id stringlengths 36 36 | source stringclasses 15
values | formatted_source stringclasses 13
values | text stringlengths 2 7.55M |
|---|---|---|---|
d2f40a7e-c7e2-4048-8009-4edc8052cf76 | trentmkelly/LessWrong-43k | LessWrong | Paper club: He et al. on modular arithmetic (part I)
In this post we’ll be looking at the recent paper “Learning to grok: Emergence of in-context learning and skill composition in modular arithmetic tasks” by He et al. This post is partially a sequel to my earlier post on grammars and subgrammars, though it can be rea... |
c5394d87-0abe-4fbc-a14f-35c592e26272 | trentmkelly/LessWrong-43k | LessWrong | Politics Discussion Thread February 2013
1. Top-level comments should introduce arguments; responses should be responses to those arguments.
2. Upvote and downvote based on whether or not you find an argument convincing in the context in which it was raised. This means if it's a good argument against the argume... |
43b9c47f-9141-45d5-b004-a158ccd84131 | trentmkelly/LessWrong-43k | LessWrong | The optimizer won’t just guess your intended semantics
A desirable property of an AI’s world model is that you as its programmer have an idea what’s going on inside. It would be good if you could point to a part of the world model and say, “This here encodes the concept of a strawberry; here is how this is linked with... |
68b5ca0d-5afd-48f5-9da3-a033fbcb87a6 | StampyAI/alignment-research-dataset/eaforum | Effective Altruism Forum | Chaining the evil genie: why "outer" AI safety is probably easy
When discussing the threat of AGI, especially in introductory resources, a very common argument is the “evil genie” or “king Midas” problem. The idea is that if you put a goal function into a machine, such as “maximize paperclips” or “find a cure for canc... |
49104d96-354f-49a1-b6d4-2f9984fa3dee | StampyAI/alignment-research-dataset/aisafety.info | AI Safety Info | How does DeepMind do adversarial training?
DeepMind explained their approach to [adversarial training](/?state=935A&question=What%20is%20adversarial%20training%3F)[^kix.qjxep5dgpfnw] in the paper [“Red Teaming Language Models with Language Models”](https://www.deepmind.com/blog/red-teaming-language-models-with-languag... |
17f528b8-3334-4197-a1c4-12f6ac74b477 | StampyAI/alignment-research-dataset/blogs | Blogs | Harry Buhrman on quantum algorithms and cryptography
 [Harry Buhrman](http://www.cwi.nl/people/552) is head of the research group ‘[Algorithms and Complexity](http://www.cwi.nl/research-groups/algorithms-and-complexity)’ at t... |
fe0258a4-df0a-4762-98aa-d10f60b60a46 | StampyAI/alignment-research-dataset/blogs | Blogs | Did life get better during the pre-industrial era? (Ehhhh)
*Click lower right to download or find on Apple Podcasts, Spotify, Stitcher, etc.*
> This is the final piece in the [Has Life Gotten Better?](https://www.cold-takes.com/has-life-gotten-better/) series.
For the last 200 years or so, [life has been get... |
a2fe131c-8355-48e2-ac65-8e7a96904256 | trentmkelly/LessWrong-43k | LessWrong | How does long-term use of caffeine affect productivity?
I am trying to figure out whether caffeine helps productivity in the long run. Looking back 10 years from now, how much more/less productive will I have been if I were to drink coffee every day, or every second day?
Reviewing what has been written on the topic b... |
3be34e3c-a139-49e0-8cae-f67a29b71c2d | StampyAI/alignment-research-dataset/eaforum | Effective Altruism Forum | The Slippery Slope from DALLE-2 to Deepfake Anarchy
OpenAI developed DALLE-2. Then StabilityAI made an open source copycat. This is a dangerous dynamic.
====================================================================================================
Stephen Casper ([scasper@mit.edu](mailto:scasper@mit.edu))
Phi... |
38280c69-aaa3-4e0e-b12c-27eb70cd9b25 | awestover/filtering-for-misalignment | Redwood Research: Alek's Filtering Results | id: post3768
A putative new idea for AI control; index here . In simple graphical form, here is the problem of divergent human preferences: Here the AI either chooses A or ¬ A , and as a consequence, the human then chooses B or ¬ B . There are a variety of situations in which this is or isn't a problem (when A or B or... |
747c0a14-cd8c-4a20-be87-6164665f02ad | trentmkelly/LessWrong-43k | LessWrong | How do you do fit-testing for KN95/N95/FFP2/FFP3 masks?
I have brought FFP3 masks half a year ago and at the time I brought them to get outside and protect myself in enviroments with allergy triggers. It didn't work. At the time I wasn't aware that fit testing is hard as writes:
> Fit-testing is a laborious task, tak... |
c62086ba-93fa-49a4-aaf0-662b280e94b7 | trentmkelly/LessWrong-43k | LessWrong | Interpreting Embedding Spaces by
Conceptualization
> In this paper, we present a novel method of understanding embeddings by transforming embedding space into a comprehensible conceptual space. We present an algorithm for deriving a conceptual space with dynamic on-demand granularity. We also show a method for transfe... |
9d41c6ad-5a35-4896-8852-3e6e7d646385 | trentmkelly/LessWrong-43k | LessWrong | HLAI 2018 Talks
The keynote and general interest talks from HLAI 2018 are available online, along with a few others. Sadly the panel I participated in was not fully recorded, but there is a short video of an excerpt of it. If you want to watch just one video I think the most relevant one to folks here will be the pane... |
21b8cb0b-09d8-4a39-a86e-376402551c77 | trentmkelly/LessWrong-43k | LessWrong | On Authority
Stephen Randy Waldman (Interfluidity) writes about the process by which authority is created. His definition of authority applies to information in a social context. Namely, information that is authoritative is information which we take to be true by our actions, regardless of whether that information is ... |
3beb94fc-c526-4292-a751-09498c6466dd | trentmkelly/LessWrong-43k | LessWrong | [SEQ RERUN] Nonsentient Optimizers
Today's post, Nonsentient Optimizers was originally published on 27 December 2008. A summary (taken from the LW wiki):
> Discusses some of the problems of, and justification for, creating AIs that are knowably not conscious / sentient / people / citizens / subjective experiencers... |
e85066f7-460f-45f4-8e56-1cc67ae0616d | trentmkelly/LessWrong-43k | LessWrong | The Four States: action, cognitive, emotional, relational
During a discussion on how to make the rationalist community more "Hufflepuff", the idea was brought up to define what "more Hufflepuff" even means. Words like cooperation, loyalty, diligence, and empathy were written down. As the list grew longer, I realized t... |
dcf13634-1f0a-4f10-8862-420fd9387d08 | trentmkelly/LessWrong-43k | LessWrong | Another cooperative rationality exercise
I don't know how well this is going to work, but I mention it here because it's actually going to be done in a few weeks time at a day-long meeting of the research group that I work with. (Not my idea. I don't know which of us thought it up.)
> Keyword game: explaining a scien... |
e91bdd54-b71a-44a6-8948-ff0f9cfb4b07 | StampyAI/alignment-research-dataset/blogs | Blogs | List of Analyses of Time to Human-Level AI
This is a list of most of the substantial analyses of AI timelines that we know of. It also covers most of the arguments and opinions of which we are aware.
Details
-------
The list below contains substantial publically available analyses of when human-level AI will appea... |
ccea9b2b-0d96-4aaa-90b5-7b7840825eb8 | trentmkelly/LessWrong-43k | LessWrong | My AI Alignment Research Agenda and Threat Model, right now (May 2023)
TLDR
Fairly short timelines, mildly fast takeoffs, and medium-high uncertainties. --> Looking for abstractions to help cognition-steering and value-loading. --> Grasping at / reacting to related/scary/FOMO lines of research.
Threat Model
An AI sy... |
f6e38c75-b48f-49a1-a5bc-c9771e2ff798 | trentmkelly/LessWrong-43k | LessWrong | [MLSN #2]: Adversarial Training
As part of a larger community building effort, I am writing a safety newsletter which is designed to cover empirical safety research and be palatable to the broader machine learning research community. You can subscribe here or follow the newsletter on twitter here.
-------------------... |
32b2a645-73cd-4351-a33d-34758f474e56 | trentmkelly/LessWrong-43k | LessWrong | A reason behind bad systems, and moral implications of seeing this reason
(Epistemic status: Pretty sure the premises are all correct and the metaphor is accurate. Exploring the implications and conclusions.)
A programmer was given an aging piece of software to use to complete a project. To the company, using the old... |
75543b56-77fa-4b7f-8a69-0823a722e276 | trentmkelly/LessWrong-43k | LessWrong | Link: 50% effective malaria vaccine developed
A study on over 15000 children in 7 African countries shows GlaxoSmithKline's anti-malaria vaccine halves the risk of contracting malaria. The trials were run on children between 6 and 12 weeks old, and between 5 and 17 months old.
Guardian article: http://www.guardian.co... |
75a5caf3-bce7-429c-a855-5236fa6dcc5d | trentmkelly/LessWrong-43k | LessWrong | A Critique of Functional Decision Theory
A Critique of Functional Decision Theory
NB: My writing this note was prompted by Carl Shulman, who suggested we could try a low-time-commitment way of attempting to understanding the disagreement between some folks in the rationality community and academic decision theorists ... |
5eed1b80-2060-428b-9776-aa9e8539d9fc | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Are minimal circuits deceptive?
*This post is part of research I did at OpenAI with mentoring and guidance from Paul Christiano.*
One possible [inner alignment](https://arxiv.org/abs/1906.01820) technique is to structure your learning algorithm’s inductive biases such that it is far more likely for it to produce an ... |
bd5834aa-9668-4cc0-bcbe-c7ec34c2067e | trentmkelly/LessWrong-43k | LessWrong | SI visiting fellow Diego Caleiro gives a TED talk on Friendly AI
Diego Caleiro, a past SI visiting fellow and current leader of a transhumanist institution in Brazil, recently delivered this TED@SaoPaulo talk on Friendly AI and Effective Altruism. If it goes viral, it has a chance to be selected for TED Global. (Th... |
a72a945d-1bd9-42de-8520-cd4cdf2dd847 | trentmkelly/LessWrong-43k | LessWrong | Summary of the sequences / Lesson plans for rationality
Does anyone know where I can find a summary of the sequences, at any level of quality? (Has anyone made their own they're willing to share?)
Or any materials prepared to help teach the sequences or similar content to an unfamiliar audience? I am starting a rati... |
32753dd5-ec14-4c98-b1b5-6f8cc9062345 | trentmkelly/LessWrong-43k | LessWrong | How to Build a Third Place on Focusmate
Introduction
Focusmate changed my life. I started using it mid-2023 and have been a power user since 2023. Here are the high-level stats:
* I've spent 2,456 hours on Focusmate.
* I've met 1,447 people.
* I have 10+ friends with whom I've spent between 30 and 300 hours.
* I'... |
95e72244-ef45-44f2-b40c-6fecb9f48a9c | trentmkelly/LessWrong-43k | LessWrong | New Hackathon: Robustness to distribution changes and ambiguity
EffiSciences is proud to announce a new hackathon organized with challenge-data-ens in France.
The objective of this event is to address a sub-problem of alignment that is particularly actionable: namely value extrapolation in the case of distribution ch... |
2a825499-229f-4ded-bb7a-7bf0ad3d214e | trentmkelly/LessWrong-43k | LessWrong | An example of self-fulfilling spurious proofs in UDT
Benja Fallenstein was the first to point out that spurious proofs pose a problem for UDT. Vladimir Nesov and orthonormal asked for a formalization of that intuition. In this post I will give an example of a UDT-ish agent that fails due to having a malicious proof se... |
3ee164fa-29f9-48b4-8822-f79298a84ae9 | trentmkelly/LessWrong-43k | LessWrong | Multiplying 10-digit numbers using Flickr: The power of recognition memory
http://people.csail.mit.edu/andyd/rec_method.pdf
Not sure what this falls under, but it's cool.
|
6462670a-f259-4c17-9abd-48db32be024a | StampyAI/alignment-research-dataset/special_docs | Other | README-by Vael Gates-date 20220509
# Interview with AI Researchers by Vael Gates
Table of Contents
=================
\*\*[Table of Contents](#table-of-contents) 1\*\*
\*\*[Interview Information](#interview-information) 2\*\*
> [Individually-selected](#individually-selected) 2
>
> [NeurIPS-or-ICML](#neurips-or-icml... |
4a7ed347-4293-4cac-b2a3-44d571dbfab1 | trentmkelly/LessWrong-43k | LessWrong | Using the LessWrong API to query for events
Taymon Beal was pinging us about some help with using the LessWrong API to construct a sidebar for SSC with upcoming meetups. I figured I would write up my response publicly so that other people might be able to learn from it.
As I understand the thing correctly (though Tay... |
3be60493-cb4d-4515-a96e-bf28f01b8922 | StampyAI/alignment-research-dataset/blogs | Blogs | Visible Thoughts Project and Bounty Announcement
(**Update Jan. 12, 2022**: We released an [FAQ](https://docs.google.com/document/d/1sxNOxvcBWw7XC6MSUUpAjO4Rbu3VcUmCy7PJWQ9Vh5o/edit) last month, with more details. Last updated Jan. 7.)
(**Update Jan. 19, 2022**: We now have an example of a [successful partial run](h... |
3d78897a-befe-45d4-80cd-0d74f7262c25 | trentmkelly/LessWrong-43k | LessWrong | Estimating the kolmogorov complexity of the known laws of physics?
In the post Complexity and Intelligence, Eliezer says that the Kolmogorov Complexity (length of shortest equivalent computer program) of the laws of physics is about 500 bits:
> Suppose you ran a Turing machine with unlimited tape, so that, starting f... |
435abc0e-b94a-4829-aef8-6257edc2887b | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Announcing the Alignment Research Center
(Cross-post from [ai-alignment.com](https://ai-alignment.com/announcing-the-alignment-research-center-a9b07f77431b))
I’m now working full-time on the [Alignment Research Center](http://alignmentresearchcenter.org) (ARC), a new non-profit focused on intent alignment research.
... |
ed2aa997-562b-4101-b60c-f39ff5fa04be | trentmkelly/LessWrong-43k | LessWrong | Too Soon
It was a cold and cloudy San Francisco Sunday. My wife and I were having lunch with friends at a Korean cafe.
My phone buzzed with a text. It said my mom was in the hospital.
I called to find out more. She had a fever, some pain, and had fainted. The situation was serious, but stable.
Monday was a normal d... |
e10337cf-35f9-4cd5-8804-b853878ab196 | trentmkelly/LessWrong-43k | LessWrong | Be careful with thought experiments
Thagard (2012) contains a nicely compact passage on thought experiments:
> Grisdale’s (2010) discussion of modern conceptions of water refutes a highly influential thought experiment that the meaning of water is largely a matter of reference to the world rather than mental represen... |
33d3ee9a-aae3-4270-a48a-723c68ee26fc | StampyAI/alignment-research-dataset/arxiv | Arxiv | A Framework for Improving Scholarly Neural Network Diagrams
1 Introduction
---------------
Neural networks are often used in Artificial Intelligence (AI) systems. Two important application domains are Natural Language Processing (NLP) and Computer Vision (CV), specialising in the creation of systems to perform task... |
6f398a97-1b99-451b-a0e0-cf0388dabdc5 | trentmkelly/LessWrong-43k | LessWrong | Weekly LW Meetups: Austin, Berlin, Champaign, Melbourne, Rio de Janeiro, Springfield MO, Washington DC
This summary was posted to LW Main on Nov 9th. The following week's summary is here.
There are upcoming irregularly scheduled Less Wrong meetups in:
* Rio de Janeiro Meetup: 09 November 2012 08:00PM
* (Springfiel... |
162bda10-ed84-48d3-bb4d-14c3b2422970 | StampyAI/alignment-research-dataset/lesswrong | LessWrong | Requirements for a STEM-capable AGI Value Learner (my Case for Less Doom)
To be an [Artificial General Intelligence](https://www.lesswrong.com/tag/artificial-general-intelligence-agi) (AGI) an AI needs to meet or exceed skilled human performance for most tasks, specifically including Science/Technology/Engineering/Ma... |
5dc3820f-f805-4cdd-a23f-0f41017f28e5 | trentmkelly/LessWrong-43k | LessWrong | "The Conditional Fallacy in Contemporary Philosophy"
> Split from "Against Utilitarianism: Sobel's attack on judging lives' goodness" for length.
Robert K. Shope, back in his 1978 paper "The Conditional Fallacy in Contemporary Philosophy", identified a kind of argument that us transhumanists will find painfully famil... |
6b43f898-b9df-4a18-b0a8-2b0f0971c096 | trentmkelly/LessWrong-43k | LessWrong | Roundabout Strategy
Epistemic status: shilling for Big Concept
In the middle of the forest was a library. In the middle of the library was a book. The book’s ideas were usually not very good. At least, they were pretty debatable.
In the middle of the book was a gem.
Roundabout
The book’s author, Mark Spitznagel, s... |
f785a4e2-81b7-41fe-8ad5-38ded4437c81 | trentmkelly/LessWrong-43k | LessWrong | Why write down the basics of logic if they are so evident?
In My Bayesian Enlightenment, Eliezer writes that he was born a Bayesian. That he decided to become a Bayesian no more than fish prefer to breathe water.
Maybe all people are born Bayesian? Although, in that case, why doesn't everyone use Bayesian statistics?... |
c692d26a-0ac1-4237-aa6d-7b65fc644d25 | trentmkelly/LessWrong-43k | LessWrong | AI Risk for Epistemic Minimalists
Financial status: This is independent research, now supported by a grant. I welcome further financial support.
Epistemic status: This is an attempt to use only very robust arguments.
----------------------------------------
Outline
* I outline a case for concern about AI that doe... |
c992ae45-cf08-4e33-9765-f79efd9b72c3 | trentmkelly/LessWrong-43k | LessWrong | Checking for the Programming Gear
People on this board have talked about programming as a gear in your brain that, to a first approximation, you have or you don't. I'm wondering if there's some well put-together resource you can direct someone with zero experience and just a web-browser to and say "if you're having fu... |
6f4bc46d-7e57-458a-a99d-121fce182fd9 | trentmkelly/LessWrong-43k | LessWrong | Derivative AT a discontinuity
But this is impossible by definition
The title may seem like a contradiction. How can you differentiate something that’s not even continuous?
The usual definition of the derivative of a function f at a point a is given by the limit:
f′(a):=limε→0f(a+ε)−f(a)ε
If f is differentiab... |
8681b1a4-a8dd-4348-90c8-e8c8d043f40c | trentmkelly/LessWrong-43k | LessWrong | Questions for old LW members: how have discussions about AI changed compared to 10+ years ago?
I wasn't on LessWrong 10 years ago, during the pre-ChatGPT era, so I'd like to hear from people who were here back in the day. I want to know how AI-related discussions have changed and how people's opinions have changed.
... |
efc92415-6b81-4a68-9272-326b03d6f296 | trentmkelly/LessWrong-43k | LessWrong | Anthropics and the Universal Distribution
(Cross-posted from Hands and Cities. Content warning: especially niche topic.)
Some readers of my recent sequence on anthropics suggested that I consider an approach that they find especially plausible — namely, UDASSA (or the “Universal Distribution” plus the “Absolute Self-... |
14138e08-5603-4768-b715-14d73046bb67 | trentmkelly/LessWrong-43k | LessWrong | Political Roko's basilisk
Why has there never been a "political Roko's basilisk", i.e. a bill or law that promises to punish any member of parliament who voted against it (or more generally any individual with government power, e.g. judge or bureaucrat, who did not do everything in their capacity to make it law)?
Eve... |
1f61d5f2-a8aa-45b2-8850-9dab122f3b77 | StampyAI/alignment-research-dataset/blogs | Blogs | botched alignment and alignment awareness
2022-05-09 edit: i have found out that this idea is more thoroughly explored [here](https://reducing-suffering.org/near-miss/) and kinda [here](https://arbital.com/p/hyperexistential_separation/).
botched alignment and alignment awareness
------------------------------------... |
a6ebfb3d-6314-49ed-a83f-24661101a113 | trentmkelly/LessWrong-43k | LessWrong | LW For External Comments?
There are several blogs which (a) cross-post to LessWrong and (b) have a relatively quiet comment section on their "canonical" sites: Meteuphoric ( canonical, LW), Don't Worry About The Vase ( canonical, LW), Sideways View ( canonical, LW), etc.
In general, I'm much happier to comment on LW,... |
174ced70-2660-4e34-af72-8bdde2f27d65 | trentmkelly/LessWrong-43k | LessWrong | Death, Existentialism, and AI
A while back, I posited two questions, which are listed below, and tried to expand upon them. I am inserting this text as 'quote text' because I realize the writing and thinking are likely flawed in some places. (For example, the assertion of something like life-bias. I have no articles ... |
b2047f9f-6cb1-466d-adbb-993ebf032d99 | trentmkelly/LessWrong-43k | LessWrong | Texas Freeze Retrospective: meetup notes
This article is a writeup of the conversation at a meetup hosted by Austin Less Wrong on Saturday, February 27, 2021. The topic was the winter weather and infrastructure crisis that took place the previous week. There were a total of 13 participants, including 8 people who were... |
21774acd-96a1-405a-b989-aeab0d913c69 | trentmkelly/LessWrong-43k | LessWrong | Addendum: basic facts about language models during training
This post is a follow-up to our previous work on language model internals through training based on great additional suggestions by commenters. If you haven’t read the original post, please read it first. This post is part of the work done at Conjecture.
Ou... |
e35d208e-eef7-4435-9eb9-6327e6c62fab | StampyAI/alignment-research-dataset/arxiv | Arxiv | Revisiting the Calibration of Modern Neural Networks
1 Introduction
---------------
Neural networks, especially vision models, are increasingly used in safety-critical applications such as autonomous driving (Bojarski et al., [2016](#bib.bib1)), medical diagnosis (Esteva et al., [2017](#bib.bib9); Jiang et al., [20... |
3fba622b-81d9-48c4-b68f-c400281063b8 | trentmkelly/LessWrong-43k | LessWrong | Minds that make optimal use of small amounts of sensory data
In That alien message, Eliezer made some pretty wild claims:
> My moral - that even Einstein did not come within a million light-years of making efficient use of sensory data.
>
> Riemann invented his geometries before Einstein had a use for them; the phys... |
afe115ca-b8c2-4667-ac11-3b71bad0059f | StampyAI/alignment-research-dataset/eaforum | Effective Altruism Forum | Preparing for AI-assisted alignment research: we need data!
*Thanks to Neel Nanda, Madeleine Chang, Michael Chen, Oliver Zhang, and Aidan O’Gara for their early feedback on this post.*
*Summary: Alignment researchers could record the intermediate outputs of their work in anticipation of using the data to train future... |
a1ca1492-d889-444c-8ca8-b2e0d2f16669 | trentmkelly/LessWrong-43k | LessWrong | The trouble with Bayes (draft)
Prerequisites
This post requires some knowledge of Bayesian and Frequentist statistics, as well as probability. It is intended to explain one of the more advanced concepts in statistical theory--Bayesian non-consistency--to non-statisticians, and although the level required is much less... |
16813b10-325b-4488-9f68-846661f58bb0 | trentmkelly/LessWrong-43k | LessWrong | In defence of ignorant thinking
Suppose you want to contribute to the understanding of some subject, but you are presently ignorant about it. Should you do something closer to (a) read everything that’s been written so far, then join in, or (b) think about it yourself a lot before you even look at the basics of what o... |
b7b5f705-a17b-4743-9d69-a287b0608294 | trentmkelly/LessWrong-43k | LessWrong | Speedrunning my Morning Makes the Coffee Taste Weird
For the first time in I can't remember how long, I failed to set a wakeup alarm on a workday. I've been covering an emergency vacation for a co-worker, and just forgot to set the alarm on a day I don't normally work. Fortunately, I also have a backup plan: I have a... |
ea64b3f3-26f4-432c-939a-33f2fb1dba87 | trentmkelly/LessWrong-43k | LessWrong | What Can We Learn About Human Psychology from Christian Apologetics?
A couple months ago I set up a Skype meeting Robin Hanson to chat about the book he's working on. But the first thing he wanted to talk wasn't directly related to the book. He'd read some of my work critiquing Christian apologetics, and said somethin... |
c4cae403-7baf-4394-b449-8ae8b9dab0fa | trentmkelly/LessWrong-43k | LessWrong | Introducing Shrubgrazer
Since starting using Mastodon a month ago, a few different things haven't been a good fit for how I like to use social media:
* I'm relatively free with following people, but there are some people whose posts I don't want to miss. I'd like these to go at the top of my feed.
* Fediverse re... |
0149c7b6-955b-420f-8f92-ccf1b44f3e01 | trentmkelly/LessWrong-43k | LessWrong | Superposition through Active Learning Lens
Introduction:
'How can we control whether superposition and polysemanticity occur?’ was one of the open questions mentioned by the authors of ‘Toy Models of Superposition’ and this work was an attempt to do so with the help of Active Learning.
The idea behind this work was ... |
08947bc0-9436-4a40-9b35-f59933ad8277 | trentmkelly/LessWrong-43k | LessWrong | Robustness as a Path to AI Alignment
[Epistemic Status: Some of what I'm going to say here is true technical results. I'll use them to gesture in a research direction which I think may be useful; but, I could easily be wrong. This does not represent the current agenda of MIRI overall, or even my whole research agenda.... |
83e89786-9bb4-4823-b54e-80dbc83aa267 | trentmkelly/LessWrong-43k | LessWrong | Site Guide: Personal Blogposts vs Frontpage Posts
Posts on practically any topic are welcomed on LessWrong [1]. I (and others on the team) feel it is important that members are able to “bring their entire selves” to LessWrong and are able to share all their thoughts, ideas, and experiences without fearing whether the... |
fe7e9d01-a77b-4788-9daf-7ab8c5f9ad41 | trentmkelly/LessWrong-43k | LessWrong | How does electricity work literally?
What I know about it from high school and general articles on the net doesn't satisfy. Maybe because I have critical holes in my knowledge.
From what I think I know: we're having AC running in the lines. AC means if we zoom down, we'll see that an electron is zipping along this di... |
fb690ea7-b7f2-48f5-a317-fd61ef1afd48 | trentmkelly/LessWrong-43k | LessWrong | Learning values versus indifference
A putative new idea for AI control; index here.
Corrigibility should allow safe value or policy change. Indifference allows the agent to accept changes without objecting. However, an indifferent agent is similarly indifferent to the learning process.
Classical uncertainty over val... |
df5549f0-38c7-47bf-aa92-99a6b2f5806b | trentmkelly/LessWrong-43k | LessWrong | Underappreciated content on LessWrong
It's likely that there is some mismatch between your preferences and the preferences of the LessWrong community about which posts or comments are good or deserving of upvotes. In addition to these differences, good content sometimes doesn't receive much attention from the communit... |
0ce877cc-b477-4aca-b20f-29ab0c2a2114 | trentmkelly/LessWrong-43k | LessWrong | Truthful and honest AI
By Owain Evans, with input from Jacob Hilton, Dan Hendrycks, Asya Bergal, Owen Cotton-Barratt, and Rohin Shah
Summary
We would like to see research aimed at creating advanced AI systems that are highly competent and do not lie to humans. To operationalize this goal, we introduce the concept of... |
f7c35128-a3f4-49bf-843f-26854510cde1 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Draft report on AI timelines
Hi all, I've been working on some AI forecasting research and have prepared a draft report on timelines to [transformative AI](https://www.openphilanthropy.org/blog/some-background-our-views-regarding-advanced-artificial-intelligence#Sec1). I would love feedback from this community, so I'v... |
5fd02229-5dda-4324-bb74-9efbe8c52b38 | trentmkelly/LessWrong-43k | LessWrong | [AN #157]: Measuring misalignment in the technology underlying Copilot
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 in t... |
371fab83-8dab-475f-86b9-08d13337362b | trentmkelly/LessWrong-43k | LessWrong | My mother is now in cryostasis
I hope this post will help to those who think about cryonics.
My mother was diagnosed with stomach cancer on 1 Jules 2010.
It was on terminal stage and she died 24 September 2010.
Before, in March she publically said that she is against cryonics on the party at her home.
But several... |
c3294160-2bf4-4cb7-af9d-d1c4f505e78d | trentmkelly/LessWrong-43k | LessWrong | Why are certain trends so precisely exponential?
I was reading a post on the economy from the political statistics blog FiveThirtyEight, and the following graph shocked me:
This, according to Nate Silver, is a log-scaled graph of the GDP of the United States since the Civil War, adjusted for inflation. What amazes me... |
b968897c-5fb4-4d33-97cd-ede930e17290 | trentmkelly/LessWrong-43k | LessWrong | Land Reclamation is in the 9th Circle of Stagnation Hell
Land reclamation is a process where swamps, wetlands, or coastal waters are drained and filled to create more dry land.
Despite being complex and technologically intensive, land reclamation is quite old and was common in the past. The reclamation of the Dutch l... |
80de4b69-7453-409b-b5f0-61a540265707 | trentmkelly/LessWrong-43k | LessWrong | Off Topic Thread: May 2009
Here's your space to talk about anything totally unrelated to being Less Wrong |
16be43d8-9f97-40a2-adbd-4eb55671958e | trentmkelly/LessWrong-43k | LessWrong | [Link] On Saving the World and Other Delusions
A blog post by Athrelon on More Right.
> 1.
>
> I recently had a conversation with a friend of mine who suffered a crisis of faith of sorts. His startup, which initially had an extremely ambitious, world-altering business plan, had to retrench and start to find a more ... |
8aa1d0e8-3a0e-42b4-905a-5bd06ba5efa4 | trentmkelly/LessWrong-43k | LessWrong | A solvable Newcomb-like problem - part 3 of 3
This is the third part of a three post sequence on a problem that is similar to Newcomb's problem but is posed in terms of probabilities and limited knowledge.
Part 1 - stating the problem
Part 2 - some mathematics
Part 3 - towards a solution
----------------... |
b5bdd3c9-e181-4432-9eca-27c600d15379 | StampyAI/alignment-research-dataset/lesswrong | LessWrong | Fighting Akrasia: Incentivising Action
**Related To:** [Incremental Progress and the Valley](/lw/7k/incremental_progress_and_the_valley/), [Silver Chairs, Paternalism, and Akrasia](/lw/9w/silver_chairs_paternalism_and_akrasia/), [How a pathological procrastinator can lose weight](/lw/am/how_a_pathological_procrastino... |
2aa5dcb9-b26c-4a00-82e5-2d3d9a3ec0d5 | StampyAI/alignment-research-dataset/arbital | Arbital | Irrational number
An irrational number is a [https://arbital.com/p/-4bc](https://arbital.com/p/-4bc) that is not a [https://arbital.com/p/-4zq](https://arbital.com/p/-4zq). This set is generally denoted using either $\mathbb{I}$ or $\overline{\mathbb{Q}}$, the latter of which represents it as the [complement](https://... |
8463f5b3-ce02-4f50-a323-948202692366 | trentmkelly/LessWrong-43k | LessWrong | How Bad Is QWERTY, Really? A Review of the Literature, such as It Is
Some six or so years ago, not long after I had started working as a programmer, I developed a bad case of repetitive strain injury (RSI) and for a while couldn't use my left hand for typing, but had to rely on my right hand and dictation. This situa... |
31ffef03-7399-448e-adac-f81120c4a0c0 | trentmkelly/LessWrong-43k | LessWrong | Why study perturbative adversarial attacks ?
This post summarizes and comments on Motivating the Rules of the Game for Adversarial Example Research
Summary of paper
Despite the amount of recent work done, human-inperceptible perturbation adversarial attacks (Example: One Pixel Attack) are not as useful as the resear... |
4c980ab7-a30d-4ee2-87de-f3c307f3dfa1 | trentmkelly/LessWrong-43k | LessWrong | MIRI's Approach
MIRI's summer fundraiser is ongoing. In the meantime, we're writing a number of blog posts to explain what we're doing and why, and to answer a number of common questions. This post is one I've been wanting to write for a long time; I hope you all enjoy it. For earlier posts in the series, see the bott... |
7850c872-0783-45e2-929f-98d868dc3931 | trentmkelly/LessWrong-43k | LessWrong | An EPub of Eliezer's blog posts
Update 2015-03-21: I would now strongly recommend reading Rationality: From AI to Zombies over this. Though the blog posts I collected here are the starting point for that book, considerable work has gone into selecting and arranging the essays as well as adding thoughtful new material ... |
2c47c553-344f-4db2-92c2-786a1e3d47b6 | LDJnr/LessWrong-Amplify-Instruct | LessWrong | "See also: Challenging the Difficult and Tips and Tricks for Answering Hard Questions.
From Michael Nielsen's Reinventing Discovery: In January 2009, [mathematician Tim] Gowers decided to use his blog to run a very unusual social experiment. He picked out an important and difficult unsolved mathematical problem, a prob... |
944d91d2-e7c7-4ded-b361-b276c2dc913d | trentmkelly/LessWrong-43k | LessWrong | Pareto improvements are rarer than they seem
this is surely not an original insight, but I haven't seen it before
A Pareto improvement is where you make one party better off and no parties worse off.
Suppose Adam has a rare baseball card. He assigns no intrinsic value to baseball cards. Adam likes Beth, and somewhat... |
9930224f-b1c3-4896-b4c1-0fab60045a19 | trentmkelly/LessWrong-43k | LessWrong | Alignment Newsletter #16: 07/23/18
Highlights
Seedbank — discover machine learning examples (Michael Tyka): Seedbank provides interactive machine learning examples in Colab notebooks (think Jupyter notebooks in the cloud). This makes it really easy to just run example code without any setup, and even to modify it to p... |
670005dc-32ab-4ee8-a886-1c0d229f47ab | trentmkelly/LessWrong-43k | LessWrong | Steelmaning AI risk critiques
At some point soon, I'm going to attempt to steelman the position of those who reject the AI risk thesis, to see if it can be made solid. Here, I'm just asking if people can link to the most convincing arguments they've found against AI risk.
EDIT: Thanks for all the contribution! Keep t... |
c91f9615-8537-4cde-803b-240d0a29f470 | trentmkelly/LessWrong-43k | LessWrong | Scroll Snapping
It's been several years since paying attention to the web platform was a big part of my full time job, and I was curious if I'd missed any interesting additions. I did some poking around with Claude and ended up finding something neat, but a bit different than I was looking for: scroll-snap-type.
Scro... |
ef4e0655-6ee2-456d-927d-b9eaa45ddb8f | trentmkelly/LessWrong-43k | LessWrong | Apology of (a form of) authoritarianism and inequality from an egalitarian
Human population and energy consumption has risen exponentially in the last century, and is reaching various of the limits of its environment (in terms of number, but even more significantly in term of impact). Beyond traditional comparisons wi... |
6a33a896-ff0a-4601-a5d4-5db95e51097d | trentmkelly/LessWrong-43k | LessWrong | Why good things often don’t lead to better outcomes
Many people believe that when something good happens, positive outcomes naturally follow — but I'm here to explain why that's often not the case. In complex systems, improvements or advancements frequently set off a chain of reactions that can undermine the original ... |
f5d999df-457b-4599-98fb-7bf87ce61efc | trentmkelly/LessWrong-43k | LessWrong | Bay area OB/LW meetup, today, Sunday, March 29, at 5pm
Eliezer and Michael Vassar will be there, as will many other exciting LW-ers. Robin Gane-McCalla will be leading us in some rationality-related games. More information here.
Whether or not you can come today, you may want to sign up on our meet-up page, if you'... |
7a76fae4-dc52-421a-be8b-d536460a96d2 | StampyAI/alignment-research-dataset/lesswrong | LessWrong | Critiques of prominent AI safety labs: Conjecture
*Cross-posted from the EA Forum. See the* [*original here*](https://forum.effectivealtruism.org/posts/gkfMLX4NWZdmpikto/critiques-of-prominent-ai-safety-labs-conjecture)*. Internal linking has not been updated for LW due to time constraints and will take you back to th... |
5e26d9a8-a470-4f3d-8a1f-8b4369c60537 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Modulating sycophancy in an RLHF model via activation steering
*Produced as part of the* [*SERI ML Alignment Theory Scholars Program*](https://serimats.org/) *- Summer 2023 Cohort, under the mentorship of Evan Hubinger. Thanks to Alex Turner for his feedback and ideas.*
This is a follow-up post to ["Reducing sycophan... |
95ce6598-2817-4b34-b7c3-56ddb56c2e35 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | By Default, GPTs Think In Plain Sight
*Epistemic status: Speculation with some factual claims in areas I’m not an expert in.*
*Thanks to Jean-Stanislas Denain, Charbel-Raphael Segerie, Alexandre Variengien, and Arun Jose for helpful feedback on drafts, and thanks to janus, who shared related ideas.*
Main claims
====... |
4226fab6-c08e-49e0-ab5e-b3592cb879db | trentmkelly/LessWrong-43k | LessWrong | Selection Effects in estimates of Global Catastrophic Risk
Here's a poser that occurred to us over the summer, and one that we couldn't really come up with any satisfactory solution to. The people who work at the Singularity Institute have a high estimate of the probability that an Unfriendly AI will destroy the world... |
26c7bbb0-5c35-4f04-95c3-694a0c50a615 | trentmkelly/LessWrong-43k | LessWrong | LW September WebDiplomacy Games: Starting soon!
Check out this post for background details.
The final list for the 48 hour game is:
Prismattic
ahartell
Kutta
Randaly
GuySrinivasan
prase
RobertLumley
The final list for the 24 hour game is:
folkTheory
printing-spoon
handoflixue
Vaniver
ahartell
shokwave
Randaly
In or... |
cdca9307-cda9-40ba-84fa-6108348dd5e7 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | AI Safety Research Project Ideas
This post contains project ideas in AI Safety from Owain Evans and Stuart Armstrong (researchers at [FHI](https://www.fhi.ox.ac.uk/research/research-areas/#aisafety_tab)). Projects are aimed at students, postdocs or summer research fellows who would be interested in collaborating on th... |
4c89088c-acaf-4b11-9cc4-83e3448b3e4d | trentmkelly/LessWrong-43k | LessWrong | Does non-access to outputs prevent recursive self-improvement?
Most existing large-scale models don't seem to have access to their outputs (except in the sense that some of their outputs show up in the inputs of the next generation of the model). Does this prevent recursive self-improvement - or at least limit the spe... |
635aa913-1514-4cfb-913b-d326a865db7b | trentmkelly/LessWrong-43k | LessWrong | NonGoogleables
Recently in another topic I mentioned the "two bishops against two knights" chess endgame problem. I claimed it was investigated over two decades ago by a computer program and established that it is a win situation for the two bishops' side. Then I was unable to Google a solid reference for my claim.
I... |
dd77d0dd-3225-496b-8745-1901639bad68 | 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... |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.