id
stringlengths
36
36
source
stringclasses
15 values
formatted_source
stringclasses
13 values
text
stringlengths
2
7.55M
0b18032c-3e6f-4648-abae-b0651f83a6c4
trentmkelly/LessWrong-43k
LessWrong
Testing the FLEDGE Prototype (I work at Google, but not on Chrome. Speaking only for myself.) Chrome has proposed TURTLEDOVE to support remarketing ads without sending your browsing history to advertisers. (discussed in why I work on ads). Since the overall proposal is very complicated and requires large changes, the...
0ebc7d94-259b-40d9-8884-69874d9359ca
StampyAI/alignment-research-dataset/lesswrong
LessWrong
Deceptive Alignment is <1% Likely by Default *Thanks to Wil Perkins, Grant Fleming, Thomas Larsen, Declan Nishiyama, and Frank McBride for feedback on this post. Thanks also to Paul Christiano, Daniel Kokotajlo, and Aaron Scher for comments on the original post that helped clarify the argument. Any mistakes are my own...
1fda0abb-d1d0-45e5-ac8b-c15b1d7293dd
StampyAI/alignment-research-dataset/lesswrong
LessWrong
Rational Agents Cooperate in the Prisoner's Dilemma *This is a crosspost from* [*my blog*](https://outsidetheasylum.blog/rational-agents-cooperate-in-the-prisoners-dilemma/)*. Its intention is to act as a primer on non-causal decision theories, and how they apply to the prisoner's dilemma in particular, for those who ...
9e5827a7-ead3-4059-ad5e-63252d1ab4a6
trentmkelly/LessWrong-43k
LessWrong
A Talmudic Rationalist Cautionary Tale The following is a translated famous passage from the Babylonian Talmud (Bava Metzia 59a–b) that I believe has a good implicit rationalist intepretation. Here is the source to read it yourself.  Everything in [] was added by the publisher of the translation and () were added by ...
e87372c2-e353-4c2f-8808-012866b5b007
trentmkelly/LessWrong-43k
LessWrong
Splitting Concepts For anyone who's gone through some form of primary education, I'd expect numbers to be a tight neural category. Numbers feel like things, they are pretty much the same, and there's little ambiguity over if something is a number or not. So it might come to one's surprise when they learn about the or...
571b495a-7196-4271-9dd5-7e8d8f8763c0
trentmkelly/LessWrong-43k
LessWrong
GPT-powered EA/LW weekly summary Skip to the summaries: LW | EAF Originally posted on the EA forum Zoe Williams used to manually do weekly summaries of the EA Forum and LessWrong, but now she doesn't. Hamish strung together a bunch of google apps scripts, google sheets expressions, graphQL queries, and D3.js to auto...
4eeafd38-cd16-4dcb-86e0-262a2e03d45f
trentmkelly/LessWrong-43k
LessWrong
Eric Michaud on the Quantization Model of Neural Scaling, Interpretability and Grokking Eric is a PhD student in the Department of Physics at MIT working with Max Tegmark on improving our scientific/theoretical understanding of deep learning -- understanding what deep neural networks do internally and why they work s...
99559cd3-adc6-4de8-b486-644b1b596490
trentmkelly/LessWrong-43k
LessWrong
Debate series: should we push for a pause on the development of AI? In March of this year, 30,000 people, including leading AI figures like Yoshua Bengio and Stuart Russell, signed a letter calling on AI labs to pause the training of AI systems. While it seems unlikely that this letter will succeed in pausing the deve...
5b9355ed-188f-4d70-a16e-e3cdd546dd49
trentmkelly/LessWrong-43k
LessWrong
Categorization of Meta-Ethical Theories (a flowchart) Hi folks! Been a LessWrong lurker for a while. Here's a little project I'm excited about, which has been useful in organizing my thoughts on meta-ethics. This piece is a walk-through of an original flowchart (made with lots of help from friends) that categorizes t...
17b7cedb-f1c2-4611-8250-c1194bbc9970
trentmkelly/LessWrong-43k
LessWrong
New LW Meetup: Lund This summary was posted to LW Main on September 18th. The following week's summary is here. New meetups (or meetups with a hiatus of more than a year) are happening in: * First meetup in Lund: 19 October 2015 06:00PM Irregularly scheduled Less Wrong meetups are taking place in: * Oslo Lesswro...
75caa25b-b4c8-4202-ac14-0c04dd4e0e78
trentmkelly/LessWrong-43k
LessWrong
Recursive Self-Improvement Followup to:  Life's Story Continues, Surprised by Brains, Cascades, Cycles, Insight, Recursion, Magic, Engelbart: Insufficiently Recursive, Total Nano Domination I think that at some point in the development of Artificial Intelligence, we are likely to see a fast, local increase in capabil...
1f84835c-2b26-461f-987c-d6bd3c613d50
trentmkelly/LessWrong-43k
LessWrong
How Specificity Works This is Part II of the Specificity Sequence You saw what mayhem we brought forth when we activated the first power of specificity, the power to demolish bad arguments, and hopefully your curiosity is piqued to see what’ll happen when we activate all the other powers. But first, let’s pause here...
c41eb905-5ee3-4e0b-a327-f5c945c94c06
trentmkelly/LessWrong-43k
LessWrong
Examples in Mathematics   After reading Luke's interview with Scott Aaronson, I've decided to come back to an issue that's been bugging me. Specifically, in the answer to Luke's question about object-level tactics, Scott says (under 3): > Sometimes, when you set out to prove some mathematical conjecture, your first...
3234edcc-67ea-47fa-8933-9ee1b834de7b
trentmkelly/LessWrong-43k
LessWrong
Can Ads be GDPR Compliant? I think the online ads ecosystem is most likely illegal in Europe, and as more decisions come out it will become clear that it can't be reworked to be within the bounds of the GDPR. This is a strong claim, but before I get into backing it up here's some background on me: * I'm not a lawy...
d33f746e-4b9e-4339-b23c-edccdeee2711
awestover/filtering-for-misalignment
Redwood Research: Alek's Filtering Results
id: post162 The Open Philanthropy has just launched a large new Request for Proposals for technical AI safety research . Here we're sharing a reference guide , created as part of that RFP, which describes what projects we'd like to see across 21 research directions in technical AI safety. This guide provides an opinio...
4c3f872d-30d3-408d-83f0-9cdc0f9f99b1
trentmkelly/LessWrong-43k
LessWrong
Personal imitation software The question is very simple: could you train a powerful computer program to act like an accurate copy of yourself?  Some LW members may have already tried this, but it would require super powerful software. Even if it's only text based, GPT-3 wouldn't be enough. Maybe GPT-13. If such a pro...
f52c393f-8c3f-495e-b396-f13c3ce4b74e
trentmkelly/LessWrong-43k
LessWrong
Forecasting progress in language models Note: this post was cross-posted to Metaculus over a week ago as part of their new Metaculus journal.  Here, I describe a way of measuring the performance of language models, and extrapolate this measure using publicly available data on benchmarks. The result is a (surprisingly...
6283c66d-06e1-4cbc-996a-d5de323576ee
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
Research request (alignment strategy): Deep dive on "making AI solve alignment for us" We might be able to train AI alignment assistants that massively accelerate/improve the alignment research that gets done. These assistants need not have (strongly) superhuman capabilities or be highly agentic, they just need to be ...
f68aefe3-b738-4c1e-bec1-8c8c71f5833d
trentmkelly/LessWrong-43k
LessWrong
How I accidentally discovered the pill to enlightenment but I wouldn’t recommend it. Main post: http://bearlamp.com.au/how-i-accidentally-discovered-the-pill-to-enlightenment-but-i-wouldnt-recommend-it/ Brief teaser... ---------------------------------------- Eastern enlightenment is not what you think.  I mean, m...
d02dab70-c502-45ef-9db3-d9ec4cfddf90
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
EIS VII: A Challenge for Mechanists Part 7 of 12 in the [Engineer’s Interpretability Sequence](https://www.alignmentforum.org/s/a6ne2ve5uturEEQK7). Thanks to Neel Nanda. I used some very nicely-written code of his from [here](https://bit.ly/neelgrokking). And thanks to both Chris Olah and Neel Nanda for briefly discu...
eba6daa1-c184-481c-b1d8-e0ec5018b215
StampyAI/alignment-research-dataset/lesswrong
LessWrong
What are some good examples of incorrigibility? The idea of corrigibility is roughly that an AI should be aware that it may have faults, and therefore allow and facilitate human operators to correct these faults. I'm especially interested in scenarios where the AI system controls a particular input channel that is sup...
d8555c88-bd3d-4d7a-94d4-4a0d9369204a
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
Emergent Deception and Emergent Optimization *[Note: this post was drafted before Sydney (the Bing chatbot) was released, but Sydney demonstrates some particularly good examples of some of the issues I discuss below. I've therefore added a few Sydney-related notes in relevant places.]* I’ve previously argued that ma...
79554c4f-3c01-471b-b0a8-42d475b6f30c
trentmkelly/LessWrong-43k
LessWrong
Why is bad teaching attached to uni certification? When most things are certified, like coffee or wood or insanity, the stuff is produced by one party, then someone else judges it. University is meant to be a certification of something or another, so a nagging question for all those who can think of a zillion better w...
04107859-392e-4317-9946-79ccde7d94dc
trentmkelly/LessWrong-43k
LessWrong
Against Maturity I remember the moment of my first break with Judaism.  It was in kindergarten, when I was being forced to memorize and recite my first prayer.  It was in Hebrew.  We were given a transliteration, but not a translation.  I asked what the prayer meant.  I was told that I didn't need to know - so long as...
99e699be-58cb-4e19-ab8f-b9e9e1ddf867
StampyAI/alignment-research-dataset/lesswrong
LessWrong
Exploring non-anthropocentric aspects of AI existential safety **Epistemic status:** These are some initial thoughts. **Goal:** Anthropocentric approaches dominate the field of AI existential safety. In particular, AI existential safety is usually formulated in terms of *alignment to human goals and values*. It migh...
f1e26ced-549e-47a2-a147-58647c18dfb8
trentmkelly/LessWrong-43k
LessWrong
Coaching AI: A Relational Approach to AI Safety Epistemic status: Design sketch. This post continues a broader enquiry into trust, affordances and distributed architectures. While those ideas are still in development, this post explores how real-time, relational interventions might offer a complementary path to AI saf...
cf99231a-e81d-4405-b931-b761d3af954c
trentmkelly/LessWrong-43k
LessWrong
Stampy's AI Safety Info - New Distillations #3 [May 2023] Hey! This is another update from the distillers at the AI Safety Info website (and its more playful clone Stampy). Here are a couple of the answers that we wrote up over the last month (May 2023). As always let us know in the comments if there are any questio...
a33b7d50-ae60-4d59-a5b5-64568122ac7b
trentmkelly/LessWrong-43k
LessWrong
Meetup : Bielefeld Meetup, October 9th Discussion article for the meetup : Bielefeld Meetup, October 9th WHEN: 09 October 2012 06:00:00PM (+0200) WHERE: Grill/Bar Verve, Klosterplatz 13, Bielefeld The first meetup in this area ever. There is lots of rationality stuff to talk about, and I will prepare one or two a...
107ca4b4-ff36-4f0a-b5cc-de772651105b
awestover/filtering-for-misalignment
Redwood Research: Alek's Filtering Results
id: post3754 A putative new idea for AI control; index here . This is the first of three articles about limitations and challenges in the concept of corrigibility (see articles 1 and 2 ). The desiderata for corrigibility are: A corrigible agent tolerates, and preferably assists, its operators in their attempts to alte...
dcbdd065-fade-4ad8-a75a-9d5a9e83e3ab
StampyAI/alignment-research-dataset/lesswrong
LessWrong
LM Situational Awareness, Evaluation Proposal: Violating Imitation Motivation ---------- “[Playing the training game](https://www.lesswrong.com/posts/pRkFkzwKZ2zfa3R6H/without-specific-countermeasures-the-easiest-path-to#While_humans_are_in_control__Alex_would_be_incentivized_to__play_the_training_game_)” most likely...
e20d90dd-d818-4751-9d1e-52390f28c3f7
StampyAI/alignment-research-dataset/blogs
Blogs
Counterarguments to the basic AI x-risk case *Katja Grace, 31 August 2022* This is going to be a list of holes I see in the basic argument for existential risk from superhuman AI systems[1](https://aiimpacts.org/counterarguments-to-the-basic-ai-x-risk-case/#easy-footnote-bottom-1-3345 "That is, systems that are some...
434ba879-1790-4c66-9934-174c85522352
trentmkelly/LessWrong-43k
LessWrong
The silence is deafening – Devon Zuegel > Imagine you're at a dinner party, and you're getting into a heated argument. As you start yelling, the other people quickly hush their voices and start glaring at you. None of the onlookers have to take further action—it's clear from their facial expressions that you're being ...
9721fc1e-b186-499d-a2c7-a7671160b4bd
trentmkelly/LessWrong-43k
LessWrong
Cryonic resurrection - an ethical hypothetical At some point in the future, we hope, brains which have been cryonically preserved may be resurrected, by some process of neural reconstruction (most likely either as nanotech, reconstituted wetware, or virtual simulation). Imagine that the technology has just come avail...
6ff6fa3b-daf2-4c5a-96f3-4159c0528420
StampyAI/alignment-research-dataset/blogs
Blogs
On the Sizes of OpenAI API Models OpenAI hasn't officially said anything about their API model sizes, which naturally leads to the question of just how big they are. Thankfully, we can use [eval harness](https://github.com/EleutherAI/lm-evaluation-harness) to evaluate the API models on a bunch of tasks and compare to ...
89c2db9b-034e-4e11-beee-58fa38abe534
StampyAI/alignment-research-dataset/lesswrong
LessWrong
"Heretical Thoughts on AI" by Eli Dourado Abstract ======== Eli Dourado presents the case for scepticism that AI will be economically transformative near term[[1]](#fnwscp0ijpsn). For a summary and or exploration of implications, skip to "[My Take](https://www.lesswrong.com/posts/nWCokT9xbrY4p98co/heretical-thoughts...
488ca8f3-d34b-47be-91c3-045a313756e3
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
Announcing AXRP, the AI X-risk Research Podcast Happy holidays! Today, I’m launching AXRP (pronounced axe-urp), the AI X-risk Research Podcast. The episodes involve me interviewing a researcher about a paper they’ve written, talking about the ideas in the paper and why they matter. The first three guests are [Adam Gle...
4ab107b3-f939-4a5a-8854-c88344204e70
trentmkelly/LessWrong-43k
LessWrong
Post Request Thread This thread is another experiment roughly in the vein of the Boring Advice Repository and the Solved Problems Repository. There are some topics I'd like to see more LW posts on, but I feel underqualified to post about them relative to my estimate of the most qualified LWer on the topic. I would gu...
3b46ee01-7f83-47dc-bd04-fd17c4965181
trentmkelly/LessWrong-43k
LessWrong
[SEQ RERUN] Higher Purpose Today's post, Higher Purpose was originally published on 23 January 2009. A summary (taken from the LW wiki):   > Having a Purpose in Life consistently shows up as something that increases stated well-being. Of course, the problem with trying to pick out "a Purpose in Life" in order to mak...
89274161-0bd7-4e3c-86f0-b327ed3451d5
StampyAI/alignment-research-dataset/lesswrong
LessWrong
AGI-level reasoner will appear sooner than an agent; what the humanity will do with this reasoner is critical The recent advancements in language models, [Gato](https://www.deepmind.com/publications/a-generalist-agent), [Minerva](https://ai.googleblog.com/2022/06/minerva-solving-quantitative-reasoning.html), etc. sugg...
13335ba4-6850-4694-b08e-2291c7a44e6f
trentmkelly/LessWrong-43k
LessWrong
Rationality 101 videotaped presentation with link to slides in description (from our LessWrong meetup introductory event)
324f7e33-c4d9-419f-b8f9-8b82e18caff6
trentmkelly/LessWrong-43k
LessWrong
System 2 as working-memory augmented System 1 reasoning The terms System 1 and System 2 were originally coined by the psychologist Keith Stanovich and then popularized by Daniel Kahneman in his book Thinking, Fast and Slow. Stanovich noted that a number of fields within psychology had been developing various kinds of ...
6b428d54-5984-4648-a36e-dd3623657267
StampyAI/alignment-research-dataset/lesswrong
LessWrong
Grinding slimes in the dungeon of AI alignment research This post aims to describe some differing views on challenges in AI alignment research (as I understand them) using an analogy to a video game with multiple levels. I then draw some conclusions about the potential upsides and downsides of doing and publishing d...
1e973d67-39e0-4384-9a8f-8e0b4af2f065
trentmkelly/LessWrong-43k
LessWrong
To what extent is AI safety work trying to get AI to reliably and safely do what the user asks vs. do what is best in some ultimate sense? Trying to get a rough estimate for some related research I’m doing. Specifically, I’m wondering if anyone could give a rough percentage of current AI safety work primarily targete...
87a21c72-2440-4e52-84ba-204a6f6dc682
trentmkelly/LessWrong-43k
LessWrong
Switching to a Yamaha P-121 Keyboard The keyboard is a bit of an awkward instrument to travel with. It's quite large, to the point that you have to give up at least one seat in a typical car. What makes this especially frustrating is that I don't actually use the whole 88 keys: The very lowest notes tend to be b...
a9cbd0b3-1c35-4532-928c-2ef910335cab
trentmkelly/LessWrong-43k
LessWrong
Vanity and Ambition in Mathematics In my time in the mathematical community I've formed the subjective impression that it's noticeably less common for mathematicians of the highest caliber to engage in status games than members of the general population do. This impression is consistent with the modesty that comes acr...
6fb7888e-8419-41fc-9383-a777e8f60250
trentmkelly/LessWrong-43k
LessWrong
Rationality and Decorating From Wired: > Thomas’s sumptuous designs led people to spend as they’d never spent before, and, in the years since the Bellagio was completed, research has supported the psychological assumptions that went into its creation. Karen Finlay is a professor at the University of Guelph, in Ontari...
e0d13b5e-9dab-4d67-a515-9371430ca957
trentmkelly/LessWrong-43k
LessWrong
James Martin Postdoctoral Research Fellowship: Socio-economic Impacts of Technological Change We are pleased to announce a new vacancy at the Programme on the Impacts of Future Technology. Please forward to any who would be interested.   James Martin Postdoctoral Research Fellowship: Socio-economic Impacts of Techn...
b9374b77-c370-4159-8bb0-1636250c289e
trentmkelly/LessWrong-43k
LessWrong
Complement Luke's Mega-Course for Aspiring Philosophers Luke has mentioned much of the research that aspiring philosophers ought to read here. In fact, he delineated a basis upon which good philosophy can be build, a worldview brought by science and experimentation that relates to, and informs, the kinds of facts whi...
30e20b06-55fe-4318-a10c-9a6ca028d095
trentmkelly/LessWrong-43k
LessWrong
Training Regime Day 5: TAPs Introduction TAP stands for Trigger Action Pattern/Planning. "Pattern" is a descriptive term, i.e. it's describing what's actually happening. "Planning" is what I hope to teach you how to do, i.e. how to create certain Trigger Action Patterns in yourself. In the literature, these are somet...
4cf2f418-24bb-4c64-b23f-4fab1c4d6f5c
trentmkelly/LessWrong-43k
LessWrong
[Video] How having Fast Fourier Transforms sooner could have helped with Nuclear Disarmament - Veritasium Veritasium presents an argument that if we'd had Fast Fourier Transforms sooner, this would have made it possible to detect underground nuclear tests without on-site audits, somewhat better nuclear non-proliferati...
91c730bc-0150-46a7-b3df-0def5e1f2e7d
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
[AN #134]: Underspecification as a cause of fragility to distribution shift Alignment Newsletter is a weekly publication with recent content relevant to AI alignment around the world. Find all Alignment Newsletter **[resources here](http://rohinshah.com/alignment-newsletter/)**. In particular, you can look through **[...
49fc2806-492b-4de5-8089-ac7a8cce31f5
trentmkelly/LessWrong-43k
LessWrong
Many Reasons I'm here to teach you a new phobia. The phobia concerns phrases such as "for many reasons". Rational belief updating is a random walk without drift. If you expect your belief to go up (down) in response to evidence, you should instead make it go up (down) right now. If you're not convinced, read about co...
d0ff7528-19cf-475d-a1b1-d5c98baa0d02
trentmkelly/LessWrong-43k
LessWrong
Welcome to Baltimore Lesswrong 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...
1f029155-626a-484c-b0f7-472d98206969
trentmkelly/LessWrong-43k
LessWrong
[SEQ RERUN] The Quotation is not the Referent Today's post, The Quotation is not the Referent was originally published on 13 March 2008. A summary (taken from the LW wiki):   > It's very easy to derive extremely wrong conclusions if you don't make a clear enough distinction between your beliefs about the world, and ...
573d51f8-ffa7-41a3-8711-420c3a08bc3c
trentmkelly/LessWrong-43k
LessWrong
Do you do weekly or daily reviews? What are they like? * What system(s) do you use to keep yourself organized and working toward your goals? I'm interested in technologies, but more interested in what ontology you use to organize your tasks, events, goals, and plans. * Did you build your system slowly over time, or ...
2725e2d0-4780-4469-af06-7bf8ca0625d9
trentmkelly/LessWrong-43k
LessWrong
Sophie Grouchy on A Build-Break Model of Cooperation (This blog is partially a response to/ built on the portion of this essay covering Stag Hunts and White/ Black Knights.) When we’re making models of the world we want to simplify things as much as possible. But if you simplify things too much your model is no good ...
1d3fffc5-c294-4657-a380-356ac8f2f304
StampyAI/alignment-research-dataset/eaforum
Effective Altruism Forum
Max Tegmark: Risks and benefits of advanced artificial intelligence --- *How can we benefit from modern AI while avoiding the risks? Max Tegmark explores this question in this EA Global: San Francisco 2016 talk.* *In the future, we may post a transcript for this talk, but we haven't created one yet. If you'd like to...
1d4cc0d7-5b30-47f1-8bf8-c41a5d115e01
trentmkelly/LessWrong-43k
LessWrong
Taking Social Initiative (This is a post from a personal daily blogging project, on social skills, taking social initiative, and "networking without being a terrible person". And I think this is relevant to the interests of LessWrong readers!) Introduction A theme I’ve touched on pretty heavily in previous posts is ...
c5a5ca59-9515-48de-945a-2541f099f1f0
trentmkelly/LessWrong-43k
LessWrong
Zoom Technologies, Inc. vs. the Efficient Markets Hypothesis The efficient markets hypothesis (or EMH for short) is the idea "that asset prices reflect all available information". Price changes in a liquid market are understood to be unpredictable—anti-inductive. Suppose some stock has the ticker symbol LW. If you wan...
91300812-668f-43ef-a275-54ceb7131396
StampyAI/alignment-research-dataset/eaforum
Effective Altruism Forum
The Navigation Fund launched + is hiring a program officer to lead the distribution of $20M annually for AI safety! Full-time, fully remote, pay starts at $200k New foundation, funded by billionaire [Jed McCaleb,](https://www.forbes.com/profile/jed-mccaleb/?sh=7adb975476bf) led by [David Coman-Hidy](https://www.linked...
d8d38505-32e4-4220-89ce-f01e72516faa
StampyAI/alignment-research-dataset/arbital
Arbital
Cyclic group summary(technical): A [group](https://arbital.com/p/3gd) $G$ is **cyclic** if it has a single [generator](https://arbital.com/p/generator_mathematics): there is one [element](https://arbital.com/p/element_mathematics) $g$ such that every element of the group is a [power](https://arbital.com/p/) of $g$. ...
f9d08524-49f8-4bb7-8ddf-e93e2c7f94d9
trentmkelly/LessWrong-43k
LessWrong
When to assume neural networks can solve a problem Note: the original article has been split into two since I think the two points were only vaguely related, I will leave it as is here, since I'd rather not re-post stuff and I think the audience on LW might see the "link" between the two separate ideas presented here....
1d1a6688-ced2-4386-a238-cbe51870dcd9
trentmkelly/LessWrong-43k
LessWrong
One: a story The earliest living creatures were single cells, battered around by their environments. Over time they clumped together, to form multi-celled organisms capable of complex self-regulation. Those organisms formed mating pairs, to better spread their genes; then they formed social groups, to protect and shel...
fa1e6ace-dbac-440f-82dc-6d61d278279b
trentmkelly/LessWrong-43k
LessWrong
What do you read? What would "desk archeology" produce? I would like to do a kind of poll: Which books/articles do you read now, which ones are on your reading list? What would a "desk archeologist" find when digging up your desk? Some links: "A Stratigraphic Analysis of Desk Debris", "Are we able to think clearl...
1c07887e-4673-4a4b-a540-5d2bf49ad91e
trentmkelly/LessWrong-43k
LessWrong
Keyboard Gremlins I had a great time playing Amherst last night with Cecilia, except for the keyboard gremlins. I just got a new keyboard (more on that later!) and it worked great when I tried it at home. At the dance, however, there were strange crackles. Even weirder, when I played the highest notes on the piano, I ...
b3d19103-f23f-4e0e-b09c-b13d9d6fa05c
StampyAI/alignment-research-dataset/special_docs
Other
Quantifying Differences in Reward Functions. Published as a conference paper at ICLR 2021 QUANTIFYING DIFFERENCES IN REWARD FUNCTIONS Adam Gleave1,2Michael Dennis1Shane Legg2Stuart Russell1Jan Leike3 1UC Berkeley2DeepMind3OpenAI gleave@berkeley.edu ABSTRACT For many tasks, the reward function is inaccessible to intro...
69be2a2c-06d6-4381-bc84-c37ab13a81d2
trentmkelly/LessWrong-43k
LessWrong
Fake qualities of mind There’s a thing where you’d like to have one “quality of mind”, but it’s not available, but you substitute it with a kind of a fake or alternative version of the same. Which is fine as long as you realize you’re doing it, but becomes an issue if you forget that what’s happening. For example, yo...
8c34109f-537c-4972-8e91-f6fa1ead33dd
StampyAI/alignment-research-dataset/blogs
Blogs
2016 ESPAI Narrow AI task forecast timeline This is an interactive timeline we made, illustrating the median dates when respondents said they expected a 10%, 50% and 90% chance of different tasks being automatable, in the [2016 Expert Survey on progress in AI](http://aiimpacts.org/2016-expert-survey-on-progress-in-ai/...
db872f39-71da-49d5-ab62-014dff9e3f5f
StampyAI/alignment-research-dataset/lesswrong
LessWrong
Lightning Post: Things people in AI Safety should stop talking about This is experimenting with a new kind of post which is meant to convey a lot of ideas very quickly, without going into much detail for each.  Things I wish people in AI Safety would stop talking about ================================================...
15ad57c8-2ed2-4dc3-a150-fd2dd7d6e676
trentmkelly/LessWrong-43k
LessWrong
What does GPT-3 understand? Symbol grounding and Chinese rooms With thanks to Rebecca Gorman for helping develop these ideas and this post. Tricking GPT-3 I recently pulled a mean trick on GPT-3. I gave it the following seed text to extend: > These are the instructions to be followed by any person or algorithm aimi...
4c57ac82-75f4-4038-b4ee-16e7329fc299
StampyAI/alignment-research-dataset/blogs
Blogs
Daniel Roy on probabilistic programming and AI ![Daniel Roy portrait](http://intelligence.org/wp-content/uploads/2014/08/Roy__h692_w540.jpg) [Daniel Roy](http://danroy.org/) is an Assistant Professor of Statistics at the University of Toronto. Roy earned an S.B. and M.Eng. in Electrical Engineering and Computer Scienc...
31a95446-a73b-48d5-aed7-7014f0b6525b
trentmkelly/LessWrong-43k
LessWrong
Why does anxiety (?) make me dumb? Something about anxiety and my brain shutting off. Disclaimer: I’m grappling with something I hardly understand at all, and I’m not all that used to the exercise. One of our goals here is to improve our thinking. That’s fairly obvious. And I have another obvious statement: The proc...
498bafe6-c81f-4bde-8e48-be07014c527b
trentmkelly/LessWrong-43k
LessWrong
A concise version of “Twelve Virtues of Rationality”, with Anki deck In an effort to internalise the Twelve Virtues of Rationality, I created an Anki deck. It's already been done, so the reason I'm posting is to share a condensed version of the article (created as a side effect of my making the deck). Hopefully it wi...
e1974cef-85eb-49f5-918a-0d14871f7faa
trentmkelly/LessWrong-43k
LessWrong
Memory in the microtubules A recent article in PloS Computational Biology suggests that memory is encoded in the microtubules. "Signaling and encoding in MTs and other cytoskeletal structures offer rapid, robust solid-state information processing which may reflect a general code for MT-based memory and information pro...
b4374378-c23f-47c3-888f-1e958efe3bd4
trentmkelly/LessWrong-43k
LessWrong
PredictionBook: A Short Note   New Updates PredictionBook has been updated and the speed improvements are massive. I was considering abandoning it because navigating the site was so slow (especially on my Android smartphone), but now I'm here to stay. Since there is no news feed on the website itself, it seemed appr...
92e75a36-9a3c-4d7d-8bab-868f14a04180
StampyAI/alignment-research-dataset/lesswrong
LessWrong
The Utility and Solubility of Anthropic Capture *Epistemic status: Heavy speculation, but I find the arguments convincing and formalizing them has lowered my credence in doom due to anthropic capture-related failure modes.* TL;DR ===== * Anthropic capture (AC)-like behavior can probably be induced by beliefs other t...
93d273fc-db01-4286-94f1-49367470b76b
trentmkelly/LessWrong-43k
LessWrong
Covid 3/31/2022: More of the Same BA.2 became the majority strain in the United States this week, as expected. Mostly things are continuing as expected, with unsurprising news on a variety of fronts. The biggest surprise is that the second booster shot was approved. Given I expect to be asked about that every so often...
f6be4bb3-11ea-4f85-ba89-de6c92bde011
trentmkelly/LessWrong-43k
LessWrong
Marketing Failure The amount of money spent worldwide on advertisements is astronomical, reaching about 550 billion dollars annually and is expected to continue growing rapidly. In 2016, the top 20 companies with the biggest annual advertising budget spend between 2.7 to 8.3 billion dollars annually. Is this a result...
42783095-9cdf-4e23-bb63-953a2ca9d5cc
trentmkelly/LessWrong-43k
LessWrong
Why is the surprisingly popular answer correct? In Nature, there's been a recent publication arguing that the best way of gauging the truth of a question is to get people to report their views on the truth of the matter, and their estimate of the proportion of people who would agree with them. Then, it's claimed, the...
4167ea58-d9dc-4ec4-bd8f-d3c86fd755cf
trentmkelly/LessWrong-43k
LessWrong
Vacancy at the Future of Humanity Institute: Academic Project Manager The Future of Humanity Institute* recently secured funding for a new Research Collaboration with Amlin Insurance focusing on systemic risks associated with risk modelling. We're looking for someone with an academic background or interests and manage...
1ca95440-1903-4a72-8acb-4e8f35bff78c
trentmkelly/LessWrong-43k
LessWrong
[SEQ RERUN] Building Weirdtopia Today's post, Building Weirdtopia was originally published on 12 January 2009. A summary (taken from the LW wiki):   > Utopia and Dystopia both confirm the moral sensibilities you started with; whether the world is a libertarian utopia of government non-interference, or a hellish dyst...
3b2759b1-0840-4d58-9604-79e00caa8f08
trentmkelly/LessWrong-43k
LessWrong
How I repeatedly failed to use Tobit modelling on censored data This is a description of my work on a data science project, lightly obfuscated and fictionalized to protect the confidentiality of the organization I handled it for (and also to make it flow better). I focus on the high-level epistemic/mathematical issues...
73b60a66-8946-469d-a9a5-c3997887db53
trentmkelly/LessWrong-43k
LessWrong
"Slow" takeoff is a terrible term for "maybe even faster takeoff, actually" For a long time, when I heard "slow takeoff", I assumed it meant "takeoff that takes longer calendar time than fast takeoff." (i.e. what is now referred to more often as "short timelines" vs "long timelines."). I think Paul Christiano populari...
caccab7a-292e-4c76-bf18-d707a6a04b4f
trentmkelly/LessWrong-43k
LessWrong
Datapoint: median 10% AI x-risk mentioned on Dutch public TV channel I am Dutch, so wanted to share this as a datapoint regarding public perception of existential AI risk, since it probably won't be noticed here otherwise. 3 days ago (23th march 2023) a Dutch AI "science communicator" who regularly appears on Dutch t...
de10c0b2-525b-491d-ada8-dbfcecba7eaa
trentmkelly/LessWrong-43k
LessWrong
Constructing Neural Network Parameters with Downstream Trainability I have recently done some preliminary experiments related to mechanistic interpretability. It seems researchers in this subfield often post their results in the AI Alignment Forum / LessWrong Forum (e.g. causal scrubbing, emergent world representation...
a922a975-d4a0-4ece-bf8a-c3e86f481b36
trentmkelly/LessWrong-43k
LessWrong
Coordination attempt: Facilitating conversational splinters in Gathertown Someone once wrote a LessWrong post saying that people, individually polled, will prefer small group conversations at events – but in practice form large clumps where the majority of people are disengaged but can't coordinate to leave in chunks,...
d250d976-bd9b-47e8-a070-06cc5c12c648
trentmkelly/LessWrong-43k
LessWrong
Nokens: A potential method of investigating glitch tokens Summary * We can probably replicate glitch token-like behaviour by simply sampling random points in the embedding space, which I call 'nokens'. * We may be able to find a large number of tokens which cause unwanted behaviour more easily than with the ' petert...
9e290b71-499f-48a5-b112-615aa60f2843
StampyAI/alignment-research-dataset/lesswrong
LessWrong
[Fiction] IO.SYS <https://www.datapacrat.com/IO.SYS.html> --- [IO.SYS](https://www.datapacrat.com/IO.SYS.html) ================================================ #### [by DataPacRat](https://blog.datapacrat.com/about/) --- If this has all been a VR, I’d like to be connected to reality when I wake up, please....
f451faf6-176d-4d5b-bbc9-9554a2ca9d88
trentmkelly/LessWrong-43k
LessWrong
Acausal Trade and the Ultimatum Game In the ultimatum game you have two participants. Let's call them Adam and Becky. You offer Adam 100 dollars and tell him he has to split it with Becky. Adam makes an offer to Becky, and Becky can either accept or reject the offer. If she rejects the offer they both get nothing, oth...
ec8bb5f0-46ef-4007-986c-4294daaa4834
trentmkelly/LessWrong-43k
LessWrong
Joseph Bloom on choosing AI Alignment over bio, what many aspiring researchers get wrong, and more (interview) This is a section of an interview I conducted with my brother, Joseph. We'll soon publish the interview sections about his research on Decision Transformers and mech interp opinions. Preview Snippets > I do...
2ad29717-b2ff-410d-8cf9-a1722e091aea
StampyAI/alignment-research-dataset/youtube
Youtube Transcripts
9 of the Best Bing (GPT 4) Prompts Sam Altman tweeted that writing a really great prompt for a chatbot Persona is an amazingly High leverage school and an early example of programming in a little bit of an actual language this video is about proving that that's correct with these game-changing prompts I was genuinely ...
bc7caa83-dd98-42cd-bdc8-e3ded853ee6e
trentmkelly/LessWrong-43k
LessWrong
Wirehead your Chickens TL;DR: If you care about farm animal welfare, work on minimizing actual animal suffering, not a human proxy for animal suffering. Epistemic status: had a chat about this with a couple of local EA enthusiasts who attended the EA Global 2018 in San-Francisco, and apparently this was not high on t...
997119a6-f838-4b7e-a2c6-f9e851dc4806
trentmkelly/LessWrong-43k
LessWrong
What is Your Goal Hive? This is a post about plans, goals, and the problems with trying to build something positive out of pieces which are negative.
02dec6fa-f91e-4ad1-a0ef-7d166f475616
trentmkelly/LessWrong-43k
LessWrong
[Preprint for commenting] Fighting Aging as an Effective Altruism Cause "Fighting Aging as an Effective Altruism Cause: A Model of the Impact of the Clinical Trials of Simple Interventions" Abstract: The effective altruism movement aims to save lives in the most cost-effective ways. In the future, technology will all...
20a280b4-1e87-4e8c-87e8-cc5309b32767
trentmkelly/LessWrong-43k
LessWrong
Meetup : Cambridge, MA first-Sundays meetup Discussion article for the meetup : Cambridge, MA first-Sundays meetup WHEN: 01 April 2012 02:00:00PM (-0400) WHERE: 25 Ames St, Cambridge, MA 02139 We're meeting at our usual place at MIT this Sunday, April 1st at 2pm. Since this is taking place on All Fools' Day, pranki...
546974c2-bc99-4785-a054-79c585eb0a51
trentmkelly/LessWrong-43k
LessWrong
Predictors exist: CDT going bonkers... forever I've been wanting to get a better example of CDT (causal decision theory) misbehaving, where the behaviour is more clearly suboptimal than it is in the Newcomb problem (which many people don't seem to accept as CDT being suboptimal), and simpler to grasp than Death in Dam...
812cfe2b-ed31-4fc0-bde2-0367b43427fe
trentmkelly/LessWrong-43k
LessWrong
Deep neural networks are not opaque. In his List of Lethalities, Eliezer writes that "matrices [in neural networks] are opaque". But I think our understanding of how neural networks actually work has improved significantly last year, with the deep learning theory book. I will briefly outline the general idea of the bo...
166acec0-47ae-4abf-80ac-32aa94e1880d
trentmkelly/LessWrong-43k
LessWrong
"Natural is better" is a valuable heuristic tl;dr Until we understand how complex and chaotic systems[1] in nature work, "natural is better" is a valuable heuristic. It applies wherever science does not understand things well enough, especially when the stakes are high. If you throw a stick into a forest it will biod...
844c4395-e43a-4174-9861-309969f615d1
trentmkelly/LessWrong-43k
LessWrong
Call for cognitive science in AI safety Epistemic status: High expected utility, but also very high variance The more I realise that AI take off is something that actually might happen, the more I am pulled towards this problem: * What are human preferences really? * What is the generator of human preferences? ...
6e9e2474-6a12-40c1-8d10-71d36ab14861
trentmkelly/LessWrong-43k
LessWrong
What are the actual arguments in favor of computationalism as a theory of identity? A few months ago, Rob Bensinger made a rather long post (that even got curated) in which he expressed his views on several questions related to personal identity and anticipated experiences in the context of potential uploading and emu...