id stringlengths 36 36 | source stringclasses 15
values | formatted_source stringclasses 13
values | text stringlengths 2 7.55M |
|---|---|---|---|
4398b184-cc21-4d71-b182-65c4bf9d9684 | awestover/filtering-for-misalignment | Redwood Research: Alek's Filtering Results | id: post3753
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 3 ). The desiderata for corrigibility are: A corrigible agent tolerates, and preferably assists, its operators in their attempts to alte... |
7049b4bc-8496-4627-b6d3-75cb714ff2de | trentmkelly/LessWrong-43k | LessWrong | Meetup : Cambridge, UK LW Meetup [Reading Group, HAEFB-03]
Discussion article for the meetup : Cambridge, UK LW Meetup [Reading Group, HAEFB-03]
WHEN: 24 February 2013 11:00:00AM (+0000)
WHERE: Trinity JCR, Cambridge, UK
Meetup! This week, we'll be continuing our reading group session of the Shiny New Sequence, Hig... |
60e0bfa8-dafc-41fc-ad43-28bae253fca4 | trentmkelly/LessWrong-43k | LessWrong | I played as AI in AI Box, and it was generally frustrating all around.
This morning I played against an anonymous gatekeeper (GK), and I lost.
The game went 2 hours and 20 minutes, and It was such a frustrating 2 hours. I felt like I was dealing with a bureaucracy! I was constantly surprised by how much time was gone... |
7fd0110b-c386-4d51-8be8-4f81c8c54ebc | trentmkelly/LessWrong-43k | LessWrong | Open thread, May. 1 - May. 7, 2017
If it's worth saying, but not worth its own post, then it goes here.
----------------------------------------
Notes for future OT posters:
1. Please add the 'open_thread' tag.
2. Check if there is an active Open Thread before posting a new one. (Immediately before; refresh the li... |
96c008aa-9c2f-4250-bdaf-d65a25a56abb | trentmkelly/LessWrong-43k | LessWrong | A New Interpretation of the Marshmallow Test
I've begun to notice a pattern with experiments in behavioral economics. An experiment produces a result that's counter-intuitive and surprising, and demonstrates that people don't behave as rationally as expected. Then, as time passes, other researchers contrive different ... |
bbd40bc5-10e5-4c4e-af63-afbcfb0e1a56 | trentmkelly/LessWrong-43k | LessWrong | Contra Musician Gender II
In the comments of yesterday's post on the most booked callers and bands, several people were interested in the gender composition of the bands. I looked at this in 2018, covering four years of data (2014 through 2017); what does it look like now?
I took my stats for 2023 and 2024 [1] and at... |
0d567a0e-3553-40c2-ac50-25d22bacdbb4 | trentmkelly/LessWrong-43k | LessWrong | Meetup : The Psychology of Video Games
Discussion article for the meetup : The Psychology of Video Games
WHEN: 17 August 2014 03:30:00PM (-0400)
WHERE: Cambridge, MA
Kelly MacNeill will talk about the psychology of video games, starting at 4pm at MIT's Building 66
Cambridge/Boston-area Less Wrong meetups start at ... |
d1e2fde4-212d-49c9-a24b-3bcb7232a34c | trentmkelly/LessWrong-43k | LessWrong | Meetup : Umbc meetup
Discussion article for the meetup : Umbc meetup
WHEN: 24 March 2015 04:56:00PM (-0500)
WHERE: 1000 Hilltop cir Catonsville
We will meet in the westhill community center, if you don't know where that is, feel free to call or text 2014066972.
Discussion article for the meetup : Umbc meetup |
99054acd-0cb2-49b7-ab73-d8543257b3e5 | trentmkelly/LessWrong-43k | LessWrong | Book Review: Being You by Anil Seth
Overview
Consciousness is "deeply inscribed into the wider patterns of nature."
This book is a good non-technical synopsis on the cutting edge of consciousness research. However, this book offers little new insight. Seth manages to represent all sides of the argument fairly without... |
3d6cfa1e-a952-490f-a4ef-edfc83e64328 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | AXRP Episode 17 - Training for Very High Reliability with Daniel Ziegler
[Google Podcasts link](https://podcasts.google.com/feed/aHR0cHM6Ly9heHJwb2RjYXN0LmxpYnN5bi5jb20vcnNz/episode/NGU1NTQ4OTAtODEyNC00ODFjLWIxZmEtMTc2NmI2MGVjMDg5)
Sometimes, people talk about making AI systems safe by taking examples where they fai... |
a08ecabd-2538-481e-a8dd-649a3eb1f5cf | trentmkelly/LessWrong-43k | LessWrong | Upvotes, knowledge, stocks, and flows
This is a short post inspired by "EXEMPLIFYING EQUITABLE GROWTH: MR. GOOGLE SERVES ME A BAKER'S HALF-DOZEN FROM THE WCEG WEBSITE, AND WHAT I LEARN THEREBY...". Delong looks at google results for his blog archive, and finds they are very weakly predictive of what he thinks is usefu... |
8708af5a-5b98-42c8-b5d2-2c76f25b668f | StampyAI/alignment-research-dataset/arxiv | Arxiv | The Met Dataset: Instance-level Recognition for Artworks
1 Introduction
---------------
Classification of objects can be done with categories defined at different levels of granularity.
For instance, a particular piece of art is classified as the “Blue Poles” by Jackson Pollock, as painting, or artwork, from the po... |
50b19cc2-0854-4f14-8809-7d6d58f0d98e | StampyAI/alignment-research-dataset/lesswrong | LessWrong | The Linguistic Blind Spot of Value-Aligned Agency, Natural and Artificial
The paper by [Travis LaCroix](https://arxiv.org/search/cs?searchtype=author&query=LaCroix%2C+T), July 2022.
Abstract:
> The value-alignment problem for artificial intelligence (AI) asks how we can ensure that the 'values' (i.e., objective fu... |
681d73cf-524e-4b9e-bfcb-22efd05d8a2a | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Overcoming Clinginess in Impact Measures
> It may be possible to use the concept of a causal counterfactual ([as formalized by Pearl](https://www.amazon.com/Causality-Reasoning-Inference-Judea-Pearl/dp/052189560X)) to separate some intended effects from some unintended ones. Roughly, "follow-on effects" could be de fi... |
ed356add-7208-4158-89b8-f8c8bf29155f | trentmkelly/LessWrong-43k | LessWrong | Schneier talks about The Dishonest Minority [Link]
Evolution. Morality. Strategy. Security/Cryptography. This hits so many topics of interest, I can't imagine it not being discussed here. Bruce Schneier blogs about his book-in-progress, The Dishonest Minority:
> Humans evolved along this path. The basic mechanism can... |
ff7e2a52-f363-4ddf-a9f3-4be4a796ad16 | trentmkelly/LessWrong-43k | LessWrong | Which rationality posts are begging for further practical development?
Which posts seem to you like they have really important practical implications for how we ought to think and act, but don't seem to come with enough information about how to use the skills, or to develop them, or even what exactly the relevant skil... |
a5483719-110e-4efc-8ec2-ce5400ee5af4 | trentmkelly/LessWrong-43k | LessWrong | The difference between Determinism & Pre-determination
1. Scope
There are two arm-waving views often expressed about the relationship between “determinism/causality” on the one hand and “predetermination/predictability in principle” on the other. The first treats them as essentially interchangeable: what is causall... |
41538e26-0f6a-4956-a3e2-717645eba5b2 | StampyAI/alignment-research-dataset/lesswrong | LessWrong | The Problem with AIXI
**Followup to**: [Solomonoff Cartesianism](/lw/jg1/solomonoff_cartesianism/); [My Kind of Reflection](/lw/s2/my_kind_of_reflection/)
**Alternate versions**: [Shorter, without illustrations](http://wiki.lesswrong.com/wiki/The_Problem_with_AIXI)
---
[AIXI](http://wiki.lesswrong.com/wiki... |
7b14a293-0460-4b06-8fec-43bceff85d8a | trentmkelly/LessWrong-43k | LessWrong | Modeling the impact of AI safety field-building programs
Summary
AI safety field-building is a large space, with many possible programs. The Center for AI Safety (CAIS) has commissioned cost-effectiveness models to support decisions regarding how to prioritize between these programs. Dan provided high-level guidance e... |
c04c91b1-a3ab-41d3-bbd9-fc35ccb54b15 | trentmkelly/LessWrong-43k | LessWrong | Demystifying Born's rule
Under the many-worlds interpretation (MWI) of quantum physics, born probabilities are a bit mysterious and hard to deal with. How do you do Solomonoff induction when you keep splitting? There are ways to do this, but they end up equivalent to another interpretation of quantum mechanics that is... |
5fa33ff2-74ec-4e4e-9fa1-388af8377a86 | StampyAI/alignment-research-dataset/eaforum | Effective Altruism Forum | How does one find out their AGI timelines?
*This question is on my list of things I should ask but think are too dumb to.*
A lot of people have estimates and forecasts on when they think a transformative artificial intelligent/superintelligent system will come into existence. Yet it doesn't seem clear how they come t... |
57a1bad9-2219-477a-b680-c644b69eba51 | trentmkelly/LessWrong-43k | LessWrong | Volunteer needed to change the front page 'Featured Articles' each week
Less Wrong needs a volunteer to change the 'Featured Articles' once a week. I have detailed instructions on what to do (which wiki entry to edit, which articles have already been featured, which articles to feature in the future, etc.) I've been d... |
f16ac64d-6a1e-459b-bb65-3b05a5e85823 | trentmkelly/LessWrong-43k | LessWrong | Conversation, event loops, and error handling
My previous post Debugging the student got me thinking about conversations. Particularly conversations where one person is trying to explain things to the other person. In that post I talked about some ways I see this go south. In this post I want to expand on these ideas ... |
562c3863-4ea6-4a6f-8de2-6cefca7fb4c1 | StampyAI/alignment-research-dataset/eaforum | Effective Altruism Forum | Google could build a conscious AI in three months
*Summary: Theories of consciousness do not present significant technical
hurdles to building conscious AI systems. Recent advances in AI relate to
capacities that aren't obviously relevant to consciousness. Satisfying major
theories with current technology has been and... |
b4f2b745-933b-4408-9872-39fdd729d14a | trentmkelly/LessWrong-43k | LessWrong | What if we stopped making GPUs for a bit?
I generally concur with MIRI's latest on safety. AI progress seems bad.
AI development is pretty tied to GPU (TPU?) production at this point? If GPU production dropped by 95% for a few years, or if GPUs stopped getting faster for a few years, I wonder if that would affect AI... |
c93dabee-0a7b-4da6-aed9-a5f7e22c983a | StampyAI/alignment-research-dataset/lesswrong | LessWrong | New OpenAI Paper - Language models can explain neurons in language models
Summary by OpenAI: We use GPT-4 to automatically write explanations for the behavior of neurons in large language models and to score those explanations. We release a dataset of these (imperfect) explanations and scores for every neuron in GPT-2... |
721ec154-284c-4e53-8b29-83982a10e5e2 | StampyAI/alignment-research-dataset/arxiv | Arxiv | Poison Frogs! Targeted Clean-Label Poisoning Attacks on Neural Networks
1 Introduction
---------------
Before deep learning algorithms can be deployed in high stakes, security-critical applications, their robustness against adversarial attacks must be put to the test. The existence of adversarial examples in deep n... |
c41e2f2c-9819-4e56-a597-9dae8f2f627d | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | A shift in arguments for AI risk
*The linked post is work done by Tom Adamczewski while at FHI. I think this sort of expository and analytic work is very valuable, so I'm cross-posting it here (with his permission). Below is an extended summary; for the full document, see his linked blog post.*
Many people now work o... |
1e8ac640-07dc-4703-8ae6-f51c96c473ed | trentmkelly/LessWrong-43k | LessWrong | Positive Queries - How Fetching
> Help, having a brain blank. I can come up w examples of times something happened, but not times something didnt-happen. What heuristic?
>
> — Kate Donovan (@donovanable) April 29, 2014
If I tell 100 people not to think of an elephant, what's the single thing they're all most like... |
7314d33c-828f-459e-a883-b427d8079b17 | trentmkelly/LessWrong-43k | LessWrong | Why safety is not safe
June 14, 3009
Twilight still hung in the sky, yet the Pole Star was visible above the trees, for it was a perfect cloudless evening.
"We can stop here for a few minutes," remarked the librarian as he fumbled to light the lamp. "There's a stream just ahead."
The driver grunted assent as he pul... |
a9b4dbe3-cba3-4b13-b078-5cf1f1c92262 | StampyAI/alignment-research-dataset/lesswrong | LessWrong | Solving For Meta-Ethics By Inducing From The Self
### How do we determine what is good? By what method/metric? What is the argument? In this paper I will attempt to argue why and how I ground my ethics in an understanding of the self and it's values. To give a very broad and insufficient tl;dr: I'll try to create a de... |
1b924dfb-ab1e-4d4b-89ab-7e61614a05da | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Doing oversight from the very start of training seems hard
**TLDR:** We might want to use some sort of oversight techniques to avoid inner misalignment failures. Models will be too large and complicated to be understandable by a human, so we will use models to oversee models (or help humans oversee models). In many pr... |
9b74fbb6-452d-4ba6-b6e1-a1dfff792e96 | trentmkelly/LessWrong-43k | LessWrong | Three mental images from thinking about AGI debate & corrigibility
Here are three mental images I’ve used when sporadically struggling to understand the ideas and prospects for AI safety via debate, IDA, and related proposals. I have not been closely following the discussion, and may well be missing things, and I don’... |
2f6bf8f5-93e8-4f82-afe7-99ac86fca8c6 | StampyAI/alignment-research-dataset/eaforum | Effective Altruism Forum | Enabling more feedback
**Acknowledgements**
*Thank you to Frances Lorenz for writing this post and Daniel Braun for his comments.*
**Introduction**
AI Safety has a pool of research talent working via three main avenues: industry, academia, and independent research. Unfortunately, independent researchers and, de... |
0c8f84a2-e8f3-4c54-8891-27fd8384c5bc | trentmkelly/LessWrong-43k | LessWrong | Slaves to fashion signaling
I understand the signaling theory of clothes fashions to be something like this: wearing fashionable clothing in spite of fashion changing all the time demonstrates that you are socially well connected and know what is up (not to mention that you have money and are not too weird, much like ... |
c78959e1-7b20-41ab-8b9d-befd48fd83d1 | trentmkelly/LessWrong-43k | LessWrong | Loose thoughts on AGI risk
My thoughts here are speculative in nature, and should not be taken as strongly-held beliefs. As part of an experiment in memetic minimization, only the tldr version of this post exists. There are no supporting links. Let me know what you think in the comments.
Summary: the ultimate threat ... |
1b1f50b5-2540-417b-8602-c736470fcd2e | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Self-regulation of safety in AI research
In many industries, but especially those with a potentially adversarial relationship to society like advertising and arms, self regulatory organizations (SROs) exist to provide voluntary regulation of actors in those industries to assure society of their good intentions. For ex... |
a5970412-2baf-4d08-afd6-0656b27a13e0 | trentmkelly/LessWrong-43k | LessWrong | Defensible positions
A useful concept when discussing philosophy is the concept of defensible positions or alternatively "hills to die on". In philosophy, some positions are more defensible than others or at least appear that way at first glance. Take for example discussions on morality.
Utilitarianism and absolute d... |
dba176c0-4deb-43c8-913d-4f544fd303d8 | trentmkelly/LessWrong-43k | LessWrong | Thomas Kwa's research journal
Thomas Kwa
Why I'm writing this
Research feedback loops for junior AI safety researchers are pretty poor right now. They're better than in the past due to the recent explosion in empirical work, but AI safety is still a very new field whose methodology is not quite nailed down, and whic... |
9292a7a4-5377-41eb-886b-32afda8c322d | StampyAI/alignment-research-dataset/lesswrong | LessWrong | Beyond Hyperanthropomorphism
A philosophical argument against "the AI-fear". |
a5227563-89b1-4960-8686-387a1a5cc6c8 | trentmkelly/LessWrong-43k | LessWrong | Thinking Bayesianically, with Lojban
"Do not walk to the truth, but dance. On each and every step of that dance your foot comes down in exactly the right spot. Each piece of evidence shifts your beliefs by exactly the right amount, neither more nor less. What is exactly the right amount? To calculate this you must stu... |
f4968086-e6a4-498f-832c-2c9f5f662163 | trentmkelly/LessWrong-43k | LessWrong | On being downvoted
Meta: I don't think I have anything even moderately good to say here. Still, 1) I observe and endorse the policy of "Posts on practically any topic are welcomed on LessWrong", and 2) I think it might be helpful to start a conversation about this, in part so that others can say useful things, and in ... |
fab0d595-c2d1-458d-ae8c-c5ec27baf809 | trentmkelly/LessWrong-43k | LessWrong | Severe control over AI agents as a tool for mass-surveillance
Critique and Extension of the “IDs for AI Systems” proposal written by A. Chan, N. Kolt, P. Wills, U. Anwar, C. Schroeder de Witt, N. Rajkumar, L. Hammond, D. Krueger, L. Heim, and M. Anderljung
IDs for AI agents
Soon, the internet will be populated with... |
5c3bafac-90d7-4087-8bb5-eec4aef5c376 | trentmkelly/LessWrong-43k | LessWrong | Agreeable ways to disable your children
Should parents purposely have deaf children if they prefer them, by selecting deaf embryos?
Those in favor argue that the children need to be deaf to partake in the deaf culture which their parents are keen to share, and that deafness isn’t really a disability. Opponents point ... |
4a0c9d06-1db9-4992-9bbd-ab4b57d5ccba | trentmkelly/LessWrong-43k | LessWrong | Human performance, psychometry, and baseball statistics
I. Performance levels and age
Human ambition for achievement in modest measure gives meaning to our lives, unless one is an existentialist pessimist like Schopenhauer who taught that life with all its suffering and cruelty simply should not be. Psychologists stu... |
0fb47de6-522b-4e37-bbf6-742d0e389d2a | trentmkelly/LessWrong-43k | LessWrong | Should CA, TX, OK, and LA merge into a giant swing state, just for elections?
As Americans know, the electoral college gives disproportionate influence to swing states, which means a vote in the extremely blue state of California was basically wasted in the 2024 election, as are votes in extremely red states like Texa... |
de2f14a3-4263-417d-8c31-803f3b2eaa25 | StampyAI/alignment-research-dataset/arbital | Arbital | Faithful simulation
The safe simulation problem is to start with some dynamical physical process $D$ which would, if run long enough in some specified environment, produce some trustworthy information of great value, and to compute some *adequate* simulation $S_D$ of $D$ faster than the physical process could have run... |
2a9d0047-cba0-4883-a91a-e67961ffaf21 | trentmkelly/LessWrong-43k | LessWrong | Mirror Organisms Are Not Immune to Predation
EDIT: See djbinder's comment below for counterarguments.
Core Argument
The barrier to preying on "mirror-image" life is likely lower than presumed due to:
1. Immediate energy from achiral fats.
2. Existing L-sugar pathways.
3. Existing D-amino acid racemases.
1. Fat... |
b438215a-c04d-4854-9e8a-300b73443061 | StampyAI/alignment-research-dataset/eaforum | Effective Altruism Forum | What is the role of Bayesian ML for AI alignment/safety?
I started a Ph.D. in Bayesian ML because I thought it was a relevant approach to AI safety. Currently, I think that this is unlikely and other paths within ML are more promising.
The main purpose of this post is to spark a discussion, get feedback and collect ... |
3875c83b-dd0e-4686-a38a-53924df6836b | StampyAI/alignment-research-dataset/lesswrong | LessWrong | The AI governance gaps in developing countries
As developed countries rapidly become more equipped in the governance of safe and beneficial AI systems, developing countries are slackened off in the global AI race and standing at risk of extreme vulnerabilities. By examining not only “how we can effectively govern AI” ... |
3e8dea34-aed6-420d-8212-a0b2a3d9aea1 | trentmkelly/LessWrong-43k | LessWrong | In the age of modern AI (LLMs and beyond), is data still the new oil?
In the age of big data and machine learning, it was common to say that data was the new oil. Every big company spent a lot of money creating their Hadoop instances, hiring data scientists, to discover new things from the data they had. It's easy to... |
e14652d9-fd5e-409c-9cc0-6ec4f0f3104a | LDJnr/LessWrong-Amplify-Instruct | LessWrong | "I asked why humanity took so long to do anything at the start, and the Internet gave me its thoughts. Here is my expanded list of hypotheses, summarizing from comments on the post, here, and here.Inventing is harder than it looksInventions are usually more ingenious than they seem. Relatedly, reality has a lot of deta... |
d3786ed8-3169-4702-ae42-e903e80d884a | StampyAI/alignment-research-dataset/blogs | Blogs | Bet with Zvi about Omicron
[Zvi Mowshowitz](https://thezvi.wordpress.com/) and I have agreed to the following bet:
* If at least 75% of the USA COVID-19 cases between 1/1/22 and 2/28/23 (inclusive) occur between 1/1/22 and 2/28/22 (inclusive), I pay Zvi $40.
* Otherwise, Zvi pays me $60.
* This bet is intended to... |
59e30a05-4d04-4a25-bfb8-45f83cb81063 | StampyAI/alignment-research-dataset/blogs | Blogs | Decisions are for making bad outcomes inconsistent
[](https://intelligence.org/2017/03/18/new-paper-cheating-death-in-damascus/)Nate Soares’ recent [decision theory](http://lesswrong.com/lw/gu1/decision_theory_faq/) paper with Ben Levinst... |
18e1ac8b-83cd-43ee-bf97-f00313c24745 | trentmkelly/LessWrong-43k | LessWrong | Meetup : Effective Altruism @ LessWrong Tel Aviv
Discussion article for the meetup : Effective Altruism @ LessWrong Tel Aviv
WHEN: 29 September 2015 07:00:00PM (+0300)
WHERE: Yigal Alon 98, Tel Aviv
We will meet in Google Israel (Electra Tower) on floor 29, as always.
We will have two talks on Effective Altruism: ... |
6d4822a3-7664-407a-bda1-34b708149810 | trentmkelly/LessWrong-43k | LessWrong | Improving the Welfare of AIs: A Nearcasted Proposal
Overview and Summary
In this document, I’ll discuss and propose a nearcasted plan for improving the algorithmic welfare of powerful AI systems, given no further technical or philosophical progress. I broadly endorse many of the propositions in Propositions Concerning... |
c67743c7-8dc5-4498-b98b-7c40cbaccdd5 | trentmkelly/LessWrong-43k | LessWrong | What are some open exposition problems in AI?
Basically, topics or ideas that are not well explained, but could benefit from good expositions. |
b8fd18f4-6e9d-424a-b902-de6933c44ffb | trentmkelly/LessWrong-43k | LessWrong | CIV: a story
The room was cozy despite its size, with wood-lined walls reflecting the dim lighting. At one end, a stone fireplace housed a roaring fire; in the middle stood a huge oak table. The woman seated at the head of it rapped her gavel. “I hereby call to order the first meeting of the Parliamentary Subcommittee... |
c650fb14-fbc0-40bb-a11f-357d3b09bd31 | trentmkelly/LessWrong-43k | LessWrong | The decline of violence as a lens for understanding effective altruism
Greetings all! There's a puzzle that I'm working on and I'm interested to see what the members of this community have to say about it.
I am an electrical engineer that is currently working on a master's in counseling. One of the big questions I ... |
c7f69f6b-46ed-4b29-9619-ed9398eb3a42 | trentmkelly/LessWrong-43k | LessWrong | Snoo App Thoughts
With our youngest child we got a Snoo automated bassinet, and overall we've been very happy with it. On the other hand, the corresponding phone app has several missing features. Since all of this is software, it should be fixable, but the current state is a hassle.
When the baby is fussy, one of the... |
170266cf-9be5-416a-be30-88ee817e5d5e | trentmkelly/LessWrong-43k | LessWrong | Meetup : Princeton NJ Meetup
Discussion article for the meetup : Princeton NJ Meetup
WHEN: 22 February 2014 01:00:00PM (-0500)
WHERE: Small World Coffee, 14 Witherspoon St. Princeton, NJ 08540
We're going to have the second ever Princeton NJ meetup! Even better than the first! Feel free to come whether or not you c... |
56392730-5d79-45e6-a817-c50258942bf8 | StampyAI/alignment-research-dataset/eaforum | Effective Altruism Forum | The AI revolution and international politics (Allan Dafoe)
---
Artificial intelligence (AI) is rapidly improving. Superhuman AI in strategically relevant domains is likely to arrive in the next several decades; some experts think two. This will transform international politics, could be profoundly destabilizing, and ... |
5234d7ee-ec86-4da1-babf-1b80d90a1864 | trentmkelly/LessWrong-43k | LessWrong | Six reasons why objective morality is nonsense
Note: I didn't write this essay, nor do I own the blog where it came from. I'm just sharing it. The essay text is displayed below this line.
Whether morality is an objective property of the universe, or instead the subjective opinion of humans, is one of the longest runn... |
9e125c75-31a5-4719-93b0-e282126118ef | trentmkelly/LessWrong-43k | LessWrong | The Heuristic About Representativeness Heuristic
(x-posted from my blog)
Some people think that the problem with representativeness heuristic is a base rate neglect. I hold that this is incorrect, and the problem is deeper than that, and simple use of a base rate isn't going to fix it. This makes the idea about "... |
b50ba42a-bd9c-4946-bd49-ca32aa72da0b | trentmkelly/LessWrong-43k | LessWrong | Advice please: Cognitive distortion preventing me from accomplishing anything
The cognitive distortion is called "catastrophizing", I think.
I'm afraid of unexpected, strongly negative events occurring to me without warning. Nothing specific, just a generalized fear. That fear is crippling me. Worse, there's a par... |
12adda5c-087c-4444-9dca-c36a3b8e9048 | StampyAI/alignment-research-dataset/lesswrong | LessWrong | My Alignment Timeline
*Epistemic status: If I don't post this now, I'll have a harder time treating it as something to build on.*
Which chapter do we end on?
---------------------------
Which chapter of doomsday are *you* on, right now?
I think of AI-alignment-going-well as a process, where AI-causes-human-extin... |
f407a96d-d2b4-405d-a4c6-95b5668c9d23 | trentmkelly/LessWrong-43k | LessWrong | What Would You Do Without Morality?
To those who say "Nothing is real," I once replied, "That's great, but how does the nothing work?"
Suppose you learned, suddenly and definitively, that nothing is moral and nothing is right; that everything is permissible and nothing is forbidden.
Devastating news, to be sure—and ... |
ddf0a226-3c92-4f7b-a2f7-fc792a77229a | trentmkelly/LessWrong-43k | LessWrong | How well can Claude write coding questions?
I'm curious as to how well Claude can write interesting coding and mathematics problems. This post is a partial product of that exploration.
It is a much harder skill to come up with a good problem than to solve one of equal difficulty. While there is a large corpus of da... |
696e3567-0879-441b-83ce-7310686c402f | trentmkelly/LessWrong-43k | LessWrong | A first success story for Outer Alignment: InstructGPT
OpenAI has given us a first taste of what success looks like in outer alignment, and that's InstructGPT. More specifically, InstructGPT actually is more aligned in two important ways: Truthfulness and hallucinations.
Specifically, OpenAI used RLHF to make GPT-3'... |
f2887ae2-c986-4400-985f-0f6bd8701e24 | trentmkelly/LessWrong-43k | LessWrong | Proving Too Much
The fallacy of Proving Too Much is when you challenge an argument because, in addition to proving its intended conclusion, it also proves obviously false conclusions. For example, if someone says “You can’t be an atheist, because it’s impossible to disprove the existence of God”, you can answer “That ... |
d6ef793e-a5ba-4edd-804b-5f06ea2a249d | trentmkelly/LessWrong-43k | LessWrong | Bragging Thread, July 2014
Your job, should you choose to accept it, is to comment on this thread explaining the most awesome thing you've done this since June 1st. You may be as blatantly proud of yourself as you feel. You may unabashedly consider yourself the coolest freaking person ever because of that awesome thin... |
a79171a9-e992-4ab6-baea-c7d3e288f1e1 | trentmkelly/LessWrong-43k | LessWrong | How four guys helped redirect Japan's coronavirus policy
In late March, popular consensus was that Japan was managing its coronavirus epidemic well, and was unlikely to be seriously affected. Patrick McKenzie, a software entrepreneur in Japan, thought that that was wrong, and on March 23rd through 25th he and three ot... |
b27ed885-c87f-442f-89ba-5c917c258eea | trentmkelly/LessWrong-43k | LessWrong | How far can AI jump?
I went to the Singularity Summit recently, organized by the Singularity Institute for Artificial Intelligence (SIAI). SIAI’s main interest is in the prospect of a superintelligence quickly emerging and destroying everything we care about in the reachable universe. This concern has two components. ... |
1cc7edba-a72d-4dab-8e65-3e9ea72089cd | StampyAI/alignment-research-dataset/lesswrong | LessWrong | Natural Abstraction: Convergent Preferences Over Information Structures
The natural abstraction hypothesis claims (in part) that a wide variety of agents will learn to use the same abstractions and concepts to reason about the world.
What's the simplest possible setting where we can state something like this formall... |
db3d2a8b-644e-4772-b73c-5926776025ec | trentmkelly/LessWrong-43k | LessWrong | How good are chatbots?
Recently, I tried some chatbots, like Mitsuku, rose, a.l.ic.e. and some more.
I read in wiki about loebner prize, and it looks that these chatbots should be good(they took part in the competitions and won in some of them)
But in practice, they are not so good - at list, not the online version,... |
e7743606-3054-48d7-93a7-f851b76f1afc | trentmkelly/LessWrong-43k | LessWrong | No convincing evidence for gradient descent in activation space
How is it that GPT performs better at question-answering tasks when you first prompt it with a series of positive examples? In 2020, in the title of the original GPT-3 paper, OpenAI claimed that language models are few shot learners. But they didn't say w... |
a3500be0-4575-41c1-8786-77ec0007c715 | trentmkelly/LessWrong-43k | LessWrong | Laughing strategy
People who believe that a certain group of other people deserve higher relative status often refuse to laugh at jokes about that group of people. Unfortunately (for them) this tends to make them look like uptight goody-goodies who don’t have a sense of humor; a group whom almost everyone agrees shoul... |
805c73b0-65ba-48aa-927d-1935ca4dfa5d | trentmkelly/LessWrong-43k | LessWrong | Preferences over non-rewards
In this penultimate post on "learning human values" series, I just want to address some human values/preferences/rewards that don't fit neatly into the (p, R) model where p in the planning algorithm and R the actual reward.
Preferences over preferences and knowledge
Most people have pref... |
57a50605-1a56-4c30-8d7b-da16b54b01a3 | trentmkelly/LessWrong-43k | LessWrong | Conversational canyons
Conversations are streams: they pass by without leaving much behind. But if you add a notebook—if you write down short summaries of what you talk about so you can return to it in later conversations and expand—then you have a landscape for the conversation to flow through, and this changes it in... |
e114452d-0d24-4f39-a962-de5f8e0200c4 | StampyAI/alignment-research-dataset/lesswrong | LessWrong | Neural program synthesis is a dangerous technology
*Crossposted from [my Medium blog](https://medium.com/@honnibal/neural-program-synthesis-is-a-dangerous-technology-677b1e0cfd5d)*
Today’s computer viruses can copy themselves between computers, but they can’t mutate. That might change very soon, leading to a serious ... |
f26a1d2e-dda7-4ce2-9643-916913a0bc40 | trentmkelly/LessWrong-43k | LessWrong | July 2010 Southern California Meetup
There will be a meetup for people from lesswrong in Los Angeles, on Friday July 9th, 2010 at 3PM. Roundtrip carpooling from San Diego is definitely available and other carpooling options may develop. The time and location are designed to make it possible for people from lesswron... |
7955b7ca-51d1-47e6-94aa-9f98d8144967 | trentmkelly/LessWrong-43k | LessWrong | Donating while in temporary debt (i.e. as a student)
Topic: I will be in debt for several years, but will eventually have a disposable income. Should I donate now or later?
Here's my situation: I am a student, with student loans and no income. I can take out more loans than I need. Grad PLUS loans have a fixed intere... |
b616c3ea-6771-4678-bcfb-de7160362b32 | trentmkelly/LessWrong-43k | LessWrong | Meetup : Moscow: Ernst Mach philosophy, Analysis of competing hypotheses, Paranoid Zendo and other games
Discussion article for the meetup : Moscow: Ernst Mach philosophy, Analysis of competing hypotheses, Paranoid Zendo and other games
WHEN: 17 July 2016 02:00:00PM (+0300)
WHERE: Москва, Большая Дорогомиловская, д.... |
1663ac6a-c8f5-4c9f-aa14-7603f0228f81 | trentmkelly/LessWrong-43k | LessWrong | Meetup : Brussels meetup
Discussion article for the meetup : Brussels meetup
WHEN: 11 August 2012 12:00:00PM (+0200)
WHERE: Rue des Alexiens 55 1000 Bruxelles
We're trying out a new location this time as the museum cafeteria can be a bit loud sometimes. We're meeting at 'La Fleur en Papier Doré' close to the Brusse... |
6eb07311-782d-4526-9b6b-76d5f46daefb | trentmkelly/LessWrong-43k | LessWrong | Finance as a career option
As a part of our research for Cognito Mentoring, Vipul Naik and I compiled a draft of a page on finance as a career option. Because some Less Wrongers are planning on earning to give and finance is a commonly considered career option for those who are earning to give, I thought that it might... |
782ebf7d-0c14-41eb-adb7-4cdaf17237ab | trentmkelly/LessWrong-43k | LessWrong | Accountability Sinks
This is a cross-post from https://250bpm.substack.com/p/accountability-sinks
> Back in the 1990s, ground squirrels were briefly fashionable pets, but their popularity came to an abrupt end after an incident at Schiphol Airport on the outskirts of Amsterdam. In April 1999, a cargo of 440 of the ... |
c0428e95-0c85-41b0-8539-a7923208d289 | trentmkelly/LessWrong-43k | LessWrong | Capital and inequality
In Thomas Piketty’s masterpiece, Capital in the 21st century, he showed the strong association between the rate of return on capital (capital here means assets like companies and land) and the levels of inequality in a society. The key mechanism at play here is that capital is much much more co... |
2162228f-2e71-4b80-bcbd-052cd194a40d | trentmkelly/LessWrong-43k | LessWrong | Five Hinge‑Questions That Decide Whether AGI Is Five Years Away or Twenty
For people who care about falsifiable stakes rather than vibes
TL;DR
All timeline arguments ultimately turn on five quantitative pivots. Pick optimistic answers to three of them and your median forecast collapses into the 2026–2029 range; pick... |
37e6ccdf-b830-4fa2-a5f5-821a3d249028 | trentmkelly/LessWrong-43k | LessWrong | Time to Exit the Sandbox
|
354de46c-a81a-4913-a8d0-a4e8fa478342 | awestover/filtering-for-misalignment | Redwood Research: Alek's Filtering Results | id: post2174
Redwood Research's current project is to train a model that completes short snippets of fiction without outputting text where someone gets injured. I'm excited about this direction and wanted to explain why. ( Disclaimer: I originally proposed this project, and am on Redwood’s board. So I may be biased in... |
f7ac310a-bd8a-4d64-a033-5640b643b4ae | trentmkelly/LessWrong-43k | LessWrong | Tracking Compute Stocks and Flows: Case Studies?
Posted in my personal capacity
The AGI governance community has recently converged on compute governance[1] as a promising lever for reducing existential risks from AI.
One likely building block for any maximally secure compute governance regime is stock and flow acco... |
98206acf-1d36-4cbd-9b8d-28744ee549bb | trentmkelly/LessWrong-43k | LessWrong | Which counterfactuals should an AI follow?
Thanks to Rebecca Gorman for her help with this post.
In the past, I've talked about counterfactual incentives. In that, the agent was motivated to maximise not an actual reward, but the (counterfactual) reward it would have gotten if things had been different.
"If things h... |
1ffeebd6-c8b3-42a4-a644-7f89e743431e | trentmkelly/LessWrong-43k | LessWrong | An Intuitive Explanation of Sparse Autoencoders for Mechanistic Interpretability of LLMs
Sparse Autoencoders (SAEs) have recently become popular for interpretability of machine learning models (although SAEs have been around since 1997). Machine learning models and LLMs are becoming more powerful and useful, but they ... |
4275ee03-19de-4da8-9ed7-06901b46f476 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | AXRP Episode 9 - Finite Factored Sets with Scott Garrabrant
[Google Podcasts link](https://podcasts.google.com/feed/aHR0cHM6Ly9heHJwb2RjYXN0LmxpYnN5bi5jb20vcnNz/episode/NjhiNGU4ZGQtODRlZi00NmZmLWI1ZDMtN2ZiNDJlMzRjZmY2)
This podcast is called AXRP, pronounced axe-urp and short for the AI X-risk Research Podcast. Here... |
e1491cb6-7d92-4ba9-a43a-f87e538744c0 | trentmkelly/LessWrong-43k | LessWrong | The Importance of Self-Doubt
[Added 02/24/14: After I got feedback on this post, I realized that it carried unnecessary negative connotations (despite conscious effort on my part to avoid them), and if I were to write it again, I would have framed things differently. See Reflections on a Personal Public Relations Fail... |
b8478a4d-bd98-44ff-9e18-bdcef5ce197e | trentmkelly/LessWrong-43k | LessWrong | Meetup : Houston Meetup
Discussion article for the meetup : Houston Meetup
WHEN: 25 March 2012 01:00:00PM (-0500)
WHERE: 2010 Commerce St, Houston, Tx. 77002
The Houston LW meetup group will be gathering this Sunday at 1PM to discuss the 4th chapter of ET Jaynes "The Logic of Science". We will also have one of our ... |
e9683b68-7e85-4eef-9a3e-ab5b8c757fc6 | trentmkelly/LessWrong-43k | LessWrong | Why pessimism sounds smart
Pessimists sound smart. Optimists make money.
–Nat Friedman (quoted by Patrick)
I’ve realized a new reason why pessimism sounds smart: optimism often requires believing in unknown, unspecified future breakthroughs—which seems fanciful and naive. If you very soberly, wisely, prudently stick ... |
e6849f79-b3ad-49de-a7a1-1b283527090d | StampyAI/alignment-research-dataset/arxiv | Arxiv | Hierarchical Reinforcement Learning via Advantage-Weighted Information Maximization
1 Introduction
---------------
Reinforcement learning (RL) has been successfully applied to a variety of tasks, including board games (Silver et al., [2016](#bib.bib36)), robotic manipulation tasks (Levine et al., [2016](#bib.bib23)... |
263fe03f-6a09-4d08-bfe9-2053578f3e7f | trentmkelly/LessWrong-43k | LessWrong | Learning values versus learning knowledge
I just thought I'd clarify the difference between learning values and learning knowledge. There are some more complex posts about the specific problems with learning values, but here I'll just clarify why there is a problem with learning values in the first place.
Consider th... |
3c726168-b71a-48a4-940c-2c9b26c35cb0 | StampyAI/alignment-research-dataset/eaforum | Effective Altruism Forum | Request for Assistance - Research on Scenario Development for Advanced AI Risk
Hi all,
I would greatly appreciate your perspectives on potential AI paths and outcomes through a brief survey. All questions are multiple-choice and will remain completely anonymous.
* [Surveymonkey version](https://www.surveymonkey.com... |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.