id stringlengths 36 36 | source stringclasses 15
values | formatted_source stringclasses 13
values | text stringlengths 2 7.55M |
|---|---|---|---|
2d89ebf3-9a5e-4d89-bb8e-10b64a6ab217 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | [Intro to brain-like-AGI safety] 5. The “long-term predictor”, and TD learning
5.1 Post summary / Table of contents
====================================
*Part of the* [*“Intro to brain-like-AGI safety” post series*](https://www.alignmentforum.org/s/HzcM2dkCq7fwXBej8)*.*
In [the previous post](https://www.alignmentfo... |
26fe558a-9cac-432c-93a8-3e856fc47b11 | trentmkelly/LessWrong-43k | LessWrong | Responsible Advanced Artificial Intelligence Act
Earlier today (4/9), the Center for AI Policy released and promoted a bill titled the "Responsible Advanced Artificial Intelligence Act." It calls for many limits on AI models.
The bill was featured on Politico Pro. Here's a paragraph:
> The draft bill is so far witho... |
ff4020be-57e2-4bbe-ac43-ada4cb1df25a | trentmkelly/LessWrong-43k | LessWrong | Memory Decoding Journal Club: Systems consolidation reorganizes hippocampal engram circuitry
Join Us for the Memory Decoding Journal Club!
A collaboration of the Carboncopies Foundation and BPF Aspirational Neuroscience
This time, we’re diving into a groundbreaking paper:
"Systems consolidation reorganizes hippocamp... |
3a48af47-f30f-4508-b62d-0eeb4512bcf1 | trentmkelly/LessWrong-43k | LessWrong | Group Rationality Diary, September 13-26
This is the public group rationality diary for September 13th - 26th, 2015. It's a place to record and chat about it if you have done, or are actively doing, things like:
* Established a useful new habit
* Obtained new evidence that made you change your mind about some beli... |
423deddf-015d-4454-8faf-bc790f722032 | trentmkelly/LessWrong-43k | LessWrong | A hundredth of a bit of extra entropy
There are two ways to calculate the amount of information in one term of a continued fraction:
* The entropy of the Gauss-Kuzmin distribution is about 3.4325 bits.
* Twice the logarithm of the Khinchin-Lévy constant is about 3.4237 bits.
These differ by about 0.0088 bits. It t... |
03ddb246-d49a-4da2-bd18-7bb15043a290 | trentmkelly/LessWrong-43k | LessWrong | Notes on Simplicity
This post examines the virtue of simplicity. It is meant mostly as an exploration of what others have learned about this virtue, rather than as me expressing my own opinions about it, though I’ve been selective about what I found interesting or credible, according to my own inclinations. I wrote th... |
a2d37451-ca59-4815-aad8-560030de4a8c | trentmkelly/LessWrong-43k | LessWrong | Less Lightning Links at Length #1
These are the things that I couldn’t fit into a Covid post even in the Not Covid section, but which aren’t big enough to deserve their own spin-off posts, and which are not time-sensitive.
With Covid winding down, I had the opportunity to work though some of that backlog this week. R... |
9a4fc3c4-79d5-4bbd-9d7f-db8304c35bcb | trentmkelly/LessWrong-43k | LessWrong | Hoodwinked: Evaluating Deception Capabilities in Large Language Models
This is a linkpost for https://arxiv.org/abs/2308.01404
Evaluating the hazardous capabilities of AI systems is useful for prioritizing technical research and informing policy decisions. To better evaluate the deception and lie detection abilities... |
f6813376-dd51-434f-9a7c-c192ab76d946 | StampyAI/alignment-research-dataset/eaforum | Effective Altruism Forum | Is AI forecasting a waste of effort on the margin?
*[Epistemic status: I am optimising for providing arguments worth considering, which means I'm not trying to make certain every argument is strictly valid. I'm not trying to be safe to defer to, so I'm not minimising false-positives. I am just trying to expand the ran... |
a05f7353-9331-42f8-abb1-d6c3df5c7355 | trentmkelly/LessWrong-43k | LessWrong | Christiano decision theory excerpt
A transcribed excerpt I found interesting from the decision theory outtake from 80,000 Hours' second interview of Paul Christiano (starting at 14:00):
----------------------------------------
Robert Wiblin: It seems like philosophers have not been terribly interested in these heter... |
a211de51-3761-4666-a40a-e1e08b7591e9 | StampyAI/alignment-research-dataset/lesswrong | LessWrong | “Cheating Death in Damascus” Solution to the Fermi Paradox
TL;DR: The Great Filter from the Fermi paradox could be escaped by choosing a random strategy. However, if all civilizations acted randomly, this could be the actual cause of the Fermi paradox. Using a meta-random strategy solves this.
----
"Death in Damas... |
e280dcd4-eb04-48b8-8cab-67b39dcdc1a5 | trentmkelly/LessWrong-43k | LessWrong | I was asked about how to resolve strong emotional fear concerns around the basilisk and this is my instructions as to how to do that.
So you just read something about the basilisk and it freaked you out. There’s some emotional argument around being simulated, tortured and extorted by a future AI. You read it, feel q... |
0528feb6-e580-411f-9065-828aa9a261da | trentmkelly/LessWrong-43k | LessWrong | Altman firing retaliation incoming?
"Anonymous sources" are going to journalists and insisting that OpenAI employees are planning a "counter-coup" to reinstate Altman, some even claiming plans to overthrow the board.
It seems like a strategy by investors or even large tech companies to create a self-fulfilling prophe... |
c161d811-24a0-4f3e-aaa5-518894054a5e | trentmkelly/LessWrong-43k | LessWrong | Owen Cotton-Barratt on GPP
I interviewed Owen Cotton-Barratt about the Global Priorities Project, as part of a ‘shallow investigation’ (inspired by Givewell) into cause prioritization which will be released soon. Notes from the conversation are cross-posted here from the 80000hours blog, and also available in other fo... |
c2f06527-b1f4-497b-9f5c-45ee206aaa8c | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | [AN #95]: A framework for thinking about how to make AI go well
Alignment Newsletter is a weekly publication with recent content relevant to AI alignment around the world. Find all Alignment Newsletter **[resources here](http://rohinshah.com/alignment-newsletter/)**. In particular, you can look through **[this spreads... |
6c43b37c-a4df-4ef0-8d93-631af1681b05 | trentmkelly/LessWrong-43k | LessWrong | Meetup : West LA
Discussion article for the meetup : West LA
WHEN: 31 July 2014 04:05:00AM (-0700)
WHERE: 11066 Santa Monica Blvd, Los Angeles, CA
VOID
Discussion article for the meetup : West LA |
e10e0759-42d5-4db7-94f6-49ccd6b5da3b | trentmkelly/LessWrong-43k | LessWrong | Put Dirty Dishes in the Dishwasher
> "It's not worth using the dishwasher: by the time you've gotten your dishes clean enough to put in, you might as well just finish washing them yourself."
People have a wide range of ways they deal with dishes before putting them in the dishwasher:
1. Rinse well. The dishes sho... |
d2e9bc44-a30f-4616-98b2-c079b092ef8e | trentmkelly/LessWrong-43k | LessWrong | The Takeoff Speeds Model Predicts We May Be Entering Crunch Time
Thanks to Ashwin Acharya, David Schneider-Joseph, and Tom Davidson for extensive discussion and suggestions. Thanks to Aidan O’Gara, Alex Lintz, Ben Cottier, James Sanders, Jamie Bernardi, Rory Erlich, and Ryan Greenblatt for feedback.
Part 1: Executiv... |
7296a45c-1207-4a7e-97d7-e3f6251728d3 | trentmkelly/LessWrong-43k | LessWrong | [Feedback please] New User's Guide to LessWrong
The LessWrong team is currently thinking a lot about what happens with new users: both the bar of their contributions being accepted, how we deliver feedback and restriction of not good contributions, but also most importantly, how we get them onboarded onto the site
Th... |
50b4f1ab-feaf-4e9b-88cc-d0aa033f8aa6 | trentmkelly/LessWrong-43k | LessWrong | When Should the Fire Alarm Go Off: A model for optimal thresholds
In this post I consider a model where there is a possible bad outcome, and the chance of it happening is related to something we can measure. We need to be able to set a threshold for these measurements, such that if a measurement exceeds a threshold th... |
a2140dea-8a13-4d4c-997a-450077ef79ef | trentmkelly/LessWrong-43k | LessWrong | Ed Boyden on the State of Science
I just listened to Tyler Cowen's interview with Ed Boyden (link and transcript). The second half contained a lot of questions about current scientific infrastructure, and Boyden had a lot of interesting comments, so I've reproduced a few particular quotes here and added headings.
(Th... |
59d76131-94a6-4b79-bffd-2f9171a202fd | trentmkelly/LessWrong-43k | LessWrong | Harry Potter and the Methods of Rationality discussion thread, January 2015, chapter 103
New chapter, and the end is now in sight!
This is a new thread to discuss Eliezer Yudkowsky’s Harry Potter and the Methods of Rationality and anything related to it. This thread is intended for discussing chapter 103.
There is a... |
a75a7dac-1a9d-4aee-9f1c-91bb77d4b780 | trentmkelly/LessWrong-43k | LessWrong | Is there an analysis of the common consideration that splitting an AI lab into two (e.g. the founding of Anthropic) speeds up the development of TAI and therefore increases AI x-risk?
I'm asking because I can think of arguments going both ways.
Note: this post is focused on the generic question "what to expect from a... |
2d690713-3e50-40ce-a5dc-72223e6149f7 | trentmkelly/LessWrong-43k | LessWrong | What are the thoughts of Less Wrong on property dualism?
I did a search about property dualism and I couldn't see much written here on this site.
1. What is your opinion on this topic?
2. Are there any articles on this site that I should read, including articles that aren't directly about property dualism? I'll add... |
bacbc50e-64f6-42bf-af28-e8d6a6078f8f | LDJnr/LessWrong-Amplify-Instruct | LessWrong | "Here’s a pattern I’d like to be able to talk about. It might be known under a certain name somewhere, but if it is, I don’t know it. I call it a Spaghetti Tower. It shows up in large complex systems that are built haphazardly.Someone or something builds the first Part A.Later, someone wants to put a second Part B on t... |
64128858-53aa-4c01-a24d-ce48466c4022 | trentmkelly/LessWrong-43k | LessWrong | Weekly LW Meetups
This summary was posted to LW Main on September 9th. The following week's summary is here.
Irregularly scheduled Less Wrong meetups are taking place in:
* Munich Meetup in September: 17 September 2016 04:00PM
The remaining meetups take place in cities with regular scheduling, but involve a change... |
4b576995-9d12-48f1-b7d0-fb440c6d934d | StampyAI/alignment-research-dataset/lesswrong | LessWrong | A concerning observation from media coverage of AI industry dynamics
tl:dr: there are indications that ML engineers will migrate to environments with less AI governance in place, which has implications for the tech industry and global AI governance efforts.
=========================
I just wanted to raise something ... |
60802bbd-dc30-46d2-a103-26dc09a84403 | trentmkelly/LessWrong-43k | LessWrong | Don't align agents to evaluations of plans
Another stab at explaining Don't design agents which exploit adversarial inputs. This is not the follow-up post mentioned therein. That post will come next.
More precise title: "Don't try directing a superintelligence to maximize your valuations of their plans using a cons... |
0353f9fe-0945-4d1d-9efb-fc20b2fa90ed | trentmkelly/LessWrong-43k | LessWrong | Why is Kleros valued so low?
Kleros provides a system that allows smart contract in natural language to be written by letting randomly drafted juries evaluate the contract. It seems to me that this principle is incredibly powerful for all sorts of real world applications whether it's arbitrating the outcome of a predi... |
20c95eb2-4630-4aee-a360-87853aedb246 | StampyAI/alignment-research-dataset/eaforum | Effective Altruism Forum | Humanity’s vast future and its implications for cause prioritization
Summary
-------
* Humanity could last millions of years and spread beyond Earth, resulting in an unimaginably large number of people being born. Even under conservative assumptions, ***at least 100 trillion people*** could exist in the future.
* The... |
25237302-a07f-4644-8782-7bb68d7114f9 | trentmkelly/LessWrong-43k | LessWrong | Confound it! Correlation is (usually) not causation! But why not?
> It is widely understood that statistical correlation between two variables ≠ causation. But despite this admonition, people are routinely overconfident in claiming correlations to support particular causal interpretations and are surprised by the resu... |
11d9055f-bbea-4863-abfa-027b919420c3 | trentmkelly/LessWrong-43k | LessWrong | Evidence on why abstract research is or isn’t respected
I previously suggested an explanation for very abstract research sometimes not being well respected: very abstract thought often looks superficially similar to very basic confusion, which looks amusingly silly. For instance, thinking about paraconsistent logic lo... |
64d7b84e-9753-4c8f-8442-86842a9ec250 | StampyAI/alignment-research-dataset/special_docs | Other | Ethical Reflections on Artificial Intelligence
6(2)/2018 ISSN 2300-7648 ( print ) / ISSN 2353-5636 ( online ) 9 6(2)/2018, 9–31Received: June 1, 2018. Accepted: August 24, 2018 DOI: http://dx.doi.org/10.12775/SetF.2018.015
Ethical Reflections
on Artificial Intelligence\*
BRIAN PATRICK GREEN
Director of Technology Eth... |
d20d933d-2c32-49bd-9dab-6b170fba260d | StampyAI/alignment-research-dataset/youtube | Youtube Transcripts | GPT3: An Even Bigger Language Model - Computerphile
rob welcome back to computer file in
these strange times that we find
ourselves recording in you've got the
green screen up there we're having a few
laggy problems with the communications
what are you going to talk about today
then
uh yeah i thought uh today it would... |
b3675cb8-3190-4c0b-9dc4-a28d155ac954 | trentmkelly/LessWrong-43k | LessWrong | On corrigibility and its basin
Post somewhat inspired by Eliezers "well why didn't anyone else write it then?"
Corrigibility has various slightly different definitions, but the general rough idea is of an AI that does what we want. An AI that doesn't trick us. An AI that doesn't try to break out of its box. An AI th... |
0c34f89b-5829-4666-816d-e9ad4683ad9e | trentmkelly/LessWrong-43k | LessWrong | A taxonomy of coordination problems
The LessWrong coordination/cooperation tag describes coordination as:
> the challenge of distinct actors being able to jointly choose their actions to achieve a favorable outcome. [...] A closely related concept is that of cooperation – multiple actors choosing their actions in way... |
e564b644-9e44-46ee-88f2-0fe0f743991d | trentmkelly/LessWrong-43k | LessWrong | Announcing My Free Online Course "Original Seeing With a Focus On Life"
In May and June of 2021, I ran an online nature study course with about two dozen participants. I had three goals:
1. Help people create or deepen a personal connection with nature.
2. Learn how to run month-long online courses with over a doze... |
c6a4442b-08b2-4f8a-b288-c56b54a611bd | trentmkelly/LessWrong-43k | LessWrong | Romance is magical
People seem to generally believe they have high romantic standards, and that they aren’t strongly influenced by things like looks, status and money. Research says our standards aren’t that high, that they drop if the standard available drops for a single evening, and that superficial factors make mo... |
ef187ab1-b0c3-4c92-af29-71fc0edbdf88 | trentmkelly/LessWrong-43k | LessWrong | Consent Isn't Always Enough
After the Time article on sexual harassment and abuse within Effective Altruism and Owen Cotton-Barratt's resignation there's been a lot of discussion around the extent to which EA culture did or didn't contribute. This has included proposals that EA try to be more conventional, such as dis... |
6e194fb9-5962-4dd4-9457-e6249b931a7c | StampyAI/alignment-research-dataset/special_docs | Other | Demonstrating the Impact of Prior Knowledge in Risky Choice.
Demonstrating the Impact of Prior Knowledge in Risky Choice
Mathew D. Hardy (mdhardy@princeton.edu)
Department of Psychology, Princeton University
Thomas L. Griffiths (tomg@princeton.edu)
Departments of Psychology and Computer Science, Princeton University
Ab... |
0996420a-df6c-429c-a4c6-e81808869ece | trentmkelly/LessWrong-43k | LessWrong | The wandering rationalist: an update
As per my original post here, and the more recent one by shokwave here, here's my attempt to nail down some dates for when I'll be in various places (both in the States and elsewhere). All dates are provisional, especially between Colorado and Washington, and my West Coast tour in ... |
98ff3931-8511-4fea-b077-1ab034ecd283 | trentmkelly/LessWrong-43k | LessWrong | Infrafunctions Proofs
Proposition 1:
Given some compact metric space of options X, if U:X→R is a bounded function, {μ|∀ν∈ΔX:μ(U)≥ν(U)}=Δ{x|∀y∈X:U(x)≥U(y)}
PROOF: To do this, we must show that anything in the first set is in the second, and vice-versa. So, assume μ is in the first set. Then for any ν, ν(U)≤μ(U). Howe... |
f01fcf05-157f-432f-9052-f143c1db0447 | trentmkelly/LessWrong-43k | LessWrong | On Developing a Mathematical Theory of Interpretability
If the trajectory of the deep learning paradigm continues, it seems plausible to me that in order for applications of low-level interpretability to AI not-kill-everyone-ism to be truly reliable, we will need a much better-developed and more general theoretical a... |
ab1f5cb6-12b0-4ecf-b4c7-851942c0c6d8 | trentmkelly/LessWrong-43k | LessWrong | [Linkpost] Towards a Theoretical Understanding of the 'Reversal Curse' via Training Dynamics
The excerpts below seem to me like a slight update towards arguments about the weakness of one-forward-passes in transformers and for agendas like externalized reasoning and translucent thoughts. They also suggest out-of-conte... |
e23963bc-490f-490a-839b-861c9ffb672a | trentmkelly/LessWrong-43k | LessWrong | Yeshua's Basilisk
Suppose you’re an AI researcher trying to make AIs which are conscious and reliably moral, so they’re trustworthy and safe for release into the real world, in whatever capacity you intend.
You can’t, or don’t want to manually create them; it’s more economical, and the only way to ensure they’re cons... |
713d9e8d-8b8c-4c3b-9b6d-931e8b3573c4 | trentmkelly/LessWrong-43k | LessWrong | A short critique of Omohundro's "Basic AI Drives"
Summary
Omohundro, in his paper, assumes goal-oriented AIs with well-defined utility functions, but contemporary models like GPT-3.5 and Claude maximally exhibit behaviours without explicit goals. Advanced models such as GPT-o1 and Claude-3.5, however, have shown preli... |
eb7e7bc2-8b17-4bf9-a1d3-8827a505fbaa | trentmkelly/LessWrong-43k | LessWrong | Emergent Misalignment: Narrow finetuning can produce broadly misaligned LLMs
This is the abstract and introduction of our new paper. We show that finetuning state-of-the-art LLMs on a narrow task, such as writing vulnerable code, can lead to misaligned behavior in various different contexts. We don't fully understand ... |
056bd1ff-44a9-44ae-8d75-923c938e3de6 | trentmkelly/LessWrong-43k | LessWrong | What TMS is like
There are two nuclear options for treating depression: Ketamine and TMS; This post is about the latter.
TMS stands for Transcranial Magnetic Stimulation. Basically, it fixes depression via magnets, which is about the second or third most magical things that magnets can do.
I don’t know a whole lot a... |
c7737f14-2c67-40c1-b5bb-ff0dcb1d470c | StampyAI/alignment-research-dataset/lesswrong | LessWrong | GPT-4: What we (I) know about it
OpenAI released a [press release](https://openai.com/research/gpt-4), [research statement](https://cdn.openai.com/papers/gpt-4.pdf), and [system card](https://cdn.openai.com/papers/gpt-4-system-card.pdf) about GPT-4 approximately one eternity (24 hours) ago. The general public can’t us... |
675572ff-18e9-4fe2-a3be-d929c6e581d8 | trentmkelly/LessWrong-43k | LessWrong | Book summary: Selfish Reasons to Have More Kids
Bryan Caplan's Selfish reasons to have more kids: why being a great parent is less work and more fun than you think was recommended to me as a strongly-science-supported take on childrearing. I found it valuable; here are my take-aways.
TLDR: Variation in parenting (wit... |
25d0b1ac-d0c2-4f0e-908d-4bc8396ae3ce | StampyAI/alignment-research-dataset/lesswrong | LessWrong | Brief summary of ai-plans.com
At AI-Plans.com (in beta), we are working on a contributable compendium of alignment plans and the criticisms against them. Currently, newcomers to the field of AI Alignment often struggle to understand what work is being done, who is doing it, and the assumptions, strengths, and weakness... |
1f7075bf-43d1-494e-85e5-24b960200c44 | trentmkelly/LessWrong-43k | LessWrong | Hammertime Postmortem
Intro.
A bit less than two months ago, I set out to write about instrumental rationality every day for thirty days. In this post, I will quickly evaluate how well I felt I did along each of my four stated objectives. I will simultaneously evaluate all the Hammertime techniques and ideas by their... |
20256942-0084-44fc-8dbc-38e816f5ca69 | trentmkelly/LessWrong-43k | LessWrong | ...Recursion, Magic
Followup to: Cascades, Cycles, Insight...
...4, 5 sources of discontinuity.
Recursion is probably the most difficult part of this topic. We have historical records aplenty of cascades, even if untangling the causality is difficult. Cycles of reinvestment are the heartbeat of the modern economy... |
2d9ac4ed-09db-4665-8c38-b82454a27097 | StampyAI/alignment-research-dataset/blogs | Blogs | AISC 1: Research Summaries
The 2018 Gran Canaria AI safety camp teams have worked hard in the preparation of the camp and in the 10 day sprint. Each team has written a brief summary of the work they did during the camp:
#### Irrationality
Team: *Christopher Galias, Johannes Heidecke, Dmitrii Krasheninnikov, Jan Ku... |
2827135b-f4bd-41b4-8987-f6437d59f600 | trentmkelly/LessWrong-43k | LessWrong | Looking for a likely cause of a mental phenomenon
I'm currently testing a promising direction for a possible collection of units at CFAR. (For those who have attended some CFAR events or test sessions, this is a collection of refinements to the fudoshin/"panic" unit.) I've hit on what I think is a key puzzle whose ans... |
741ceecd-3b4c-447f-b0e9-bae0b89ffe6a | StampyAI/alignment-research-dataset/aisafety.info | AI Safety Info | What is "biological cognitive enhancement"?
“Biological cognitive enhancement” refers to technologies to change human biology to increase people’s cognitive capabilities. There may be genes or molecules that can be modified to improve general intelligence. Researchers [have already done this in mice](https://pubmed.nc... |
99da6553-cc98-4aa4-9444-a440d1f810a1 | trentmkelly/LessWrong-43k | LessWrong | Controlling Intelligent Agents The Only Way We Know How: Ideal Bureaucratic Structure (IBS)
Here's my claim:
"Ideal Bureaucratic Structure" (or IBS for short as I imagine this claim is likely to meet serious groans, maybe even deep from within your bowels) has an unparalleled ability to control and direct individual... |
a520739a-042d-42a7-bcb3-cd33a1339a24 | trentmkelly/LessWrong-43k | LessWrong | Demon Threads
tldr: a Demon Thread is a discussion where everything is subtly warping towards aggression and confusion (i.e. as if people are under demonic influence), even if people are well intentioned and on the same 'side.' You can see a demon thread coming in advance, but it's still hard to do anything about.
("... |
a475da66-9a95-414e-927d-abc3164ea574 | trentmkelly/LessWrong-43k | LessWrong | No Clickbait - Misalignment Database
This is a database of cases of Misalignment - classified by Type of Misalignment, Type of AI, etc.
Link to add more:
https://docs.google.com/forms/d/e/1FAIpQLSfE7ZeSV6W_YmKYrgy7BaiFKj90dBJ2qDUaYXzbpi_ILEs9sQ/viewform?usp=sf_link
Link to the DB: https://docs.google.com/spreadshee... |
e41e1244-d47d-488f-9968-8ca544c9031d | trentmkelly/LessWrong-43k | LessWrong | Checking for Threshold Number of Rationalists in São Paulo
Hello, this post is only to check if there are enough people in São Paulo to make it worth it to arrange a meetup for São Paulo Less Wrongers.
Please reply saying "Paulista" if you are living in São Paulo area and would like to join a meeting (not necessarily... |
5fe2bc79-4014-4e67-82d2-edc9895721da | trentmkelly/LessWrong-43k | LessWrong | Somewhat against "just update all the way"
Sometimes, a person's probability in a proposition over a long timespan follows a trajectory like this:
I.e. the person gradually finds the proposition more plausible over time, never really finding evidence against it, but at the same time they update very gradually, rather... |
2d5981b8-e423-45a6-bede-6a3291b1b86d | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Selection Theorems: A Program For Understanding Agents
What’s the type signature of an agent?
For instance, what kind-of-thing is a “goal”? What data structures can represent “goals”? Utility functions are a common choice among theorists, but they [don’t seem quite right](https://www.lesswrong.com/posts/RQpNHSiWaXTvD... |
45ec238f-ed72-4afe-978d-6c296e35b020 | StampyAI/alignment-research-dataset/distill | Distill Scientific Journal | Differentiable Image Parameterizations
Neural networks trained to classify images have a remarkable — and surprising! — capacity to generate images.
Techniques such as DeepDream , style transfer, and feature visualization leverage this capacity as a powerful tool for exploring the inner workings of neural networks, a... |
7db34a6d-3aef-4a41-9fb7-4e39aae6d589 | trentmkelly/LessWrong-43k | LessWrong | A summary of the Hanson-Yudkowsky FOOM debate
In late spring this year, Luke tasked me with writing a summary and analysis of the Hanson-Yudkowsky FOOM debate, with the intention of having it eventually published in somewhere. Due to other priorities, this project was put on hold for the time being. Because it doesn't... |
ccd25401-5a14-44c0-9b06-91e2f8671e24 | StampyAI/alignment-research-dataset/lesswrong | LessWrong | Wentworth and Larsen on buying time
*This post summarizes a discussion between Thomas Larsen and John Wentworth. They discuss efforts to buy time, outreach to AGI labs, models of institutional change, and other ideas relating to a post that Thomas, Olivia Jimenez, and I released:*[*Instead of technical research, more ... |
0c4cfd76-1d18-424d-817c-6e8ffe7a283c | trentmkelly/LessWrong-43k | LessWrong | GPT-4 solves Gary Marcus-induced flubs
TLDR: GPT-4 succeeds at 15 problems from Gary Marcus that exposed failures of GPT-3.
I enjoyed reading the ACX post "My Bet: AI Size Solves Flubs" last year. Here are some excerpts:
> Here’s the basic structure of an AI hype cycle:
>
> 1. Someone releases a new AI and demonst... |
d7c80dd4-429a-4fb7-8bbe-8780ae7b6b1d | trentmkelly/LessWrong-43k | LessWrong | Competitive Truth-Seeking
In many domains, you can get better primarily by being correct more frequently. If you’re managing a team, or trying to improve your personal relationships, it’s very effective to improve your median decision. The more often you’re right, the better you’ll do, so people often implicitly optim... |
575a5f4d-5bdd-4932-9f80-9ae6fb6e07cb | trentmkelly/LessWrong-43k | LessWrong | The absurdity of un-referenceable entities
> Whereof one cannot speak, thereof one must be silent.
>
> Ludwig Wittgenstein, Tractatus Logico-Philosophicus*
Some criticism of my post on physicalism is that it discusses reference, not the world. To quote one comment: "I consider references to be about agents, not abou... |
2ce5a9ee-24c3-4ff0-969e-c20f11f7c882 | StampyAI/alignment-research-dataset/special_docs | Other | A Psychoanalytic Approach to the Singularity: Why We Cannot Do Without Auxiliary Constructions
Chapter 12
A Psychoanalytic Approach
to the Singularity: Why We Cannot DoWithout Auxiliary Constructions
Graham Clarke
All of us should ask ourselves what we can do now to improve
the chances of reaping the bene fits and avoi... |
67c89a24-994f-432a-9f5a-5a55e44207cf | trentmkelly/LessWrong-43k | LessWrong | Making Reasoning Obviously Locally Correct
> x = y
>
> x2 = x*y
>
> x2 - y2 = x*y - y2
>
> (x+y)(x-y) = y(x-y)
>
> x+y = y
>
> y+y = y
>
> 2*y = y
>
> 2 = 1
The above is an incorrect "proof" that 2=1. Even for those who know where the flaw is... |
3ac7c6e3-c902-427f-a50a-c72e71502c95 | trentmkelly/LessWrong-43k | LessWrong | Null-boxing Newcomb’s Problem
The leering statue of the unnamed trickster god stared down at him from its sacred alcove, motionless as always. Maxwell sighed. He had been praying at the base of the statue for the past hour, and like every time previous, he hadn’t felt anything. The priests had promised salvation from ... |
203f2a88-79b5-4d6a-9dd0-03adcbb1268b | LDJnr/LessWrong-Amplify-Instruct | LessWrong | "Epistemic Status: Confident about the problem, solution and exact mechanism is more of an untested hypothesis.There's a saying for communities: if you're not gaining members, you're losing members. Sometimes you hit just the right size and you'd prefer things stay exactly as they are. But in practice, some people will... |
086f7395-1740-4410-bd2f-9e698fec6d21 | trentmkelly/LessWrong-43k | LessWrong | A Taxonomy of Bias: Mindware Problems
This is the third part in a mini-sequence presenting content from Keith E. Stanovich's excellent book What Intelligence Tests Miss: The psychology of rational thought. It will culminate in a review of the book itself.
Noting that there are many different kinds of bias, Keith Stan... |
1485b95e-3df4-4277-acfa-6273549f7a3f | trentmkelly/LessWrong-43k | LessWrong | Chess and cheap ways to check day to day variance in cognition
Lately I've been playing chess. I'm not very good at it. I started because our four-year-old got into it, and I preferred not to lose to her before she had earned it. I play online on chess.com, which has instantaneous matchups with strangers at approximat... |
7105fba2-fdb5-4629-86a1-3f64c30e5208 | StampyAI/alignment-research-dataset/lesswrong | LessWrong | Validator models: A simple approach to detecting goodharting
*Crossposted from my personal* [*blog*](https://www.beren.io/2023-02-19-validator-models-simple-approach-detecting-goodhearting/)*.*
A naive approach to aligning an AGI, and an abstract version of what is currently used in SOTA approaches such as RLHF, is t... |
512ca39e-3e59-4d27-9c30-ff570529c508 | trentmkelly/LessWrong-43k | LessWrong | Real-time hiring with prediction markets
Epistemic status: speculation
I have a couple of assumptions about hiring:
* It is difficult and often unpleasant.
* No one is very good at it.
* It has a lot of hidden costs.
I have a further assumption about the need for labor:
* The people doing the work now have the... |
2d20254f-c48a-46e8-92d3-05e75ad3f7cb | StampyAI/alignment-research-dataset/lesswrong | LessWrong | Superintelligence 28: Collaboration
*This is part of a weekly reading group on [Nick Bostrom](http://www.nickbostrom.com/)'s book, [Superintelligence](http://www.amazon.com/Superintelligence-Dangers-Strategies-Nick-Bostrom/dp/0199678111). For more information about the group, and an index of posts so far see the [anno... |
8e4d7a93-49f8-4133-a48d-fcd443bf724e | trentmkelly/LessWrong-43k | LessWrong | Movie Review: Megan
I will follow my usual procedure here – a fully spoiler-free 1-bit review, then a mostly spoiler-free short review to determine if you should see it, then a fully detailed spoiler-filled analysis.
Fully Spoiler-Free 1-Bit Review – Should You See Megan In Theaters, Yes or No?
Yes.
Mostly Spoiler... |
ce19e68e-73a9-4279-b23b-49f9c86bd62f | trentmkelly/LessWrong-43k | LessWrong | [New] Rejected Content Section
tl:dr;
* The LessWrong reviews all first time submissions from new users
* In the last month, we've built a new process where we reject some of those submissions (with explanation) – you can now see what was rejected from the main site on the new Rejected Content section
* New users s... |
f6f943b7-9b7b-4af2-ba1c-78a4e36c7999 | trentmkelly/LessWrong-43k | LessWrong | Jobs: Help scale up LM alignment research at NYU
NYU is hiring alignment-interested researchers!
* I'll be working on setting up a CHAI-inspired research center at NYU that'll focus on empirical alignment work, primarily on large language models. I'm looking for researchers to join and help set it up.
* The alignme... |
1f5c8ad4-869c-44db-83b1-34e6c0c87f3d | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | We have achieved Noob Gains in AI
TL;DR I explain why I think AI research has been slowing down, not speeding up, in the past few years.
How have your expectations for the future of AI research changed in the past three years?
Based on recent posts in this forum, it seems that results in text generation, protein fol... |
80c08d34-034d-4be9-9d54-b0c38779fd6a | trentmkelly/LessWrong-43k | LessWrong | Mental minimalism
Hypothesis:
It is nice to work in other people’s spaces relative to mine, because I am not supposed to interact with their stuff. Having fewer affordances reduces mental noise, in the same way that minimalism does. You can have something like the mental effects of minimalism even in a pigsty, just ... |
b54aa7f5-4d9d-43ae-9b6c-361cd2fbbf9a | trentmkelly/LessWrong-43k | LessWrong | Link: My something-like-Friendliness-research blog
If you want to know what that something-like-Friendliness is, you can read the blog! You can find it here: http://willnewsome.wordpress.com/ (About page here: http://willnewsome.wordpress.com/about/ )
Anyone who's bothered to notice the trend of my posts and comments... |
ecf2db03-9d0d-4bd8-9be3-64fb073532a2 | trentmkelly/LessWrong-43k | LessWrong | Transfer Learning in Humans
I examine the literature on transfer learning in humans. Far transfer is difficult to achieve, best candidate interventions are to practice at the edge of one's ability and make many mistakes, evaluate mistakes after one has made them, learn from training programs modeled after expert tacit... |
bcaf1a44-1552-4381-87fe-10232d1de018 | trentmkelly/LessWrong-43k | LessWrong | Dwarkesh Patel on Continual Learning
A key question going forward is the extent to which making further AI progress will depend upon some form of continual learning. Dwarkesh Patel offers us an extended essay considering these questions and reasons to be skeptical of the pace of progress for a while. I am less skeptic... |
8e59076e-a4aa-46af-9ba0-218207068d6a | trentmkelly/LessWrong-43k | LessWrong | Meetup : San Antonio Meetup!
Discussion article for the meetup : San Antonio Meetup!
WHEN: 29 November 2015 02:00:00PM (-0600)
WHERE: 12651 Vance Jackson Rd #118, San Antonio, TX 78230
Bubble tea, frozen yogurt, and discussion at Yumi Berry! All are welcome! New Meetup to discuss rationality and all things LessWron... |
f1fe38c6-35a3-4f25-9884-5c8a5f968f34 | trentmkelly/LessWrong-43k | LessWrong | Argument Maps Improve Critical Thinking
Charles R. Twardy provides evidence that a course in argrument mapping, using a particular software tool improves critical thinking. The improvement in critical thinking is measured by performance on a specific multiple choice test (California Critical Thinking Skills Test). Thi... |
8e3492b1-a636-411b-b0bc-f0a11457ba0f | trentmkelly/LessWrong-43k | LessWrong | Metaculus Launches Conditional Cup to Explore Linked Forecasts
Game out scenarios and bring greater clarity to the relationships between important events in Metaculus's new Conditional Cup. Every week, flex your conditional forecasting skills on a new conditional pair.
The tournament has launched with two conditional... |
b28e09fd-b874-40e7-b137-b162f4071b8e | StampyAI/alignment-research-dataset/youtube | Youtube Transcripts | AI Safety Reading Group (Session 40)
hello and welcome to the 40s session of
the AI safety reading group and today we
have read an article called racing
through precipice a model of artificial
intelligence development by Joe
Armstrong Nick Bostrom and cars children
they are all working at the future of
humanity instit... |
347ef5be-afeb-4774-9bf1-edcb7673380b | trentmkelly/LessWrong-43k | LessWrong | Rational Project Management
There's a lot of discussion on the site about akrasia (failure of willpower), but most of it focuses on the individual. Obviously, though, organizations also can be said to suffer from akrasia. Even when there is clear agreement among the leaders or members of a business, charity, political... |
aabeaa68-b96c-4918-95ac-aa8328832324 | trentmkelly/LessWrong-43k | LessWrong | OpenPhil's "Update on Cause Prioritization / Worldview Diversification"
OpenPhil recently posted an update on their overall cause prioritization strategy. In particular, exploring the concept of worldview diversification:
> When choosing how to allocate capital, we need to decide between multiple worldviews. We use “... |
ac06368a-48ab-46e7-968d-b540f134e886 | StampyAI/alignment-research-dataset/lesswrong | LessWrong | A library for safety research in conditioning on RLHF tasks
There is a couple of discussions regarding conditioning / decision transformer training. "Conditioning" as in, placing your reward model's reward in the prompt. We then train our language model to create a completion that follows the specified reward.
See [S... |
c3dd9d6b-94df-42c4-b5b0-9d4e8829ff55 | trentmkelly/LessWrong-43k | LessWrong | Boltzmann Brains and Within-model vs. Between-models Probability
Why Boltzmann brains?
Back in the days of Ludwig Boltzmann (before the ~1912 discovery of galactic redshift), it seemed like the universe could be arbitrarily old. Since the second law of thermodynamics says that everything tends to a state of high entr... |
82dd785c-0f22-455d-be88-d8b670ff305c | trentmkelly/LessWrong-43k | LessWrong | Curiosity Killed the Cat and the Asymptotically Optimal Agent
Here's a new(ish) paper that I worked on with Marcus Hutter.
If an agent is exploring enough to be guaranteed strong performance in the limit, that much exploration is enough to kill it (if the environment is minimally difficult and dangerous). It's nothin... |
88c5945b-7acc-4484-84e0-c5e7c6374ff1 | trentmkelly/LessWrong-43k | LessWrong | BASALT: A Benchmark for Learning from Human Feedback
Copying the abstract of the paper:
> The last decade has seen a significant increase of interest in deep learning research, with many public successes that have demonstrated its potential. As such, these systems are now being incorporated into commercial products. ... |
83d54e39-a1a8-45c0-b13f-0e1bf92eb9a5 | trentmkelly/LessWrong-43k | LessWrong | Is math subjective?
I was arguing with a friend over whether reality is objective or subjective. He says subjective, because of the double-slit experiment, and that reality changes depending on whether we observe it or not, so it only exists in our minds.
I know this is right.. But something still seems off to me.
I... |
fe7a8505-36d9-4e45-a7c3-740cc04bf644 | StampyAI/alignment-research-dataset/blogs | Blogs | You Can’t Predict a Game of Pinball
*Jeffrey Heninger, 29 March 2023*
Introduction
------------
When thinking about a new idea, it helps to have a particular example to use to gain intuition and to clarify your thoughts. Games are particularly helpful for this, because they have well defined rules and goals. Many ... |
19a719bc-5b62-456e-95ab-8a658de8dac9 | trentmkelly/LessWrong-43k | LessWrong | Don't Use Your Favorite Weapon
Why your favorite arguments aren't most likely to land
If forced to condense my political ideology onto the usual spectrum, I’ve always been left of center. I think that the government can and should work to improve wellbeing, and this generally entails change. It’s called “progressivi... |
70136e12-a055-4a7d-98e7-a5a670c7096d | trentmkelly/LessWrong-43k | LessWrong | An academic journal is just a Twitter feed
This was inspired by a tweet exchange with Michel Nivard:
There’s a lot of thinking going on about how to rework academic journals and the peer review process. It’s easy to succumb to featuritis. Can we automate publication using a pre-analysis plan? We need to be on the b... |
6966ef16-cd85-4aa9-a2a6-1bcf710bd7df | trentmkelly/LessWrong-43k | LessWrong | Discuss: Original Seeing Practices
What can we practice to help us think original thoughts? How can we see beyond our cached responses?
This is a place to share life patterns, techniques, or exercises that help you (either occasionally or regularly) think new thoughts. |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.