id stringlengths 36 36 | source stringclasses 15
values | formatted_source stringclasses 13
values | text stringlengths 2 7.55M |
|---|---|---|---|
c0074d3f-38a4-4652-85d9-afe547b65a5f | StampyAI/alignment-research-dataset/arbital | Arbital | Finite set
summary(Technical): A finite [set](https://arbital.com/p/3jz) $X$ is a set which is not infinite: that is, there is some $n \in \mathbb{N}$ such that the [https://arbital.com/p/-4w5](https://arbital.com/p/-4w5) of $X$ is equal to $n$. Examples: $\{ 1,2 \}$ and $\{ \mathbb{N} \}$. Non-examples: $\mathbb{N}$,... |
5ab56b13-33a0-45b7-9d16-60ba2840895f | trentmkelly/LessWrong-43k | LessWrong | Rational vs. real utilities in the cousin_it-Nesov model of UDT
In this article cousin_it describes an algorithm, using a Turing Oracle, which gives a working model of UDT. This algorithm seems to work well in the case of a utility function which takes natural numbers or rationals as values but, as I will explain, the... |
b3340aa0-d78c-41ef-bcac-8a40cf2aadf7 | trentmkelly/LessWrong-43k | LessWrong | [inactive] £2000 bounty - contraceptives (and UTI) literature review
edit: No longer active
I would like if there was a well-researched LessWrong post on the pros and cons of different contraceptives. - Same deal with a good post on how to treat or prevent urinary tract infection, although I'm less excited about that... |
9a72b3ac-9f00-4620-8621-940e144a2b48 | trentmkelly/LessWrong-43k | LessWrong | GPT-3 Gems
GPT-3 is captivating, and not just because of its potential reasoning abilities. This post will be a living collection of my favorite experiences with the network.
Bold text is my input; square brackets contain my commentary. As of 8/19/20, AIDungeon's Custom mode's first completion is sampled from GPT-2.... |
49cfabca-d92c-48c8-bd18-320e539a1f3e | trentmkelly/LessWrong-43k | LessWrong | Hindsight Devalues Science
This essay is closely based on an excerpt from Meyers’s Exploring Social Psychology; the excerpt is worth reading in its entirety.
Cullen Murphy, editor of The Atlantic, said that the social sciences turn up “no ideas or conclusions that can’t be found in [any] encyclopedia of quotations . ... |
8b82ca0d-572c-4074-86d2-ed35468fef5c | StampyAI/alignment-research-dataset/lesswrong | LessWrong | Decisions are not about changing the world, they are about learning what world you live in
[Cross-posted from my blog](https://donerkebabphilosophy.wordpress.com/2018/07/28/decisions-are-not-about-changing-the-world-they-are-about-learning-what-world-you-live-in/).
***Epistemic status: Probably discussed to death in ... |
70574e2b-b32a-4b4a-abda-6da3861439e6 | trentmkelly/LessWrong-43k | LessWrong | Covid 12/15/22: China’s Wave Begins
Cases increased this week, as expected. Deaths also increased, which I did not expect so soon, indicating a bigger and faster winter wave. That was not the big event in the West this week with respect to Covid.
The big event this week in the West’s Covid experience was a memory of ... |
e434d2c8-fe8c-4964-8eae-a8e20ab7fef8 | trentmkelly/LessWrong-43k | LessWrong | Book Notes: Scaling, Why is Animal Size so Important?
TL;DR: Movie!Ant-Man is unrealistic. Also, the Internet Archive has an ebook library!
Movie!Ant-Man can be the size of an ant…
…and the size of an airplane…
…but throughout it all, his own body’s shape and structure doesn’t change. Without background objects (... |
dfa7431c-7ffc-47c1-9b83-c4e100863434 | trentmkelly/LessWrong-43k | LessWrong | [Intro to brain-like-AGI safety] 9. Takeaways from neuro 2/2: On AGI motivation
(Last revised: July 2024. See changelog at the bottom.)
9.1 Post summary / Table of contents
Part of the “Intro to brain-like-AGI safety” post series.
Most posts in the series thus far—Posts #2–#7—have been primarily about neuroscience.... |
2a694ae9-5e1c-4ab9-bc53-078909c5edad | trentmkelly/LessWrong-43k | LessWrong | On fundamental solitude
A quote from Aldous Huxley that has stuck in my mind more than perhaps any other over the years:
> “We live together, we act on, and react to, one another; but always and in all circumstances we are by ourselves. The martyrs go hand in hand into the arena; they are crucified alone. Embraced, t... |
5159428c-b06b-48ec-a641-5edd97e93a7c | trentmkelly/LessWrong-43k | LessWrong | Timeless Modesty?
Looking back at my description of policy-level thinking, one might charge it of the same crime as outside view; namely, being overly likely to lead to modest epistemology:
> Policy-level thinking, on the other hand, helps you to not get lost in the details. It provides the rudder which can keep you ... |
6d8a1b72-7da7-4bdf-96e6-753154f203f3 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Goodhart Taxonomy
[Goodhart’s Law](https://en.wikipedia.org/wiki/Goodhart%27s_law) states that "any observed statistical regularity will tend to collapse once pressure is placed upon it for control purposes." However, this is not a single phenomenon. I propose that there are (at least) four different mechanisms throug... |
cbd206f9-0428-48d0-a41e-da7d93b5b155 | trentmkelly/LessWrong-43k | LessWrong | Problems in AI Alignment: A Scale Model
After trying too hard for too to make sense about what bothers me with the AI alignment conversation, I have settled, in true Millenial fashion, on a meme:
Explanation:
The Wikipedia article on AI Alignment defines it as follows:
> In the field of artificial intelligence (A... |
0c1a0ae4-8932-436f-b0db-590e49d94a93 | StampyAI/alignment-research-dataset/arxiv | Arxiv | Decision Transformer: Reinforcement Learning via Sequence Modeling.
1 Introduction
---------------
Recent work has shown transformers [vaswani2017attention] can model high-dimensional distributions of semantic concepts at scale, including effective zero-shot generalization in language [brown2020gpt3] and out-of-dis... |
3422bb83-aba8-4e6d-823a-b8091f617046 | StampyAI/alignment-research-dataset/arxiv | Arxiv | Policy Gradient Bayesian Robust Optimization for Imitation Learning.
1 Introduction
---------------
We consider the following question: How should an intelligent agent act if it has epistemic uncertainty over its objective function? In the fields of reinforcement learning (RL) and optimal control, researchers and p... |
cabc49b2-0f60-4610-b368-eb43cf55089b | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Conversation on technology forecasting and gradualism
This post is a transcript of a multi-day discussion between Paul Christiano, Richard Ngo, Eliezer Yudkowsky, Rob Bensinger, Holden Karnofsky, Rohin Shah, Carl Shulman, Nate Soares, and Jaan Tallinn, following up on the Yudkowsky/Christiano debate in [1](https://www... |
f26d6e16-9a8b-4555-bd18-273bc567cbb6 | trentmkelly/LessWrong-43k | LessWrong | Your AI Safety focus is downstream of your AGI timeline
Cross-posted from Substack
Feeling intellectually understimulated, I've begun working my way through Max Lamparth's CS120 - Introduction to AI Safety. I'm going to use this Substack as a kind of open journaling practice to record my observations on the ideas pre... |
ad1b214f-80dd-4db1-a80d-6f378e712a3b | StampyAI/alignment-research-dataset/lesswrong | LessWrong | The Absent-Minded Driver
This post examines an attempt by professional decision theorists to treat an example of time inconsistency, and asks why they failed to reach the solution (i.e., [TDT](http://wiki.lesswrong.com/wiki/Timeless_decision_theory)/[UDT](http://wiki.lesswrong.com/wiki/Updateless_decision_theory)) tha... |
681c41dd-7de4-45b5-97f8-ac47fcc61620 | trentmkelly/LessWrong-43k | LessWrong | Monthly Shorts 1&2/23
Apologies for missing last month’s, and the delay on this one. I attended EAG, and enjoyed it. Wrote a few small pieces for the EA Forum, and am in the process of writing a few more. In the meantime…
Politics and Possibilities
The FTC is proposing a ban on all non-compete clauses.
For the obvi... |
c9bb4376-f9df-47ba-b6ff-3c02006b00f3 | trentmkelly/LessWrong-43k | LessWrong | True Stories of Algorithmic Improvement
In May 2020, OpenAI released a report on algorithmic efficiency improvements in deep learning. Main headline:
> Compared to 2012, it now takes 44 times less compute to train a neural network to the level of AlexNet (by contrast, Moore’s Law would yield an 11x cost improvement o... |
4f92c935-d60a-43e7-bca4-48efcc94a0cc | trentmkelly/LessWrong-43k | LessWrong | The benevolence of the butcher
A few days ago I published this post on the risks of powerful transformative AGI (by which I meant AGI that takes off fast and pretty much rules the world in no time), even if aligned. Among the comments there was one by Paul Christiano which I think was very interesting, but also focuse... |
8c6120cf-600f-4569-8312-98fecb4f9ac1 | trentmkelly/LessWrong-43k | LessWrong | Give it a google
I'm a programmer. In programming, people talk about how the ability to Google is such an important skill. No one knows everything. In practice, people are always looking things up and running into weird situations that they have to figure out.
That's not what this post is about. Not quite. This post ... |
803b25d8-ff79-439e-9edf-67e6467779f6 | trentmkelly/LessWrong-43k | LessWrong | What convincing warning shot could help prevent extinction from AI?
> - Tell me father, when is the line
>
> where ends everything good and fine?
>
> I keep searching, but I don't find.
>
> - The line my son, is just behind.
>
> Camille Berger
There is hope that some “warning shot” would help humanity g... |
96ce69ce-863d-4efc-8458-4961aaaec8b2 | trentmkelly/LessWrong-43k | LessWrong | Thinking Toys
Original post at http://garybasin.com/thinking-toys/
My mind seems to mostly act as an emotional problem-solving machine. It's trying to find paths to get to pleasure and avoid pain. Of course, this often requires solving abstract problems, as well. Watching myself, and others, identify and solve proble... |
819f7fdd-40f9-43e1-98e2-8fdf1aa5fab3 | StampyAI/alignment-research-dataset/lesswrong | LessWrong | The limits of corrigibility
I previously wrote a [critique](https://www.lesswrong.com/posts/ZyyMPXY27TTxKsR5X/problems-with-amplification-distillation) of Paul Christiano's [Amplification/Distillation](https://ai-alignment.com/iterated-distillation-and-amplification-157debfd1616) idea. Wei Dei [pointed out](https://ww... |
4c824ef4-c3c0-4c05-8693-67b6cd7379c6 | trentmkelly/LessWrong-43k | LessWrong | Personal musings on Individualism and Empathy
One type of the more common accusations that people throw at me can be described like this: "You're an egoist. You don't think about others." To my teenage self this sounded amusing, and I would respond with a smug "Why yes, I am an egoist, thank you very much!"
Here I ha... |
11d7d715-48dd-4d20-8934-d0901bdcb503 | trentmkelly/LessWrong-43k | LessWrong | Mere Messiahs
Yesterday I discussed how the halo effect, which causes people to see all positive characteristics as correlated—for example, more attractive individuals are also perceived as more kindly, honest, and intelligent—causes us to admire heroes more if they're super-strong and immune to bullets. Even though,... |
436b3cd3-3a9d-4c89-ba69-825cfaf7c1ae | trentmkelly/LessWrong-43k | LessWrong | What will the scaled up GATO look like? (Updated with questions)
Demis Hassabis mentioned a few months ago that Deepmind is in the middle of scaling up its generalist agent GATO. Based on this, I would except it to come out either by the end of the year or in the beginning of the next year. The original model had just... |
045b3a93-749b-4e50-824e-f7dff412b158 | StampyAI/alignment-research-dataset/arxiv | Arxiv | Evaluating Explainable AI: Which Algorithmic Explanations Help Users Predict Model Behavior?
1 Introduction
---------------
Interpretable machine learning is now a widely discussed topic Rudin ([2019](#bib.bib49 "Stop Explaining Black Box Machine Learning Models for High Stakes Decisions and Use Interpretable Model... |
ef8ea892-6751-48df-a71e-d81638220287 | StampyAI/alignment-research-dataset/arxiv | Arxiv | Super-NaturalInstructions: Generalization via Declarative Instructions on 1600+ Tasks..
1 Introduction
---------------
Building models that can follow language instructions and generalize to *unseen* tasks has received increased attention in the recent literature mishra2021cross; wei2021finetuned; sanh2021multitask... |
0ecb87a9-0e7e-4331-a723-ae9a11f1056e | trentmkelly/LessWrong-43k | LessWrong | The Values-to-Actions Decision Chain
Update 2022: Looking back on younger Remmelt's writing on decision-making, it is useful but cocky. It explained what I now call 'monitoring values' for carving out states of the world to decide between, but not how choice arises from 'deep embedded values'.
-----------------------... |
4f0ebec1-7f42-418c-834c-f2f035d21a75 | StampyAI/alignment-research-dataset/special_docs | Other | How Would Catastrophic Risks Affect Prospects for Compromise?
How Would Catastrophic Risks Affect Prospects for Compromise?
=============================================================
29 August 2015
by [Brian Tomasik](https://longtermrisk.org/author/brian-tomasik/ "Posts by Brian Tomasik")
First written: 24 Feb.... |
a0fde67d-cec6-4817-83ce-6027846eaa9d | StampyAI/alignment-research-dataset/lesswrong | LessWrong | ChatGPT Plays 20 Questions [sometimes needs help]
*Cross posted* [*from New Savanna*](https://new-savanna.blogspot.com/2023/10/chatgpt-plays-20-questions-sometimes.html)*.*
One of the questions that keeps coming up about LLMs, and certainly about ChatGPT, is: Can they reason? Well, it depends on what you mean by re... |
1e0eb653-17cc-4963-9d15-d4f746098ae0 | trentmkelly/LessWrong-43k | LessWrong | Another take on agent foundations: formalizing zero-shot reasoning
After spending more time thinking about MIRI’s agenda, I’ve come upon another framing of it, which I’m coining zero-shot reasoning. [1]
This post is a distillation of my intuitions around zero-shot reasoning. These views should be taken as my own, and... |
7b91b5c7-9d4f-40b4-8445-737621b358f2 | trentmkelly/LessWrong-43k | LessWrong | "Membranes" is better terminology than "boundaries" alone
We've been thinking about the core concept behind «boundaries» extensively, and going forward we're going to stop using the term "boundaries" (alone) but use the term "«boundaries/membranes»" instead.
Summary: We've found that whenever try to use the word "bou... |
bd9d053f-e0e9-43e4-8888-2a9135589f59 | trentmkelly/LessWrong-43k | LessWrong | Just Pivot to AI: The secret is out
Despite the 'proof' being public the scaling hypothesis remained a Thielian secret until recently. It took chat-GPT to convince Facebook/Meta AI research that scaled-up prosaic AI was a plausible path to beyond human intelligence. They recently had internal meetings sounding the ala... |
bdf6c908-0203-4815-b14a-99935a77ebc8 | trentmkelly/LessWrong-43k | LessWrong | Bring Back the Sequences?
Given that
1. Deciding to read and actually reading the sequences is 'work'
2. Reading the latest frontpaged article on LessWrong is 'fun'
3. We frequently have gaps in the posting rate of articles that make it to the front page
4. There are many people who joined this community after th... |
15d018f7-47a2-4f3d-a4bd-19025ffea5c5 | trentmkelly/LessWrong-43k | LessWrong | Newcomb's Problem and Regret of Rationality
The following may well be the most controversial dilemma in the history of decision theory:
> A superintelligence from another galaxy, whom we shall call Omega, comes to Earth and sets about playing a strange little game. In this game, Omega selects a human being, sets dow... |
617283f5-455c-43d5-af9d-2f8334d57700 | trentmkelly/LessWrong-43k | LessWrong | [Linkpost] Some high-level thoughts on the DeepMind alignment team's strategy
Update: The original title "DeepMind alignment team's strategy" was poorly chosen. Some readers seem to have interpreted the previous title as meaning that this was everything that we had thought about or wanted to say about an "alignment pl... |
81d2b3f9-adda-4355-a3b3-f29ab3df19f4 | trentmkelly/LessWrong-43k | LessWrong | Project Proposal: Considerations for trading off capabilities and safety impacts of AI research
There seems to be some amount of consensus that people working on AI safety (at least within fairly mainstream ML / AI paradigms) shouldn't worry much about the effects of their projects on AI capabilities. New researchers ... |
8b48299a-b97f-4efd-b5a7-bdb05ab7498c | trentmkelly/LessWrong-43k | LessWrong | How to Defend the Indefensible
Image by Dall-E 3
----------------------------------------
I’ve got bad news and good news. The bad news is, you’ve painted yourself into a corner by committing yourself to a proposition which seems overwhelmingly unlikely to be true. It’s been so long since that trespass that you’v... |
5c2b6964-3d21-4e1e-a41a-69b3332251c2 | trentmkelly/LessWrong-43k | LessWrong | Meetup : Melbourne social meetup
Discussion article for the meetup : Melbourne social meetup
WHEN: 24 August 2012 07:00:00PM (+1000)
WHERE: Ben's house; see mailing list, Carlton VIC 3053, Australia
Melbourne's next social meetup is on Friday 24th August, 6:30 for 7pm, at my house. If you have any trouble working ou... |
23d21174-6e0c-40b9-8e08-a77631ec436d | trentmkelly/LessWrong-43k | LessWrong | An inquiry into the thoughts of twenty-five people in India
Sometimes I get excited about running surveys. Here is a Positly one from November 2020 in which I asked the following questions, to participants from India:
1. What are you looking forward to this week?
2. What do you think of as the most important thing ... |
64536eb4-db20-4908-83ef-3556ec2efa79 | trentmkelly/LessWrong-43k | LessWrong | Finite Factored Sets
This is the edited transcript of a talk introducing finite factored sets. For most readers, it will probably be the best starting point for learning about factored sets.
Video:
(Lightly edited) slides: https://intelligence.org/files/Factored-Set-Slides.pdf
1. Short Combinatorics Talk
1... |
48b16aef-0ef6-4564-be65-e3d1fb6dc2d8 | trentmkelly/LessWrong-43k | LessWrong | How tall is the Shard, really?
According to Wikipedia, the Shard (the tallest building in the UK) stands "309.6 meters (1,016 feet) high". I put this in my Anki deck as "Height of the Shard / 310m", but I was saying "height" to mean "tallness" (because I don't much like that word) and I had assumed Wikipedia was using... |
dfcc9668-66d4-4f89-bbd5-5a10bee06841 | trentmkelly/LessWrong-43k | LessWrong | Limitations of eyewitness testimony
From wikipedia:
Eyewitness testimony isn't reliable-- it degrades rapidly with time (significant fading in 20 minutes), is easily overridden by circumstances (people are apt to assume that the guilty person is in a line-up unless they're specifically told the guilty person might no... |
3e680d88-93df-4314-8ca6-55609f4d2551 | trentmkelly/LessWrong-43k | LessWrong | Unbendable Arm as Test Case for Religious Belief
In Materialist Conceptions of God, I wrote about how one can interpret religious claims as hyperstititions, beliefs that become true as a result of you believing in them. Since then, I’ve stumbled on a very simple, practical example of a hyperstition that you can test o... |
9ef8dc36-6215-4ecc-b8d6-b3dce201dec4 | trentmkelly/LessWrong-43k | LessWrong | An example and discussion of extension neglect
I recently used an automatic tracker to learn how I was spending my time online. I learned that my perceptions were systemically biased: I spend less time than I thought on purely non-productive sites, and far more time on sites that are quasi-productive.
For example, I ... |
3a497c88-8825-4663-9ed9-8e2cb0e4c299 | trentmkelly/LessWrong-43k | LessWrong | Good ways to monetarily profit from the increasing demand for power?
There have been several good posts about how to profit from a world of AI takeoff.
However, when it comes to individual investment recommendations, most of them are for direct AI companies or builders of GPUs and compute infrastrucutre
Recently, it... |
913c3299-7202-49dd-aaec-4f2fa4989d43 | StampyAI/alignment-research-dataset/lesswrong | LessWrong | Current AI harms are also sci-fi
A common argument I've seen for focusing on current AI harms over future ones is that the future ones are sci-fi (i.e. the ideas originated in the [science fiction genre](https://en.wikipedia.org/wiki/Science_fiction)). This argument is fallacious though, because many (perhaps all?) cu... |
586b1deb-9d27-4af7-a043-69a3e40aefda | trentmkelly/LessWrong-43k | LessWrong | Vector semantics and “Kubla Khan,” Part 2
Cross posted from New Savanna.
I want to continue the line of thought I opened up on March 14, Vector semantics and the (in-context) construction of meaning in Coleridge’s “Kubla Khan.”
There are at least two reasons why vector semantics is a useful tool for thinking about... |
c07837a6-4658-47ee-a4b3-cd6eac66350b | trentmkelly/LessWrong-43k | LessWrong | Link: More visibility for EA movement
"Effective Altruists Are a New Type of Nice Person"
http://www.vice.com/en_uk/read/why-are-young-people-being-so-altruistic-209 |
720e038b-1efb-492b-8273-fd7f21a556a6 | trentmkelly/LessWrong-43k | LessWrong | Open thread, Sep. 19 - Sep. 25, 2016
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... |
8efe250b-f417-4060-998b-78b3a5d5e2cf | trentmkelly/LessWrong-43k | LessWrong |
Has attempt been made to identify all mechanisms which allowed human species to come this in terms of value alignment?
My question is close to the point of view of Great Filter argument. I am looking for the list of structural and behavioral features that allowed development of complexity all the way to the current s... |
df36ee1c-83f7-4368-b204-e7aaf2decd7c | StampyAI/alignment-research-dataset/lesswrong | LessWrong | AI Safety is Dropping the Ball on Clown Attacks, and Mind Control in General
Epistemic status: High confidence (~>70%) that clown attacks are prevalent, and deliberately weaponized by governments and/or intelligence agencies in particular. Very high confidence (~>90%) that the human brain is highly vulnerable to clown... |
674d5cf9-4d7e-462a-a9e8-d62eb2e93b66 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Anthropic's Responsible Scaling Policy & Long-Term Benefit Trust
*I'm delighted that Anthropic has formally committed to [our responsible scaling policy](https://www-files.anthropic.com/production/files/responsible-scaling-policy-1.0.pdf).
We're also sharing more detail about the [Long-Term Benefit Trust](https://www.... |
98cd25b4-5936-4e21-9d39-a9ca2530a257 | trentmkelly/LessWrong-43k | LessWrong | Wormy the Worm
I.
Planarian worms are perfect regenerators. That is, they’re immortal. If you cut off a chunk of one, it will grow back, and the worm will continue as before.
This also means that a large enough chunk will grow into a separate worm. And in fact, some species of planaria reproduce this way, instead of... |
7deefd50-20df-4447-a4b2-efadd4169ac2 | trentmkelly/LessWrong-43k | LessWrong | One path to coherence: conditionalization
[Thanks to guy-whose-name-I-forgot working on paths to coherence for the conversation at Newspeak House after EAG London that prompted this thought, and Jozdien and eschatropic for some related chit-chat.]
Many agents start with some level of incoherence in their preferences... |
84691049-446d-446f-8c27-cfff1279159e | trentmkelly/LessWrong-43k | LessWrong | Learning doesn't solve philosophy of ethics
A putative new idea for AI control; index here.
This post will use the formalism of this post to illustrate some well known philosophical thought experiments and show why learning algorithms are not sufficient to solve them.
Examples
Death and life extension
A human con... |
9d7ca3a1-90cb-43b6-98e1-c5951c690b19 | trentmkelly/LessWrong-43k | LessWrong | Bounty: Diverse hard tasks for LLM agents
Update 3/14/2024: This post is out of date. For current information on the task bounty, see our Task Development Guide.
Summary
METR (formerly ARC Evals) is looking for (1) ideas, (2) detailed specifications, and (3) well-tested implementations for tasks to measure perfo... |
623adb59-00c7-4936-bd08-1f0246eae184 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | My take on Jacob Cannell’s take on AGI safety
Jacob Cannell wrote some blog posts about AGI safety / alignment and neuroscience between 2010 and 2015, which I read and enjoyed quite early on when I was first getting interested in the same topics a few years ago. So I was delighted to see him reappear on Lesswrong a ye... |
b3db9261-d4ab-4146-b11e-0e5d5838ffd0 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Motivations, Natural Selection, and Curriculum Engineering
*Epistemic status: I am professionally familiar with contemporary machine learning theory and practice, and have a long-standing amateur interest in natural history. But (despite being a long-term lurker on LW and AF) I feel very uncertain about the nature and... |
ca3bfd11-7db8-4a72-b9b5-9db8e80a04c4 | trentmkelly/LessWrong-43k | LessWrong | Why I Quit Social Media
Epistemic Status: Personal
I’m not a Puritan. I eat dessert, enjoy a good cocktail, and socialize a lot. I like fun, and I don’t think fun is wrong.
So I really understand the allergy to messages of ascetic self-denial.
But lately I’ve found it necessary to become more like this... and less ... |
4966211a-0663-4724-8e5c-87c33cd052a8 | StampyAI/alignment-research-dataset/blogs | Blogs | Product safety is a poor model for AI governance
*Rick Korzekwa, February 1, 2023*
*Note: This post is intended to be accessible to readers with relatively little background in AI safety. People with a firm understanding of AI safety may find it too basic, though they may be interested in knowing which kinds of poli... |
fa954b08-d372-42f5-be70-652f8c498065 | StampyAI/alignment-research-dataset/lesswrong | LessWrong | Goal-misgeneralization is ELK-hard
Consider an adversarial training-scheme for solving goal-misgeneralization, ( here i consider Redwood Research’s work on [“Adversarial Training for High-Stakes Reliability”](https://arxiv.org/pdf/2205.01663.pdf)).
Consider a model that was trained to perform a specific task. To guar... |
64643408-4bb5-45c4-8412-02324b804aad | StampyAI/alignment-research-dataset/lesswrong | LessWrong | Pivotal acts using an unaligned AGI?
Introduction
============
A recent post ([AGI Ruin: A List of Lethalities](https://www.lesswrong.com/posts/uMQ3cqWDPHhjtiesc/agi-ruin-a-list-of-lethalities)) gives an argument for humanity’s likely inability to survive the invention of superintelligent AGI. The argument goes very ... |
fb9a5d57-4096-40af-86dd-8226fdee2125 | trentmkelly/LessWrong-43k | LessWrong | Algo trading is a central example of AI risk
I suspect this observation is far from original, but I didn't see it explicitly said anywhere, so it seemed worthwhile spelling it out.
The paperclip maximizer is a popular example of unfriendly AI, however it's not the most realistic one (obviously it wasn't meant to be).... |
d6840082-6f5e-42d5-92a1-6ec225dc6bdb | trentmkelly/LessWrong-43k | LessWrong | New article on in vitro iterated embryo selection
The article by Robert Sparrow:
> A series of recent scientific results suggest that, in the not-too-distant future, it will be possible to create viable human gametes from human stem cells. This paper discusses the potential of this technology to make possible what I ... |
cf19178f-1779-4543-986b-087ad44aa292 | trentmkelly/LessWrong-43k | LessWrong | Attempt at a gears model of getting infected with COVID-19
I read a bit but don't know how good my understanding is. For reasoning about TAFFIX or the RaDVaC vaccine it would be good to have a decent gears model. I'm writing my current model so that others can check it for errors and not for others believing my model:... |
ba505bc3-1d26-43e8-be34-1f058aea997f | trentmkelly/LessWrong-43k | LessWrong | The usefulness of correlations
I sometimes wonder just how useful probability and statistics are. There is the theoretical argument that Bayesian probability is the fundamental method of correct reasoning, and that logical reasoning is just the limit as p=0 or 1 (although that never seems to be applied at the meta-lev... |
0a7eb346-dca0-4e88-ae1c-84b26b0a988d | trentmkelly/LessWrong-43k | LessWrong | The impact merge
(Cross-posted from Hands and Cities)
Lots of people want to do big things — start a big company, write a bestselling book, participate in an important project. I’ll call this the “accomplishment desire.”
Often this is centrally tied to social status; the relevant type of “bigness” is highly correlat... |
5db753de-1a7f-4d13-88ed-592c805b1044 | trentmkelly/LessWrong-43k | LessWrong | [LINK] On the unlikelihood of intelligent life
"The Planet-of-the-Apes Hypothesis" Revisited --Will Intelligence be a Constant in the Universe?
> If intelligence is good for every environment, we would see a trend in the encephalization quotient among all organisms as a function of time. The data does not show that. ... |
002e0914-ac45-4b76-a4d5-81c3f2ea6e99 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Gradient hacking is extremely difficult
**Epistemic Status**: *Originally started out as a comment on* [*this post*](https://www.lesswrong.com/posts/2sTTEkzvscWCPBQAk/gradient-filtering) *but expanded enough to become its own post. My view has been formed by spending a reasonable amount of time trying and failing to ... |
bd7c698f-198c-4a39-819e-c7eee14eabce | StampyAI/alignment-research-dataset/eaforum | Effective Altruism Forum | Critique of Superintelligence Part 4
This is part 4 of a 5-part sequence:
[Part 1: summary of Bostrom's argument](https://forum.effectivealtruism.org/posts/A8ndMGC4FTQq46RRX/criqitue-of-superintelligence-part-1)
[Part 2: arguments against a fast takeoff](https://forum.effectivealtruism.org/posts/BGWmAqrk64q2w6JjM/cr... |
fe0a4cfd-ca07-4bd2-8768-f469853562a6 | trentmkelly/LessWrong-43k | LessWrong | 3. Improve Cooperation: Better Technologies
Having read the manual for civilization and analyzed previous rounds of the game, let's turn to how to make the most of future rounds in this chapter. Over the course of civilization, we developed rights, contracts and various other mechanisms and institutions to help us coo... |
4f3317e3-f82b-4961-baea-f550204dd5c2 | StampyAI/alignment-research-dataset/lesswrong | LessWrong | Conversation with Eliezer: What do you want the system to do?
*This is a write-up of a conversation I overheard between Eliezer and some junior alignment researchers. Eliezer reviewed this and gave me permission to post this, but he mentioned that "there's a lot of stuff that didn't get captured well or accurately." I... |
d3183de8-6a8c-40ca-a874-6e65d885fce9 | trentmkelly/LessWrong-43k | LessWrong | One Argument Against An Army
I talked about a style of reasoning in which not a single contrary argument is allowed, with the result that every non-supporting observation has to be argued away. Here I suggest that when people encounter a contrary argument, they prevent themselves from downshifting their confidence by ... |
b7cfeaac-4daa-4237-85a5-3cc0c0a7bb93 | trentmkelly/LessWrong-43k | LessWrong | Why I got the smallpox vaccine in 2023
A common figure cited for the death toll of smallpox is 500 million, as in the title of the excellent essay and adaptation. Half a billion human lives lost is difficult to internalize. The more you think about it, the more it becomes clear that its eradication was among humanity'... |
81b72790-fc08-445c-bf69-a8536f95ba19 | trentmkelly/LessWrong-43k | LessWrong | The inescapability of knowledge
Financial status: This is independent research, now supported by a grant. I welcome further financial support.
Epistemic status: This is in-progress thinking.
----------------------------------------
Friends, I know that it is difficult to accept, but it just does not seem tenable th... |
61fc1a19-89ff-40eb-9022-4ea663a890f5 | StampyAI/alignment-research-dataset/arxiv | Arxiv | Expressing Robot Incapability
1. Introduction
----------------
As robots become increasingly capable, they may unintentionally mislead humans to overestimate their capabilities (Cha
et al., [2015](#bib.bib7)). Thus, it is important for a robot to communicate when it is incapable of accomplishing a task. There are ... |
4fdcdfd4-e758-4339-9fbb-b9da09e70ce7 | trentmkelly/LessWrong-43k | LessWrong | "AIXIjs: A Software Demo for General Reinforcement Learning", Aslanides 2017
|
7d7738ed-89d7-4eb3-a956-b355594f3772 | trentmkelly/LessWrong-43k | LessWrong | "If You're Not a Holy Madman, You're Not Trying"
I've been reading the hardcover SSC collection in the mornings, as a way of avoiding getting caught up in internet distractions first thing when I get up. I'd read many of Scott Alexander's posts before, but nowhere near everything posted; and I hadn't before made any a... |
2fd24242-379f-4104-a68a-aa4a15015bb9 | trentmkelly/LessWrong-43k | LessWrong | Apply to Visiting Fellows at Constellation, due June 13
Applications just opened for Constellation’s Visiting Fellows program.
Constellation is a research center in Berkeley. We host people who are working to help the world safely navigate the development of transformative AI. Our office houses great orgs and researc... |
d8cb26f8-1c39-410c-b852-cb478411a625 | trentmkelly/LessWrong-43k | LessWrong | [LINK] Productivity Ninja: 5 Powerful Tips For Getting More Stuff Done
From the blog [Bakadesuyo](http://www.bakadesuyo.com/2013/10/productivity-ninja/):
>1) Know When You’re At Your Best
>And plan accordingly. To be a productivity ninja focus less on time management, and more on managing your energy.
>Charlie Mung... |
fb80e746-d52e-4c68-8d6b-e25f173cd16d | StampyAI/alignment-research-dataset/lesswrong | LessWrong | Biases are engines of cognition
*This work was done while at* [*Conjecture*](https://conjecture.dev)*.*
Humans are not perfect reasoners. **Cognitive biases are not biases; they are engines of cognition, each exploiting regularities of the universe. Combining enough of these “cognitive biases” was enough to create a... |
91f0af0f-4e7e-40f0-ba30-da450b3ecac2 | trentmkelly/LessWrong-43k | LessWrong | Paraddictions: unreasonably compelling behaviors and their uses
I've been thinking about something I will clumsily call paraddictions: desires that don't quite rise to the level of an addiction, but that have a disproportionate and hard-to-moderate influence over your behavior. Can they be used as a tool for motivatio... |
f72f3d3b-854c-4fed-a313-c7b2861aed75 | trentmkelly/LessWrong-43k | LessWrong | Stampy's AI Safety Info soft launch
Stampy’s AI Safety Info is a project to create an interactive FAQ about existential risk from AI, started by Rob Miles. Our goal is to build a single resource aimed at informing all audiences, whether that means giving them a basic introduction to the concepts, addressing their obje... |
ce6a3cbc-c80d-4b59-87ce-32cf3f97f580 | trentmkelly/LessWrong-43k | LessWrong | Chapter 40: Pretending to be Wise, Pt 2
Harry, holding the tea cup in the exactly correct way that Professor Quirrell had needed to demonstrate three times, took a small, careful sip. All the way across the long, wide table that was the centerpiece of Mary's Room, Professor Quirrell took a sip from his own cup, making... |
5132b872-e512-4424-b088-730ae6bc4efd | trentmkelly/LessWrong-43k | LessWrong | Three signs you may be suffering from imposter syndrome
You do not have imposter syndrome, because imposter syndrome does not exist.
In most professional domains, no one knows how competent they are, exactly. Unless you’re in a field where there is an agreed upon, systematic method of ranking success, like sports, yo... |
dafd5242-c215-476b-bb22-033a82569903 | trentmkelly/LessWrong-43k | LessWrong | What Should I Do Today?
Skip to the last 3 paragraphs if too long.
What should I do today? This infamous question. I find the start of the day sets the precedent for the remainder—consciously or unconsciously. When you’re not severely unemployed, at least you have somewhere to be most days or at least some kind... |
42a6e579-f6a1-47e0-a84b-4ccf9ac82e07 | StampyAI/alignment-research-dataset/lesswrong | LessWrong | Are AI developers playing with fire?
**By** [**Marcus Arvan**](https://www.marcusarvan.net/)**, Associate Professor of Philosophy, The University of Tampa**
I write this post as a concerned citizen and philosopher based on my best understanding of how AI chatbots (i.e. large language models) function and my [previous... |
5722ca3e-0185-4813-b3bd-5996fafb37db | StampyAI/alignment-research-dataset/eaforum | Effective Altruism Forum | Crazy ideas sometimes do work
(On Aaron Gertler's suggestion, I am sharing something that I originally wrote on [my Facebook page](https://www.facebook.com/permalink.php?story_fbid=1523116118051987&id=100010608396052). Perhaps it might inspire some people to pick up one or two of their crazier ideas and actually do so... |
bc5a8e18-1980-4063-b285-77336e7c1043 | trentmkelly/LessWrong-43k | LessWrong | The positional embedding matrix and previous-token heads: how do they actually work?
tl;dr: This post starts with a mystery about positional embeddings in GPT2-small, and from there explains how they relate to previous-token heads, i.e. attention heads whose role is to attend to the previous token. I tried to make the... |
9feaf0c5-22d2-4596-a283-c0fe944103aa | trentmkelly/LessWrong-43k | LessWrong | A taxonomy of Oracle AIs
Sources: An old draft on Oracle AI from Daniel Dewey, conversation with Dewey and Nick Beckstead. See also Thinking Inside the Box and Leakproofing the Singularity.
Can we just create an Oracle AI that informs us but doesn't do anything?
"Oracle AI" has been proposed in many forms, but mo... |
e20e34eb-e4cd-4eb4-935a-789bf1c44e8e | trentmkelly/LessWrong-43k | LessWrong | Should people build productizations of open source AI models?
In early 2021, before I had developed any opinions on AI safety, my startup cofounder and I were contacted by a group of people who wanted an alternative to an AI CYOA platform, AIDungeon. They were frustrated with the site's current management and especial... |
9e3183cb-e60d-46bd-b73e-482f7df5e27a | trentmkelly/LessWrong-43k | LessWrong | indexical uncertainty and the Axiom of Independence
I’ve noticed that the Axiom of Independence does not seem to make sense when dealing with indexical uncertainty, which suggests that Expected Utility Theory may not apply in situations involving indexical uncertainty. But Googling for "indexical uncertainty" in combi... |
8a0f014e-85df-431f-8c07-1337b57ec3e3 | StampyAI/alignment-research-dataset/arxiv | Arxiv | The Challenge of Value Alignment: from Fairer Algorithms to AI Safety
Forthcoming in The Oxford Handbook of Digital Ethics The Challenge of Value Alignment: from Fairer Algorithms to AI Safety Iason Gabriel and Vafa Ghazavi 1. Introduction There has long been a view among observers of artificial intelligence (AI... |
cf9b462d-2863-42b8-9e67-5fe8de0f54ad | trentmkelly/LessWrong-43k | LessWrong | [Book Review] "The Vital Question" by Nick Lane
There is a singularity at the beginning of biology. We know what happened before biology―that's just physics. We know what happened after the first cells: evolution. But we don't know how the first cells formed out of nonliving matter.
The most popular theory is called ... |
dcf72aeb-d686-4a51-9212-cfcb67f88025 | trentmkelly/LessWrong-43k | LessWrong | A Year of Spaced Repetition Software in the Classroom
Last year, I asked LW for some advice about spaced repetition software (SRS) that might be useful to me as a high school teacher. With said advice came a request to write a follow-up after I had accumulated some experience using SRS in the classroom. This is my rep... |
5d3ca36d-73b0-4f14-920a-aa7a7d3f8a36 | trentmkelly/LessWrong-43k | LessWrong | Bitcoin and ESG Investing
How do cryptocurrencies fit in Environmental, Social, and Governance Investing?
Original Post
Environmental, Social, and Governance (ESG) Investing is a practice that evolved from the practice of excluding equities based on moral values. ESG Investing is broad and constantly evolving, and ke... |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.