id stringlengths 36 36 | source stringclasses 15
values | formatted_source stringclasses 13
values | text stringlengths 2 7.55M |
|---|---|---|---|
df40e361-dc68-414b-b50b-8388e7baedc0 | trentmkelly/LessWrong-43k | LessWrong | 19th-century progress studies
I wrote earlier about an 1857 plan for the transcontinental railroad (as yet unbuilt), and how it advocated funding the project through the sale of stock to a broad base of public investors. However, another aspect of the plan was even more striking: it explicitly advocating studying the ... |
bbb08214-bc23-4ff5-9ed3-bbbefecfb331 | StampyAI/alignment-research-dataset/agisf | AGI Safety Fund | Model evaluation for extreme risks
1 Introduction
---------------
As AI progress has advanced, general-purpose AI systems have tended to display new and hard-to-forecast capabilities – including harmful capabilities that their developers did not intend (Ganguli2022-wp). Future systems may display even more dangerou... |
86aa69e5-302d-460f-a2cf-865ab40195dc | awestover/filtering-for-misalignment | Redwood Research: Alek's Filtering Results | id: post2695
Co-authored with Rebecca Gorman . In section 5.2 of their Arxiv paper, " The Incentives that Shape Behaviour ", which introduces structural causal influence models and a proposal for addressing misaligned AI incentives, the authors present
the following graph: The blue node is a "decision node", defined a... |
3605306b-f2f9-42f2-ac19-9283b54dbf38 | trentmkelly/LessWrong-43k | LessWrong | Rewilding the Gut VS the Autoimmune Epidemic
Did you know that an autoimmune disease now affects one in five Americans? Of these, 80% are women; autoimmune disease is the fifth leading cause of death in women under 65. Most autoimmune diseases are being diagnosed in increasing numbers, but it is still unclear if all t... |
ae2f7728-9cfd-4759-bd92-177c7496ae0d | trentmkelly/LessWrong-43k | LessWrong | Qualitative differences
Probably I am making a huge mistake resuming a ten years old discussion, especially because this is my first post. Anyway, let’s make this experiment.
One common argument for choosing “dust specks” over “torture” is that the experience of torture is qualitatively different form the experience ... |
8aee4fd0-7654-4aef-b68c-711f65cdf46d | trentmkelly/LessWrong-43k | LessWrong | To Whom I May Have Concerned: A standard explanation of my disagreement
EDIT: I put this here so that it can be easily on the Internet so I can link people to it. When I link it to people, they'll be here on Less Wrong. The good it does to people who are already here, I don't really know. I feel that that makes it app... |
4ed95bc3-58cd-4481-808e-25fa9367cda9 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Reflection Mechanisms as an Alignment target: A survey
This is a product of the [2022 AI Safety Camp](https://aisafety.camp/). The project has been done by [Marius Hobbhahn](https://www.mariushobbhahn.com/aboutme/) and [Eric Landgrebe](https://www.linkedin.com/in/eric-landgrebe-959698169/) under the supervision of [Be... |
2fd1fa48-e938-4c97-9841-da0986163529 | trentmkelly/LessWrong-43k | LessWrong | Meetup : Bratislava Meetup X.
Discussion article for the meetup : Bratislava Meetup X.
WHEN: 24 February 2014 06:00:00PM (+0100)
WHERE: Bistro The Peach, Heydukova 21, Bratislava
The same place; discussion about HP:MoR, and possibly other topics.
Na rovnakom mieste a v rovnakom čase ako zvyčajne, okrúhle desiate b... |
74c82e19-2a99-4b4d-911c-26766f9b9286 | trentmkelly/LessWrong-43k | LessWrong | Brief thoughts on inefficient writing systems
A phonemic language is one where how you write is very similar to how things sound. Essentially, if you know how to say something and you know the alphabet, then you can write/read the same thing. Serbian is an example of a phonemic language. A non-phonemic language is one... |
03796a32-36a8-4d28-b9c8-4f9a477c7a16 | trentmkelly/LessWrong-43k | LessWrong | Some Vacation Photos
In the past three months, I've flown overnight from San Francisco to the UK/Europe three times.
Three times, I got a window seat on the left side of the plane.
And three times... well, see for yourself.
First tripSecond tripThird trip
Key thing to know: these aurora did not look like this to th... |
b2040c73-a454-4bf1-bb29-b8fda2b374a9 | trentmkelly/LessWrong-43k | LessWrong | How do I start a programming career in the West?
I was expecting great things when I graduated from high school. I scored 35 on the ACT, was messing around with crypto, and fully expected to drop out of college to start a business. Things were looking up.
Instead, due to a combination of depression and procrastinatio... |
b96ce1b1-5a87-4d22-85d1-54fafcbd7560 | trentmkelly/LessWrong-43k | LessWrong | The unexpected difficulty of comparing AlphaStar to humans
This is crossposted from the AI Impacts blog.
Artificial intelligence defeated a pair of professional Starcraft II players for the first time in December 2018. Although this was generally regarded as an impressive achievement, it quickly became clear that not... |
201a9614-114e-4fba-a8f9-67a2122bb162 | trentmkelly/LessWrong-43k | LessWrong | Toward a Broader Conception of Adverse Selection
“I refuse to join any club that would have me as a member” -Marx[1]
Adverse Selection is the phenomenon in which information asymmetries in non-cooperative environments make trading dangerous. It has traditionally been understood to describe financial markets in which ... |
bd9e3ca2-3d91-4337-9ab2-5d32fc75e69d | trentmkelly/LessWrong-43k | LessWrong | Contest for outlining rules for this contest.
Each round of the contest will produce the rules governing the next round. Any more information would be too much of a hint. This is dumb and not worth spending time on; I nonetheless look forward to any submissions. |
ef6f906f-e5cf-420c-b2ab-a845140e329c | trentmkelly/LessWrong-43k | LessWrong | On Meaning and Machines
Cross-posted from mybrainsthoughts.com
Merriam Webster:
meaning \ˈmē-niŋ \ noun 1 a the thing one intends to convey especially by language b the thing that is conveyed especially by language 2 something meant or intended 3 significant quality 4 a the logical connotation of a word or phrase b t... |
4d0469b4-d9bc-4415-b867-a7e7e37a74d5 | trentmkelly/LessWrong-43k | LessWrong | Time and Effort Discounting
Related to: Akrasia, hyperbolic discounting, and picoeconomics
If you're tired of studies where you inevitably get deceived, electric shocked, or tricked into developing a sexual attraction to penny jars, you might want to sign up for Brian Wansink's next experiment. He provided secretarie... |
fcb06c35-d821-4a61-a7df-0fbbe7f1a3b5 | trentmkelly/LessWrong-43k | LessWrong | Democratic Currency
As an economics major, I have often heard it pronounced (largely correctly) that the great thing about money and the free market is that money is an extremely nimble decision-making process; money provides a way to measure how much value people can gain from something being created, or some service... |
16bec031-2713-42b2-b827-30fa35f44644 | StampyAI/alignment-research-dataset/blogs | Blogs | ML Living Library Opening
**Update Jan. 2021**: We’re no longer seeking applications for this position.
---
The Machine Intelligence Research Institute is looking for a very specialized autodidact to keep us up to date on developments in machine learning—a “living library” of new results.
ML is a fast-moving a... |
90b03e38-436d-4bec-8ebb-72aef499b31d | StampyAI/alignment-research-dataset/lesswrong | LessWrong | A summary of current work in AI governance
A summary of current work in AI governance
==========================================
If you’d like to learn more about AI governance, apply to the [AI Safety Fundamentals: Governance Track](https://www.agisafetyfundamentals.com/ai-governance-curriculum), a 12-week, part-tim... |
cd830953-11ab-4f7e-af54-b6836c7c9e60 | StampyAI/alignment-research-dataset/arbital | Arbital | B-Class
A B-Class page is fairly comprehensive, with good prose and structure; but needs further work to reach the A-Class requirement for engaging, professional, intuitive explanations, checked by asking for detailed feedback from both the target audience and [reviewers](https://arbital.com/p/5ft).
B-Class pages are... |
aca0b914-3e87-4c7f-b50b-0cb3e271fae1 | trentmkelly/LessWrong-43k | LessWrong | Q&A with experts on risks from AI #1
[Click here to see a list of all interviews]
Brandon Rohrer
Sandia National Laboratories
Cited by 536
Education
PhD, Mechanical Engineering, Massachusetts Institute of Technology, 2002.
Neville Hogan, Advisor and Thesis Committee Chair.
MS, Mechanical Engineering, Massachusett... |
6b4d360b-95bc-48e0-a4c1-26d488459ed3 | trentmkelly/LessWrong-43k | LessWrong | Three car seats?
I recently was linked to Car Seats as Contraception, which argues that car seat laws have led to parents choosing to have fewer children because "physical limitations of many cars precludes the use of three car seats in the back seat." Now, this is a correlational study, so I'm skeptical that this is ... |
43b49fd4-2a61-4417-bddc-962a483d0304 | trentmkelly/LessWrong-43k | LessWrong | Optimal predictors and propositional calculus
This is a writeup of the stuff Benja and I thought up during the logical uncertainty workshop in May.
An optimal predictor for (D,μ) allows assigning probabilities to logical sentences of the form x∈D but in general it doesn't allow assigning probabilities to propositiona... |
b8c2db5d-d099-4ee5-af21-07359fe179ae | trentmkelly/LessWrong-43k | LessWrong | [LINK] Father of Cryonics "Dies" at age 92
http://www.dailytech.com/Father+of+Cryonics+Dies+at+92+Frozen+in+Cryonics+Institute+/article22273.htm
Via Hacker News (comments: http://news.ycombinator.com/item?id=2813021) |
f190ee43-0391-4262-b6a1-fcc8fb8373b2 | trentmkelly/LessWrong-43k | LessWrong | What Is The Alignment Problem?
So we want to align future AGIs. Ultimately we’d like to align them to human values, but in the shorter term we might start with other targets, like e.g. corrigibility.
That problem description all makes sense on a hand-wavy intuitive level, but once we get concrete and dig into technic... |
b5c5e1fd-16d1-43e1-b306-9e42577427a1 | StampyAI/alignment-research-dataset/special_docs | Other | U.S. AI Workforce: Policy Recommendations
October 2021
U.S. AI Workforce: Policy Recommendations
CSET Policy Brief
AUTHORS Diana Gehlhaus Luke Koslosky Kayla Goode Claire Perkins
Center for Security and Emerging Technology | 1 Executive Summary The U.S. artificial intelligence workforce, which stood at 14 millio... |
56174adb-5be8-4f3e-aeca-c98b3b4046ab | trentmkelly/LessWrong-43k | LessWrong | Winners of AI Alignment Awards Research Contest
This post describes the winning submissions of AI Alignment Awards. We offered prizes for novel contributions to the problems of goal misgeneralization and corrigibility. (For more context, see this post and our site.)
Summary
We received 118 submissions. Of these, 5 ... |
326e839f-c6ff-4c4f-958a-ee1edbb6c510 | awestover/filtering-for-misalignment | Redwood Research: Alek's Filtering Results | id: post715
This is Section 6 of my report “ Scheming AIs: Will AIs fake alignment during training in order to get power? ”. There’s also a summary of the full report here (audio here ). The summary covers most of the main points and technical terms, and I’m hoping that it will provide much of the context necessary to... |
700b5229-a4a8-44bc-82f5-acd7b84be220 | trentmkelly/LessWrong-43k | LessWrong | Announcing AISIC 2022 - the AI Safety Israel Conference, October 19-20
ALTER, with the University of Haifa and the Technion, is excited to announce the details about its upcoming conference introducing AI safety in Israel. We’re going to be hosting Stuart Russell, as well as several other AI safety researchers, to spe... |
a35e3f03-767a-4723-8da6-446f47225c45 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Agents Over Cartesian World Models
Thanks to Adam Shimi, Alex Turner, Noa Nabeshima, Neel Nanda, Sydney Von Arx, Jack Ryan, and Sidney Hough for helpful discussion and comments.
Abstract
========
We analyze agents by supposing a Cartesian boundary between agent and environment. We extend partially-observable Marko... |
3f388efb-f9b4-4455-bfd6-e3c4359578a6 | trentmkelly/LessWrong-43k | LessWrong | How much to worry about the US election unrest?
Political tensions in the US are high. There have been a lot of mass protests in recent months, some of which have turned violent (often after police provocation). Donald Trump and Mike Pence have not to promise to respect the election result if they lose. Many on the le... |
b7e7a419-7747-4749-8d3f-852fb16d6868 | trentmkelly/LessWrong-43k | LessWrong | Why is trying to save the world (tm) psychologically taxing?
This is a bit of a brain dump, not much new. I'm putting them down to see if any solutions come to mind.
Difficulty of problems
The problems we have are hard. You don't get the dopamine hit of solving something or getting money from a product. You have to ... |
1f72ef2c-ff0d-4c8e-8ed2-ae9c3d81f3a2 | trentmkelly/LessWrong-43k | LessWrong | Am I anti-social if I get vaccinated now?
In the West, or at least in my home country (Switzerland), we have ample vaccine supply, and are - at the moment - controlling Covid more or less fine.
I read of urges to provide vaccines to poor countries where even health workers could not get vaccinated due to lack of vacc... |
e5d85cc0-9f36-496d-962f-ce2073ee7ef4 | StampyAI/alignment-research-dataset/blogs | Blogs | New paper: “Defining human values for value learners”
[](https://intelligence.org/files/DefiningValuesForValueLearners.pdf)MIRI Research Associate Kaj Sotala recently presented a new paper, “**[Defining Human Values for Value Learners... |
6ec6e206-223f-4276-977d-f90d92965f52 | trentmkelly/LessWrong-43k | LessWrong | LW-ish meetup in Boulder, CO
This Saturday at the Hellems Arts and Sciences building, room 185 at the University of Colorado at Boulder, there will be a presentation on Intelligence Explosion dynamics.
Hope to see anyone there that can make it. |
2f214259-d5bb-4c91-a7e7-7369249c4884 | trentmkelly/LessWrong-43k | LessWrong | Kahneman's Planning Anecdote
Followup to: Planning Fallacy
From "Timid Choices and Bold Forecasts: Cognitive Perspective on Risk Taking" by Nobel Laureate Daniel Kahneman and Dan Lovallo, in a discussion on "Inside and Outside Views":
> In 1976 one of us (Daniel Kahneman) was involved in a project designed to devel... |
1f882357-3169-4e29-9406-2a8c6541ccd8 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Is Agent Simulates Predictor a "fair" problem?
It's a simple question, but I think it might help if I add in context. In the paper introducing [Functional Decision Theory](https://arxiv.org/pdf/1710.05060.pdf), it is noted that it is impossible to design an algorithm that can perform well on all decision problems sinc... |
cf7b603e-b328-4094-aabd-6092129a1988 | trentmkelly/LessWrong-43k | LessWrong | Small Go Boards
I'm a big fan of playing go on small boards: 5x5, 7x7, 9x9. The smaller the board the shorter the game, and when you're new playing lots of games quickly is better for learning and keeping it fun. I also like that on smaller boards you're more able to substitute reasoning for go-specific intuition, giv... |
b0065abc-bf07-4dbd-94a0-bbc259f30eb8 | trentmkelly/LessWrong-43k | LessWrong | How Doomed are Large Organizations?
We now take the model from the previous post, and ask the questions over the next several posts. This first answer post asks these questions:
1. Are these dynamics the inevitable results of large organizations?
2. How can we forestall these dynamics within an organization?
3. To... |
31014aef-aa3a-4b75-a047-2d11cbf40702 | StampyAI/alignment-research-dataset/blogs | Blogs | In-context Learning and Induction Heads
### Contents
[Key Concepts](#key-concepts)
* [In-context Learning](#in-context-learning-key-concept)
* [Induction Heads](#definition-of-induction-heads)
* [Per-Token Loss Analysis](#per-token-loss-intro)
[Arguments Summary](#toc-arguments)
[Argument 1 (Macroscopic co-occurre... |
4f033f2e-a480-4525-90b5-1f75a3305de3 | trentmkelly/LessWrong-43k | LessWrong | A quick note on weirdness points and Solstices [And also random other Solstice discussion]
Common knowledge is important. So I wanted to note:
Every year on Solstice feedback forms, I get concerns about songs like "The X days of X-Risk" or "When I Die" (featuring lines including 'they may freeze my body when I die'),... |
c0e53264-9405-4bf7-8aae-6e90cc369e92 | trentmkelly/LessWrong-43k | LessWrong | Meetup : London Social Meetup, 21/12/2014
Discussion article for the meetup : London Social Meetup, 21/12/2014
WHEN: 21 December 2014 02:00:00PM (+0000)
WHERE: Shakespeare's Head, Africa House, 64-68 Kingsway, London WC2B 6BG, UK-68 Kingsway, London WC2B 6BG, UK
LessWrong London is having another meetup this Sunday... |
36de1ca5-92a2-4e41-a426-e055a01da98a | trentmkelly/LessWrong-43k | LessWrong | The Lighthaven Campus is open for bookings
Lightcone Infrastructure (the organization that grew from and houses the LessWrong team) has just finished renovating a 7-building physical campus that we hope to use to make the future of humanity go better than it would otherwise.
We're hereby announcing that it is general... |
99ec7dff-75ad-4a59-88a5-8bbcaba215e3 | trentmkelly/LessWrong-43k | LessWrong | Cast it into the fire! Destroy it!
We should only use AGI once to make it so that no one, including ourselves, can use it ever again.
I'm terrified of both getting atomized by nanobots and of my sense of morality disintegrating in Extremistan. We don't need AGI to create a post-scarcity society, cure cancer, solve cl... |
dcc453ad-2cee-454c-a4ca-e2a082039bf7 | trentmkelly/LessWrong-43k | LessWrong | Go Try Things
So this isn't quite done, and its late here so I don't quite trust my judgements about writing at this hour. I've never done a top-level post before, so I wanted to get some feedback first.
----------------------------------------
Failure isn’t that Bad
You’ve probably read about how to properly tu... |
cb0728c3-ef5a-4cf9-81ac-8a511d527dba | trentmkelly/LessWrong-43k | LessWrong | Two Zendo-inspired games
LW has often discussed the inductive logic game Zendo, as a possible way of training rationality. But I couldn't find any computer implementations of Zendo online.
So I built two (fairly similar) games inspired by Zendo; they generate rules and play as sensei. The code is on GitHub, along wit... |
6ce3930f-3655-4592-b26c-c36219ca971e | trentmkelly/LessWrong-43k | LessWrong | Will people be motivated to learn difficult disciplines and skills without economic incentive?
In a recent interview with ABC news, Sam Altman emphasised multiple times that he sees individual AI tutoring as one of the greatest applications of AI:
> I'm most excited about is the ability to provide individual learning... |
9727ebf1-1d5c-4501-aced-e26dd57bf282 | trentmkelly/LessWrong-43k | LessWrong | At what point does disease spread stop being well-modeled by an exponential function?
Early in the spread of a new disease, its growth rate is exponential. But obviously, no disease can maintain that rate forever, it will soon run out of new people to infect. It has to level off at some point.
At what point (at what ... |
844ce62d-4353-4789-bd87-f9e8fbd5ab6f | StampyAI/alignment-research-dataset/eaforum | Effective Altruism Forum | A Guide to Forecasting AI Science Capabilities
*The following contains resources that I (Eleni) curated to help the* [*AI Science team*](https://docs.google.com/document/d/1J6aadVK6b2nQGaVNPu0bV13ki0ijEy576CFGF0qV3oo/edit#heading=h.9lmc73wscx1r) *of AI Safety Camp 2023 prepare for the second half of the project, i.e.,... |
c83cdd50-35b3-49f2-afaa-b8644ea47d1e | trentmkelly/LessWrong-43k | LessWrong | Some criteria for sandwiching projects
I liked Ajeya's post a lot, and I think the alignment community should try to do sandwiching projects along the lines she describes. Here I wanted to flesh out some potential criteria for a good sandwiching project; there's not too much original thinking here but I found it helpf... |
56773814-d942-4a86-b0fe-693912634956 | trentmkelly/LessWrong-43k | LessWrong | Catastrophic Goodhart in RL with KL penalty
TLDR: In the last two posts, we showed that optimizing for a proxy can fail to increase true utility, but only when the error is heavy-tailed. We now show that this also happens in RLHF with a KL penalty.
This post builds on our earlier result with a more realistic setting ... |
7f802e46-3e39-47d8-84e4-9a482870c6b2 | trentmkelly/LessWrong-43k | LessWrong | Ex nihilo
As a quick announcement, Dr. Martin Burckhardt has started a new blog on Substack called Ex nihilo. Dr. Burckhardt is a German cultural philosopher well-known in Germany for his out-of-the-box thinking on the Question of the Machine, how it relates to Artificial Intelligence, and the cultural implications of... |
9b25ab7a-0ee9-4198-be0d-60353ff36e77 | trentmkelly/LessWrong-43k | LessWrong | Fixing science via a basic income
I ran across Ed Hagen’s article “Academic success is either a crapshoot or a scam”, which pointed out that all the methodological discussion about science’s replication crisis is kinda missing the point: yes, all of the methodological stuff like p-hacking is something that would be va... |
0691b9f2-273d-4267-b9a1-08d8341b0533 | trentmkelly/LessWrong-43k | LessWrong | Build knowledge base first, or backchain?
Specifically for AI alignment (small field, preparadigmatic, plausibly-short-timelines), but a general principle decision-mechanism could apply to other fields.
What I mean by this is, what ratio of pre-learning to filling-in-gaps, leads to deeper and quicker insights? (Note ... |
ee479c67-d7a5-4ab9-a2ad-cba63906da46 | trentmkelly/LessWrong-43k | LessWrong | RLHF is the worst possible thing done when facing the alignment problem
Epistemic status: The title might not be literally true in the sense that e.g. if the inventors of RLHF hadn't come up with it then someone else probably would, so the counterfactual effect is small, or e.g. that the worst possible thing you could... |
8bc50b23-12e8-4eb1-b241-3e9579c4d0bf | trentmkelly/LessWrong-43k | LessWrong | [SEQ RERUN] Crisis of Faith
Today's post, Crisis of Faith was originally published on 10 October 2008. A summary (taken from the LW wiki):
> When an idea has become deeply entrenched in your mind, it requires an awful lot of work to doubt that idea properly.
Discuss the post here (rather than in the comments to ... |
7e663cac-8a14-4ab6-ba4e-a3cdbadf9bfc | StampyAI/alignment-research-dataset/lesswrong | LessWrong | Newcomb's Problem: A problem for Causal Decision Theories
**This is part of a sequence titled, "Introduction to decision theory"**
**The previous post is ["An introduction to decision theory"](/lw/2l5/an_introduction_to_decision_theory/ "An introduction to decision theory")**
In the previous post I introduced ev... |
f9623690-e686-425b-bb49-cb7d3227c905 | trentmkelly/LessWrong-43k | LessWrong | I'd like advice from LW regarding migraines
So, I read a post a little while ago saying that asking the community for advice on personal problems was okay, and no one seemed to disagree strongly with this. Therefore, I'll just ask for some advice, and hope that I'm not accidentally going past some line. If I do, I apo... |
56e94f78-9b50-4f0c-855b-b4a6b6f8c8bc | LDJnr/LessWrong-Amplify-Instruct | LessWrong | "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. Add other people to your sensorium. Make conversation with others a big ... |
77a4d33d-0a01-432c-98e0-c7523aa8c62d | trentmkelly/LessWrong-43k | LessWrong | Why don't governments seem to mind that companies are explicitly trying to make AGIs?
Epistemic Status: Quickly written, uncertain. I'm fairly sure there's very little in terms of the public or government concerned about AGI claims, but I'm sure there's a lot I'm missing. I'm not at all an expert on government or poli... |
42d52c83-0728-432c-88db-6987252789cb | trentmkelly/LessWrong-43k | LessWrong | Rationality Quotes 18
Q. Why does "philosophy of consciousness/nature of reality" seem to interest you so much?
A. Take away consciousness and reality and there's not much left.
-- Greg Egan, interview in Eidolon 15
"But I am not an object. I am not a noun, I am an adjective. I am the way matter behaves wh... |
b2635dfe-1b6b-40e9-8498-b4b9fbd11a6e | trentmkelly/LessWrong-43k | LessWrong | A "Bitter Lesson" Approach to Aligning AGI and ASI
TL;DR: I discuss the challenge of aligning AGI/ASI, and outline an extremely simple approach to aligning an LLM: train entirely on a synthetic dataset that always shows the AI acting aligned (even when the humans behave badly), and use a conditional training/inference... |
f9095d78-0349-4d1c-8e64-432e4e041fce | StampyAI/alignment-research-dataset/lesswrong | LessWrong | I Vouch For MIRI
Another take with more links: [AI: A Reason to Worry, A Reason to Donate](https://putanumonit.com/2017/12/10/worried-about-ai/)
I have made a $10,000 donation to the [Machine Intelligence Research Institute (MIRI)](https://intelligence.org/2017/12/01/miris-2017-fundraiser/) as part of their winter f... |
9941b008-7c10-40a0-a04a-47c086f5be10 | trentmkelly/LessWrong-43k | LessWrong | A Gentle Introduction to Risk Frameworks Beyond Forecasting
This was originally posted on Nathaniel's and Nuno's substacks (Pending Survival and Forecasting Newsletter, respectively). Subscribe here and here!
Discussion is also occurring on the EA Forum here (couldn't link the posts properly for technical reasons).
... |
e09bbeb9-7b0f-4dfc-913d-eac17349acf2 | trentmkelly/LessWrong-43k | LessWrong | "Meetup" proposal: Google+
Not all of us live near big cities. I know of only one Less Wronger in the Grand Rapids, MI area apart from myself, and the nearest existing meetup is in Madison, WI, across Lake Michigan. I'd like to go to a Less Wrong meetup with more than just a couple of people, but I don't want to have ... |
240ea61d-adeb-46ca-a75e-5da0fc13aeb8 | trentmkelly/LessWrong-43k | LessWrong | Can you get AGI from a Transformer?
UPDATE IN 2023: I wrote this a long time ago and you should NOT assume that I still agree with all or even most of what I wrote here. I’m keeping it posted as-is for historical interest.
Introduction
I want to share my thoughts about the calculations that Transformers (such as GPT... |
803f943d-7a3d-4650-a72a-0d984072126f | trentmkelly/LessWrong-43k | LessWrong | Meta-level adversarial evaluation of oversight techniques might allow robust measurement of their adequacy
Summary: Many proposed AGI alignment procedures involve taking a pretrained model and training it using rewards from an oversight process to get a policy. These procedures might fail when the oversight procedure ... |
ea8c1ffc-faca-4edd-8e66-5df2998d691b | trentmkelly/LessWrong-43k | LessWrong | Shame: Authentic & Applied/Manufactured Shame
I've been chewing on the book The Language of Emotions by Karla McLaren a lot, lately. It gives a surprising and (to me) unique perspective on emotions and how they behave when they're able to act in their healthy un-stuck ways ("Flowing" as the book calls it).
Shame in p... |
b55ffcae-a300-4917-aa84-ec694326a008 | StampyAI/alignment-research-dataset/distill | Distill Scientific Journal | Distill Update 2018
#### [Things that Worked Well](#successes)
* [Interfaces for Ideas](#interfaces)
* [Engagement as a Spectrum](#engagement)
* [Software Engineering Best Practices for Scientific Publishing](#best-practices)
#### [Challenges & Improvements](#challenges)
* [The Disti... |
b11c1efa-e52f-4033-bcac-de3d3c127aee | trentmkelly/LessWrong-43k | LessWrong | Seek Fair Expectations of Others’ Models
Epistemic Status: Especially about the future.
Response To (Eliezer Yudkowsky): There’s No Fire Alarm for Artificial General Intelligence
It’s long, but read the whole thing. Eliezer makes classic Eliezer points in classic Eliezer style. Even if you mostly know this already,... |
31618735-2dd4-4869-921d-cbfbf954b47f | trentmkelly/LessWrong-43k | LessWrong | Contra double crux
Summary: CFAR proposes double crux as a method to resolve disagreement: instead of arguing over some belief B, one should look for a crux (C) which underlies it, such that if either party changed their mind over C, they would change their mind about B.
I don't think double crux is that helpful, pri... |
b05386ed-7839-4f4c-baee-5233f87ec154 | trentmkelly/LessWrong-43k | LessWrong | AI catastrophes and rogue deployments
[Thanks to Aryan Bhatt, Ansh Radhakrishnan, Adam Kaufman, Vivek Hebbar, Hanna Gabor, Justis Mills, Aaron Scher, Max Nadeau, Ryan Greenblatt, Peter Barnett, Fabien Roger, and various people at a presentation of these arguments for comments. These ideas aren’t very original to me; m... |
d1d2d1fd-b9d5-4993-9dcf-bc2e74148b36 | trentmkelly/LessWrong-43k | LessWrong | How useful is Corrigibility?
Epistemic Status: My best guess (I'm new to AI safety)
We don't know how to formally define corrigibility and this is part of the reason why we haven't solved it so far. Corrigibility is defined in Arbital as:
> [the agent] doesn't interfere with what we would intuitively see as attempts... |
77dac415-4139-4ddd-bc44-95753c05e5e7 | trentmkelly/LessWrong-43k | LessWrong | The problem of pseudofriendliness
The Friendly AI problem is complicated enough that it can be divided into a large number of subproblems. Two such subproblems could be:
1. The problem of goal interpretation – This means that the human expectation of the results of an AI implementing a goal differ from the results t... |
01131e75-4730-417b-85b2-cb4f8b01b52a | trentmkelly/LessWrong-43k | LessWrong | Meetup : First meetup in Lund
Discussion article for the meetup : First meetup in Lund
WHEN: 19 October 2015 06:00:00PM (+0200)
WHERE: Sölvegatan 18, 223 62, Lund, Sweden
Hello, this is going to be the first meetup I know about in Skane area. Since I don't know any other LessWrongers here, I'll be waiting in front ... |
440cbbac-7762-442f-a468-e39a37c86283 | awestover/filtering-for-misalignment | Redwood Research: Alek's Filtering Results | id: post143
TL;DR : The Google DeepMind AGI Safety team is hiring for Applied Interpretability research scientists and engineers . Applied Interpretability is a new subteam we are forming to focus on directly using model internals-based techniques to make models safer in production . Achieving this goal will require d... |
e5b4cc6d-2aa5-43ff-affd-2d9598e84aa1 | trentmkelly/LessWrong-43k | LessWrong | Progress links and tweets, 2022-07-27
Opportunities
* Arcadia Science launches an Entrepreneur-in-Residence (EIR) program (thread from @ArcadiaScience)
* The Longevity Prize goes to “outstanding progress in longevity research” (announcement from @allisondman)
* Nicole Ruiz is making a list of opportunities in front... |
26e6fa0b-29c1-46b9-a010-0a00265e0b43 | StampyAI/alignment-research-dataset/blogs | Blogs | clippy in panpsychia
clippy in panpsychia
--------------------
[panpsychism](https://en.wikipedia.org/wiki/Panpsychism) is the view that mindstuff is the fundamental substrate of the cosmos, and what appears to us like material reality is generated or hallucinated by those minds.
but importantly, these hallucinati... |
4f488403-f390-43af-9a5a-0a90d60853fd | trentmkelly/LessWrong-43k | LessWrong | Selection Pressures on LM Personas
TLDR:
* Though we tend to have LMs play the role of a helpful assistant, they can in fact generate the responses of any persona you ask for
* It’s not just explicit — the persona you get is also steered by implicit cues, and confined by what it’s easy for the LM to generate
* The... |
2845f21e-d821-4b63-87af-e8b83bb9f6a3 | trentmkelly/LessWrong-43k | LessWrong | Does the body have an almost infinite number of potential positions?
Edited to add: It was the wording not the concept that was niggling. Thank you for all the replies.
> The body is capable of innumerable positions.
Better, but still subject to further consideration and I've realised I was getting side-tracked.
I... |
2dd47ba6-28c2-4bfd-9c60-416d84f82aa9 | StampyAI/alignment-research-dataset/lesswrong | LessWrong | Proposal: labs should precommit to pausing if an AI argues for itself to be improved
As evidence that compute is a major bottleneck on capabilities has accumulated, many people have become more skeptical of extremely fast takeoff speeds. One major reason for this is that if humans were trying to stop it, it would prob... |
f0867c96-bcf6-44c3-bd45-c48e4627e61b | trentmkelly/LessWrong-43k | LessWrong | Status for status sake is a fact of political life
Epistemic status: Exploring a new area (new to me)
Recently I was listening to Political Scientist Bruce Bueno De Mesquita on the Econtalk podcast. BdM had just established that Jefferson Davis, when offered the presidency of the confederacy, already knew the South w... |
4fc2e45f-7204-4625-a2a5-1683f61e1c00 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Apply for research internships at ARC!
**(Update: we are no longer accepting applications for interns.)**
In addition to [hiring full-time researchers](https://www.alignmentforum.org/posts/dLoK6KGcHAoudtwdo/arc-is-hiring), ARC is also hiring for 1-3 month internships. We’re open to interns for summer 2022 in particul... |
ef60e43d-788b-4c2f-bbc3-7ccf236b51de | StampyAI/alignment-research-dataset/lesswrong | LessWrong | You won’t solve alignment without agent foundations
The following is a part of a response to a Vivek Hebbar's problem (the questions in *cursive*) in my SERI MATS application. I looked at why an alignment proposal doesn't work at an AGI level, in connection to only two of [Eliezer's doom arguments](https://www.lesswro... |
7d8bce60-adcd-4e0a-af84-20e023e5dcec | StampyAI/alignment-research-dataset/arxiv | Arxiv | On Avoiding Power-Seeking by Artificial Intelligence
On Avoiding Power-Seeking
by Articial Intelligence
/a.sc /d.sc/i.sc/s.sc/s.sc/e.sc/r.sc/t.sc/a.sc/t.sc/i.sc/o.sc/n.sc /p.sc/r.sc/e.sc/s.sc/e.sc/n.sc/t.sc/e.sc/d.sc
/b.sc/y.sc
A/l.sc/e.sc/x.sc/a.sc/n.sc/d.sc/e.sc/r.sc M/a.sc/t.sc/t.sc T/u.sc/r.sc/n.sc/e.sc/r.sc
/t.sc... |
37311f35-cee5-4def-ae3a-0d234d298360 | trentmkelly/LessWrong-43k | LessWrong | New(ish) AI control ideas
The list of posts is getting unwieldy, so I'll post the up-to-date stuff at the beginning:
Humans inconsistencies:
* Bias in rationality is much worse than noise
* Learning values, or defining them?
* Rationality and overriding rewards: combined model
* Humans can be assigned any values... |
396ecff3-ea66-47d7-aa49-02181820ae1f | trentmkelly/LessWrong-43k | LessWrong | Selection processes for subagents
This post consists of rough thoughts on possible training processes that select for multi-agent architectures, or “subagents”, in neural networks. I deliberately ignore delegative subagents that are instantiated by mesa-optimizers and instead focus on subagent architectures that arise... |
67e22d8e-1270-4282-ad47-851bf480991d | trentmkelly/LessWrong-43k | LessWrong | News: AGI on the rise at MIT
I've noticed recently that MIT is taking more notice of AGI.
There is an on-line course and recently announced MIT Intelligence Quest and a news item on it.
Worth keeping an eye on.
|
e617df61-1cf3-4c16-8abf-6a059a31b827 | trentmkelly/LessWrong-43k | LessWrong | CDT vs. EDT on Deterrence
This seems like a natural question, but I haven't been able to find this topic discussed online.
We are in a world with two continents on opposite sides. Each continent contains a single nation with nuclear weapons. Each side has one person with a button in front of them that will launch nuc... |
231d3f5a-0f39-42e3-9a38-2924b28cead0 | StampyAI/alignment-research-dataset/eaforum | Effective Altruism Forum | Good policy ideas that won’t happen (yet)
Key points:
-----------
* Over the last year the Centre for Effective Altruism has had over a dozen meetings with UK policymakers and tested the waters on a range of policies we thought might have significant positive benefits for the world.
* In most cases we quickly found ... |
dcd2d457-4f23-4aa2-bbf5-2046e1eabae1 | trentmkelly/LessWrong-43k | LessWrong | In balance and flux
Someone more familiar with ecology recently noted to me that it used to be a popular view that nature was ‘in balance’ and had some equilibrium state, that it should be returned to. Whereas the new understanding is that there was never an equilibrium state. Natural systems are always changing. Anot... |
e61685b9-4c55-4274-800f-2d52963e5839 | trentmkelly/LessWrong-43k | LessWrong | AI #95: o1 Joins the API
A lot happened this week. We’re seeing release after release after upgrade.
It’s easy to lose sight of which ones matter, and two matter quite a lot.
The first is Gemini Flash 2.0, which I covered earlier this week.
The other is that o1, having turned pro, is now also available in the API.
... |
4a129368-0374-4233-82e8-d14e85436830 | StampyAI/alignment-research-dataset/blogs | Blogs | Subsystem Alignment
You want to figure something out, but you don’t know how to do that yet.
You have to somehow break up the task into sub-computations. There is no atomic act of “thinking”; intelligence must be built up... |
a57a12e2-f034-4846-a346-ec41b0a4dcfe | trentmkelly/LessWrong-43k | LessWrong | Find a study partner - May 2014 Thread
For reasons mentioned in So8res article as well as for other reasons: studying with a partner can be very good.
So if you're looking for a study partner for an online course, reading a manual or else (whether it's in the MIRI course list or not) tell others in the comment sectio... |
b09804b5-0a93-45b7-b736-bf30728c1d5b | trentmkelly/LessWrong-43k | LessWrong | Less Laptop Velcro
A year ago I broke my laptop screen, and took the opportunity to build something I've always wanted: a monitor that folds vertically so I don't have to bend my neck:
A few months ago my cracked-screen laptop finished dying, and I got a new one. I use the stacked monitor a lot less now, since f... |
c7447be2-0f41-409e-a497-847cdbdc4c3d | StampyAI/alignment-research-dataset/arxiv | Arxiv | Definitions of intent suitable for algorithms
1 Introduction
---------------
The establishment of criminal intent, the volitional component of mens rea in the someone accused of committing a crime, is a necessary task in proving that a crime was committed. An autonomous algorithm with agency can perform actions whi... |
2047f19c-a5aa-4bf1-9f51-8f2f00ffbd7a | trentmkelly/LessWrong-43k | LessWrong | A defense on QI
I dont know if this is the right place to post but i need help finding a debunking of this article https://arxiv.org/ftp/arxiv/papers/0902/0902.0187.pdf I am sure many worlds DOES imply immortality but i cant seem to adress all the big words the article uses .if this is already done please just link th... |
c7e0689d-d35b-4d27-b1e2-1357d344328f | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | The “no sandbagging on checkable tasks” hypothesis
*(This post is inspired by*[*Carl Shulman’s recent podcast with Dwarkesh Patel*](https://www.dwarkeshpatel.com/p/carl-shulman?nthPub=521#details)*, which I highly recommend. See also discussion from Buck Shlegeris and Ryan Greenblatt*[*here*](https://www.lesswrong.com... |
cdad4a35-0f85-4128-b554-c73f74825f31 | trentmkelly/LessWrong-43k | LessWrong | The Practitioner's Path 2.0: A new framework for structured self-improvement
The Guild of the ROSE is proud to announce the full launch of the Practitioner's Path 2.0, a new framework for structured self-improvement.
Two years ago, we founded the Guild to provide structure and community for people interested in self-... |
a33ac17a-3c76-4bc7-9ecd-887e9f2971d2 | trentmkelly/LessWrong-43k | LessWrong | Equilibria in adversarial supervised learning
Note: This describes an idea of Jessica Taylor's.
Introduction
An important aim in safe AI development is to make systems whose submodules analyze how they might go wrong. One model for this in academic machine learning research is training on adversarial examples.
In o... |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.