id stringlengths 36 36 | source stringclasses 15
values | formatted_source stringclasses 13
values | text stringlengths 2 7.55M |
|---|---|---|---|
868fb8f6-2c9f-419d-95ae-afe89d56acd1 | trentmkelly/LessWrong-43k | LessWrong | New book on atheism, transhumanism, and x-risk
Phil Torres is the creative force behind the highly enjoyable folk music of Baobab, and he also writes philosophy papers (under the name "Philippe Verdoux").
His forthcoming book may be of interest to LWers: A Crisis of Faith: Atheism, Emerging Technologies, and the Futu... |
c2687f0d-8839-4acf-84bd-9613035c561b | trentmkelly/LessWrong-43k | LessWrong | Does the ChatGPT (web)app sometimes show actual o1 CoTs now?
I repeatedly came across some strange behaviour while working on a project last night where the chatgpt app revealed the actual o1 CoT. The CoT had a arrowhead button next to it clicking which revealed the summary, similar to the standard "Thought about <top... |
cc3d2484-9cb3-40b9-90e0-20513a3d20e7 | trentmkelly/LessWrong-43k | LessWrong | Safe AI and moral AI
[Note: This post is an excerpt from a longer paper, written during the first half of the Philosophy Fellowship at the Center for AI Safety. This post is something of a companion piece to my deontological AI post; both were originally written as parts of a single paper. (There's a small amount of o... |
8c0d353d-25cc-446b-9de0-939139631870 | trentmkelly/LessWrong-43k | LessWrong | The necklaces are known basically for their deep faith based connotations.
The jewelry are known generally for their deep faith based connotations. They are put on around the neck while protective pandora 2010 versus evil powers and many types of dangers. They are also extremely necessary for fashion make-up. Nonethe... |
dc3d983d-15b0-4c7b-b458-008f6eb20bad | trentmkelly/LessWrong-43k | LessWrong | Meetup : London Social - August 24th
Discussion article for the meetup : London Social - August 24th
WHEN: 24 August 2014 02:00:00PM (+0100)
WHERE: Holborn, London
The next London meetup will be Sunday, August 24th. This will be a social meetup with no special content. Come along and discuss all the weird LessWrong... |
4520970b-fd4b-4c67-8e56-e6b1ce21b1ea | trentmkelly/LessWrong-43k | LessWrong | Cracked wrote an Article about Life-Extension
7 Random Things You Won't Believe Are Shortening Your Life
TL;DR
1. Starting School Early
2. Being the Eldest Child
3. Having a Small Ass
4. Flying Often
5. Living Closer to Sea-Level
6. Not Going to Church.
7. Retiring Early.
I'm glad a high-traffic site such lik... |
610dcb31-7011-44e8-b4a8-7f2edd19ce57 | trentmkelly/LessWrong-43k | LessWrong | Other Papers About the Theory of Reward Learning
This is the seventh post in the theoretical reward learning sequence, which starts in this post. Here, I will provide shorter summaries of a few additional papers on the theory of reward learning, but without going into as much depth as I did in the previous posts (but ... |
a09cf05f-f31f-4ce7-a17f-4dc871ab4961 | trentmkelly/LessWrong-43k | LessWrong | Obvious identity fail
Paul Graham points out something important: religion and politics are generally unfruitful topics of discussion because people have identities tied to them.
An implication:
> The most intriguing thing about this theory, if it’s right, is that it explains not merely which kinds of discussions to... |
868a0514-d96f-42eb-a92f-989d995df806 | awestover/filtering-for-misalignment | Redwood Research: Alek's Filtering Results | id: post717
This is Section 4.4-4.7 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 necess... |
44139b32-ceb4-42f8-a43f-15022131586b | trentmkelly/LessWrong-43k | LessWrong | The Garden of Eden
[This is the text of the sermon given by Pastor James Windrow on Sunday, July 14, 2024.]
Greetings, brothers and sisters. Today, we reflect on humanity’s origin in the Garden of Eden and its fall and devolution thereafter.
When Christians used to think about the Garden of Eden, it was common to im... |
bb0bd757-c1e6-4952-b840-49a2da6ec2bb | trentmkelly/LessWrong-43k | LessWrong | Trapping AIs via utility indifference
If you wanted to check whether your creations were prone to misbehaving, one way would be to leave a big juicy apple hanging tantalisingly from a very visible tree, turn your back for a while, and see what they get up to.
That would work if you were God and your creations were es... |
4197f60d-ec14-4420-849d-97882d8a92c6 | trentmkelly/LessWrong-43k | LessWrong | Bag-Drying Clips
With the amount of buying from the commercial supply chain that we've been doing, I've been needing to divide up a lot of large containers. We typically use ziplock bags, and they're sturdy enough that they could be washed and reused. We mostly haven't been reusing them, though, because while I don't ... |
66712b97-d5a1-4f31-85d4-9a647e2a6b50 | StampyAI/alignment-research-dataset/arxiv | Arxiv | Resource Allocation to Agents with Restrictions: Maximizing Likelihood with Minimum Compromise
1 Introduction
---------------
There are many practical contexts where a set of agents
must be suitably matched with a set of resources.
Examples of such contexts include matching classes with classrooms [[24](#bib.bib24)... |
f5a3f5b0-a5f4-4314-a760-f5acfef2ddf8 | trentmkelly/LessWrong-43k | LessWrong | T-Mobile: Spurious Account Takeover Warning
This morning, reading in bed, I got a very worrying notification:
This is the notification you would receive if someone was in the process of taking control of your phone number, which could then give them access to other accounts where you had used that phone number as a... |
cebb3f18-fa8b-4f66-aac6-2aec1f43c6f2 | StampyAI/alignment-research-dataset/lesswrong | LessWrong | Alignment 101 - Ch.1 - AGI
Overview
========
1. **Foundation Models**: The chapter begins with an exploration of how contemporary approaches in machine learning lean towards the development of centralized foundation models. The section elaborates the merits and drawbacks of such a paradigm.
2. **Leveraging Computatio... |
f4a07c38-578a-4009-b821-3d1bdc61a390 | trentmkelly/LessWrong-43k | LessWrong | Why you should attend EA Global and (some) other conferences
Many of you know about Effective Altruism and the associated community. It has a very significant overlap with LessWrong, and has been significantly influenced by the culture and ambitions of the community here.
One of the most important things happening in... |
b441482c-56dd-4363-99a1-dfd0503daef3 | trentmkelly/LessWrong-43k | LessWrong | Trajectory
Position:
What's your goal? What do you want? What do you want different in this world? What should be different in this world?
Where are you now?
Velocity:
Where are you headed? Where are your actions pointing towards? Which actions are pushing towards your goal? Which actions are pulling away?
What's ... |
0d96bbc0-b853-489a-926e-c4f0a9779d5c | StampyAI/alignment-research-dataset/lesswrong | LessWrong | Would a halfway copied brain emulation be at risk of having different values/identity?
When I was thinking about the concept of human brain emulation recently, a disturbing idea occurred to me. I have never seen anyone address it, so I suspect it is probably caused by my being deeply confused about either human neurob... |
c26da33f-410a-40d3-a65c-fe949df1c339 | trentmkelly/LessWrong-43k | LessWrong | Explained: Gödel's theorem and the Banach-Tarski Paradox
I want to share the following explanations that I came across recently and which I enjoyed very much. I can't tell and don't suspect that they come close to an understanding of the original concepts but that they are so easy to grasp that it is worth the time if... |
a6928086-419b-4b42-9698-10cc61bbb2cd | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Announcement: AI alignment prize round 3 winners and next round
We (Zvi Mowshowitz and Vladimir Slepnev) are happy to announce the results of the third round of the [AI Alignment Prize](https://www.lesswrong.com/posts/SSEyiHaACSYDHcYZz/announcement-ai-alignment-prize-round-2-winners-and-next), funded by Paul Christian... |
68e8eb40-df44-4a39-bac8-debd274771c9 | trentmkelly/LessWrong-43k | LessWrong | "The Perfect is the Enemy of the Good", 80/20 Rule, INTPish, and Rationality
http://www.reddit.com/r/INTP/comments/i27i1/does_anyone_agree_that_this_is_the_biggest/
What are your thoughts on this thread? |
a1c62cdf-5f39-4bfb-8520-7a221e0fef07 | trentmkelly/LessWrong-43k | LessWrong | Communities: A single moderator is often superior to the wisdom of crowds
Less Wrong is a community where I believe that most of the people would accept that "knowledge is power". In our modern age, the internet provides some of the best tools for distributing this information. Many tools, such as Reddit, StackExchang... |
644ef3cc-5145-4989-a715-5b67c6e4edcd | trentmkelly/LessWrong-43k | LessWrong | What Caplan’s "Missing Mood" Heuristic Is Really For
Bryan Caplan wrote and quoted the idea of a "proper reaction," the "specific package of moods" that a "reasonable person" holding an intellectual position would have. If this mood is missing, it is grounds for suspicion, and we can "learn a lot" from it. It's a "val... |
df1cbd32-c9ba-414b-a4af-cf32c1f8862f | trentmkelly/LessWrong-43k | LessWrong | Could city design impact spread of infections?
I was looking at the difference in reported performance between Los Angeles and NY City. NYC looks a lot worse on all counts I think.
What is producing the different results I wondered. I suspect things like population density will matter. Certainly population behavior a... |
9056e5b9-ab2f-4f58-819e-d75609ed2e17 | trentmkelly/LessWrong-43k | LessWrong | Knowledge Base 8: The truth as an attractor in the information space
Hypothesis: The truth, defined as the information with 100% credibility calculated by the HITS algorithm in accordance with the consensus theory of truth, is an attractor in the information space.
Interface between all intelligent agents
The post K... |
8a71a794-7383-449f-aecc-e86497169440 | trentmkelly/LessWrong-43k | LessWrong | Survey - Psychological Impact of Long-Term AI Engagement
As part of the AI Safety, Ethics and Society Course, I’m conducting a survey to better understand the psychological and emotional effects of long-term engagement with AI technologies, particularly within the AI safety community. This is an invitation for you to ... |
1aea4eea-8c4d-4d7a-9b05-7cbc609dae85 | trentmkelly/LessWrong-43k | LessWrong | Solving Two-Sided Adverse Selection with Prediction Market Matchmaking
Market making might maximize meaningful matches, like this pair of pears.
0: Navigation
I’m aiming for a reader who knows what “prediction markets” and “adverse selection” are, who likes the first and not the second, and who enjoys systems that h... |
c5707962-3f8d-4d3c-b985-a819146bc191 | trentmkelly/LessWrong-43k | LessWrong | Boston Solstice 2018
Six years ago Julia and I took the bus down to NYC with a group from the Boston LessWrong meetup. We were headed to a Solstice celebration and megameetup Ray was organizing. It was a good weekend: lots of talking to people I'd only met online, various proto-EA conversations, friendly and relaxed a... |
396bd14e-3440-4db6-8586-c68087f464e1 | trentmkelly/LessWrong-43k | LessWrong | What Is a Major Chord?
In the discussion on my minimal chords post, someone commented:
> I have no idea what any of this means (what is a chord? what is a major chord? what is a note? what is a first/fourth/fifth note? is there a 65th note? what is a scale? what is a major scale? what does it mean that a note is of a... |
8cb67fcb-32ae-4da9-936f-e8503182ec9c | StampyAI/alignment-research-dataset/arxiv | Arxiv | Enabling Automatic Certification of Online Auctions
1 Introduction
---------------
Trust is an important issue in building up markets for rational participants. At the heart of markets lie the mechanism that spells out the rules of engagements for any participant. If the rules lack robustness, then paticipants may ... |
f4812178-8ed2-4c87-8871-ae03c73c09f8 | trentmkelly/LessWrong-43k | LessWrong | AISC project: SatisfIA – AI that satisfies without overdoing it
This project is part of the upcoming round of AI Safety Camp.
SatisfIA – AI that satisfies without overdoing it
Summary
This project will make a contribution to some fundamental design aspects of AI systems. We will explore novel designs for generic ... |
d1bacda0-9aab-440c-a8e0-f9a657d5c443 | trentmkelly/LessWrong-43k | LessWrong | Nootropics and Cognitive-enhancement Discussion Area
Many people on LW have expressed interest in nootropics and cognitive enhancement. However, since people are presumably interested in the long-term efficacy, side effects and withdrawal symptoms of using various supplements, open-threads (which quickly become hard t... |
6f3e62a4-e342-47fd-a5c0-75d7902982cf | StampyAI/alignment-research-dataset/special_docs | Other | The errors, insights and lessons of famous AI predictions – and what they mean for the future
Introduction Predictions about the future development of artificial intelligence (AI 1 ) are as confident as they are diverse. Starting with Turing's initial estimation of a 30% pass rate on Turing test by the year 2000 (Tur... |
9db20838-5518-4e6e-b646-129a30bc4cba | trentmkelly/LessWrong-43k | LessWrong | There's probably a tradeoff between AI capability and safety, and we should act like it
I think something like the following is probably true under fairly weak assumptions:[1]
> For almost any goal X, and almost any agent that achieves X safely, there's an agent that achieves even more X unsafely.
If it is true, it ... |
08f45271-959a-4b2b-a59f-4ba8cc3272c0 | trentmkelly/LessWrong-43k | LessWrong | Reveal Culture
(crossposted from malcolmocean.com; originally published June 2015)
I have things to say about the Ask/Guess/Tell Cultures model, and an addition/amendment to propose: Reveal Culture. Shifting cultures is hard, so what you’re about to read is not going to have a quality of “let’s all go do this!” I do ... |
578face1-1db8-4c12-a6fa-c9bbb8f8e658 | trentmkelly/LessWrong-43k | LessWrong | Feedback on promoting rational thinking about one's career choice to a broad audience
I'd appreciate feedback on optimizing a blog post that promotes rational thinking about one's career choice to a broad audience in a way that's engaging, accessible, and fun to read. I'm aiming to use story-telling as the driver of t... |
bdf94efd-14e5-46ec-90db-80e52ea6c2ee | trentmkelly/LessWrong-43k | LessWrong | What are some articles that updated your beliefs a lot on an important topic?
The content doesn't need to be hosted on LessWrong.
It does need to have changed your beliefs personally. If it changed other people's beliefs, please put this as a comment instead.
Also, the article should be more than simply updating you... |
55d5e573-03a2-4690-b622-d503f470e418 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Clarifying Consequentialists in the Solomonoff Prior
I have spent a long time being confused about [Paul’s post](https://ordinaryideas.wordpress.com/2016/11/30/what-does-the-universal-prior-actually-look-like/) on consequentialists in the Solomonoff prior. I now think I understand the problem clearly enough to engage ... |
b47b95d3-d993-4061-9551-40571bb38224 | trentmkelly/LessWrong-43k | LessWrong | AI Safety Oversights
The field of AI Safety at large is making four key oversights:
1. LLMs vs. Agents. AI safety researchers have been thorough in examining safety concerns from LLMs (bias, deception, accuracy, child safety, etc). Agents powered by LLMs, however, are more dangerous and dangerous in different ways ... |
2fbb5415-f629-4b41-a1a1-e9eec7e66c3b | trentmkelly/LessWrong-43k | LessWrong | Averaging value systems is worse than choosing one
A continuation of Only humans can have human values. Revised late in the evening on April 30.
Summary: I will present a model of value systems, and show that under it, the "averaged value system" found by averaging the values of all the agents,
* (RESTATED:) has m... |
8a8d37e0-844f-4c2f-84af-22caa6b6f1e6 | trentmkelly/LessWrong-43k | LessWrong | [Linkpost] Review: Happiness
This is a link post for https://thefangirlverdict.com/2022/02/10/review-happiness/
A few months ago, I watched a Korean TV show called Happiness. While it had several stretches of logic, I found its treatment of the psychology and dynamics of severe disease outbreaks quite interesting.
D... |
d297ae42-1e4e-4a2e-a8ef-7e40c1df9fe7 | trentmkelly/LessWrong-43k | LessWrong | DeepMind and Google Brain are merging [Linkpost]
The new organization, called Google DeepMind, will be led by Demis Hassabis (current CEO of DeepMind). Jeff Dean (co-founder of Google Brain) will be the chief scientist of the new organization.
See also:
* Hooray for stepping out of the limelight
* DeepMind's CEO ... |
e30cdfee-cf50-4ec7-aafc-d7e0afabdf6b | StampyAI/alignment-research-dataset/lesswrong | LessWrong | Using PICT against PastaGPT Jailbreaking
> This post uses [PICT](https://www.lesswrong.com/posts/HJinq3chCaGHiNLNE/pict), and is inspired by [GPT-Eliezer](https://www.lesswrong.com/posts/pNcFYZnPdXyL2RfgA/using-gpt-eliezer-against-chatgpt-jailbreaking#Eliezer_and_ChatGPT_jailbreaking)
>
>
Produced as part of the... |
c6d9388a-8854-4368-ae58-e5cb13961ad0 | trentmkelly/LessWrong-43k | LessWrong | Iterated Sleeping Beauty and Copied Minds
Before I move on to a summation post listing the various raised thought experiments and paradoxes related to mind copying, I would like to cast attention to a particular moment regarding the notion of "subjective probability".
In my earlier discussion post on the subjective e... |
da00bfb7-6de5-4b2a-a8c9-8780579fcda0 | StampyAI/alignment-research-dataset/lesswrong | LessWrong | My Updating Thoughts on AI policy
*My* [*conversation on policy and AI with Richard*](https://www.lesswrong.com/posts/DJB82jKwgJE5NsWgT/some-cruxes-on-impactful-alternatives-to-ai-policy-work) *was over a year ago, so for Daniel Kokotajlo's LW writing day I thought I’d write down new thoughts I’ve thought since then (... |
901977f3-4c39-4ab8-b961-4ff9e2d34940 | StampyAI/alignment-research-dataset/youtube | Youtube Transcripts | 245. Democratising Risk 2
hello and welcome to session
245 in the ai safety.com reading group
tonight we will be talking about the
article democratizing risk in search for
methodology to study existing risk and
this is the second part
the authors are still carla so prima and
new
chem and we are focusing on section 4
t... |
ce280609-7abc-4abb-9463-49f4b9702439 | trentmkelly/LessWrong-43k | LessWrong | Value Learning is only Asymptotically Safe
I showed recently, predicated on a few assumptions, that a certain agent was asymptotically “benign” with probability 1. (That term may be replaced by something like “domesticated” in the next version, but I’ll use “benign” for now).
This result leaves something to be desire... |
5ccee779-b080-4c9a-93eb-bdbc5e3f1f1a | trentmkelly/LessWrong-43k | LessWrong | Is behavioral safety "solved" in non-adversarial conditions?
I’m trying to crystalize something I was said to a friend recently: I think that techniques like RLHF and Constitutional AI seem to be sufficient for making large language models (LLMs) “behaviorally safe” in non-adversarial conditions.
Let’s define my term... |
1cbbf99f-4407-435e-a38a-07bbf2ab8f4a | StampyAI/alignment-research-dataset/arxiv | Arxiv | Making Efficient Use of Demonstrations to Solve Hard Exploration Problems
1 Introduction
---------------
Reinforcement learning from demonstrations has proven to be an effective strategy for attacking problems that require sample efficiency and involve hard exploration. For example, Aytar et al. ([2018](#bib.bib2))... |
a3603b1e-4568-408b-8e1b-ad227f4b5af4 | LDJnr/LessWrong-Amplify-Instruct | LessWrong | "Originally posted on The Roots of Progress, January 6, 2018From the time that humans began to leave their nomadic ways and live in settled societies about ten thousand years ago, we have needed to build structures: to shelter ourselves, to store our goods, to honor the gods.The easiest way to build is with dirt. Mud, ... |
eb999491-5dd4-44e1-9e9f-d4acbbe2643f | trentmkelly/LessWrong-43k | LessWrong | MRI tracers
MRI scans and PET scans are different methods for medical imaging, and they observe different things:
* MRI scans can see which nuclei are present in regions. Sometimes bond types of particular elements can be distinguished, eg in 1H NMR and fMRI.
* PET scans add a short-lived radioactive tracer (usual... |
69a62a95-2288-438b-ae49-881e95af0189 | trentmkelly/LessWrong-43k | LessWrong | Introducing… The Less Wrong Forum!
A while ago, some posters mentioned that it would be good if LW had a forum. Earlier this week I was planning a role-playing game with some other users and decided that a forum would be a good place for us and other groups to play online. And the Less Wrong Forum was born.
Currently... |
867428e8-10c9-4503-a572-be6d51544bf9 | trentmkelly/LessWrong-43k | LessWrong | Meetup : Washington DC: Goals, Political Advocacy as Effective Altruism, and Stories
Discussion article for the meetup : Washington DC: Goals, Political Advocacy as Effective Altruism, and Stories
WHEN: 10 November 2013 03:00:00AM (-0500)
WHERE: National Portrait Gallery, Washington, DC 20001, USA
So, we came up wi... |
9c11aabc-c232-49d3-99ee-ad90a4d67b9d | trentmkelly/LessWrong-43k | LessWrong | Introducing WAIT to Save Humanity
The EA/rationality community has struggled to identify robust interventions for mitigating existential risks from advanced artificial intelligence. In this post, I identify a new strategy for delaying the development of advanced AI while saving lives roughly 2.2 million times [-5 time... |
f9f12d95-b586-4a2c-a2c2-7abae1b4be76 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Abstraction, Evolution and Gears
*Meta:* [*this project*](https://www.lesswrong.com/s/ehnG4mseKF6xALmQy) *is wrapping up for now. This is the second of probably several posts dumping my thought-state as of this week.*
It is an empirical fact that we can predict the day-to-day behavior of the world around us - positio... |
c54f5607-a73d-4150-b266-235958577f85 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Information Loss --> Basin flatness
*This work was done under the mentorship of Evan Hubinger through the SERI MATS program. Thanks to Lucius Bushnaq, John Wentworth, Quintin Pope, and Peter Barnett for useful feedback and suggestions.*
In this theory, the main proximate cause of flat basins is a type of **informatio... |
1b1e3526-7f86-42f5-81e2-60e07ca22726 | trentmkelly/LessWrong-43k | LessWrong | Covid 8/19: Cracking the Booster
The news about Covid-19 is now essentially on a few distinct tracks: Vaccine effectiveness and booster shots, vaccination mandates, mandates for NPIs and other Covid-19 related crippling of the living of life, and the actual path of the pandemic. One could argue for splitting those fur... |
ad7609af-3fc6-4a23-bc74-7b98a0505f52 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Future directions for ambitious value learning
To recap the sequence so far:
* [Ambitious value learning](https://www.alignmentforum.org/s/4dHMdK5TLN6xcqtyc/p/5eX8ko7GCxwR5N9mN) aims to infer a utility function that is safe to maximize, by looking at human behavior.
* [However](https://www.alignmentforum.org/s/4dHMdK... |
ba80fbcf-59a5-4788-b65e-152878256bd4 | StampyAI/alignment-research-dataset/arxiv | Arxiv | Big Self-Supervised Models are Strong Semi-Supervised Learners
1 Introduction
---------------
Figure 1: Bigger models yield larger gains when fine-tuning with fewer labeled examples.
![Refer to caption](/htm... |
45fc8e4c-eaad-4747-aa5b-72d77a089a04 | trentmkelly/LessWrong-43k | LessWrong | Building habits: requesting advice on installing mental software
I'd like to figure out how to create habits more effectively and systematically, especially mental habits.
For example I might want to develop habits like
* Noticing when I'm getting distracted and thinking about what I could do to reduce it
* Notici... |
50b3a6ee-33f6-4801-8c05-9e0b48a962df | awestover/filtering-for-misalignment | Redwood Research: Alek's Filtering Results | id: post3112
In the last few months, some AI safety researchers have signaled what Wei Dai called a "wave of optimism" for using transparency tools to catch deceptive behavior in our machine learning models. By a transparency tool, I mean any technique that allows us to peek inside a model to get a better idea of what... |
e7e57698-1b1b-470b-89ba-72bb9d266c98 | trentmkelly/LessWrong-43k | LessWrong | Unpacking "Shard Theory" as Hunch, Question, Theory, and Insight
I read several shard theory posts and found the details interesting, but I couldn't quite see the big picture. I'm used to hearing "theory" refer to a falsifiable generalization of data. It is typically stated as a sentence, paragraph, list, mathematical... |
c211d618-0cd7-47f5-9de8-716eee2f04cb | trentmkelly/LessWrong-43k | LessWrong | Broken Benchmark: MMLU
Phillip over at the AI Explained channel has been running some experiments on his SmartGPT framework against the MMLU benchmark and discovered a not-insignificant amount of issues with the problem set.
Among them:
* Crucial context missing from questions (apparently copy-paste errors?)
* Amb... |
72a772d6-a454-4430-a277-4ba11e543a26 | trentmkelly/LessWrong-43k | LessWrong | Coffee trucks: a brilliant idea that someone should do?
I sometimes wonder if the world should have coffee trucks, like ice cream trucks, roaming the street. Especially when half the population is working from home.
Coffee seems ideal for this because:
1. you can probably prepare it fresh in a briefly stopping vehi... |
f5e902a5-a528-4b09-bd53-a3dcc82ac888 | trentmkelly/LessWrong-43k | LessWrong | An overall schema for the friendly AI problems: self-referential convergence criteria
A putative new idea for AI control; index here.
After working for some time on the Friendly AI problem, it's occurred to me that a lot of the issues seem related. Specifically, all the following seem to have commonalities:
* Nasty... |
ee2e5329-14e0-4d66-b52a-d5c22391f5ef | trentmkelly/LessWrong-43k | LessWrong | AI Reading Group Thoughts (2/?): Reconstructive Psychosurgery
Previously on AI Reading Group Thoughts
In a highly unofficial meeting of the AI Reading Group (most of the participants happened to be in a room and we started talking about AI; there was no formal meeting time, no reading homework, and no dessert), we me... |
acf49644-f139-44de-80ca-6b4944e6d28f | StampyAI/alignment-research-dataset/youtube | Youtube Transcripts | The AI revolution and international politics | Allan Dafoe | EAG 2017 Boston
next up it is my honor to introduce
Professor Alan Defoe he is an assistant
professor of political science at Yale
University and a research associate at
the future of humanity Institute at
Oxford where he is involved in building
the AI polit... |
21482838-93c2-4460-9425-e26db0cc314b | StampyAI/alignment-research-dataset/youtube | Youtube Transcripts | AI Research Considerations for Existential Safety
hello
uh my name is jack and welcome to ai
research considerations for existential
safety
i am excited to introduce you to andrew
kritsch
andrew is a research scientist at the
center for human compatible ai
at uc berkeley and is also the founder
and president of the be... |
31a3d177-f0d4-452d-a9e6-0f4cba160a0c | StampyAI/alignment-research-dataset/distill | Distill Scientific Journal | Visualizing memorization in RNNs
Memorization in Recurrent Neural Networks (RNNs) continues to pose a challenge
in many applications. We’d like RNNs to be able to store information over many
timesteps and retrieve it when it becomes relevant — but vanilla RNNs often struggle to do this.
Several network archite... |
fbf5039c-354a-401b-8bab-fb88570a16c3 | trentmkelly/LessWrong-43k | LessWrong | What organizations other than Conjecture have (esp. public) info-hazard policies?
I believe Anthropic has said they won't publish capabilities research?
OpenAI seems to be sort of doing the same (although no policy AFAIK).
I heard FHI was developing one way back when...
I think MIRI sort of does as well (default to no... |
befef392-e54b-42b6-b1e0-1657ba698d9c | trentmkelly/LessWrong-43k | LessWrong | Explaining SolidGoldMagikarp by looking at it from random directions
Summary
I conducted an experiment that provides evidence that many of the weird tokens demonstrating the SolidGoldMagikarp phenomenon are interior points in the token embedding cloud. This would be sufficient to make them “unspeakable”.
My experimen... |
63cb3407-4778-49d3-b5f1-023c023ae0e9 | trentmkelly/LessWrong-43k | LessWrong | Long Bets by Confidence Level
If you want to make a long-term bet one of your options is to register your bet with the Long Now Foundation as a Long Bet. They have some rules, which are roughly:
* Both parties put up the same amount, at least $200/each.
* Long Bets effectively runs a donor-advised fund (DAF).
* Wh... |
4ecb825a-37c3-4124-8a86-f895541121ec | trentmkelly/LessWrong-43k | LessWrong | $500 bounty for best short-form fiction about our near future world; $100 for recommending winning piece: new “Art of Near Future World” quarterly art project
I’m launching a new project called “Art of the Near Future World” to celebrate and encourage short-form fiction about the world a few months to a few years from... |
df05d3c5-cdb4-47af-ad83-bb2f0357e514 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | What does it mean to apply decision theory?
*Based on discussions with Stuart Armstrong and Daniel Kokotajlo.*
There are two conflicting ways of thinking about foundational rationality arguments such as the VNM theorem.
1. As direct arguments for normative principles. The axioms are supposed to be premises which y... |
be7ee008-dc0f-48a2-9445-1d7a77400ca4 | trentmkelly/LessWrong-43k | LessWrong | The Benefits of Taking Cold Showers: A 3 Month Trial
For the format that’s easy on the eyes: https://medium.com/@John_Greer/the-benefits-of-taking-cold-showers-a-3-month-trial-7e14ca25d79
There are many reported health benefits of cold showers like increased testosterone levels and better circulation. I haven’t done ... |
592e65c7-833d-415a-835e-13f5c8ea45c2 | trentmkelly/LessWrong-43k | LessWrong | $100 for the best article on efficient charity: the finalists
Part of the Efficient Charity Article competition. Several people have written articles on efficient charity --
* Throwawayaccount_1 has an excellent article hidden away in a comment
* Waitingforgodel has an article in discussion: "How Greedy Bastard... |
ead0f6c5-ac65-4ebd-aff2-9ab59620d61b | trentmkelly/LessWrong-43k | LessWrong | [SEQ RERUN] Say It Loud
Today's post, Say It Loud was originally published on 19 September 2008. A summary (taken from the LW wiki):
> If you're uncertain about something, communicate that uncertainty. Do so as clearly as you can. You don't help yourself by hiding how confused you are.
Discuss the post here (rat... |
3ff3082c-57ac-4ca2-bade-8ef9fd711120 | trentmkelly/LessWrong-43k | LessWrong | 2025 Q1 Pivotal Research Fellowship (Technical & Policy)
We’re excited to announce that applications are now open for our 2025 Q1 Pivotal Research Fellowship, a 9-week program designed to enable promising researchers to produce impactful research and accelerate their careers in technical AI safety, AI governance, and ... |
ed7d4d85-8202-46fc-aa30-f11f305fd68d | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Universality and Hidden Information in Concept Bottleneck Models
Summary
-------
I use the CUB dataset to finetune models to classify images of birds, via a layer trained to predict relevant concepts (e.g. "has\_wing\_pattern::spotted"). These models, when trained end-to-end, naturally put additional relevant informa... |
05180281-8b79-4a21-ba03-e48ffd92541b | trentmkelly/LessWrong-43k | LessWrong | Meetup : Fort Collins, Colorado Meetup Thursday 7pm
Discussion article for the meetup : Fort Collins, Colorado Meetup Thursday 7pm
WHEN: 12 July 2012 07:00:00PM (-0600)
WHERE: 144 North College Avenue, Fort Collins, CO 80524
Freshly back from Less Wrong Central, copt and EvelynM have new games and stories of wizard... |
79d027a5-2776-493a-b12c-dbce588f9e35 | trentmkelly/LessWrong-43k | LessWrong | What can we call the mistaken belief that something can't be measured?
In his book How to Measure Anything, Douglas W. Hubbard defines measurement as a quantitative reduction in uncertainty. He also shows how anything can be measured, i.e. how it's always possible to quanitatively reduce uncertainty.
However, it's co... |
104a9ae1-5897-415f-b34c-562844980c93 | trentmkelly/LessWrong-43k | LessWrong | Preference For (Many) Future Worlds
Followup to: Quantum Russian Roulette; The Domain of Your Utility Function
The only way to win is cheat
And lay it down before I'm beat
and to another give my seat
for that's the only painless feat.
Suicide is painless
It brings on many changes
and I can take or leave it if I plea... |
18f88e79-56fd-449c-97fc-94573f1602c3 | trentmkelly/LessWrong-43k | LessWrong | Projects-in-Progress Thread
From Raemon's Project Hufflepuff thread, I thought it might be helpful for there to be a periodical check-in thread where people can post about their projects in the spirit of cooperation. This is the first one. If it goes well, maybe we can make it a monthly thing.
If you're looking for a... |
b2211084-1bbd-44b6-b607-22975e4fdb1c | trentmkelly/LessWrong-43k | LessWrong | My thoughts on the social response to AI risk
A common theme implicit in many AI risk stories has been that broader society will either fail to anticipate the risks of AI until it is too late, or do little to address those risks in a serious manner. In my opinion, there are now clear signs that this assumption is fals... |
03e68391-9ffb-41d4-bb18-c4fa9d4b7fbe | trentmkelly/LessWrong-43k | LessWrong | Where Are We the Weakest?
As rationalists, we should be able to consistently and accurately make predictions that enable us to act effectively.
As humans, we don't. At least not perfectly.
We need to improve. Many of us have, or at least believe we have. However, it's a notably hacked improvement. PredictionBook is ... |
264ab695-c06c-46d1-9de7-8b9e487f1ee9 | trentmkelly/LessWrong-43k | LessWrong | The Pinnacle
In another universe, in another time, the stars don’t shine like ours. In fact, they act in a manner quite more familiar to you and I as that of a black hole. Take Los Ral as an example. An average star, whose future in this universe is rare, and storied.
Like its peers, Los Ral does not emit light. Inste... |
d7053235-6266-4c9e-9537-14beb3d101fa | trentmkelly/LessWrong-43k | LessWrong | Measuring lethality in reduced expected heartbeats
Some of us here are already familiar with Micromorts - a unit that stands for a 1 in a million chance of dying. The wikipedia page lists a number of sample values. One obvious example is that smoking 1.4 cigarettes is one micromort. This is a good tool for comparing t... |
ab85ebb1-261f-48f2-a2a0-1baefe0cff89 | trentmkelly/LessWrong-43k | LessWrong | Anthropic Decision Theory V: Linking and ADT
A near-final version of my Anthropic Decision Theory paper is available on the arXiv. Since anthropics problems have been discussed quite a bit on this list, I'll be presenting its arguments and results in this, subsequent, and previous posts 1 2 3 4 5 6.
Now that we've se... |
22222542-c42f-4182-ba82-a2d01308c5a4 | trentmkelly/LessWrong-43k | LessWrong | Horizontal vs vertical generality
This is a note/definition I want to put out there. While the term may sound similar to breadth vs depth of knowledge, it is importantly distinct.
Let's define intelligence as the ability to process information and make decisions. (This is a very broad definition and includes not just... |
0e686bb2-e668-4560-bb25-60dff9ed19a1 | trentmkelly/LessWrong-43k | LessWrong | Meetup : Less Wrong NH Meetup
Discussion article for the meetup : Less Wrong NH Meetup
WHEN: 26 January 2016 07:00:00PM (-0500)
WHERE: 269 Pearl St Manchester NH 03104
The eleventh NH meet-up is Tuesday, 1/26, in Manchester, NH at 7 pm at a private residence. Light refreshments will be provided.
Have you read Rati... |
15099405-84b2-4ab9-9096-01238e4cd90d | trentmkelly/LessWrong-43k | LessWrong | Stop learning, start thinking [LINK]
Stop Learning, by Jacob Barnett. This is an 18 minute video, and I think there's a lot to be said for getting the material in the order given.
However, if you'd rather have text, here it is in rot13. Wnpbo Oneargg pbzrf bss nf naablvat. Ybhq, ohzcgvbhf, naq ynhtuf ng zbfg bs uvf b... |
c408a1e5-3a87-4c20-be78-ca304acbc41d | trentmkelly/LessWrong-43k | LessWrong | How has the total amount of gain-of-function research worldwide grown/shrunk over time?
Given the advance in underlying biological tools/technology, I would not find it surprising if gain-of-function research had increased 10x over the last decade. (This is just based on priors; getting an actual estimate is the point... |
b19cf167-0689-47ab-b16e-45ec9754f209 | trentmkelly/LessWrong-43k | LessWrong | Resource gathering and pre-corriged agents
A putative new idea for AI control; index here.
Resource-gathering agent
It will often be useful to have a model of a “pure” resource gathering agent – one motivated only to gather resources, accumulated power, spread efficiently, and so on. This model could be used as beha... |
e0732367-5875-4d2f-b935-e72082bcab85 | StampyAI/alignment-research-dataset/blogs | Blogs | MIRI’s November 2013 Newsletter
| | |
| --- | --- |
|
| |
| --- |
|
[Machine Intelligence Research Institute](http://intelligence.org)
|
|
|
| | |
| --- | --- |
|
| |
| --- |
|
Dear friends,
We’re experimenting with a new, ultra-brief newsletter style. To let us know what you think of it, simply repl... |
fe261780-5c27-452d-bbce-f72778d44f6f | trentmkelly/LessWrong-43k | LessWrong | Section 7: Foundations of Rational Agency
This post is part of the sequence version of the Effective Altruism Foundation's research agenda on Cooperation, Conflict, and Transformative Artificial Intelligence.
7 Foundations of rational agency
We think that the effort to ensure cooperative outcomes among TAI systems w... |
c30995fd-3fce-44ac-bc2c-1a6e987618bd | StampyAI/alignment-research-dataset/blogs | Blogs | January 2021 Newsletter
#### MIRI updates
* MIRI’s Evan Hubinger uses a notion of optimization power to define whether AI systems are [compatible with the strategy-stealing assumption](https://www.lesswrong.com/posts/WwJdaymwKq6qyJqBX).
* MIRI’s Abram Demski discusses debate approaches to AI safety [that don’t rely ... |
7d4a0e6c-3568-4f73-844c-6e64168e13d8 | StampyAI/alignment-research-dataset/lesswrong | LessWrong | Operationalizing Newcomb's Problem
The standard formulation of Newcomb's problem has always bothered me, because it seemed like a weird hypothetical designed to make people give the wrong answer. When I first saw it, my immediate response was that I would two-box, because really, I just don't believe in this "perfect ... |
ba85e2c3-01cf-458e-9154-57419b093521 | trentmkelly/LessWrong-43k | LessWrong | The Person As Input
I. Humans are emotion-feeling machines.
I don’t mean that humans are machines that happen to feel emotions. I mean that humans are machines whose output is the feeling of emotions—“emotion-feeling” is the thing of value that we produce.
Not just “being happy." Then wireheading is the ultimate go... |
e8abc82e-7023-4e26-ae73-fd462a89237f | trentmkelly/LessWrong-43k | LessWrong | Meetup : Durham NC meetup: Rationality Checklist
Discussion article for the meetup : Durham NC meetup: Rationality Checklist
WHEN: 29 November 2012 07:00:00PM (-0500)
WHERE: Whole Foods Market, 621 Broad St., Durham NC 27705
We'll be reading, discussing, and checking the Checklist of Rationality Habits (http://lessw... |
17c97068-c0c9-4dcb-a3ee-81e48f4d897a | StampyAI/alignment-research-dataset/special_docs | Other | Competence-Aware Systems.
Artificial Intelligence 00 (2022) 1–34Artificial Intelligence
Competence-Aware Systems 1
Connor Basich∗a, Justin Svegliatob, Kyle H. Wrayc, Stefan Witwickic, Joydeep Biswasd, 2
Shlomo Zilbersteina
3
aUniversity of Massachusetts Amherst, Amherst, MA, USA {cbasich, shlomo}@cs.umass.edu 4
bUnive... |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.