id stringlengths 36 36 | source stringclasses 15
values | formatted_source stringclasses 13
values | text stringlengths 2 7.55M |
|---|---|---|---|
0c6fbd02-109a-4768-8c18-4f4b8965eacb | StampyAI/alignment-research-dataset/lesswrong | LessWrong | Apply for MATS Winter 2023-24!
[Applications are now open](https://airtable.com/appxum3Sqh7TdDvdg/shrtfHWhRFZdkhaIM) for the Winter 2023-24 cohort of [MATS](https://www.matsprogram.org/home) (previously SERI MATS). Mentors include [Adrià Garriga Alonso](https://www.lesswrong.com/users/rhaps0dy), [Alex Turner](https://... |
a108bc20-eb1d-483f-9034-704d5dbe2a20 | trentmkelly/LessWrong-43k | LessWrong | Collaborative project: New rationality materials page
As you are probably aware, we have a new front page featuring a graphic of a brain. One of the links on the front page, "A source of edited rationality materials," links to the Less Wrong meetup group resources page. A number of users have suggested that this isn't... |
cfb0e7cc-830e-4c9b-91a7-1da064d2e334 | trentmkelly/LessWrong-43k | LessWrong | Non-Transactional Compliments
I used to dislike receiving compliments.
I used to dislike giving compliments.
One day, I sat down and reflected on the nature of compliments.
What other things do you “give” and “receive”?
Material goods, gifts, capital. Items of value.
I realized I’d been thinking of compliments as a... |
76bd22ab-183d-4378-8e5d-1a25a8b06b4f | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Epistemological Vigilance for Alignment
*This post is part of the work done at* [*Conjecture*](https://conjecture.dev)*.*
Nothing hampers Science and Engineering like unchecked assumptions.
As a concrete example of a field ridden with hidden premises, let's look at sociology. Sociologist must deal with the feedback ... |
20db76db-da30-4cbe-bfc3-d0f18df6bcdb | trentmkelly/LessWrong-43k | LessWrong | Logical predictors cannot dominate their own complexity class
One good property for a logical predictor (A Turing machine which tries to guess the output of an environment E that is too complex to simulate) is that it is able to to do asymptotically at least as well as any other logical predictor in some class.
We sa... |
88a6a3e1-ffa1-4234-b47c-1af21f38cd59 | StampyAI/alignment-research-dataset/blogs | Blogs | Historic trends in chess AI
The Elo rating of the best chess program measured by the Swedish Chess Computer Association did not contain any greater than 10-year discontinuities between 1984 and 2018. A four year discontinuity in 2008 was notable in the context of otherwise regular progress.
Details
-------
This c... |
498fe684-5059-479e-b71c-46f83712060f | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | History of the Development of Logical Induction
.mjx-chtml {display: inline-block; line-height: 0; text-indent: 0; text-align: left; text-transform: none; font-style: normal; font-weight: normal; font-size: 100%; font-size-adjust: none; letter-spacing: normal; word-wrap: normal; word-spacing: normal; white-space: nowr... |
4a6f4057-5464-40a7-beaf-9023748f3b0c | trentmkelly/LessWrong-43k | LessWrong | Is there a LessWrong Index?
Pardon me, please, if this is not the way to go about asking such questions (it's all I know). Is this more for LessWrong itself, or for LessWrong Discussion?
Is there some kind of comprehensive organization by subject of LessWrong posts?
I know there are the sequences, but also a lot ... |
7038c412-f946-4273-9168-7cbc63decb0d | trentmkelly/LessWrong-43k | LessWrong | PaLM API & MakerSuite
This seems overshadowed by the GPT-4 news, but apparently Google is also releasing PaLM and an API along with "MakerSuite, a tool that lets developers start prototyping quickly and easily."
Not many details yet (perhaps they rushed this out to compete with the GPT-4 announcement?), but some feat... |
a8ee0a87-2cec-4b3f-8741-dc073e0f61d0 | trentmkelly/LessWrong-43k | LessWrong | When you are your own favorite charity
If you think that SIAI is the best charity, and you work for SIAI, should you give any money to charity? |
d2b08202-134e-47fe-9159-0b73ca47610b | trentmkelly/LessWrong-43k | LessWrong | Computational Thread Art
This post describes the iterative process I went through while creating the thread art which you can see featured on my website. This is also crossposted to my personal blog.
Black & White Algorithm
The basic version of the algorithm is pretty straightforward. The image is rescaled so that w... |
93e96dc8-2153-48fa-9343-66cf6afa2228 | trentmkelly/LessWrong-43k | LessWrong | Reality as Category-Theoretic State Machines: A Mathematical Framework
Epistemic Status: Exploratory theoretical Mathematical foundations are solid, physical interpretations are speculative but grounded in established theoretical approaches._
Overview This post proposes treating reality as an interconnected system of... |
d43e5eb5-05b1-492f-9321-b6dc40441f40 | trentmkelly/LessWrong-43k | LessWrong | What is the most impressive game LLMs can play well?
Epistemic status: This is an off-the-cuff question.
~5 years ago there was a lot of exciting progress on game playing through reinforcement learning (RL). Now we have basically switched paradigms, pretraining massive LLMs on ~the internet and then apparently doing ... |
eb3c83a5-a005-4995-8123-b6e0bfdd5a4b | StampyAI/alignment-research-dataset/lesswrong | LessWrong | Reframing misaligned AGI's: well-intentioned non-neurotypical assistants
I think when people imagine misaligned AGI's, they tend to imagine a superintelligent agent optimizing for something other than human values (e.g. paperclips, or a generic reward signal), and mentally picture them as adversarial or malevolent. I ... |
bbbe7e84-57f9-4ebe-a2e9-925c52aded36 | trentmkelly/LessWrong-43k | LessWrong | Importing a Python File by Name
Let's say you have a python script:
2024-01-24--evaluate-chimeras.py
And you want to pull a section of it out into a separate file:
2024-01-25--strand-split-artifacts.py
You would hope you could just do something like:
import "/path/to/2024-01-25--strand-split-artifacts.py"
But... |
fe73b023-9cdd-4192-bbf4-73d0a6a5a581 | trentmkelly/LessWrong-43k | LessWrong | Meetup : Munich Meetup in December
Discussion article for the meetup : Munich Meetup in December
WHEN: 13 December 2014 03:00:00PM (+0100)
WHERE: Gast, Rosenheimer Straße 5, 81667 Munich, Germany
The next Munich meetup will take place on December 13th. We will play the calibration game and discuss recent articles p... |
59b7ef3d-5a0a-4879-8c40-6741bd0184a4 | trentmkelly/LessWrong-43k | LessWrong | Announcing Atlas Computing
Atlas Computing is a new nonprofit working to collaboratively advance AI capabilities that are asymmetrically risk-reducing. Our work consists of building scoped prototypes and creating an ecosystem around @davidad’s Safeguarded AI programme at ARIA (formerly referred to as the Open Agency A... |
69acab33-889e-4548-8d62-73c2a53fc7d4 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Distinguishing AI takeover scenarios
*Epistemic status: lots of this involves interpreting/categorising other people’s scenarios, and could be wrong. We’d really appreciate being corrected if so. [ETA: so far, no corrections.]*
**TLDR**: see the [summary table](https://www.lesswrong.com/posts/qYzqDtoQaZ3eDDyxa/disti... |
b8e3e220-4549-4143-a8f8-11bf4b18a6ea | trentmkelly/LessWrong-43k | LessWrong | Scope Insensitivity Judo
It's easy to bemoan scope insensitivity, a special case of that phenomenon where we mere humans end up caring more about the death of one person than one hundred, better remember the last bite of a meal than the first dozen, and think less is more and more is less. After all, if we didn't negl... |
88a80c5f-1847-4c3a-b6bc-7ec04f5d781e | trentmkelly/LessWrong-43k | LessWrong | Need help with an MLP fanfiction with a transhumanist theme.
EDIT: I am now taking arguments for alicornism. Alicornism being the placeholder term I've given to the stance that all ponies should be alicorns. Please PM me or post here if you have a good one, or an argument against one of anti-alicornism's strongest ... |
7cff98b9-e04f-4b8e-8550-627bb378d65d | trentmkelly/LessWrong-43k | LessWrong | Chair Hanging Peg
Our house is now back up to full occupancy, which meant some things that had been hanging out in unused rooms needed better homes. Four of those things were folding chairs, and now they are up and out of the way:
I made this out of three pieces of scrap wood: two bits of 2x6 and a 1 1/4 dowel. Her... |
690699db-06df-42f1-86c7-a4a1f78e128c | trentmkelly/LessWrong-43k | LessWrong | Should we publish mechanistic interpretability research?
TL;DR: Multiple people have raised concerns about current mechanistic interpretability research having capabilities externalities. We discuss to which extent and which kind of mechanistic interpretability research we should publish. The core question we want to ... |
90578756-5e6c-4de1-aa0c-e46ac4caf3e2 | trentmkelly/LessWrong-43k | LessWrong | Weekly Roundup #2
Experimenting with pushing the non-Covid roundup on Friday instead of Thursday, and trying out dropping non-Covid from the title, and I plan to start pulling more focused posts out of the bank of written stuff in a week or two. Over time I’ll figure out how much belongs in a digest, how much is time ... |
fb32049c-a00a-4665-81e5-4cf72fbf4311 | trentmkelly/LessWrong-43k | LessWrong | Does the lottery ticket hypothesis suggest the scaling hypothesis?
The lottery ticket hypothesis, as I (vaguely) understand it, is that artificial neural networks tend to work in the following way: When the network is randomly initialized, there is a sub-network that is already decent at the task. Then, when training ... |
c46ff3ce-a19c-4275-ad37-526b6a18590e | trentmkelly/LessWrong-43k | LessWrong | The term 'altruism' in group selection
The way I see it, altruism has been the big selling point for group selection. The only way altruism would have been able to evolve is through group selection, so the presence of altruism is strong evidence for the existence of group selection.
Group selectionists have been (rig... |
ab48818c-3124-411b-b457-2c6fc8b37674 | trentmkelly/LessWrong-43k | LessWrong | Stupid Questions, December 2015
This thread is for asking any questions that might seem obvious, tangential, silly or what-have-you. Don't be shy, everyone has holes in their knowledge, though the fewer and the smaller we can make them, the better.
Please be respectful of other people's admitting ignorance and don't ... |
b0e7951c-6e7e-4ea1-b504-18171c913cc4 | trentmkelly/LessWrong-43k | LessWrong | Passing the Recursive Buck
Followup to: Artificial Addition, The Ultimate Source, Gödel, Escher, Bach: An Eternal Golden Braid
Yesterday, I talked about what happens when you look at your own mind, reflecting upon yourself, and search for the source of your own decisions.
Let's say you decided to run into a burnin... |
50f2c94d-dab4-4fff-8970-58d2ebd8b2ec | trentmkelly/LessWrong-43k | LessWrong | Pseudorandomness contest, Round 1
I’ve decided to run a pseudorandomness contest — a reverse Turing test of sorts, if you will. Winners will get to send some of my money to a charity of their choice! Here’s how it’s going to work:
Round 1: Pseudorandom number generation
Should you choose to take part, you will have ... |
21f10a15-0599-4578-9ea9-6c1408aced75 | trentmkelly/LessWrong-43k | LessWrong | MIRI: 2020 Updates and Strategy
Highlights from MIRI's year-end strategy update:
* MIRI executive director Nate Soares reports that the "new research directions" we initiated in 2017 have "at this point, largely failed, in the sense that neither Eliezer nor I have sufficient hope in it for us to continue focusing ou... |
bfef321f-acf0-42b0-a808-8ed2150a3bf1 | StampyAI/alignment-research-dataset/eaforum | Effective Altruism Forum | Cognitive science and failed AI forecasts
Not too long ago, the trend in cognitive science was that connectionism will fail. Many cognitive scientists thought that connectionism is limited and it is highly unlikely that we can actually assimilate the functions of human minds by imitating the neural circuitry of the h... |
62248c3d-d02f-4b72-aca1-cbadfb3882b1 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Instrumental Occam?
In *[The Real Rules have No Exceptions](https://www.lesswrong.com/posts/duxy4Hby5qMsv42i8/the-real-rules-have-no-exceptions),* Said says (apparently speaking of *instrumental* rules, not epistemic ones):
> *Prefer simplicity in your rules. Be vigilant that your rules do not grow too complex; make... |
0f382683-2cf4-478b-b3b4-5ae5c858c8b1 | trentmkelly/LessWrong-43k | LessWrong | What does the reaction to NFTs tell us about how people evaluate ecological damage?
Recently, the Ethereum network finalised the ERC-721 interface standard for Non-Fungible Tokens (NFTs). This standard lays out a protocol for the exchange and ownership of this new class of assets. An extremely simple explanation of th... |
c3bf3999-c81f-4b97-8c24-195214aa3062 | trentmkelly/LessWrong-43k | LessWrong | Levers for Biological Progress - A Response to "Machines of Loving Grace"
By Niko McCarty for Asimov Press
Dario Amodei, the CEO of Anthropic, recently published an essay called “Machines of Loving Grace.” It sketches out his vision for how AI could radically transform neuroscience, economics, diplomacy, and the mean... |
c6c96275-e4ce-447b-94a8-348dfa6f7b70 | trentmkelly/LessWrong-43k | LessWrong | Far negatives of cryonics?
In considering the pros and cons of cryonics, has anyone addressed the possibility of being revived in an unpleasant future, for instance as a "torture the infidel" exhibit in a theme park of a theocratic state? I had some thoughts on the issue but figured I would see what else has been writ... |
1ef5dc32-1e15-49db-944a-5cff6c36fa89 | trentmkelly/LessWrong-43k | LessWrong | Existential Risk II
Meta
-This is not a duplicate of the original less wrong x-risk primer. I like lukeprog's article just fine, but it works mostly as a punch in the gut for anyone who needs a wake up call. Very little of the actual research on x-risk is discussed in that article, so the gap that was there before ... |
536b70b2-e5fe-471f-8a90-71093afbbd34 | trentmkelly/LessWrong-43k | LessWrong | The Median is Less than the Average
In The Economics of Media I discussed how news is written by large organizations including businesses, governments and major political parties.
The same was true in the age of paper media except with the addition of more powerful local monopolies. Much of the United States (and lik... |
487c2102-aaad-4efa-9c8a-90f707bea83e | trentmkelly/LessWrong-43k | LessWrong | Meetup : Rational-Humanist Open Mic, NYC September 5th
Discussion article for the meetup : Rational-Humanist Open Mic, NYC September 5th
WHEN: 05 September 2012 07:00:00PM (-0400)
WHERE: 159 E Houston Street, New York, NY 10003
Last month was the first Open Mind Open Mic, a cultural event co-hosted by the Center fo... |
862aeb1c-34ff-4a92-a4a8-4ed5732021e8 | trentmkelly/LessWrong-43k | LessWrong | Deliberately trying to annoy simulators?
I regularly seek inspiration by taking long solo walks; and during my most recent such, considering what practical consequences (if any) there would be of the universe I know being a simulation - something flipped in my head, and I thought to myself, "Screw the simulators. If I... |
b2d6e693-3798-425f-b3f3-e6042cc3b099 | StampyAI/alignment-research-dataset/eaforum | Effective Altruism Forum | Could unions be an underrated driver for AI safety policy?
Disclaimer: I’m not an expert in politics unions, or AI safety policy.
It seems like a significant number of people who deem AI safety to be important feel that “AI Policy” is an important potential solution. Personally, from the outside, AI policy seems very... |
e9549e8e-f640-456d-a2c0-0b420875c04d | trentmkelly/LessWrong-43k | LessWrong | Entangled Photons
Previously in series: Decoherence as Projection
Today we shall analyze the phenomenon of "entangled particles". We're going to make heavy use of polarized photons here, so you'd better have read yesterday's post.
If a particle at rest decays into two other particles, their net momentum must add... |
fbacd2ab-91e5-4931-9d65-82c6ca391464 | StampyAI/alignment-research-dataset/eaforum | Effective Altruism Forum | When Will We Spend Enough to Train Transformative AI
The world has experienced exponential growth in the power of deep learning systems over the past 10 years. Since AlexNet in 2012, there has been continuous rapid growth. However, it is worth decomposing the causes for that growth in order to predict the power of fut... |
0a310bc9-b732-417b-836a-0f6fd2c1769c | trentmkelly/LessWrong-43k | LessWrong | Deception by cherry picking
[This is an idea I've had 'kicking around' for a long time - may as well see what LW makes of it.]
The Bayesian update procedure tacitly presupposes that the event we're updating on is not itself random. Indeed, naively updating on a random event, that turns out to be correlated with the v... |
9b447f17-a4f0-49f8-9850-56b66eef75b5 | trentmkelly/LessWrong-43k | LessWrong | Electric Grid Cyberattack: An AI-Informed Threat Model
Author’s note: This post was written during a two-week long research sprint at constellation.org. My analysis represents a preliminary attempt to formulate a detailed AI-cyber threat model of attacks against the energy grid. The conclusions reached here are there... |
65003cee-ed30-4bfe-8424-6d680060a805 | StampyAI/alignment-research-dataset/blogs | Blogs | Historic trends in book production
The number of books produced in the previous hundred years, sampled every hundred or fifty years between 600AD to 1800AD contains five greater than 10-year discontinuities, four of them greater than 100 years. The last two follow the invention of the printing press in 1492.
The re... |
c7a4a577-2be2-45d0-ba18-d5a844895adb | StampyAI/alignment-research-dataset/lesswrong | LessWrong | Superintelligence reading group
In just over two weeks I will be running an online reading group on Nick Bostrom's *[Superintelligence](http://en.wikipedia.org/wiki/Superintelligence)*, on behalf of [MIRI](http://intelligence.org/). It will be here on LessWrong. This is an advance warning, so you can [get a copy](http... |
d7c0df47-86c1-4f44-a26d-7996f1ab7c05 | trentmkelly/LessWrong-43k | LessWrong | Links to discussions on social equilibrium and human value after (aligned) super-AI?
Suppose we work hard and get a stroke of luck and unaligned AI isn't a problem.
What does society look like when human-level and super-intelligent AIs can be spun up and work way cheaper than humans do?
I'm in a bit of an existentia... |
c1b05e30-6a35-40bd-be20-fd71b5f20e90 | StampyAI/alignment-research-dataset/blogs | Blogs | New paper: “Alignment for advanced machine learning systems”
[](https://intelligence.org/files/AlignmentMachineLearning.pdf)MIRI’s research to date has focused on the problems that we laid out in our [late 20... |
60216e36-64ae-41e3-82e7-03ddbe1fae29 | StampyAI/alignment-research-dataset/eaforum | Effective Altruism Forum | Expected impact of a career in AI safety under different opinions
This table explores the ethical [expected-value](https://www.lesswrong.com/tag/expected-value) of a career in AI safety, under different opinions about AI and the long-term future. I made it for considering the robustness of the value of my possible car... |
683911ae-3828-4101-941d-9e3781e82c94 | StampyAI/alignment-research-dataset/arxiv | Arxiv | Risk-Aware Active Inverse Reinforcement Learning
Risk-Aware Active Inverse Reinforcement Learning
Daniel S. Brown, Yuchen Cui, and Scott Niekum
Department of Computer Science
University of Texas at Austin, United States
dsbrown@cs.utexas.edu, yuchencui@utexas.edu, sniekum@cs.utexas.edu
Abstract: Active learning from... |
85e77d3f-c35b-44a3-9e7f-d5a00e8c705f | trentmkelly/LessWrong-43k | LessWrong | Steganography in Chain of Thought Reasoning
Here I give a possible phenomenon of steganography in chain of thought reasoning, where a system doing multi-stage reasoning with natural language encodes hidden information in its outputs that is not observable by humans, but can be used to boost its performance on some tas... |
99923c49-3d2f-4419-ad44-3c13a3df73d0 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Alignment Newsletter #25
Highlights
==========
**[Towards a New Impact Measure](https://www.alignmentforum.org/posts/yEa7kwoMpsBgaBCgb/towards-a-new-impact-measure)** *(Alex Turner)*: This post introduces a new idea for an impact measure. It defines impact as change in our ability to achieve goals. So, to measure imp... |
7382b026-76cb-4f4b-bc3c-89e02efe286f | trentmkelly/LessWrong-43k | LessWrong | Avoiding Rationalization
Previously: Red Flags for Rationalization
----------------------------------------
It is often said that the best way to resist temptation is to avoid it. This includes the temptation to rationalize. You can avoid rationalization by denying it the circumstances in which it would take place.
... |
2fce9cea-fc99-48d8-9446-0b3b1f861d01 | StampyAI/alignment-research-dataset/blogs | Blogs | New Transcript: Eliezer Yudkowsky and Massimo Pigliucci on the Intelligence Explosion
In this 2010 conversation hosted by [bloggingheads.tv](http://www.bloggingheads.tv), [Eliezer Yudkowsky](http://yudkowsky.net/) and [Massimo Pigliucci](http://rationallyspeaking.blogspot.com/) attempt to unpack the fundamental assump... |
dcf55bb1-d31c-46ab-8e60-1e26d2b0a05c | trentmkelly/LessWrong-43k | LessWrong | Information Markets
Epistemic status: Exploratory. This post would be shorter and contain more math if I'd thought about it more first.
I don't like prediction markets, as currently described. They're similar to ordinary stock markets, which economists say are supposed to be efficient, but don't look it. People say "... |
e1f5fc36-da0d-4c32-89b2-9b439d9f1a8c | trentmkelly/LessWrong-43k | LessWrong | Meetup : Cambridge First-Sunday Meetup
Discussion article for the meetup : Cambridge First-Sunday Meetup
WHEN: 03 March 2013 02:00:00PM (-0500)
WHERE: 25 Ames St Cambridge, MA
Cambridge/Boston-area Less Wrong meetups recur on the first and third Sunday of every month, 2pm at the MIT Landau Building [25 Ames St, Bld... |
a7661aec-c669-47e4-98d8-80b2cf4d66b1 | trentmkelly/LessWrong-43k | LessWrong | Performance Trends in AI
|
d976ff8b-91f2-46c8-9b1b-3789d0e4b869 | trentmkelly/LessWrong-43k | LessWrong | What should be my triggers for initiating self quarantine re: Corona virus
[Some answer here, where I asked this question in comments.]
Is there some reliable source on the triggers for quarantining one's self? Both immediate triggers (when X happens, I will not leave the house again until y happens), and warning sig... |
a832680d-b1e3-430e-96da-6f7db6ce0a80 | trentmkelly/LessWrong-43k | LessWrong | Unpacking the Concept of "Blackmail"
Keep in mind: Controlling Constant Programs, Notion of Preference in Ambient Control.
There is a reasonable game-theoretic heuristic, "don't respond to blackmail" or "don't negotiate with terrorists". But what is actually meant by the word "blackmail" here? Does it have a place as... |
673992e5-9a41-4dc5-8037-c25eda3198a0 | trentmkelly/LessWrong-43k | LessWrong | A Correspondence Theorem
I’ve been thinking lately about formalizations of the Correspondence Principle - the idea that new theories should reproduce old theories, at least in the places where the old theories work. Special relativity reduces to Galilean relativity at low speed/energy, general relativity reduces to Ne... |
8dea8d1a-6607-442f-8d6b-8c23e7247a01 | trentmkelly/LessWrong-43k | LessWrong | [Link] Forty Days
A post from Gregory Cochran's and Henry Harpending's excellent blog West Hunter.
> One of the many interesting aspects of how the US dealt with the AIDS epidemic is what we didn’t do – in particular, quarantine. Probably you need a decent test before quarantine is practical, but we had ELISA by 198... |
4ef9c192-bfe7-4f2d-b64e-b5dee2a8be66 | trentmkelly/LessWrong-43k | LessWrong | Why real estate is the only investment that matters in AI dominated future
Thinking of best long term investments for the next decades, I came to the following conclusions.
With the advancements in AI technology in the next 10-20 years, the cost of intelligence, manpower, and construction is expected to decrease sign... |
66ef30f0-3ab7-42c4-805b-ebcbb27f7b3f | trentmkelly/LessWrong-43k | LessWrong | Who is Harry Potter? Some predictions.
Microsoft has released a paper called "who is Harry Potter" in which they claim to make a neural network forget who Harry Potter is.
https://www.microsoft.com/en-us/research/project/physics-of-agi/articles/whos-harry-potter-making-llms-forget-2/
Here are some of my predictions... |
af2f0b0e-5746-48fb-bfe7-9c1be40a1df7 | trentmkelly/LessWrong-43k | LessWrong | Anthropic's SoLU (Softmax Linear Unit)
Anthropic continues their Transformer Circuits Thread. In previous work they were unable to make much progress interpreting MLP layers. This paper is focused on addressing that limitation.
High-level Takeaways
> In this paper, we report an architectural change which appears to ... |
2b7fde8b-29ad-4411-87d3-a2b1975b809f | trentmkelly/LessWrong-43k | LessWrong | Can growth continue?
I gave a five-minute talk talk for Ignite Long Now 2022. It’s a simplified and very condensed treatment of a complex topic. The video is online; below is a transcript with selected visuals and added links.
----------------------------------------
Humanity has had a pretty good run so far. In the... |
86468669-4f06-4d42-b881-0e5928558b50 | trentmkelly/LessWrong-43k | LessWrong | the dumbest theory of everything
which is maybe to say the simplest?
abstract: in this short introductory paper, i present a not fake theory of everything. i start by extending christopher alexander’s definition of life as a status and attribute. i then posit that life is the physical interface of consciousness, refe... |
a6eaa513-eb86-494b-808a-64b7234dfda4 | trentmkelly/LessWrong-43k | LessWrong | Strengthening the foundations under the Overton Window without moving it
As I understand them, the social rules for interacting with people you disagree with are like this:
* You should argue with people who are a bit wrong
* You should refuse to argue with people who are very wrong, because it makes them seem more... |
983e9f20-50fc-42dd-8dfa-18871590661b | trentmkelly/LessWrong-43k | LessWrong | Bugs?
I got two notifications about replies to my comment on Valentine's "Kensho" post (one reply by Qiaochu_Yuan and one by Valentine.) I can read the replies, but I cannot reply to them, and when I open the original post I don't see the replies under my own comment. Help? |
ed579d0a-b70c-4c21-8d70-1c9d6d8e2f81 | trentmkelly/LessWrong-43k | LessWrong | What is the foundation of me experiencing the present moment being right now and not at some other point in time?
This question may be either trivial to answer and I just don‘t get it, or it may be highly controversial. I am trying to pose my problem as shortly as I can.
What is responsible for my subjective experien... |
b7b287d6-ec51-4ca5-a51d-6246cd9d7567 | trentmkelly/LessWrong-43k | LessWrong | Personal predictions for decisions: seeking insights
Do some of you keep a record of personal predictions?
I.e. either predictions of personal events, such as
> likelihood of getting at least a 10% raise in the next 12 months
or events conditional on decisions, actions, such as
> likelihood of getting at least 10%... |
bc3010db-d92e-40b8-8f5b-9dc93b3c85b3 | StampyAI/alignment-research-dataset/special_docs | Other | How to study superintelligence strategy
(Last updated Feb. 11, 2015.)
What could an economics graduate student do to improve our strategic picture of superintelligence? What about a computer science professor? A policy analyst at RAND? A program director at IARPA?
In the last chapter of [\*Superintelligence\*](htt... |
5a69f4e3-b20f-4975-8876-59c555e6950b | StampyAI/alignment-research-dataset/lesswrong | LessWrong | What can thought-experiments do?
Introduction
------------
What's the difference between actually being in a situation and just thinking about the situation? In other words, what's the difference between a real-life ethical dilemma and a hypothetical ethical dilemma?
There are many differences.
* Some differences a... |
98d88006-8964-4c09-a7b4-4cee62f1a1ef | trentmkelly/LessWrong-43k | LessWrong | Selection Has A Quality Ceiling
Suppose we’re working on some delightfully Hard problem - genetically engineering a manticore, or terraforming Mars, or aligning random ML models. We need very top tier collaborators - people who are very good at a whole bunch of different things. The more they’re good at, and the bette... |
6e09256b-d240-4a29-8eb0-37e898680f60 | trentmkelly/LessWrong-43k | LessWrong | The right kind of fun?
If you consider that the utility generated by working is much greater than the utility directly generated by having fun, then the main thing that you're going to optimizing when you have fun is how much motivation the memory of having that fun increases your working capabilities. This is distinc... |
8bc7da83-edbb-4aa6-b502-0534566146be | trentmkelly/LessWrong-43k | LessWrong | [SEQ RERUN] Risk-Free Bonds Aren't
Today's post, Risk-Free Bonds Aren't, was originally published on 22 June 2007. A summary (taken from the LW wiki):
> There are no risk-free investments. Even US treasury bills would fail under a number of plausible "black swan" scenarios. Nassim Taleb's own investment strategy does... |
1315efed-d4f7-4298-9a46-00dcb316e7ee | trentmkelly/LessWrong-43k | LessWrong | The Inner Alignment Problem
This is the third of five posts in the Risks from Learned Optimization Sequence based on the paper “Risks from Learned Optimization in Advanced Machine Learning Systems” by Evan Hubinger, Chris van Merwijk, Vladimir Mikulik, Joar Skalse, and Scott Garrabrant. Each post in the sequence corre... |
0a9d6ecb-0bb0-4a83-81fc-5753e1dbf510 | trentmkelly/LessWrong-43k | LessWrong | More on policy arguments and the AB problem
This is a followup to this earlier post, which I realized today didn't quite cover everything I wanted to.
To recap, the AB problem (note: I don't know who coined this name, does anyone know the origin?) is a problem that arises in arguments that leads people mischaracteriz... |
78b26a90-0580-4746-9b0a-94fa2b300101 | trentmkelly/LessWrong-43k | LessWrong | Rationality Quotes: July 2010
This is our monthly thread for collecting these little gems and pearls of wisdom, rationality-related quotes you've seen recently, or had stored in your quotesfile for ages, and which might be handy to link to in one of our discussions.
* Please post all quotes separately, so that they ... |
d56bbe01-9872-458f-8652-5dd4d30e3d34 | trentmkelly/LessWrong-43k | LessWrong | Experimental psychology on word confusion
There's plenty of experimental work about how humans make poor judgments and decisions, but I haven't yet found much about how humans make poor judgments and decisions because of confusions about words. And yet, I expect such errors are common — I, at least, encounter them fre... |
5ed9a0e0-390b-4954-982e-53572ec70e8a | StampyAI/alignment-research-dataset/arxiv | Arxiv | When Should a Decision Maker Ignore the Advice of a Decision Aid?
WHEN SHOULD A DECIS ION MAKER IGNORE THE ADVICE
OF A DECISION AID?
Paul E. Lehner
School of Information Technology
George Mason University
4400 university Drive
Fairfax, Virginia 22030 Decision Science Consortium
1895 Preston White Drive
Reston,... |
1465b079-d96d-4bc4-975a-bff162ad9ce9 | StampyAI/alignment-research-dataset/blogs | Blogs | communicating with successful alignment timelines
communicating with successful alignment timelines
-------------------------------------------------
consider the following plan for AI alignment:
1. (quantum-)generate an asymmetric [cryptographic signature keypair](https://en.wikipedia.org/wiki/Digital_signature) ... |
111a1ac6-382d-43b6-98e8-9b00ce04ea13 | trentmkelly/LessWrong-43k | LessWrong | Free Will as Unsolvability by Rivals
> Nadia wanted to solve Alonzo. To reduce him to a canonical, analytic representation, sufficient to reconfigure him at will. If there was a potential Alonzo within potential-Alonzo-space, say, who was utterly devoted to Nadia, who would dote on her and die for her, an Alonzo-solut... |
cdb47829-9f9b-4843-885b-4d23fa11c1d1 | StampyAI/alignment-research-dataset/special_docs | Other | Information Gathering Actions Over Human Internal State.
Information Gathering Actions over Human Internal State
Dorsa Sadigh S. Shankar Sastry Sanjit A. Seshia Anca Dragan
Abstract — Much of estimation of human internal state
(goal, intentions, activities, preferences, etc.) is passive : an
algorithm observes human a... |
5e8ae577-0473-4e9c-be7f-6ce93f25cebb | trentmkelly/LessWrong-43k | LessWrong | SBF's comments on ethics are no surprise to virtue ethicists
EDIT: Replaced the term "moral schizophrenia" with "internal moral disharmony" since the latter is more accurate and just.
DISCLAIMER: Although this is a criticism of the LW/EA community, I offer it in good faith. I don't mean to "take down" the community i... |
aa096007-1442-432e-bd99-140fcbc6f087 | trentmkelly/LessWrong-43k | LessWrong | [AN #114]: Theory-inspired safety solutions for powerful Bayesian RL agents
Alignment Newsletter is a weekly publication with recent content relevant to AI alignment around the world. Find all Alignment Newsletter resources here. In particular, you can look through this spreadsheet of all summaries that have ever been... |
2da75e62-a113-4ca9-98c8-2226893840d2 | trentmkelly/LessWrong-43k | LessWrong | Strengthening the Argument for Intrinsic AI Safety: The S-Curves Perspective
Epistemic Status: I'm attempting to present the strongest version (steelmaning) of a viewpoint that isn't my own, and then offer valid criticisms of it.
TL;DR: Existential risks (X-risks) arise from exponential thinking. However, if we ass... |
7c7f089d-3c13-4335-acd3-036e83536fc9 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | AI doom from an LLM-plateau-ist perspective
*(in the form of an FAQ)*
Q: What do you mean, “LLM plateau-ist”?
=======================================
**A:** As background, I think it’s obvious that there will *eventually* be **“transformative AI” (TAI)** that would radically change the world.[[1]](#fn76tsq17s1nk)
... |
424b384a-a36d-4822-8a1c-1e8b7a7e672d | trentmkelly/LessWrong-43k | LessWrong | How to write an academic paper, according to me
Disclaimer: this is entirely a personal viewpoint, formed by a few years of publication in a few academic fields. EDIT: Many of the comments are very worth reading as well.
Having recently finished a very rushed submission (turns out you can write a novel paper in a day... |
1562c386-d13e-4dae-8569-0cf2a5f08aa5 | trentmkelly/LessWrong-43k | LessWrong | A claim that Google's LaMDA is sentient
https://cajundiscordian.medium.com/is-lamda-sentient-an-interview-ea64d916d917 apparently posted by a Google engineer.
It could be an elaborate hoax, and has remnants of gwern's idea (https://www.gwern.net/fiction/Clippy) of a transformer waking up and having internal experienc... |
4114bd13-8b23-423d-bb0e-a4da712a40ac | trentmkelly/LessWrong-43k | LessWrong | AI #20: Code Interpreter and Claude 2.0 for Everyone
There is no shortage of people willing to talk about every week as a huge week in AI with tons of amazing new releases and announcements. At first this was usually right, then it mostly stopped being right. This week we had Code Interpreter and Claude 2.0 and x.AI (... |
a8624fed-1638-4706-b7a9-312b061daaa6 | StampyAI/alignment-research-dataset/eaforum | Effective Altruism Forum | Preliminary investigations on if STEM and EA communities could benefit from more overlap
**TLDR**
--------
This post discusses the potential benefits of increased overlap between the EA and STEM communities, particularly for addressing existential risks.
We (at Foresight Institute, a non-profit working to advance te... |
1f1252e2-b751-43d7-8c4e-109f0ec57040 | trentmkelly/LessWrong-43k | LessWrong | [SEQ RERUN] Is Fairness Arbitrary?
Today's post, Is Fairness Arbitrary? was originally published on 14 August 2008. A summary (taken from the LW wiki):
> When we say that a fair division of pie among N people is for each person to get 1/N of the pie, we aren't being arbitrary. We're being fair.
Discuss the post ... |
ea95b77b-fc5b-41aa-90be-b3b52ac8d7a4 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Corrigibility Via Thought-Process Deference
> *We would ideally want the agent to [behave] as if it were thinking, "I am incomplete and there is an outside force trying to complete me, my design may contain errors and there is an outside force that wants to correct them and this a good thing, my expected utility calcu... |
afbaf95e-f961-4465-b759-6a250803d22c | trentmkelly/LessWrong-43k | LessWrong | Rationality Quotes January 2013
Happy New Year! Here's the latest and greatest installment of rationality quotes. Remember:
* Please post all quotes separately, so that they can be voted up/down separately. (If they are strongly related, reply to your own comments. If strongly ordered, then go ahead and post them... |
4e8f9840-017c-4272-a1e8-75f8d80958bc | StampyAI/alignment-research-dataset/special_docs | Other | Interview with 7ujun
\*\*Interview with 7ujun, on 3/18/22\*\*
\*\*0:00:03.4 Vael:\*\* Alright. So jumping right in, my first question is, can you tell me about what area of AI you work on in a few sentences?
\*\*0:00:11.1 Interviewee:\*\* \[Interviewee describes working on natural language processing research\]
\*\... |
ccd02c55-0ce2-4c40-90a1-1e88e80075ab | trentmkelly/LessWrong-43k | LessWrong | How to solve the national debt deadlock
The US Congress is trying to resolve the national debt by getting hundreds of people to agree on a solution. This is silly. They should agree on the rules of a game to play that will result in a solution, and then play the game.
Here is an example game. Suppose there are N r... |
8545cc6b-edfb-4f58-bc15-7495249ce110 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | A Rephrasing Of and Footnote To An Embedded Agency Proposal
I wanted to clarify some comments made in [this post](https://www.alignmentforum.org/posts/TnkDtTAqCGetvLsgr/a-possible-resolution-to-spurious-counterfactuals), which proposed a way to resolve the issues brought up in the Action Counterfactuals subsection of ... |
9d782cd9-693f-4dc1-9797-59eadc70771c | trentmkelly/LessWrong-43k | LessWrong | My experience applying to MATS 6.0
The current cohort of the ML Alignment & Theory Scholars Program, MATS 6.0, had a unique application process and its broadest selection of mentors yet, with 40 mentors to apply to. I was invited to interview with twelve mentors and was accepted by five (which I later learned was an u... |
f2394388-f3bd-4b2b-8a01-e807648867af | trentmkelly/LessWrong-43k | LessWrong | Humans are not agents: short vs long term
A putative new idea for AI control; index here.
This is an example of humans not being (idealised) agents.
Imagine a human who has a preference to not live beyond a hundred years. However, they want to live to next year, and it's predictable that every year they are alive, t... |
e455dadc-44ba-433b-9fe8-04f9b13a5c6e | trentmkelly/LessWrong-43k | LessWrong | Call For Agreement: Should LessWrong have better protection against cultural collapse?
As you are probably already aware, many internet forums experience a phenomenon known as "eternal September". Named after a temporary effect where the influx of college freshmen would throw off a group's culture every September, et... |
f737999b-1643-4a02-a776-31bd6bb1212a | trentmkelly/LessWrong-43k | LessWrong | Ray Kurzweil on The Colbert Report [video embed]
The Colbert Report Mon - Thurs 11:30pm / 10:30c Ray Kurzweil www.colbertnation.com
Colbert Report Full Episodes Political Humor & Satire Blog Video Archive |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.