id
stringlengths
36
36
source
stringclasses
15 values
formatted_source
stringclasses
13 values
text
stringlengths
2
7.55M
500de043-6e43-4f01-b245-d02af958bdf5
trentmkelly/LessWrong-43k
LessWrong
AI #36: In the Background Wow, what a week. We had the Executive Order, which I read here so you don’t have to and then I have a tabulation of the reactions of others. Simultaneously there was the UK AI Summit. There was also robust related discussion around Responsible Scaling Policies, and the various filings comp...
fcdd31bb-0d4b-4f38-966d-f0ce60fa69db
trentmkelly/LessWrong-43k
LessWrong
Come up with better Turing Tests So the Turing test has been "passed", and the general consensus is that this was achieved in a very unimpressive way - the 13 year old Ukrainian persona was a cheat, the judges were incompetent, etc... These are all true, though the test did pass Turing's original criteria - and there ...
ae036d37-c6ab-4dce-b516-0aebbd6e070f
trentmkelly/LessWrong-43k
LessWrong
Activation Energies and Uncertainty From "An Equilibrium of No Free Energy" > We can see the notion of an inexploitable market as generalizing the notion of an efficient market as follows: in both cases, there’s no free energy inside the system. In both markets, there’s a horde of hungry organisms moving around tryin...
4b22b79e-6007-4f1b-aeb9-38836f669e22
trentmkelly/LessWrong-43k
LessWrong
Some additional SAE thoughts Thanks to Lee Sharkey for feedback on the first section and Lee Sharkey, Jake Mendel, Kaarel Hänni and others at the LISA office for conversations around this post. This post is a collection of a few small experiments and thoughts from the last couple of months that never really turned in...
6219373e-8e3a-4e9d-b6ea-31a52a6351ff
trentmkelly/LessWrong-43k
LessWrong
A Month's Worth of Rational Posts - Feedback on my Rationality Feed. For the last two months I have been publishing a feed of rationalist articles. Oriignally the feed was only published on the SSC discord channel SSC Discord (Be charitable, kind and don't treat the place like 4chan). For the last few days I have also...
094249df-0218-4ed2-b3b7-7d186c5fb4cc
StampyAI/alignment-research-dataset/youtube
Youtube Transcripts
$100,000 for Tasks Where Bigger AIs Do Worse Than Smaller Ones #short in AI bigger is better right large models outperform smaller ones according to scaling laws except not always sometimes bigger models can actually perform worse and sometimes in ways that might be dangerous so since we're going to be building really...
fa079017-96a7-4067-ad37-8f8b08c1df1a
StampyAI/alignment-research-dataset/eaforum
Effective Altruism Forum
I have thousands of copies of HPMOR in Russian. How to use them with the most impact? As a result of a crowdfunding campaign a couple of years ago, I printed 21k copies of [HPMOR](https://lesswrong.com/hpmor). 11k of those went to the crowdfunding participants. I need ideas for how to use the ones that are left with t...
ee2c766a-fcc5-451c-852e-44c8116ac8e7
trentmkelly/LessWrong-43k
LessWrong
Weak-to-Strong Generalization: Eliciting Strong Capabilities With Weak Supervision Links: Blog, Paper. Abstract: > Widely used alignment techniques, such as reinforcement learning from human feedback (RLHF), rely on the ability of humans to supervise model behavior—for example, to evaluate whether a model faithfully...
f98e4b79-5105-4efe-adaf-6b8a44f61977
trentmkelly/LessWrong-43k
LessWrong
How do I suggest someone for the Rationality Blogs sidebar? I've been seeing a lot of good content coming out of certain "rationality-adjacent" blogs recently and I think that there's a certain someone who ought to be listed in the Rationality Blogs sidebar but I don't think they are. (I'm not mentioning them by name ...
024889a5-3a9f-4475-83a9-ff467d4d3763
StampyAI/alignment-research-dataset/lesswrong
LessWrong
A FLI postdoctoral grant application: AI alignment via causal analysis and design of agents TL;DR: This is a proposal I wrote for the [Future of Life postdoctoral grants](https://grants.futureoflife.org/prog/postdoctoral/), on the topic of AI Safety. The topic of this proposal is being able to model human preferences ...
5b16ff51-ad3c-4573-9fc1-e117c08281f1
trentmkelly/LessWrong-43k
LessWrong
Your risk of developing long COVID is probably high Context: I wrote the following on Facebook for a general audience. I'm posting it here because I'm very concerned. I'm severely disabled from a presumable* COVID infection 17 months ago. I haven't done deep dives into (m)any of these, so I'm plausibly wrong. However,...
8190953a-5fe3-4e2e-a8a7-9662156cb938
trentmkelly/LessWrong-43k
LessWrong
Deriving probabilities from causal diagrams What this is: an attempt to examine how causal knowledge gets turned into probabilistic predictions. I'm not really a fan of any view of probability that involves black boxes. I want my probabilities (or more practically, the probabilities of toy agents in toy problems I co...
2d51b444-f546-4361-b6e2-eafb15506732
trentmkelly/LessWrong-43k
LessWrong
Brain Preservation Foundation [link] An organization calling itself the Brain Preservation Foundation is offering a cash prize, currently worth about $100,000, to anyone who can demonstrate the successful preservation of the connectome of a large mammalian brain. They also have an impressive web page. Does anyone here...
0d16712d-a5fc-4659-8150-858ca10f3fb0
trentmkelly/LessWrong-43k
LessWrong
Do you consider perfect surveillance inevitable? A lot of my recent research work focusses on: 1. building the case for why perfect surveillance is becoming increasingly hard to avoid in the future 2. thinking through the implications of this, if it happened When I say perfect surveillance, imagine everything your ...
e43a8067-edc2-4289-b1c7-380c60ab2d1a
trentmkelly/LessWrong-43k
LessWrong
How good a proxy for accuracy is precision? Accuracy = averaging to the right answer. Precision = having a low standard deviation. It's easy to get a precise but inaccurate answer- just guess 0 every time. But for situations where people aren't goodharting, are a cluster of answers that are more similar to each ...
9fd5af3a-d66e-4886-aef5-69cd79711c95
LDJnr/LessWrong-Amplify-Instruct
LessWrong
"[Epistemic status: Sharing current impressions in a quick, simplified way in case others have details to add or have a more illuminating account. Medium-confidence that this is one of the most important parts of the story.]Here's my current sense of how we ended up in this weird world where:I still intermittently run ...
55a79d3f-adc2-4a95-ae8e-dacf35763306
StampyAI/alignment-research-dataset/lesswrong
LessWrong
[Linkpost] diffusion magnetizes manifolds (DALL-E 2 intuition building) In trying to make sense of why diffusion works so well, I explored three complementary models related to magnetism, Hansel & Gretel, and landing airplanes.
d392e0c0-876b-4198-9de9-a5644b627aab
StampyAI/alignment-research-dataset/arxiv
Arxiv
Using Natural Language and Program Abstractions to Instill Human Inductive Biases in Machines.. 1 Introduction --------------- Humans are able to rapidly perform a variety of tasks without extensive experience (Ormrod, [2016](#bib.bib1)). This may be because of strong inductive biases towards abstract structured kn...
6ebfa65d-f836-49c9-b060-21421005eb70
trentmkelly/LessWrong-43k
LessWrong
Communicating effectively: form and content Effective communication techniques, particularly in written communication, are an important part of the aspiring rationalist's toolkit. Alicorn's recent post makes excellent points about niceness, and touches parenthetically on the larger issue of form versus content. The g...
0daa46a5-341b-4bcf-b479-d0f890466604
StampyAI/alignment-research-dataset/blogs
Blogs
Circuits Updates — May 2023 We report a number of developing ideas on the Anthropic interpretability team, which might be of interest to researchers working actively in this space. Some of these are emerging strands of research where we expect to publish more on in the coming months. Others are minor points we wish to...
aec01c05-8b27-45b9-8858-4d987b5610bc
trentmkelly/LessWrong-43k
LessWrong
Death in Groups What is the value of your life? Not a life, but your life. Would you consider yourself expendable? The last time I almost died, it was over a missing shotgun. I was in Afghanistan at the time, somewhere closer to the border with Pakistan. The roads are rough, particularly when there aren't any, and du...
178d0496-a0e7-41f5-9bb5-e5a20a1953de
trentmkelly/LessWrong-43k
LessWrong
On the Rationality of Deterring ASI I’m releasing a new paper “Superintelligence Strategy” alongside Eric Schmidt (formerly Google), and Alexandr Wang (Scale AI). Below is the executive summary, followed by additional commentary highlighting portions of the paper which might be relevant to this collection of readers. ...
06343e36-39ff-4a0f-898c-68b1c41607aa
trentmkelly/LessWrong-43k
LessWrong
The Anthropic Trilemma Speaking of problems I don't know how to solve, here's one that's been gnawing at me for years. The operation of splitting a subjective worldline seems obvious enough - the skeptical initiate can consider the Ebborians, creatures whose brains come in flat sheets and who can symmetrically divide...
8d20c719-39bb-4f41-ac81-3f6f932c1200
trentmkelly/LessWrong-43k
LessWrong
How did you integrate voice-to-text AI into your workflow? ChatGPT seems to have really awesome voice-to-text ability. However, it seems to only record within ChatGPT itself so can't be used to create notes or type in other programs and it's unclear to me how to best take advantage of the increased technological capab...
dbd9bd3a-250b-48f0-883f-4df2f0363d2c
trentmkelly/LessWrong-43k
LessWrong
Fermi Paradox - Reference Class arguments and Other Possibilities None
381b3b9e-9146-4d46-9b8b-c3f44fd01c47
trentmkelly/LessWrong-43k
LessWrong
Meetup : Salt Lake City Meetup (Location Changed!) Discussion article for the meetup : Salt Lake City Meetup (Location Changed!) WHEN: 07 April 2012 12:00:42PM (-0600) WHERE: Holladay Library: 2150 Murray Holladay Rd, Holladay, UT 84117 Presentation on social influence, based off Influencer, and discussion. Discu...
8f532b69-06ea-4b85-963b-228126961d53
trentmkelly/LessWrong-43k
LessWrong
Wanted: Rationalist Pushback (link) Recently I started a new blog, named "Undivided Looking: comments on physics and theology".  You can find it here: http://www.wall.org/~aron/blog/ My main goals are to: * discuss science and/or religion issues * popularize some fundamental physics ideas to folks who aren't afra...
66d620c0-c714-4692-8389-4385f7c5bb8e
trentmkelly/LessWrong-43k
LessWrong
Jane Street Inteview Guide (Notes on Probability, Markets, etc.) Context: I'm currently prepping for a Jane Street trader intern interview. They sent me this guide to look over and, while there's not a lot that's new to me or LW, it's got a lot of concepts that are popular here on LW (expected value, conditional proba...
dd4c44aa-385a-42bb-9f9a-0c79a9f7ab25
trentmkelly/LessWrong-43k
LessWrong
[AN #142]: The quest to understand a network well enough to reimplement it by hand 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 ev...
4b6f9927-12a3-45ed-be18-8b670b9f0aef
StampyAI/alignment-research-dataset/eaforum
Effective Altruism Forum
Closing the Feedback Loop on AI Safety Research. Is there a consensus among AI safety researchers that there is no way to safely study an AI agent's behavior within a simulated environment? It seems to me as if the creation of an adequate AGI Sandbox would be a top priority (if not #1) for AI safety researchers as it...
19c7e3c0-f825-4639-a994-610f1f0cf9a4
trentmkelly/LessWrong-43k
LessWrong
XX by Rian Hughes: Pretentious Bullshit SPOILER WARNING: Extensive spoilers for XX. XX was recommended by one of the comments on my Sci-Fi micro-reviews, so I read it. I can't seem to find any critical reviews of the book online so I'm going to while away a few hours of my life in a misguided attempt to reveal that t...
ba099059-aa7c-4b7c-a084-7c803b3d62e0
trentmkelly/LessWrong-43k
LessWrong
Universality and the “Filter” 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: In this post, I refine universality, in order to explain the "Filter". The Filter is a method of risk anal...
5173e2d2-0bc2-4447-a635-0b36a569af12
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
Why The Focus on Expected Utility Maximisers? Epistemic Status ---------------- Unsure[[1]](#fnd0323hkyj3), partially [noticing my own confusion](https://www.lesswrong.com/s/zpCiuR4T343j9WkcK). Hoping [Cunningham's Law](https://en.wikipedia.org/wiki/Ward_Cunningham#%22Cunningham's_Law%22) can help resolve it.    [R...
84d5f2ad-0a27-4053-b145-cbd021f7a085
trentmkelly/LessWrong-43k
LessWrong
Meetup : Denver Area LW January Meetup Discussion article for the meetup : Denver Area LW January Meetup WHEN: 03 January 2017 07:00:00PM (-0700) WHERE: 4955 S Ulster St Ste 103, Denver, Colorado 80237 Darcy's Pub Come for food, drink, and good conversation! We should be in the library in the back. Discussion ar...
b0d5fed2-0a84-4436-af81-cbc4e81ab1ec
trentmkelly/LessWrong-43k
LessWrong
Monosemanticity & Quantization In this post, I will cover Anthropic's work on monosemanticity[1]. Starting with a brief introduction to the motivation and methodology. Then move on to my ablation experiments where I train a sparse autoencoder on "gelu-2l"[2] and its quantized versions to see what insights I can gain. ...
35e87642-3f40-40d7-8623-e2d5472de353
StampyAI/alignment-research-dataset/lesswrong
LessWrong
Which parts of the existing internet are already likely to be in (GPT-5/other soon-to-be-trained LLMs)'s training corpus? In particular: * If a book is in easily accessible parts of the web, but is there as a pdf image of a scanned book rather than as text, is it likely to be in the training corpus of (GPT5/similar) ...
ddda1dbe-72e8-455b-a54b-3d04e55f866d
trentmkelly/LessWrong-43k
LessWrong
Where Do We Go From Here? [Mostly unjustified rambling and unpruned babble. Some politics, but mostly as framing. Still, I wouldn't share it here if I didn't think it had a nugget or two. Cross-posted from Grand, Unified, Empty.] I. Well. Joe Biden will in all likelihood be the next American President come January....
d6ad9dcf-ec14-400c-8c2d-1c4573c42c8f
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
state of my alignment research, and what needs work in this post i give an overview of the state of my [AI alignment](https://www.lesswrong.com/posts/T4KZ62LJsxDkMf4nF/a-casual-intro-to-ai-doom-and-alignment-1) research, as well as what i think needs to be worked on, notably for people who might want to join my effort...
75cff92b-35ae-4379-9409-e21c2aa8c035
StampyAI/alignment-research-dataset/special_docs
Other
Risk analysis and risk management for the artificial superintelligence research and development process Risk Analysis and Risk Management for the Artificial Superintelligence Research and Development Process Anthony M. Barrett and Seth D. Baum Global Catastrophic Risk Institute http://sethbaum.com \* http://tony-barre...
773f2a41-983d-46a1-9302-a4b270ab7ddd
trentmkelly/LessWrong-43k
LessWrong
How do I read things on the internet > This is a linkpost - I recommend reading it at the original URL for a better reading UX I have a somewhat elaborate process for reading things that I find on the web. I've been inspired to share it because after many a long iteration it finally feels adequate! Reading things on...
9ffd8429-25d7-44ae-a3ac-6e299fcb2877
trentmkelly/LessWrong-43k
LessWrong
Should we expect the future to be good? 1. Introduction Should we expect the future to be good? This is an important question for many reasons. One such reason is that the answer to this question has implications for what our intermediate goals should be. If we should expect the future to be good, then it would be rel...
cbe80b67-40ff-48d8-b8cb-11c90c2d5085
StampyAI/alignment-research-dataset/lesswrong
LessWrong
Understanding strategic deception and deceptive alignment *The following is the main part of a* [*blog post*](https://www.apolloresearch.ai/blog/understanding-da-and-sd) *we just published at Apollo Research. Our main goal with the post is to clarify the concept of deceptive alignment and distinguish it from strategic...
0e5f0cb3-b504-42e9-846a-473f1c9c1cd1
trentmkelly/LessWrong-43k
LessWrong
TASRA: A Taxonomy and Analysis of Societal-Scale Risks from AI Partly in response to calls for more detailed accounts of how AI could go wrong, e.g., from Ng and Bengio's recent exchange on Twitter, here's a new paper with Stuart Russell: * Discussion on Twitter... comments welcome! https://twitter.com/AndrewCrit...
e03b837f-1f98-4aec-acfa-736e412d8247
trentmkelly/LessWrong-43k
LessWrong
The Thinking Ladder - Wait But Why The Thinking Ladder is the latest post in a huge ongoing Wait But Why series called The Story of Us. > Our beliefs make up our perception of reality, they drive our behavior, and they shape our life stories. History happened the way it did because of what people believed in the past...
c305e873-a9da-48af-8ebc-f61559553906
trentmkelly/LessWrong-43k
LessWrong
Connecting the good regulator theorem with semantics and symbol grounding EDIT: For a great reappraisal of the good regulator theorem using modern terminology, see here. I've been writing quite a bit about syntax, semantics, and symbol grounding. Recently, I've discovered the "good regulator theorem" in systems scie...
69b513ba-eb0a-4ca9-8c94-2ae60fdedf1a
trentmkelly/LessWrong-43k
LessWrong
Static Place AI Makes Agentic AI Redundant: Multiversal AI Alignment & Rational Utopia (This is the result of three years of thinking and modeling hyper‑futuristic and current ethical systems. It's not the first post in the series, the ideas described in it are counterintuitive and can be accidentally dismissed too so...
c843443c-a37a-44a2-a0ef-f9fcdd522115
trentmkelly/LessWrong-43k
LessWrong
MAD Trollies
f269f201-ccb3-449a-a733-0ce19e2e6017
trentmkelly/LessWrong-43k
LessWrong
We need a career path for invention If technological progress has slowed down, what is causing it? Here is a hypothesis. Broadly speaking, there are three domains of activity important to technological progress: science, invention, and business. Science discovers new knowledge; invention creates useful machines, chem...
2bbbd62d-b6ba-4804-b88a-531bf759ddb5
trentmkelly/LessWrong-43k
LessWrong
Replicated cognitive bias list? Given the replication crisis in the social sciences, do we have, somewhere, a list of cognitive biases tied to replicated studies? I could be missing something, but I didn't see anything on the Wikipedia page indicating whether or not the results replicated.
97e6db20-c57b-4db8-b55d-bf22ff2cb361
trentmkelly/LessWrong-43k
LessWrong
[SEQ RERUN] Angry Atoms Today's post, Angry Atoms was originally published on 31 March 2008. A summary (taken from the LW wiki):   > It is very hard, without the benefit of hindsight, to understand just how it is that these little bouncing billiard balls called atoms, could ever combine in such a way as to make some...
61223c70-952c-4196-91b1-00dd6b301643
trentmkelly/LessWrong-43k
LessWrong
A digitized belief network? Hello to all, Like the rest of you, I'm an aspiring rationalist. I'm also a software engineer. I design software solutions automatically. It's the first place my mind goes when thinking about a problem. Today's problem is the fact that our beliefs all rest on beliefs that rest on beliefs....
3b8388aa-eba2-4e7b-9c0e-654ab9dd40c6
StampyAI/alignment-research-dataset/lesswrong
LessWrong
Facing the Intelligence Explosion FYI, the english site is working again: http://intelligenceexplosion.com/en/ I highly recommend reading this, or spreading it around if you're already fluent. This is incredible work. Thanks Luke!
bc892b85-8683-445a-a211-ea10a0f98a4e
trentmkelly/LessWrong-43k
LessWrong
Purchase Fuzzies and Utilons Separately Yesterday: > There is this very, very old puzzle/observation in economics about the lawyer who spends an hour volunteering at the soup kitchen, instead of working an extra hour and donating the money to hire someone... > > If the lawyer needs to work an hour at the soup kitche...
efd8a836-33ce-462d-9c44-afdb7f04f84b
trentmkelly/LessWrong-43k
LessWrong
Quick Thoughts on Immoral Mazes I finally got around to reading (most of) Zvi's sequence Immoral Mazes. I may want to turn some of the following ideas into longer posts, but I thought I'd try to get out some quick thoughts. Connections to Dictator's Handbook I've also recently finished reading Dictator's Handbook. T...
8662b2a1-14c3-49dc-abed-77622401e5a9
StampyAI/alignment-research-dataset/lesswrong
LessWrong
A tension between two prosaic alignment subgoals Written quickly rather than not at all, as I've described this idea a few times and wanted to have something to point at when talking to people. 'Quickly' here means **I was heavily aided by a language model while writing, which I want to be up-front about given recent ...
9e305af2-ba62-458f-9170-b15fc62749e6
StampyAI/alignment-research-dataset/eaforum
Effective Altruism Forum
How to Improve China-Western Coordination on EA Issues? **Summary:** I have a strong US-China skill set that allows me to effectively run projects between teams spanning China and Western countries. I'm curious what past or current projects the EA community is already working on that leverage coordination with groups ...
db306579-9adf-4901-ad23-ccb0276c34d0
trentmkelly/LessWrong-43k
LessWrong
US-China trade talks should pave way for AI safety treaty [SCMP crosspost] This is a crosspost from the South China Morning Post. Few expected it would happen this fast, but there is widespread relief at the news that China and the United States are talking again. In only a few days, renewed trade talks have resulted...
f6bb2de5-b23f-4a72-ae85-8bc9b3422761
trentmkelly/LessWrong-43k
LessWrong
Welcome to SLC SSC 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 frequently ...
5abff705-ead1-4396-85ac-f5df4ee7991f
trentmkelly/LessWrong-43k
LessWrong
Anthropic: Reflections on our Responsible Scaling Policy Last September we published our first Responsible Scaling Policy (RSP) [LW discussion], which focuses on addressing catastrophic safety failures and misuse of frontier models. In adopting this policy, our primary goal is to help turn high-level safety concepts ...
c5cc735f-df59-4fc8-8cfa-bdf73ee092cb
trentmkelly/LessWrong-43k
LessWrong
AI #76: Six Shorts Stories About OpenAI If you’re looking for audio of my posts, you’re in luck. Thanks to multiple volunteers you have two options. 1. Option one is Askwho, who uses a Substack. You can get an ElevenLabs-quality production, with a voice that makes me smile. For Apple Podcasts that means you can add ...
1973551c-2413-4de2-a340-469a3599742d
trentmkelly/LessWrong-43k
LessWrong
Consequentialism FAQ There are a lot of explanations of consequentialism and utilitarianism out there, but not a lot of persuasive essays trying to convert people. I would like to fill that gap with a pro-consequentialist FAQ. The target audience is people who are intelligent but may not have a strong philosophy backg...
660cc958-95d6-4d33-9496-cb822a499e5c
trentmkelly/LessWrong-43k
LessWrong
Who wants to run a B2B SaaS startup? A friend of mine has been bootstrapping a business-to-business software-as-a-service startup that's seeing serious growth. It needs someone who can put dedicated effort into scaling it, but my friend is near the end of their career and looking to retire. What do people do in this s...
d7114945-4a6f-43e4-af27-726b66f88072
StampyAI/alignment-research-dataset/arbital
Arbital
Needs clickbait A meta tag for pages which don't have [clickbait](https://arbital.com/p/597). Clickbait style guidelines: Yes: "First letter is capital." Yes: "Pique user's interest in the topic." Yes: "Most clickbaits are just one sentence telling user what they might find out if they read the page." Yes: "Can...
b18bfd74-6468-4de9-a354-f6c5641b84e7
trentmkelly/LessWrong-43k
LessWrong
Tim Dillon's fake business is the most influential video I have watched in the last 24 months I came across this video last year and have waited for some time to see if something else would radically change my perspective in a greater way. Most of 2-10 on the list are posts here on LessWrong, so I thought I would shar...
adbf5a20-8645-4859-83b6-7721fdfe713d
trentmkelly/LessWrong-43k
LessWrong
Meno's Paradox https://en.wikipedia.org/wiki/Meno A succinct argument form of Meno's paradox that I found online follows: We wish to make an inquiry into/about x (x being an object of epistemology, covers stuff that can be known) 1. Either we know x or we don't know x 2. If we know x, inquiry is unnecessary 3. I...
8f4f521b-5d90-4084-b2c3-7871ac0a526e
awestover/filtering-for-misalignment
Redwood Research: Alek's Filtering Results
id: post2999 Background One possible way to train an amplification model is to use an auxiliary reinforcement learning objective to help guide the training of the amplification model. This could be done either by training two separate models, an agent and a question-answerer, or a single model trained on a joint objec...
284a6af3-8a32-4029-9df4-b441e3d6aa40
trentmkelly/LessWrong-43k
LessWrong
Hugo Awards - HP:MoR (part 2) Summary: please nominate HPMoR:Podcast for consideration to the brand-new FanCast category.   The Hugo Awards, as many are probably aware, are awards given to outstanding Science Fiction/Fantasy works, as voted upon by the reading public. They work by letting anyone who attends the annu...
2766bb4c-a4db-4b76-b917-f92ab80c5830
trentmkelly/LessWrong-43k
LessWrong
What Is The Cost of Your Attention? Before Fake News, before Yellow Journalism, there was Benjamin Day, who ran a newspaper called The Sun. The Sun was started on September 3rd, 1833 with a revolutionary new business model: "Resell the attention of the audience otherwise known as advertising" (Timothy Wu, The Attentio...
70aa7a84-dbb0-4502-bb64-cdd8c0e4634d
trentmkelly/LessWrong-43k
LessWrong
Overview of strong human intelligence amplification methods How can we make many humans who are very good at solving difficult problems? Summary (table of made-up numbers) I made up the made-up numbers in this table of made-up numbers; therefore, the numbers in this table of made-up numbers are made-up numbers. ...
578183ad-f122-4bdd-9663-9a35897a5f1f
trentmkelly/LessWrong-43k
LessWrong
Moral Golems Assume Uncertainty > [T]he Christian humanist and Hebraist Johannes Reuchlin (1492) report[ed] the creation of an artificial anthropoid whose forehead bore the Hebrew inscription "YHVH Elohim emeth" (The name of God is truth) and who fell dead to the ground when the aleph, the first letter forming the Heb...
a588e583-d801-4dfc-ad17-fbc4edbbcf5f
trentmkelly/LessWrong-43k
LessWrong
Chuckling a Bit at Microsoft and the PCFG Formalism
d21af9d0-392d-4580-af2f-3490f15d7fe3
trentmkelly/LessWrong-43k
LessWrong
How should we adjust microCOVID estimates for omicron, boosters, testing? As I assess potential New Year's Eve plans with the intent of spending several days indoors immediately thereafter with my parents (early-to-mid 60s, one has mild-to-moderate asthma but probably healthier than your average 60 year old otherwise ...
35430c38-e630-48f1-9aa1-62b5670d8821
trentmkelly/LessWrong-43k
LessWrong
What to do after college? My friend is looking for some advice on what he should do after graduating from Harvey Mudd College. Some relevant bits of information about him are that he   * is not a US citizen, so he'd only be able to stay in the US if he's working or at a grad school. He's open to suggestions for oth...
9b72c7dd-5bc3-415b-b9da-abc753ee66cd
StampyAI/alignment-research-dataset/arbital
Arbital
Difficulty of AI alignment This page attempts to list basic propositions in computer science which, if they are true, would be ultimately responsible for rendering difficult the task of getting good outcomes from a [sufficiently advanced Artificial Intelligence](https://arbital.com/p/7g1). [https://arbital.com/p/auto...
c0979b23-200f-4f09-b3ec-6a852e32352a
trentmkelly/LessWrong-43k
LessWrong
Drugs are Sometimes Good, Actually A casual post in a series about what I wish someone had told me when I was younger. (Legal disclaimer: I'm not advocating you do illegal things. If you do illegal things you're on your own.) Like a lot of people who were kids in 90s America, I first learned about drugs from DARE. A...
c10860a6-fd2f-4994-9ac2-57d80fff1a95
StampyAI/alignment-research-dataset/lesswrong
LessWrong
Distilled - AGI Safety from First Principles *This is meant to be supplemental to*[*Richard Ngo’s AGI Safety from First Principles sequence*](https://www.alignmentforum.org/s/mzgtmmTKKn5MuCzFJ)*. It can serve as an overview/guide to the sequence, and an outline for navigating his core claims. It's also a submission fo...
1747fa64-0f4e-4dfc-b521-de1949984b02
trentmkelly/LessWrong-43k
LessWrong
Some Lessons Learned from Studying Indirect Object Identification in GPT-2 small To learn more about this work, check out the paper. We assume general familiarity with transformer circuits. Intro: There isn’t much interpretability work that explains end-to-end how a model is able to do some task (except for toy mode...
3149e69c-0d1e-4a7a-90be-34c453970fb3
trentmkelly/LessWrong-43k
LessWrong
Seek Upside Risk When I meet somebody new, there are two interesting things that could happen. We could get on super well, become great friends, and be part of each other’s lives for year to come. Or we don’t. Maybe we like each other, maybe we don’t, but we likely just go our separate ways. But the upside in the firs...
b48cee6d-22da-443f-87fa-a48c38c6c2c2
StampyAI/alignment-research-dataset/blogs
Blogs
the uncertainty of 2+2=4 the uncertainty of 2+2=4 ------------------------ people generally claim "it will rain tomorrow" and "2+2=4" are different *kinds* of knowledge. not only is one about the world and the other about mathematics, but the former is probabilistic and the latter is certain. i here would like to ...
fe2e4b6e-7668-4778-8db1-67212aae56b1
trentmkelly/LessWrong-43k
LessWrong
Created Already In Motion Followup to:  No Universally Compelling Arguments, Passing the Recursive Buck Lewis Carroll, who was also a mathematician, once wrote a short dialogue called What the Tortoise said to Achilles.  If you have not yet read this ancient classic, consider doing so now. The Tortoise offers Achill...
d1b4c2d3-15ef-4fc6-bc89-729072786209
trentmkelly/LessWrong-43k
LessWrong
Straight-edge Warning Against Physical Intimacy I’m someone who doesn’t do drugs or drink alcohol for a variety of reasons, the most important to me being so that I’m as “true to myself” as possible. By that, I mean that I do not want to do actions that I wouldn’t do sober (and might regret or might not identify with)...
8987bc4e-4b12-4e08-b38d-adcaf08cb550
trentmkelly/LessWrong-43k
LessWrong
After critical event W happens, they still won't believe you In general and across all instances I can think of so far, I do not agree with the part of your futurological forecast in which you reason, "After event W happens, everyone will see the truth of proposition X, leading them to endorse Y and agree with me abou...
723c8894-1c98-4668-8a03-71148c283842
trentmkelly/LessWrong-43k
LessWrong
Moved from Moloch's Toolbox: Discussion re style of latest Eliezer sequence Habryka: I considered moving this thread to meta, since I think it's highest value to separate the presentational conversation from the object-level conversation. I asked Hypothesis whether he would be fine with it, he said yes, and seemed to ...
c24491cc-9469-490c-a419-be7f4565d848
trentmkelly/LessWrong-43k
LessWrong
Open thread, June 2-16, 2012 If it's worth saying, but not worth its own post, even in Discussion, it goes here.
e0425287-fa97-4b94-ace3-bb969c6e709a
trentmkelly/LessWrong-43k
LessWrong
How can we protect economies during massive public health crises? Without the economic knock-on effects, telling people to all close up shop and stay home during a massive health crisis would be a no-brainer. Unfortunately, the risk of a nasty economic shock seems to be weighing heavily in the decisions of our leaders...
34ed41e9-8fee-4574-b1a5-b093ee84a509
StampyAI/alignment-research-dataset/lesswrong
LessWrong
Using the universal prior for logical uncertainty (retracted) (I'm not sure about any of this) (Yup! Vanessa Kowalski pointed out a crucial error, see her comment below. The idea about the universal prior picking up all computable regularities is *unproved*. I'm leaving the post as it was, but please read it as a hi...
e8e47c26-7a81-42e1-a71e-e5ba87e31858
StampyAI/alignment-research-dataset/eaforum
Effective Altruism Forum
Future Matters #3: digital sentience, AGI ruin, and forecasting track records > We could thus imagine, as an extreme case, a technologically highly advanced society, containing many complex structures, some of them far more intricate and intelligent than anything that exists on the planet today — a society which never...
a020386b-9422-4928-a6a3-347d3d55ae4a
trentmkelly/LessWrong-43k
LessWrong
What is Evidence? > The sentence “snow is white” is true if and only if snow is white. > > —Alfred Tarski   > To say of what is, that it is, or of what is not, that it is not, is true. > > —Aristotle, Metaphysics IV Walking along the street, your shoelaces come untied. Shortly thereafter, for some odd reason, you...
32115213-c653-4120-a076-67be83ec6e49
trentmkelly/LessWrong-43k
LessWrong
Weekly LW Meetups: Atlanta, Austin, Berlin, Durham, London, Montreal, Vancouver This summary was posted to LW main on March 22nd. The following week's summary is here. There are upcoming irregularly scheduled Less Wrong meetups in: * Durham HPMoR meetup, ch. 47-50: 23 March 2013 11:30AM * Vancouver: 23 March 2013 ...
2e2b8f75-6437-4ca1-9e69-d8f82109cbf7
StampyAI/alignment-research-dataset/blogs
Blogs
Medium Lights I like this [basketball substack called Medium Lights](https://mediumlights.substack.com/) that explains all kinds of cool basketball plays the announcers wouldn't necessarily comment on. Unlike most sites like this, it usually explains in enough detail that I can actually follow what's cool about the pl...
5ca70544-d52d-4e7e-9968-e8599b6e0edc
trentmkelly/LessWrong-43k
LessWrong
Most problems don't differ dramatically in tractability (under certain assumptions) Cross-post from EA Forum. Recall the importance-tractability-neglectedness (ITN) framework for estimating cost-effectiveness: * Importance = utility gained / % of problem solved * Tractability = % of problem solved / % increase in ...
b39163b6-8caf-4b38-bb5e-77d6413aae27
awestover/filtering-for-misalignment
Redwood Research: Alek's Filtering Results
id: post84 We often talk about ensuring control, which in the context of this doc refers to preventing AIs from being able to cause existential problems, even if the AIs attempt to subvert our countermeasures. To better contextualize control, I think it's useful to discuss the main countermeasures (building on my prio...
fe03d74f-334b-4445-ba17-e0c6aea1c3e3
StampyAI/alignment-research-dataset/eaforum
Effective Altruism Forum
Background for "Understanding the diffusion of large language models" *This post is one part of the sequence* [*Understanding the diffusion of large language models*](https://forum.effectivealtruism.org/s/8rYkpiFhbb4HsbzFc)*.  As context for this post, I strongly recommend reading at least the* [*5-minute summary*](ht...
5225af49-06c6-4fac-876d-08de01c70505
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
Seven Strategies for Tackling the Hard Part of the Alignment Problem Thanks to Michael Ripa for feedback.  TL;DR ===== * There are two types of problems that AI systems can have: problems that we can observe during development and problems that we can’t. The hardest part of the alignment problem involves problems we...
fdd0818e-4c18-47c2-bad5-3f584eeab719
trentmkelly/LessWrong-43k
LessWrong
Part of a THINK Meetup Group? We Want to Hear From You! Hello everybody, I've recently started as a volunteer for The High Impact NetworK, an effective altruism group with local chapters mostly in the US and UK. We're aiming to get more people, especially students, interested in effective altruism and enlarging the ne...
8f16a428-85c4-439d-bd5c-a82df78f982f
StampyAI/alignment-research-dataset/arbital
Arbital
Unforeseen maximum An unforeseen maximum of a [utility function](https://arbital.com/p/109) (or other [preference framework](https://arbital.com/p/5f)) is when, e.g., you tell the AI to produce smiles, thinking that the AI will make people happy in order to produce smiles. But unforeseen by you, the AI has an alterna...
6f3f8872-339b-46dd-838f-33b86040f4c3
trentmkelly/LessWrong-43k
LessWrong
$300 for the best sci-fi prompt > EDIT: The competition has ended. Check the results here.      We would like to find the best prompt to make GPT-4 do the following: * write the first chapter of a science fiction novel * the result should be good enough to make seasoned sci-fi readers (us) crave for a continuatio...
5b000aa4-8449-4ff8-8583-0c23ed015fe1
trentmkelly/LessWrong-43k
LessWrong
Questioning GLS-Coherence At the end of my trolling post, I said that naturalistic logical updates stand or fall with the strength of the reflective consistency argument for them, and that this has not been sufficiently articulated yet. Indeed, the choice of GLS-coherence as a model of bounded rationality is highly qu...
ac7d4c30-0690-45b2-a5e5-9478fe804498
StampyAI/alignment-research-dataset/eaforum
Effective Altruism Forum
Why people want to work on AI safety (but don’t) [*Epistemic status*](https://forum.effectivealtruism.org/posts/bbtvDJtb6YwwWtJm7/epistemic-status-an-explainer-and-some-thoughts)*: Based mainly on my own experience and a couple of conversations with others learning about AI safety. It’s very likely that I’m some overl...
fd8e866d-d7c4-49ac-95cc-0571b48a59e5
trentmkelly/LessWrong-43k
LessWrong
I Really Don't Understand Eliezer Yudkowsky's Position on Consciousness From Twitter: > I'd say that I "don't understand" why the people who worry that chickens are sentient and suffering, don't also worry that GPT-3 is sentient and maybe suffering; but in fact I do understand, it's just not a charitable understandin...