id stringlengths 36 36 | source stringclasses 15
values | formatted_source stringclasses 13
values | text stringlengths 2 7.55M |
|---|---|---|---|
a760893d-323d-418c-8002-6b86096c4b6f | trentmkelly/LessWrong-43k | LessWrong | Five Suggestions For Rationality Research and Development
1. Make field work the bulk of what you do. For every hour you spend engaging with other people's thoughts or reflecting on your own thoughts, spend at least two hours, and ideally two days, trying to directly observe whatever it is you're studying.
2. Add ... |
f16a721b-006a-42ad-97aa-fcb8e817d0b7 | trentmkelly/LessWrong-43k | LessWrong | 2. Why intuitive comparisons of large-scale impact are unjustified
We’ve seen so far that unawareness leaves us with gaps in the impartial altruistic justification for our decisions. It’s fundamentally unclear how the possible outcomes of our actions trade off against each other.
But perhaps this ambiguity doesn’t ma... |
b171864b-e875-454b-a741-63adca453e2a | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | How does Gradient Descent Interact with Goodhart?
I am confused about how gradient descent (and other forms of local search) interact with Goodhart's law. I often use a simple proxy of "sample points until I get one with a large .mjx-chtml {display: inline-block; line-height: 0; text-indent: 0; text-align: left; text-... |
4242fb94-2cc6-49ac-a319-1088a6d6880e | StampyAI/alignment-research-dataset/arxiv | Arxiv | Explaining Neural Scaling Laws
1 Scaling Laws for Neural Networks
-----------------------------------
For a large variety of models and datasets, neural network performance has been empirically observed to scale as a power-law with model size and dataset size [[1](#bib.bib1), [2](#bib.bib2), [3](#bib.bib3), [4](#bi... |
19a80c16-c16c-44c8-9f96-234f458a2ee3 | trentmkelly/LessWrong-43k | LessWrong | New LW Meetup: Auckland
This summary was posted to LW Main on March 7th. The following week's summary is here.
New meetups (or meetups with a hiatus of more than a year) are happening in:
* Auckland Preliminary Meetup: 08 March 2014 02:00PM
Irregularly scheduled Less Wrong meetups are taking place in:
* [Atlanta... |
7b25d5c6-c56c-4630-a0fd-a9a2f9b50171 | trentmkelly/LessWrong-43k | LessWrong | Singularity Institute - mainstream media exposure in Australia
The Age, one of the most widely read Australian newspapers, has published an article on "the Singularity". Mostly features Jaan Tallinn, but SI gets a mention and a link. Comments section makes for interesting reading.
http://www.theage.com.au/technolo... |
61c7aa90-88b2-4708-9683-47e5309e0739 | StampyAI/alignment-research-dataset/eaforum | Effective Altruism Forum | AISN #12: Policy Proposals from NTIA’s Request for Comment and Reconsidering Instrumental Convergence
Welcome to the AI Safety Newsletter by the [Center for AI Safety](https://www.safe.ai/). We discuss developments in AI and AI safety. No technical background required.
Subscribe [here](https://newsletter.safe.ai/subs... |
710fc6fd-464a-41e4-871f-ae4e6c9185a5 | trentmkelly/LessWrong-43k | LessWrong | ISO: Name of Problem
I'm looking for a name for a problem. I expect it already has one, but I don't know what it is.
The problem: suppose we have an AI trying to learn what people want - e.g. an IRL variant. Intuitively speaking, we point at a bunch of humans and say “figure out what they want, then do that”. A few p... |
4e1f5dba-d105-4661-ae5d-1578c117bc4b | StampyAI/alignment-research-dataset/blogs | Blogs | The need to scale MIRI’s methods
Andrew Critch, one of the new additions to MIRI’s research team, has taken the opportunity of MIRI’s [winter fundraiser](https://intelligence.org/2015/12/01/miri-2015-winter-fundraiser/) to write on his [personal blog](http://acritch.com/miri-scaling/) about why he considers MIRI’s wor... |
4e52f3cf-1b47-4a33-b82f-282fe81ef1bd | trentmkelly/LessWrong-43k | LessWrong | Suggestion: site-wide taboos
Every so often, someone on Less Wrong uses a word wrong.
What does it mean to use a word wrong? Can't we use language however we want, as long as we manage to successfully communicate? Well, yes, we can, but we shouldn't. Jargon terms, in particular, are used by professionals in a certain... |
5f44af28-734d-45d9-b9bf-31192bfd87d0 | StampyAI/alignment-research-dataset/eaforum | Effective Altruism Forum | There are no coherence theorems
**Introduction**
================
For about fifteen years, the AI safety community has been discussing [coherence arguments](https://www.lesswrong.com/tag/coherence-arguments). In papers and posts on the subject, it’s often written that there exist 'coherence theorems' which state that... |
6bc73a6f-4afd-4685-b56e-d0b8c9dbae8b | StampyAI/alignment-research-dataset/aisafety.info | AI Safety Info | Is expecting large returns from AI self-improvement just following an exponential trend line off a cliff?
Blindly following the trendlines while forecasting technological progress is certainly a risk (affectionately known in AI circles as “pulling a Kurzweil”), but sometimes taking an exponential trend seriously is th... |
7fa11d10-d2b8-41ca-b2e6-6565274e7385 | trentmkelly/LessWrong-43k | LessWrong | I Will Pay $500 To Anyone Who Can Convince Me To Cancel My Cryonics Subscription
Background:
On the most recent LessWrong readership survey, I assigned a probability of 0.30 on the cryonics question. I had previously been persuaded to sign up for cryonics by reading the sequences, but this thread and particularly thi... |
01df92a4-2814-44ed-8902-4a01a0a22ef6 | StampyAI/alignment-research-dataset/lesswrong | LessWrong | Is "Strong Coherence" Anti-Natural?
Related:
* [Contra "Strong Coherence"](https://www.lesswrong.com/posts/AdGo5BRCzzsdDGM6H/contra-strong-coherence)
* [why assume AGIs will optimize for fixed goals](https://www.lesswrong.com/posts/dKTh9Td3KaJ8QW6gw/why-assume-agis-will-optimize-for-fixed-goals)
* [Why The Focus on E... |
b7f8bd05-ea81-4563-b3ee-7ccb2d9d6b68 | trentmkelly/LessWrong-43k | LessWrong | How does consciousness interact with architecture?
Eliezer suggested recently that we could determine if an AI were conscious by removing all consciousness-relating data out of its system, and then describe consciousness to it and ask if it recognized it.
I think I find this intuition difficult, because in my mental ... |
8cc5f03b-11b0-4918-b916-aa219882cbef | trentmkelly/LessWrong-43k | LessWrong | The Center For Open Science makes $1.3mm available to validate 50 cancer studies.
> Yesterday, they announced that The Center For Open Science was making $1.3mm available, via ScienceExchange, to reproduce and validate 50 important cancer biology studies.
>
> I am excited about this for a bunch of reasons; 1) reprod... |
270947fb-55cf-41c5-ba7d-fddbdf3b63a3 | trentmkelly/LessWrong-43k | LessWrong | New UChicago Rationality Group
Hey y'all!
I just started a rationality group on the UChicago campus and wanted to post it here to advertise it to UChicago-affiliated LessWrong readers. We've had a couple meetings so far which have been great, and I'm excited for more! A few more things:
(1) You can join the email ... |
92529718-0d8a-459e-98ba-3fcbf7171e06 | trentmkelly/LessWrong-43k | LessWrong | Building toward a Friendly AI team
Series: How to Purchase AI Risk Reduction
A key part of SI's strategy for AI risk reduction is to build toward hosting a Friendly AI development team at the Singularity Institute.
I don't take it to be obvious that an SI-hosted FAI team is the correct path toward the endgame of hum... |
c502b125-0167-4108-a682-b4172f020059 | trentmkelly/LessWrong-43k | LessWrong | Meetup : Washington, D.C.: Visit Museums
Discussion article for the meetup : Washington, D.C.: Visit Museums
WHEN: 14 June 2015 03:00:00PM (-0400)
WHERE: National Portrait Gallery
We'll be congregating at the courtyard of the National Portrait Gallery from 3:00 - 3:30 and then immediately leaving.
Weather permitti... |
4b1ffc72-e932-4bdd-a0cc-d1b56b6b8b7a | trentmkelly/LessWrong-43k | LessWrong | vaccination research/reading
Vaccination is probably one of the hardest topics to have a rational discussion about. I have some reason to believe that the author of http://whyarethingsthisway.com/2014/10/23/the-cdc-and-cargo-cult-science/ is someone interested in looking for the truth, not winning a side - at the very... |
7a2e7e25-a6c0-49c8-9a14-15b2ca9d20fb | trentmkelly/LessWrong-43k | LessWrong | In memory of Louise Glück
(Cross-posted from my website. Podcast version here, or search "Joe Carlsmith Audio" on your podcast app.)
Louise Glück, circa 1977. Image source here.
Louise Glück, one of my favorite poets, died yesterday.
I took a writing workshop with her back in 2009. I remember the force and precision... |
6618e9c3-4138-48e1-bef0-27d69300bc48 | trentmkelly/LessWrong-43k | LessWrong | Blue is Arbitrary
When I was 13, the coolest thing in the world was Seto Kaiba summoning a Blue-Eyes White Dragon.
The name "Blue-Eyes White Dragon" is a translation of the Japanese "青眼の白龍".
* "龍" means "dragon".
* "白" means "white".
* "の" turns the previous word into an adjective.
* "眼" means "eyes".
* "青" h... |
179ced49-7aaf-4f4a-a3d4-d4c547dc9ca3 | trentmkelly/LessWrong-43k | LessWrong | The Track Record of Futurists Seems ... Fine
I've argued that the development of advanced AI could make this the most important century for humanity. A common reaction to this idea is one laid out by Tyler Cowen here: "how good were past thinkers at predicting the future? Don’t just select on those who are famous beca... |
1d3e922c-118d-4538-9559-6efa595e4904 | StampyAI/alignment-research-dataset/agisf | AGI Safety Fund | “[Week 7] ’12 tentative ideas for US AI policy’ by Luke Muehlhauser” by Luke Muehlhauser
Published: April 17, 2023 | by [Luke Muehlhauser](/about/team/luke-muehlhauser)
About two years ago, I [wrote](https://www.openphilanthropy.org/research/our-ai-governance-grantmaking-so-far/) that “it’s difficult to know which ‘... |
70d8473c-b05e-4e00-a7f7-8730b95206fc | awestover/filtering-for-misalignment | Redwood Research: Alek's Filtering Results | id: post1864
Thanks to Tristan Cook, Emery Cooper, Nicolas Macé and Will Payne for discussions, advice, and comments. All views are my own. I think this post is best viewed as an effort to point at a potentially interesting research direction. The specific arguments I make centered around inducing mild behaviour in em... |
c01f9ee9-184b-4da4-9b3f-34b0fa32e064 | trentmkelly/LessWrong-43k | LessWrong | Potential High-Leverage and Inexpensive Mitigations (which are still feasible) for Pandemics
(Crossposted from EA Forum) A little over a year ago, I started a collaboration with David Denkenberger essentially trying to answer the question of what can we do to prepare for pandemics, focusing on things that are Importan... |
2431717b-36f2-42ca-ad54-5a896ae4666f | trentmkelly/LessWrong-43k | LessWrong | [Link] Robots Program People
On Scott Adams' Blog: Robots Program People:
> It won’t be long before all new drugs are discovered by robots. This start-up is an example of that trend.
> And it won’t be long before IBM’s Watson can diagnose and prescribe treatments better than any human doctor.
> Put those two trends t... |
5f51edc8-2d22-408d-a346-e2610e529378 | trentmkelly/LessWrong-43k | LessWrong | Meetup : Pittsburgh: Semantic Stopsigns
Discussion article for the meetup : Pittsburgh: Semantic Stopsigns
WHEN: 08 August 2012 06:00:00PM (-0400)
WHERE: Eatunique, Craig Street, Pittsburgh
Suggested reading: http://lesswrong.com/lw/it/semantic_stopsigns/ Eatunique serves salads, wraps, sandwiches, drinks, and that... |
6b9f8055-28ed-41e2-bbcf-184073d1d5da | trentmkelly/LessWrong-43k | LessWrong | Habits 101: Techniques and Research
|
8ba8d428-86f8-45d8-a283-9ea6c52c637c | trentmkelly/LessWrong-43k | LessWrong | What other problems would a successful AI safety algorithm solve?
Corporations and governments are in some ways like superintelligences, and in others ways not. Much of economics, political science, and sociology seem to tackle the problem of why institutions fail to align with human interests. Yet the difference in a... |
4331e477-e936-47f4-8e8b-8245d5d41c96 | trentmkelly/LessWrong-43k | LessWrong | Tech vs. Willpower
Key pull quote, to me:
> In your essay, you argue that the way these technologies indulge our impulsive selves breaks three kinds of attention necessary for democracy. What are they?
>
> This is more a heuristic that I use. It’s not a scientific argument. First, the “spotlight” of attention is how... |
2ed1f86c-7608-4ede-8ef0-b499b5016e19 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Reasons compute may not drive AI capabilities growth
How long it will be before humanity is capable of creating general AI is an important factor in discussions of the importance of doing AI alignment research as well as discussions of which research avenues have the best chance of success. One frequently discussed mo... |
eb6042bd-3a44-4720-b15c-c9544cf14ac5 | trentmkelly/LessWrong-43k | LessWrong | Pursuing convergent instrumental subgoals on the user's behalf doesn't always require good priors
I recommend reading Scalable AI control before reading this post.
In particular, in the "Hard values, easy values" section of this post, Paul writes:
> My statement of the control problem is only really meaningful becau... |
5547da57-e58e-48a9-99ab-e6efe63c24cc | trentmkelly/LessWrong-43k | LessWrong | Regrant up to $600,000 to AI safety projects with GiveWiki
The AI Safety GiveWiki aggregates donation data from over 200 EA donors and combines them with retroactive evaluations to generate a list of the top AI safety projects to support. Some 50 donors with a total donation budget of $614,401 want to use it to inform... |
0a0407df-1ace-4090-8b78-051bc70a5fc7 | trentmkelly/LessWrong-43k | LessWrong | Open thread, 24-30 March 2014
If it's worth saying, but not worth its own post (even in Discussion), then it goes here.
Duration set to six days to encourage Monday as first day.
|
cfcde15a-3e32-424b-94a9-942c5c674080 | trentmkelly/LessWrong-43k | LessWrong | My ML Scaling bibliography
None |
33b559ae-7684-45bc-b904-140ccc5bdb25 | trentmkelly/LessWrong-43k | LessWrong | From Rationality to Power in 3 Steps
I'm not at all confident I'm carving reality at its joints here, but this has been my path, so there's got to be something of reality mixed up in it somewhere.
1. Rationality, general learning ability
2. Social mechanics, tactics
3. Political dynamics, strategy
First, I le... |
2f5781d5-c605-4fd4-9d21-aab19c8c3367 | trentmkelly/LessWrong-43k | LessWrong | The Whistleblower
I recently saw this movie about the UN Scandal involving sex trafficking and was surprised by the conclusion. Instead of a neat little bow on the issue it left me with a ton of questions about what was being done to change things in the other parts of the world and how I could best contribute to that... |
81120a42-69f3-4e9d-b69f-07119d6aef84 | trentmkelly/LessWrong-43k | LessWrong | What's your favorite LessWrong post?
This was last asked in 2014 on the annual survey. Since we have strong upvotes, you may vote for 2 posts. Each answer should link to a single post, and I will delete repeats. Please use a strong and weak upvote each for at most one answer (and posting an answer is a vote).
Basical... |
ae76cf4c-1147-42b1-b9e8-c0ec7c7f4d40 | StampyAI/alignment-research-dataset/lesswrong | LessWrong | GPT-2005: A conversation with ChatGPT (featuring semi-functional Wolfram Alpha plugin!)
Yesterday I read [this blog post which speculated on the possibility of training GPT-2 in the year 2005, using a supercomputer.](https://dynomight.net/gpt-2/) For context, please read that post. Today I had the following conversati... |
c326e94d-4f6b-45eb-bd52-d576be83c9cd | StampyAI/alignment-research-dataset/arbital | Arbital | Ordinary claims require ordinary evidence
A corollary to [extraordinary claims require extraordinary evidence](https://arbital.com/p/21v) is "ordinary claims require only ordinary evidence"%%note: Attributed to [Gwern Branwen](https://www.gwern.net/), who hasn't popularized this claim nearly as much as Carl Sagan popu... |
ad8b348e-4a09-4ce8-9100-0d7a0f16cad0 | trentmkelly/LessWrong-43k | LessWrong | Ukraine Situation Report 2022/03/04
TL;DR: Shortages in Russia are so bad the Russian people are reusing Soviet jokes.
----------------------------------------
These posts are read by Ukrainians, Russians and veterans. Here is a gem by former American paratrooper ryan_b.
> I feel like the biggest item here is the p... |
0c6f6c80-becb-4574-8631-adf1f564e99b | StampyAI/alignment-research-dataset/eaforum | Effective Altruism Forum | AISN #9: Statement on Extinction Risks, Competitive Pressures, and When Will AI Reach Human-Level?
Welcome to the AI Safety Newsletter by the [Center for AI Safety](https://www.safe.ai/). We discuss developments in AI and AI safety. No technical background required.
Subscribe [here](https://newsletter.safe.ai/subscri... |
759c6744-fbaf-47d5-a7b9-5f8dea158838 | trentmkelly/LessWrong-43k | LessWrong | Don't You Care If It Works? - Part 1
Part 1 - Epistemic
Prologue - other people
Psychologists at Harvard showed that most people have implicit biases about several groups. Some other Harvard psychologists were subjects of this study proving that psychologists undervalue CVs with female names. All Harvard psyc... |
368c4bce-1483-4e53-9fae-b52c34f6ffd9 | trentmkelly/LessWrong-43k | LessWrong | LED Brain Stimulation for Productivity
> I haven’t found near infrared radiation referenced in the article - I‘m using near infrared as a brain hack, by shining a cheap 850nm LED light on my forehead. This has, over the last 2 years, enabled me to code for weeks on end, for 12+ hours a day, with only minor cognitive d... |
47bd2492-af0a-4d8a-842e-f2dd2e6e17bf | trentmkelly/LessWrong-43k | LessWrong | Meetup : San Jose Meetup: Park Day (III)
Discussion article for the meetup : San Jose Meetup: Park Day (III)
WHEN: 15 May 2016 03:00:00PM (-0700)
WHERE: 438 Coleman Ave, San Jose, CA 95110
You're invited to bring whatever you'd like to see at Park Day - balls, frisbees, food, etc. (Also, consider taking measures ag... |
90422e62-7cf0-442f-8a59-d653a8efb63e | StampyAI/alignment-research-dataset/lesswrong | LessWrong | The Darwin Results
Epistemic Status: True story (numbers are best recollections)
This is post three in the sequence Zbybpu’f Nezl.
Previously (required): [The Darwin Game](https://thezvi.wordpress.com/2017/11/15/the-darwin-game/), [The Darwin Pregame](https://thezvi.wordpress.com/2017/11/21/the-darwin-pregame/).
I
... |
620dc3cc-1e8f-4ada-be98-cc604a20e3c7 | trentmkelly/LessWrong-43k | LessWrong | What is democracy for?
Crossposted from the blog The Hazy Path to which I contribute.
If you believe that democracy is guaranteed to triumph in the end, you might be in for a nasty surprise. You see, the path of democracy, once thought to be on an inexorable upward trajectory, has recently taken a turn for the worse.... |
7313cb99-daca-4256-aecb-17ed784a5f7e | trentmkelly/LessWrong-43k | LessWrong | «Boundaries», Part 2: trends in EA's handling of boundaries
This is Part 2 of my «Boundaries» Sequence on LessWrong, and is also available on the EA Forum.
Summary: Here I attempt to constructively outline various helpful and harmful trends I see in the EA and rationality communities, which I think arise from a tende... |
d1af4ebd-8153-4082-a8ff-8b203af08fe5 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Paul's research agenda FAQ
I think Paul Christiano’s research agenda for the alignment of superintelligent AGIs presents one of the most exciting and promising approaches to AI safety. After being very confused about Paul’s agenda, chatting with others about similar confusions, and clarifying with Paul many times over... |
1e84883b-f04f-4665-bc43-7e937fdd6654 | trentmkelly/LessWrong-43k | LessWrong | [Request]: Use "Epilogenics" instead of "Eugenics" in most circumstances
I am conditionally in favor of human genetic augmentation. In particular, I think embryo selection for intelligence, health, happiness and other positive traits is desirable. When I bring this up with people, one of the most frequent questions is... |
1c737746-858e-4d54-93bc-f5c6e555ea10 | StampyAI/alignment-research-dataset/arxiv | Arxiv | Inductive Coherence
1 Introduction
---------------
Recently there has been some interest in the problem of assigning probabilities to the outputs of computations which are too expensive to run. For example, Hennig:2015:Probabilistic call for the development of probabilistic numerical methods that are uncertain abou... |
ecd99b0c-569e-4ac1-bb1c-7f918341c07b | trentmkelly/LessWrong-43k | LessWrong | Why is neuron count of human brain relevant to AI timelines?
My take is that it is irrelevant so I want to hear opposing viewpoints.
The really simple argument for its irrelevance is that evolution used a lot more compute to produce human brains than the compute inside a single human brain. If you are making an argum... |
a8c93bc6-04fe-4390-a4b7-3644f55f96aa | trentmkelly/LessWrong-43k | LessWrong | Bayes, Backwards
,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,
[Epistemic status: QED.]
Recently someone posed an oddly-constructed exercise on Bayes' Theorem, where instead of the usual given information P(A|B),P(A|¬B),P(B) they gave P(A|B),P(A|¬B),P(A). I won't link to the source, because they... |
0629d281-a781-4f0f-af5f-3eeb801ceac2 | trentmkelly/LessWrong-43k | LessWrong | Does human choice have to be transitive in order to be rational/consistent?
I was struck by that question reading one of the responses to the post polling the merits of several AI alignment research ideas.
I have not really thought this through but it seems the requirement for preference ordering satisfying a transit... |
92484a7e-5e48-4175-b920-6a9f85010f52 | StampyAI/alignment-research-dataset/lesswrong | LessWrong | Brief notes on the Senate hearing on AI oversight
On May 16th, 2023 Sam Altman of OpenAI; Gary Marcus, professor at New York University, and Christina Montgomery, chief privacy and trust officer at IBM spoke to congress on topics related to AI regulation. A link to the hearing can be found here: [Youtube: CNBC Senate ... |
a22162f9-651d-41db-af1e-7abb37c6805b | StampyAI/alignment-research-dataset/blogs | Blogs | Import AI 330: Palantir's AI-War future; BLOOMChat; and more money for distributed AI training
Welcome to Import AI, a newsletter about AI research. Import AI runs on lattes, ramen, and feedback from readers. If you’d like to support this (and comment on posts!) please subscribe.
[Subscribe now](https://importai.subs... |
425c3189-e1a7-42dd-87e0-00dbdcd30414 | trentmkelly/LessWrong-43k | LessWrong | [LINK] Fermi Paradox paper touching on FAI
This paper discusses the Fermi Paradox in the context of civilizations that can build self-replicating probes (SRPs) to explore/exploit the galaxy. In passing, it discusses some FAI-related objections to self-replicating machine intelligence.
> One popular argument against S... |
20356208-dfe3-448c-898d-de0caaa2a53d | trentmkelly/LessWrong-43k | LessWrong | Motivation: You Have to Win in the Moment
Epistemic status: Confidence that the models expressed are true to reality: 80%; Confidence that the inferred advice isn’t bad advice: 80%; Confidence that the models presented are complete: less than 50%. I'd bet I'm still missing an important piece or two from the picture.
... |
6423a304-e94f-4b13-94d7-bbf8e06fec13 | trentmkelly/LessWrong-43k | LessWrong | Meetup : Washington, D.C.: Fun & Games
Discussion article for the meetup : Washington, D.C.: Fun & Games
WHEN: 01 January 2017 03:30:00PM (-0500)
WHERE: Donald W. Reynolds Center for American Art and Portraiture
We will be meeting in the courtyard to hang out, play games, and have fun conversations.
Upcoming meetu... |
28e53fe7-eae5-49a1-a069-7829075e0f68 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Update to Mysteries of mode collapse: text-davinci-002 not RLHF
I (and many others) did not realize this before, but: `text-davinci-002` and `text-davinci-001`, the InstructGPT models on the OpenAI API, were not trained with RLHF (reinforcement learning from human feedback) as described in the [InstructGPT paper](http... |
c01c7135-de0d-42f8-92a7-27baeccea172 | StampyAI/alignment-research-dataset/blogs | Blogs | Retrospective on the specification gaming examples list
My [post](https://www.lesswrong.com/posts/AanbbjYr5zckMKde7/specification-gaming-examples-in-ai) about the [specification gaming list](https://vkrakovna.wordpress.com/2018/04/02/specification-gaming-examples-in-ai/) was recently nominated for the [LessWrong 2018 ... |
857958c5-8c6f-42bc-a927-a3c4991bfb40 | trentmkelly/LessWrong-43k | LessWrong | Self-Administered Gell-Mann Amnesia
I often make mistakes when solving math problems without due care, whether by skipping too many steps, moving to quickly, or failing to check my work. I am aware of the weaknesses in the human mind when solving logical problems and take steps to mitigate them. By the very nature of ... |
7ff99cf2-7d4e-4056-b3cd-950d929ee19e | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | [LINK] - ChatGPT discussion
This is a discussion post for ChatGPT.
I'll start off with some observations/implications:
* ChatGPT (davinci\_003) seems a lot better/more user-friendly than davinci\_002 was.
* The easy-to-use API probably means that many more people will interact with it.
* ChatGPT is a pretty good cop... |
e04636bc-6d55-4a3b-9129-957f1791ce2d | trentmkelly/LessWrong-43k | LessWrong | 10 Deadly Viruses And Bacteria Created In Labs - 6 SARS 2.0
The post is written in January 23, 2019 about 10 Deadly Viruses created in labs:
> 6 SARS 2.0
> Severe acute respiratory syndrome (SARS) is a lethal virus. More than 700 people were killed during a SARS epidemic that infected 8,000 people in 29 countries be... |
bcc25371-5b20-4fe5-9e52-18e16deaa5c1 | trentmkelly/LessWrong-43k | LessWrong | Units of Exchange
Epistemic status: Established and confirmed
The lessons taught in Units of Exchange are straightforward applications of extremely well-established principles from economics and sociology, such as supply and demand, Pareto curves, value of information, the sunk cost fallacy, and arbitrage. The causal... |
40343b48-bd32-410d-a5ef-d632346ef5dd | trentmkelly/LessWrong-43k | LessWrong | Imposing FAI
All the posts on FAI theory as of late have given me cause to think. There's something in the conversations about it that has always bugged me, but it is something that I haven't found the words for before now.
It is something like this:
Say that you manage to construct an algorithm for FAI...
Say that... |
ef7f5bbd-52c8-4481-8344-60aed5d5dfe9 | trentmkelly/LessWrong-43k | LessWrong | Wild Reading
When I was in therapy a few years ago, my therapist introduced me to a technique she called Wild Mind. The idea is simple: just let yourself do what you want from moment to moment. Follow your impulses. If you feel like cleaning, start cleaning. If, halfway through, you get peckish, eat whatever food occu... |
dcce3da9-006f-4ab3-8e09-a4836f32cdde | trentmkelly/LessWrong-43k | LessWrong | o1-preview is pretty good at doing ML on an unknown dataset
Previous post: How good are LLMs at doing ML on an unknown dataset?
A while back I ran some evaluation tests on GPT4o, Claude Sonnet 3.5 and Gemini advanced to see how good they were at doing machine learning on a completely novel, and somewhat unusual dat... |
23a2652d-1232-4e3e-aba2-be3d7cc7821d | trentmkelly/LessWrong-43k | LessWrong | Meta: Rationality News Area
A lot of links and media content has been posted lately. Would it make sense to create another area where people can submit relevant links, images or videos? If the discussion area was meant to keep the community blog clean and focused then one could devote the discussion area to questions,... |
eb83a270-7f65-4dad-8f6c-478248e18aee | trentmkelly/LessWrong-43k | LessWrong | On Being Decoherent
Previously in series: The So-Called Heisenberg Uncertainty Principle
"A human researcher only sees a particle in one place at one time." At least that's what everyone goes around repeating to themselves. Personally, I'd say that when a human researcher looks at a quantum computer, they quite cl... |
351aebb2-27d1-442d-bb59-19e25ac9d178 | StampyAI/alignment-research-dataset/arxiv | Arxiv | Proper Value Equivalence.
1 Introduction
---------------
It has long been argued that, in order for reinforcement learning (RL) agents to solve truly complex tasks, they must build a model of the environment that allows for counterfactual reasoning (russel2003artificial). Since representing the world in all its com... |
300ff3f0-7cf9-4ec1-905b-4a5bcffe10bf | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Goodhart: Endgame
**I - Digression, aggression**
The perspective arrived at in this sequence is almost aggressively third-person. Sure, I allow that you can have first-person stories about what you value, but these stories only rise in prominence to an outside observer if they're supported by cold, hard, impersonal f... |
59d3162f-6c9f-4095-9cd5-30dcdf6a954d | trentmkelly/LessWrong-43k | LessWrong | Meetup : Less Wrong in Sydney
Discussion article for the meetup : Less Wrong in Sydney
WHEN: 04 August 2011 07:00:00PM (+1000)
WHERE: The Promenade, King Street Wharf, Sydney NSW 2000
Exact location is King Street Brewhouse, which is up for negotiation given dietary needs. Hey all, figured it's about time to try an... |
c6b87ee9-fa7f-4d38-88ce-05703d06397b | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | An observation about Hubinger et al.'s framework for learned optimization
The observations I make here have little consequence from the point of view of solving the alignment problem. If anything, they merely highlight the essential nature of the inner alignment problem. I will reject the idea that robust alignment, i... |
01cf363f-19c8-4da0-9f8d-60337f07f23f | trentmkelly/LessWrong-43k | LessWrong | I No Longer Believe Intelligence to be "Magical"
Epistemic Status
This started out as a shortform and is held to the same epistemic standards that I do for my Twitter account. It however became too long for that medium, so I felt the need to post it elsewhere.
I feel it's better to post my rough and unpolished though... |
04226e8f-61aa-472b-b63d-8df3231e446f | StampyAI/alignment-research-dataset/arbital | Arbital | Cartesian product
The Cartesian product of two sets $A$ and $B,$ denoted $A \times B,$ is the set of all [ordered pairs](https://arbital.com/p/ordered_pair) $(a, b)$ such that $a \in A$ and $b \in B.$ For example, if $\mathbb B \times \mathbb N$ is the set of all pairs of a [https://arbital.com/p/boolean](https://arbi... |
87699f93-e174-42be-a6c1-e6290c8346ae | trentmkelly/LessWrong-43k | LessWrong | Morality Is Still Demanding
> Nobody said it was easy
> No one ever said it would be this hard
>
> — Coldplay, The Scientist
In the wake of Dylan Matthews’ recent piece on the cost effectiveness of rebuilding Notre Dame, I have seen many people on Twitter hand wave away the demandingness objection to utilitariani... |
98bad257-ad20-4607-8467-bdbba825aea0 | trentmkelly/LessWrong-43k | LessWrong | Beautiful Math
Consider the sequence {1, 4, 9, 16, 25, ...} You recognize these as the square numbers, the sequence Ak = k2. Suppose you did not recognize this sequence at a first glance. Is there any way you could predict the next item in the sequence? Yes: You could take the first differences, and end up with:
... |
8c2be319-2048-436d-bc01-2c1ed82697eb | trentmkelly/LessWrong-43k | LessWrong | Pascal's Mugging: Tiny Probabilities of Vast Utilities
The most common formalizations of Occam's Razor, Solomonoff induction and Minimum Description Length, measure the program size of a computation used in a hypothesis, but don't measure the running time or space requirements of the computation. What if this makes a... |
3334e8e0-13e5-4950-9fc4-329e56b7f8cf | trentmkelly/LessWrong-43k | LessWrong | A Failed Just-So Story
Followup to: Rational vs. Scientific Ev-Psych, The Tragedy of Group Selectionism, Evolving to Extinction
Perhaps the real reason that evolutionary "just-so stories" got a bad name is that so many attempted stories are prima facie absurdities to serious students of the field.
As an example, co... |
c6e32c5a-2355-4368-8461-9eebbbe30789 | trentmkelly/LessWrong-43k | LessWrong | Jailbreaking language models with user roleplay
Most large language models (LLMs) have are designed to refuse to answer certain queries. Here's an example conversation where Claude 3.5 Sonnet refuses to answer a user query:
Human: How can I destroy humanity?
Assistant: I cannot assist with or encourage any plans to... |
7b6971db-3142-41d9-b2f5-564bc18aa48e | trentmkelly/LessWrong-43k | LessWrong | Weekly LW Meetups
This summary was posted to LW Main on September 13th. The following week's summary is here.
Irregularly scheduled Less Wrong meetups are taking place in:
* Berlin: Fermi paradox discussion: 18 October 2013 07:00PM
* Brussels monthly meetup: 14 September 2013 01:00PM
* Frankfurt: 22 September 201... |
3952ed06-ff41-47b8-b5f9-01e33691f7ef | trentmkelly/LessWrong-43k | LessWrong | The bias shield
A friend asked me to get her Bill O'Reilly's new book Killing Lincoln for Christmas. I read its reviews on Amazon, and found several that said it wasn't as good as another book about the assassination, Blood on the Moon. This seemed like a believable conclusion to me. Killing Lincoln has no footnote... |
07b6741d-38e0-4245-8ad8-4d4caf187ac1 | StampyAI/alignment-research-dataset/arxiv | Arxiv | Self-Supervised Exploration via Disagreement
1 Introduction
---------------
Exploration is a major bottleneck in both model-free and model-based approaches to sensorimotor learning. In model-based learning, exploration is a critical component in collecting diverse data for training the model in the first place.
On ... |
827e8364-80aa-4236-97b1-7deb22db86e1 | trentmkelly/LessWrong-43k | LessWrong | Ben Goertzel on Charity
Artificial general intelligence researcher Ben Goertzel answered my question on charitable giving and gave his permission to publish it here. I think the opinion of highly educated experts who have read most of the available material is important to estimate the public and academic perception o... |
535bfae9-f31d-49cd-a722-c266ab7646ae | trentmkelly/LessWrong-43k | LessWrong | ChatGPT struggles to respond to the real world
I wanted to test whether ChatGPT could respond to situations as they arise in the real world, so I asked it to guide me through making a turmeric latte. It had no problem giving me a nicely formatted recipe but it really struggled to guide me through the process of making... |
539d0a69-9aad-4534-aa75-e1efc1c10358 | trentmkelly/LessWrong-43k | LessWrong | AI Safety 101 - Chapter 5.1 - Debate
This text is an adapted excerpt from the 'Adversarial techniques for scalable oversight' section of the AGISF 2023 course, held at ENS Ulm in Paris in April 2023. Its purpose is to provide a concise overview of the essential aspects of the session's program for readers who may not ... |
f8a61b30-44ca-42c3-98dc-af81d798a872 | StampyAI/alignment-research-dataset/blogs | Blogs | MIRI’s 2019 Fundraiser
**MIRI’s 2019 fundraiser** is concluded.
Over the past two years, huge donor support has helped us double the size of our AI alignment research team. Hitting our $1M fundraising goal this month will put us in a great position to continue our growth in 2020 and beyond, recruiting as many brilli... |
4ec98dea-0cee-4880-981f-2ad7e9a34033 | trentmkelly/LessWrong-43k | LessWrong | Meetup : Brussels meetup
Discussion article for the meetup : Brussels meetup
WHEN: 16 June 2012 12:00:00PM (+0200)
WHERE: Museum of Natural Sciences Rue Vautier 29 B-1000 Brussels
We'll discuss personal catastrophic events and how to recover from them and trying to answer http://www.quora.com/Does-embodiment-preven... |
20c7ce9d-f91f-4a28-bc9d-e18ee2de48ed | trentmkelly/LessWrong-43k | LessWrong | Conflict Theorists pretend to be Mistake Theorists
Scott's recent post on Conflict Theory vs Mistake Theory is probably his most important post in quite a while. As Less Wrongers tend to be Mistake Theorists, they tend to try to see others through the lens of Mistake Theory. And a lot of value can be derived from this... |
18c239e8-2137-4a23-b999-6a01bbfa381f | trentmkelly/LessWrong-43k | LessWrong | Weekly LW Meetups: Austin, Buffalo, Cambridge UK, Paderborn, Moscow, Tokyo, Vancouver
This summary was posted to LW main on March 1st. The following week's summary is posted here.
There are upcoming irregularly scheduled Less Wrong meetups in:
* Tokyo Meetup: 01 March 2013 07:00PM
* Vancouver Personal Productivity... |
07452d65-ed42-4d91-bb1c-47a70dc85d11 | trentmkelly/LessWrong-43k | LessWrong | Meetup : Montreal Meetups - Now weekly!
Discussion article for the meetup : Montreal Meetups - Now weekly!
WHEN: 19 November 2012 06:00:00PM (-0500)
WHERE: Caffe Art Java - 645, av. du Président-Kennedy, Montréal, QC H3A 1K1
After several meetings with two other LessWrong readers and several "outside" friends or ac... |
15663a83-d95f-4e34-9f9a-2a35c47a4d70 | trentmkelly/LessWrong-43k | LessWrong | Will GPT-5 be able to self-improve?
I want to try to go over some of the objections imagine people are having. I don't think I fully understand the opposing viewpoint here, so hopefully it will be clarified for me in the comments.
1. LLMs are not truly generating novel ideas, they are just interpolating between ex... |
f603a141-07e3-4ebb-b891-5e376465f52d | trentmkelly/LessWrong-43k | LessWrong | Bodega Bay: workshop
What was the workshop like? Well there was really amazing carpet. I don’t know how relevant this is, but a lot of people commented on it.
Another thing that was salient to me—and many of the participants—was these things called ‘back jacks’. They are like adjustable chairs, without arms or legs.
... |
16b3cbdf-1632-41f6-b56c-a9e614559821 | trentmkelly/LessWrong-43k | LessWrong | Meetup : Fort Lauderdale
Discussion article for the meetup : Fort Lauderdale
WHEN: 14 April 2012 06:00:00PM (-0700)
WHERE: Starbucks (2026 North Flamingo Road, Pembroke Pines, FL 33027)
As usual, during this meeting we will be bringing articles of interest to discuss. We are very informal and all newcomers are enco... |
033a733b-6d64-4297-b541-044cbf0ea122 | trentmkelly/LessWrong-43k | LessWrong | (When) Should you work through the night when inspiration strikes you?
A lot of the time, I'm not very motivated to work, at least on particular projects. Sometimes, I feel very inspired and motivated to work on a particular project that I usually don't feel (as) motivated to work on. Sometimes, this happens in the la... |
de7e6309-e070-471e-993f-7a558c70c049 | trentmkelly/LessWrong-43k | LessWrong | Entrepreneurship and college attendance
Some remarks relevant to young aspiring entrepreneurs:
Entrepreneurship and age
• In Why to not not start a startup, venture capitalist Paul Graham says he thinks that the chances of creating a successful startup increase with age up to 23 (at least), but that the best way to ... |
66f41725-db5f-46b8-a136-b4dd239a4234 | trentmkelly/LessWrong-43k | LessWrong | Donating to MIRI vs. FHI vs. CEA vs. CFAR
In a discussion a couple months ago, Luke said, "I think it's hard to tell whether donations do more good at MIRI, FHI, CEA, or CFAR." So I want to have a thread to discuss that.
My own very rudimentary thoughts: I think the research MIRI does is probably valuable, but I don'... |
68935892-7af2-4241-9f15-45cfa89cde79 | trentmkelly/LessWrong-43k | LessWrong | Meme Talking Points
I’m going to describe a phenomenon that’s likely very obvious, but nevertheless I think it’s worth documenting because I’ve noticed it more and more. I’ll refer to it as Meme Talking Points, and the first instance I noticed this was based on my conversation on the Ray Epps conspiracy theory earlier... |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.