-43k | LessWrong | Why hasn't there been research on the effectiveness of zinc for Covid-19?
There was significant interest in zinc early on. It seemed like one of the most promising avenues for protecting oneself against Covid-19. Searching on both Google and Lesswrong itself, I do not see new research on it. (This is the closest thing... |
3a9d8388-7387-45f1-9a7d-0da5bfcd7799 | trentmkelly/LessWrong-43k | LessWrong | Out in the Great Northwest
One of the songs my family would sing growing up was "Out in the Great Northwest" ( mp3, 2011 recording). For me, the point of the song was the verse:
> A Scotsman went out there to live, he called his house a "hoos".
> They showed him a great big animal, they said it was a moose.
> Out in ... |
8ed2ab20-041e-4cf1-9d18-ff6917107780 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Where are intentions to be found?
*This is independent research. To make it possible for me to continue writing posts like this, please consider* [*supporting me*](https://www.alexflint.io/donate.html)*.*
---
As we build powerful AI systems, we want to ensure that they are broadly beneficial. Pinning down exactl... |
1be5d54b-a673-47c3-aec1-b100438d4764 | trentmkelly/LessWrong-43k | LessWrong | Why books don't work
A look at how books are not optimal for conveying information. Analogies to lectures, with an interesting take on cognitive models, i.e. the assumptions you make about how learning happens. Also some interesting citations on average reading time and attention span.
I really like this piece becaus... |
3c64c9fa-35b0-4e7f-b8b2-25497280378c | trentmkelly/LessWrong-43k | LessWrong | Get your gun license
Should you apply for a gun license? Most people in the rationalist community think they should not. After all, it seems that America should have more gun control. However, the correct policy for the country and the correct policy for you can be different things. I’m not going to talk about the cor... |
893f0253-2850-49fc-9a8d-173475df390b | trentmkelly/LessWrong-43k | LessWrong | Guardians of Ayn Rand
> "For skeptics, the idea that reason can lead to a cult is absurd. The characteristics of a cult are 180 degrees out of phase with reason. But as I will demonstrate, not only can it happen, it has happened, and to a group that would have to be considered the unlikeliest cult in history. It is... |
9434c0ca-52ef-486c-a236-e9580cfa99ab | trentmkelly/LessWrong-43k | LessWrong | A cognitive intervention for wrist pain
Intention
(Added 2019-03-19.)
The web contains much information about wrist pain, RSI, carpal tunnel syndrome etc. Most of it suggests that it comes from repetitive small movements, such as typing. It warns about dire consequences, and recommends improving workplace ergonomics,... |
d4862216-caa7-4f90-813e-9a1cbbfb8c94 | LDJnr/LessWrong-Amplify-Instruct | LessWrong | "I thought this video was a really good question dissolving by Richard Feynman. But it's in 240p! Nobody likes watching 240p videos. So I transcribed it. (Edit: That was in jest. The real reasons are because I thought I could get more exposure this way, and because a lot of people appreciate transcripts. Also, Paul Gra... |
0f7c3803-9cda-4978-a683-8baf6d93180e | StampyAI/alignment-research-dataset/eaforum | Effective Altruism Forum | Do EA folks want AGI at all?
I wonder if EA folks, overall, consider AGI a positive but they want it aligned as well?
Would the EA community prefer that AGI were never developed? |
f2cc142a-f175-428b-9a5f-66e6ffc6296b | StampyAI/alignment-research-dataset/aisafety.info | AI Safety Info | When do experts think human-level AI will be created?
<iframe src="https://www.youtube.com/embed/3K25VPdbAjU" title="Everything might change forever this century (or we’ll go extinct)" frameborder="0" allow="accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture; web-share" allowfulls... |
5c492754-a49f-4222-a488-db1a625d7afa | trentmkelly/LessWrong-43k | LessWrong | MIRI course list study pairs
Inspired by: On learning difficult things
In his recent post, user So8res says his number one piece of advice for learning something difficult is to have study partner to learn with you.
Since there is a decent amount of interest here in going through the MIRI course list, it might be w... |
d78fc360-fa7e-4984-99dc-616a49d1ce1d | trentmkelly/LessWrong-43k | LessWrong | Control Symmetry: why we might want to start investigating asymmetric alignment interventions
[This is a post summarizing the motivation for an AISC 2024 project: If you are interested in participating you can apply here: https://aisafety.camp/ (project 25: Asymmetric control in LLMs: model editing and steering that ... |
e30ae5a3-b3e6-4dab-a522-de8975cd6f70 | trentmkelly/LessWrong-43k | LessWrong | Newsom Vetoes SB 1047
It’s over, until such a future time as either we are so back, or it is over for humanity.
Gavin Newsom has vetoed SB 1047.
NEWSOM’S MESSAGE IN FULL
Quoted text is him, comments are mine.
> To the Members of the California State Senate: I am returning Senate Bill 1047 without my signature.
>
... |
bc56220e-8705-47ae-90f6-a5ff9ec35aca | StampyAI/alignment-research-dataset/blogs | Blogs | New paper: “Safely interruptible agents”
[](https://intelligence.org/files/Interruptibility.pdf)Google DeepMind Research Scientist Laurent Orseau and MIRI Research Associate Stuart Armstrong have written a new paper on error-tolerant ag... |
0aa6fac2-c491-4516-998b-2493cdc62f53 | trentmkelly/LessWrong-43k | LessWrong | Superintelligence 25: Components list for acquiring values
This is part of a weekly reading group on Nick Bostrom's book, Superintelligence. For more information about the group, and an index of posts so far see the announcement post. For the schedule of future topics, see MIRI's reading guide.
----------------------... |
e54b58df-be44-498b-8c5e-569595dbc2d4 | trentmkelly/LessWrong-43k | LessWrong | DeepSeek: Lemon, It’s Wednesday
It’s been another *checks notes* two days, so it’s time for all the latest DeepSeek news.
You can also see my previous coverage of the r1 model and, from Monday various reactions including the Panic at the App Store.
TABLE OF CONTENTS
1. First, Reiterating About Calming Down About ... |
f54c4039-c421-4136-9acf-4edb758d4d62 | StampyAI/alignment-research-dataset/blogs | Blogs | Cold Links: nonfiction yarns
These are longreads that I recommend because they're just good yarns. There's usually not a broader lesson. I don't read tons of these so you probably have better ones ([send them in](https://www.guidedtrack.com/programs/4kal2ue/run?posttitle=Cold%20Links%3A%20nonfiction%20yarns)!) but the... |
d8bac1d8-7f64-4e2c-bdba-5e8ea7d3981b | trentmkelly/LessWrong-43k | LessWrong | Notes on Meta's Diplomacy-Playing AI
Disclaimer: I'm not an expert at machine learning, AI safety or Diplomacy, so there may be errors here, though hopefully no major ones. For previous discussion of CICERO on here, see the comments in this post, this rundown and this commentary.
Summary
* CICERO is a new AI develo... |
02d9eed2-0698-40c6-84da-a314b2213b86 | trentmkelly/LessWrong-43k | LessWrong | The Cold War divided Science
What can we learn about science from the divide during the Cold War?
I have one example in mind: America held that coal and oil were fossil fuels, the stored energy of the sun, while the Soviets held that they were the result of geologic forces applied to primordial methane.
At least one... |
9ed1d7f3-f378-4fb6-8937-7bb2d2b0d60e | trentmkelly/LessWrong-43k | LessWrong | The spiritual benefits of material progress
The Industrial Revolution gave us abundance and comfort—but what did it do to our souls?
Recently my progress colleague Alec Stapp responded to a Twitter thread disparaging the Industrial Revolution for what it “did to humanity.” Alec’s response was basically that abundance... |
c80a0107-097c-4625-9daa-53b0ae352be2 | trentmkelly/LessWrong-43k | LessWrong | Decision Theories: A Semi-Formal Analysis, Part II
Or: Causal Decision Theory and Substitution
Previously:
0. Decision Theories: A Less Wrong Primer
1. The Problem with Naive Decision Theory
Summary of Post: We explore the role of substitution in avoiding spurious counterfactuals, introduce an implementation of Cau... |
e1e6db6c-1a23-4555-bab5-3bcc0ff8e259 | StampyAI/alignment-research-dataset/blogs | Blogs | February 2016 Newsletter
| |
| --- |
|
**Research updates**
* New at IAFF: [Thoughts on Logical Dutch Book Arguments](https://agentfoundations.org/item?id=582); [Another View of Quantilizers: Avoiding Goodhart’s Law](https://agentfoundations.org/item?id=596); [Another Concise Open Problem](https://agentfoundations.... |
a6a61993-ba92-45d5-abe5-6fedc5f3dae2 | LDJnr/LessWrong-Amplify-Instruct | LessWrong | "Whenever someone exhorts you to "think outside the box", they usually, for your convenience, point out exactly where "outside the box" is located. Isn't it funny how nonconformists all dress the same...
In Artificial Intelligence, everyone outside the field has a cached result for brilliant new revolutionary AI idea—n... |
6791db28-6240-4d54-968a-14ff9c8441c0 | trentmkelly/LessWrong-43k | LessWrong | Community Page Mini-Guide
Here's how to make a meetup event and/or group on the community page.
1. Login to your (or create a) LessWrong account.
2. Navigate to the community page via the top-left hamburger menu or the community tab on the frontpage.
3. In the centre (underneath the 'local groups' header), click th... |
cb78d596-d6e4-4be5-ae00-febd893fa618 | trentmkelly/LessWrong-43k | LessWrong | What does an imperfect agent want?
A putative new idea for AI control; index here.
I'll roughly divide ways of establishing human preferences into four categories:
#. Assume true #. Best fit #. Proxy measures #. Modelled irrationality
----------------------------------------
In some cases, these are more differenc... |
6d0fc187-a6c0-45bb-a0ff-a0496d0408c5 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Counterfactual Mugging Poker Game
Consider the following game:
Player A receives a card at random that is either High or Low. He may reveal his card if he wishes.
Player B then chooses a probability .mjx-chtml {display: inline-block; line-height: 0; text-indent: 0; text-align: left; text-transform: none; font-style:... |
335a5fce-a2de-4e89-aab0-c25f361d4d01 | trentmkelly/LessWrong-43k | LessWrong | Can't downvote
Downvote buttons seem to have just stopped working for me, saying I need at least 1 karma. I'm posting in case it leads someone to fix the problem. |
d59d45b6-b176-4e6d-bbda-3042579528dc | awestover/filtering-for-misalignment | Redwood Research: Alek's Filtering Results | id: post3491
Summary: I propose that counterfactual corrigibility can be slightly adjusted to address known problems with bad bets and defending against backstops. This adjusted version has learning-esque behaviour and hence it can have problems with overconfidence about the button state. A new proposal: lower-bounded... |
cdf67aa7-73a2-4a00-a5ee-b9ecf84f7312 | StampyAI/alignment-research-dataset/arxiv | Arxiv | Rigorous Agent Evaluation: An Adversarial Approach to Uncover Catastrophic Failures
1 Introduction
---------------
How can we ensure machine learning systems do not make catastrophic mistakes?
While machine learning systems have shown impressive results across a variety of domains (Krizhevsky et al., [2012](#bib.bi... |
0e354f82-f01e-43c1-b7cc-38c97438da5f | trentmkelly/LessWrong-43k | LessWrong | Meetup : Applied Rationality Talks: Thinking in Bayes
Discussion article for the meetup : Applied Rationality Talks: Thinking in Bayes
WHEN: 26 September 2013 07:00:00PM (-0400)
WHERE: Royal Oak 161 Laurier Avenue East, Ottawa, ON
This is the second in a series of talks on CFAR and Less Wrong rationality topics off... |
19c33b95-bc2c-4e44-837f-e4bb61bc985d | trentmkelly/LessWrong-43k | LessWrong | A Robust Natural Latent Over A Mixed Distribution Is Natural Over The Distributions Which Were Mixed
This post walks through the math for a theorem. It’s intended to be a reference post, which we’ll link back to as-needed from future posts. The question which first motivated this theorem for us was: “Redness of a mark... |
81bbbed0-f089-4729-9e4f-2a21cb1c5a69 | trentmkelly/LessWrong-43k | LessWrong | Finding Backward Chaining Circuits in Transformers Trained on Tree Search
This post is a summary of our paper A Mechanistic Analysis of a Transformer Trained on a Symbolic Multi-Step Reasoning Task (ACL 2024). While we wrote and released the paper a couple of months ago, we have done a bad job promoting it so far. As ... |
5aa8bcf1-2e6b-4100-92c6-c980e25c19f9 | trentmkelly/LessWrong-43k | LessWrong | Is "red" for GPT-4 the same as "red" for you?
Penned by Yusuke Hayashi, an independent researcher hailing from Japan, this article bears no affiliation to the authors whose scholarly works are referenced herein. Demonstrating intellectual autonomy, the analysis presented is unequivocally distinct from the cited public... |
62e6870e-5e01-45fb-819f-1d54807208a5 | trentmkelly/LessWrong-43k | LessWrong | "Manna" by Marshall Brain
Oldie but goodie. A piece of fiction describing how a computer system can do the job of human managers at fast food restaurants (scarily plausible), how this leads to a dystopia (slowly getting implausible), and how to avoid this scenario and reach utopia (give me a break). |
b3a265e9-55f4-46ce-adc0-0d5085feac76 | trentmkelly/LessWrong-43k | LessWrong | Muehlhauser-Goertzel Dialogue, Part 1
Part of the Muehlhauser interview series on AGI.
Luke Muehlhauser is Executive Director of the Singularity Institute, a non-profit research institute studying AGI safety.
Ben Goertzel is the Chairman at the AGI company Novamente, and founder of the AGI conference series.
... |
97739313-2927-481f-ae66-082ce197e632 | trentmkelly/LessWrong-43k | LessWrong | A Fervent Defense of Frequentist Statistics
[Highlights for the busy: de-bunking standard "Bayes is optimal" arguments; frequentist Solomonoff induction; and a description of the online learning framework. Note: cross-posted from my blog.]
Short summary. This essay makes many points, each of which I think is worth re... |
066636d8-b140-443d-a2c7-5a128ea071d5 | trentmkelly/LessWrong-43k | LessWrong | Alignment Newsletter #24
Starting from this week, Richard Ngo will join me in writing summaries. His summaries are marked as such; I'm reviewing some of them now but expect to review less over time.
Highlights
Introducing the Unrestricted Adversarial Examples Challenge (Tom B. Brown et al): There's a new adversarial... |
1afb6a38-7e41-47aa-9800-46b9e4e0d26c | trentmkelly/LessWrong-43k | LessWrong | Rationality Quotes August 2011
Here's the new quotes thread, for all those quotes you were going to post.
Rules:
* Please post all quotes separately, so that they can be voted up/down separately. (If they are strongly related, reply to your own comments. If strongly ordered, then go ahead and post them together.)... |
03c4af9c-aa7f-45e8-89b7-7134b2513b95 | StampyAI/alignment-research-dataset/blogs | Blogs | Hardware overhang
*Hardware overhang* refers to a situation where large quantities of computing hardware can be diverted to running powerful AI systems as soon as the software is developed.
Details
-------
### Definition
In the context of AI forecasting, *hardware overhang* refers to a situation where enough com... |
677e8e9f-5094-4eb3-ac79-1d0beb940852 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Extortion and trade negotiations
To illustrate an old point - that it's hard to [distinguish between extortion and trade negotiations](/lw/hz9/countess_and_baron_attempt_to_define_blackmail/) - here's a schematic diagram of extortion, alternating actions by player B (blackmailer/extorter//blue) and V (victim//violet):... |
1327ffb6-38cb-4463-949e-28b18abb0fb1 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Alignment Newsletter #48
Find all Alignment Newsletter resources [here](http://rohinshah.com/alignment-newsletter/). In particular, you can [sign up](http://eepurl.com/dqMSZj), or look through this [spreadsheet](https://docs.google.com/spreadsheets/d/1PwWbWZ6FPqAgZWOoOcXM8N_tUCuxpEyMbN1NYYC02aM/edit?usp=sharing) of al... |
979ff42d-7240-46e8-bf98-0494aaa2b611 | trentmkelly/LessWrong-43k | LessWrong | Humans don't generally have utility functions
First, the background:
Humans obviously have a variety of heuristics and biases that lead to non-optimal behavior. But can this behavior truly not be described by a function?
Well, the easiest way to show that utility isn't described by a function is to show the existen... |
9846e2e5-da1d-4d7e-b5d0-d0dfb7b64df8 | trentmkelly/LessWrong-43k | LessWrong | Exploring OpenAI's Latent Directions: Tests, Observations, and Poking Around
TL;DR: Interactive exploration of new directions in GPT2-SMALL. Try it yourself.
OpenAI recently released their Sparse Autoencoder for GPT2-Small. In this story-driven post, I run experiments and poke around the 325k active directions to see... |
a1ec17f3-7b2a-4fd0-ab30-126c926cdbc6 | trentmkelly/LessWrong-43k | LessWrong | ASPR & WARP: Rationality Camps for Teens in Taiwan and Oxford
The FABRIC team (formerly known as ESPR) is running two immersive applied rationality workshops for analytical students aged 16 to 19 seeking to understand themselves and the world.
Asian Spring Programme on Rationality (ASPR)
* Where: Taipei, Taiwan
* ... |
b40b0b6b-9302-4fe7-8f80-e92e0c3b0713 | trentmkelly/LessWrong-43k | LessWrong | A Kindness, or The Inevitable Consequence of Perfect Inference (a short story)
This open letter takes the form of annotations to Spearfield Teaching Hospital’s morbidity and mortality timeline for the incident on November 2, now widely known following the leak of confidential documents by some of my former colleagues.... |
a5ddd596-c738-49be-918a-51bc48c58f43 | trentmkelly/LessWrong-43k | LessWrong | Terminally ill teen won historic ruling to preserve body
|
372fe715-eceb-4223-b9d1-24d3b3940f46 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Knowledge is not just precipitation of action
Knowledge is not just precipitation of action
=============================================
*Financial status: This is independent research. I welcome* [*financial support*](https://www.alexflint.io/donate.html) *to make further posts like this possible.*
*Epistemic st... |
8b9bf21a-8e3e-4f5b-bc2e-b256e552c0e3 | trentmkelly/LessWrong-43k | LessWrong | AI CoT Reasoning Is Often Unfaithful
A new Anthropic paper reports that reasoning model chain of thought (CoT) is often unfaithful. They test on Claude Sonnet 3.7 and r1, I’d love to see someone try this on o3 as well.
Note that this does not have to be, and usually isn’t, something sinister.
It is simply that, as t... |
28f55839-6c4b-46c9-a6ee-a143fc6a58b1 | trentmkelly/LessWrong-43k | LessWrong | Beware The Believer, or a study in depth of recursion
I submit to the good people of Less Wrong this wonderful video. Is it a parody of scientifism? A parody of creationist parodies of scientifism? How deep does the recursion go? (those who already know, don't spoil the fun for the rest!). |
8b8a7c9d-14cb-4cc5-a80a-4f4fe35e917b | trentmkelly/LessWrong-43k | LessWrong | What does "meta-execution without indirection" look like?
I've been trying to understand IDA/Factored Evaluation at a deep level, and I find meta-execution especially confusing. The LW post says that it is "HCH + annotated functional programming + a level of indirection", but I'm not sure what the "level of indirectio... |
452310c6-c342-4d69-93c1-3f605228c33d | trentmkelly/LessWrong-43k | LessWrong | It's not like anything to be a bat
...at least not if you accept a certain line of anthropic argument.
Thomas Nagel famously challenged the philosophical world to come to terms with qualia in his essay "What is it Like to Be a Bat?". Bats, with sensory systems so completely different from those of humans, must have e... |
47a43212-5347-4d34-a72b-1ec4ec9e5a09 | trentmkelly/LessWrong-43k | LessWrong | An alarming fact about the anti-aging community
Past and Present
Ten years ago teenager me was hopeful. And stupid.
The world neglected aging as a disease, Aubrey had barely started spreading memes, to the point it was worth it for him to let me work remotely to help with Metuselah foundation. They had not even recei... |
38313556-8bc7-4f31-847d-e95da9cdb397 | trentmkelly/LessWrong-43k | LessWrong | Meetup : Sydney Meetup: January
Discussion article for the meetup : Sydney Meetup: January
WHEN: 22 January 2014 06:30:00PM (+1100)
WHERE: Caffe Tiamo, 374 Pitt St, Sydney
Full details TBD. We're going to try to have some discussion, a rationality exercise and a game. Details of which we're still nutting out on the... |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.