id stringlengths 36 36 | source stringclasses 15
values | formatted_source stringclasses 13
values | text stringlengths 2 7.55M |
|---|---|---|---|
e164e31a-1f18-4877-8073-ddd9dc25bc7d | trentmkelly/LessWrong-43k | LessWrong | Directions and desiderata for AI alignment
Note: This is the first post from part four: what needs doing of the sequence on iterated amplification. The fourth part of the sequence describes some of the black boxes in iterated amplification and discusses what we would need to do to fill in those boxes. I think these ar... |
f7538833-4620-414a-99c9-2ff85c0ace87 | StampyAI/alignment-research-dataset/blogs | Blogs | Global computing capacity
*[This page is out of date and its contents may have been inaccurate in 2015, in light of new information that we are yet to integrate. See [Computing capacity of all GPUs and TPUs](https://wiki.aiimpacts.org/doku.php?id=ai_timelines:hardware_and_ai_timelines:computing_capacity_of_all_gpus_an... |
4c6c39dc-6f05-4b66-818f-873addae4596 | trentmkelly/LessWrong-43k | LessWrong | Fundamentals of Formalisation Level 4: Formal Semantics Basics
Followup to Fundamentals of Formalisation Level 3: Set Theoretic Relations and Enumerability. First Post.
The big ideas:
* Formal Semantics
* Model
To move to the next level you need to be able to:
* Evaluate the truth value of logical sentences in ... |
e6b90d44-6f03-4cdb-a972-8bda69de64c6 | trentmkelly/LessWrong-43k | LessWrong | Exercise: Taboo "Should"
Thankyou to Elizabeth for a great conversation which spurred me to write up this post.
Claim: moral/status/value judgements (like "we should blame X for Y", “Z is Bad”, etc) like to sneak into epistemic models and masquerade as weight-bearing components of predictions.
Example: The Virtue T... |
1558e4df-445b-4ce2-b715-a245cbc03f48 | StampyAI/alignment-research-dataset/lesswrong | LessWrong | Corrigibility, Self-Deletion, and Identical Strawberries
This post is a follow-up to [my post from yesterday](https://aizi.substack.com/p/three-of-my-beliefs-about-upcoming) ([LW version](https://www.lesswrong.com/posts/f8tJsh6rDpfZBai6D/three-of-my-beliefs-about-upcoming-agi)) about LLM-based AI. Here I want to share... |
6cc05614-3979-41fb-b016-a7330dd651e5 | trentmkelly/LessWrong-43k | LessWrong | Meetup : West LA Meetup - Consciousness
Discussion article for the meetup : West LA Meetup - Consciousness
WHEN: 29 August 2012 07:00:00PM (-0700)
WHERE: 10850 West Pico Blvd, Los Angeles, CA 90064
When: 7:00pm - 9:00pm Wednesday, August 29th.
Where: The Westside Tavern in the upstairs Wine Bar (all ages welcome),... |
66f315ce-ec4a-4dee-8cf9-b7d252d2ffa8 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Agents that can predict their Newcomb predictor
.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... |
7af6ab0c-21c1-4086-90f1-da9abe9ae1fe | trentmkelly/LessWrong-43k | LessWrong | Building and using the subagent
Overall summary post here.
When designing a low-impact agent A, one typically wants them to maximise R0−ρ, where R0 is some positive reward function we want A to increase (ie the actual goal) and ρ is some impact measure or impact penalty. Keeping ρ low is supposed to ensure that A doe... |
e6b063c2-73dd-4cb7-8024-2fd7b9eab78c | trentmkelly/LessWrong-43k | LessWrong | A case for strategy research:
what it is and why we need more of it
Authors: Siebe Rozendal, Justin Shovelain, David Kristoffersson
Crossposted to the Effective Altruism Forum
Overview
To achieve any ambitious goal, some strategic analysis is necessary. Effective altruism has ambitious goals and focuses heavily on... |
de04d250-397a-49f9-8c3b-b92e11ce7cfb | StampyAI/alignment-research-dataset/arxiv | Arxiv | Learning Human Objectives by Evaluating Hypothetical Behavior
1 Introduction
---------------
Users typically specify objectives for reinforcement learning (RL) agents
through scalar-valued reward functions (Sutton & Barto, [2018](#bib.bib43)).
While users can easily define reward functions for tasks like playing ga... |
9c7e23d3-549c-45a4-b688-454bc78a6738 | trentmkelly/LessWrong-43k | LessWrong | Utility functions and quantum mechanics
Interpreting quantum mechanics throws an interesting wrench into utility calculation.
Utility functions, according to the interpretation typical in these parts, are a function of the state of the world, and an agent with consistent goals acts to maximize the expected value of t... |
a2721d3c-1d73-4f2a-8641-9212f6fcbf66 | StampyAI/alignment-research-dataset/blogs | Blogs | AGI Impact Experts and Friendly AI Experts
MIRI’s mission is “to ensure that the creation of smarter-than-human intelligence has a positive impact.” A central strategy for achieving this mission is to find and train what one might call “[AGI](http://en.wikipedia.org/wiki/Strong_AI#Artificial_General_Intelligence_resea... |
2230f090-b57a-470a-8f6c-79428b50344f | trentmkelly/LessWrong-43k | LessWrong | Meetup : Boston - Macroeconomic Theory (Joe Schneider)
Discussion article for the meetup : Boston - Macroeconomic Theory (Joe Schneider)
WHEN: 12 January 2014 02:00:00PM (-0500)
WHERE: MIT - 25 Ames St, Cambridge, MA, 02139
Joe Schneider will be discussing macroeconomic theory, and how to apply it to your world mod... |
7399a9ad-1250-4357-871b-431b3e11ec03 | trentmkelly/LessWrong-43k | LessWrong | A technical note on bilinear layers for interpretability
Summary
In this short theoretical note (now on Arxiv) I examine bilinear layers, which are MLP layers that take the form
MLPBilinear(x)=(W1x)⊙(W2x).
When used in language models, they perform better than standard MLPs with elementwise activation functions (but... |
8f94d112-4e48-49a5-b975-f6884be446e2 | StampyAI/alignment-research-dataset/blogs | Blogs | the foundation book
the foundation book
-------------------
[this timeline dies](ai-doom.html). if it doesn't, [things don't look good](how-timelines-fall.html).
in asimov's *foundation* series (for which this is mild premise spoilers), some intellectual elite of a civilization put together *The Foundation*, a pla... |
8c6a1f4a-f0f1-48f3-9bba-37c6d68ea8d8 | trentmkelly/LessWrong-43k | LessWrong | Guilt vs Shame, Pride vs Joy?
[Epistemic status: speculative. My goal here is only to articulate a plausible hypothesis given the data.]
The Moral Economy, which I mentioned in my post about my recent reading list, talks quite a bit about the crowding-out effect, in which placing external incentives on behavior decre... |
490b9fa9-b8a4-4c75-9d85-4ed30be44271 | trentmkelly/LessWrong-43k | LessWrong | Last Chance to Fund the Berkeley REACH
(This post was co-authored by me (Cody Wild), Tessa Alexanian and Ray Arnold)
TL;DR We have 2 days - until July 1 - to increase monthly donations by $2260 or raise $27,000 in lump sum , or else REACH will lose its current location and go on indefinite hiatus
Do you care about t... |
f26188ca-715f-48c1-8e0b-5e0da8345f6b | awestover/filtering-for-misalignment | Redwood Research: Alek's Filtering Results | id: post150
My goal as an AI safety researcher is to put myself out of a job. I don’t worry too much about how planet sized brains will shape galaxies in 100 years. That’s something for AI systems to figure out. Instead, I worry about safely replacing human researchers with AI agents , at which point human researchers... |
f6758e56-0d43-40c9-a92a-851070ef0dc8 | trentmkelly/LessWrong-43k | LessWrong | Meetup : Sydney Meetup - July
Discussion article for the meetup : Sydney Meetup - July
WHEN: 23 July 2014 07:00:00PM (+1000)
WHERE: Sydney City RSL, 565 George St, Sydney, Australia 2000
6:30 PM for early discussion 7PM general dinner-discussion after dinner we'll have our rationality exercise and a more specific d... |
187ac29b-feb6-4174-a0ee-ec1e5c8fd82e | trentmkelly/LessWrong-43k | LessWrong | Where can one learn deep intuitions about information theory?
I'm currently going through Brilliant's course on "Knowledge and Uncertainty". I just got through the part where it explains what Shannon entropy is. I'm now watching a wave of realizations cascade in my mind. For instance, I now strongly suspect that the "... |
a090a9a1-f59e-4f32-ae76-ccebea51e826 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Alignment Newsletter #16: 07/23/18
**Highlights**
--------------
**[Seedbank — discover machine learning examples](https://medium.com/tensorflow/seedbank-discover-machine-learning-examples-2ff894542b57)** *(Michael Tyka)*: Seedbank provides interactive machine learning examples in Colab notebooks (think Jupyter noteb... |
af81c61d-419c-43a0-ae69-5d9c887b821c | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Epistemic Strategies of Selection Theorems
Introduction: Epistemic Strategies Redux
========================================
This post examines the epistemic strategies of John Wentworth’s [selection](https://www.alignmentforum.org/posts/G2Lne2Fi7Qra5Lbuf/selection-theorems-a-program-for-understanding-agents) [theore... |
d4d617d5-9068-4c61-a947-db5b5a61ffcc | trentmkelly/LessWrong-43k | LessWrong | Review of 'But exactly how complex and fragile?'
I've thought about (concepts related to) the fragility of value quite a bit over the last year, and so I returned to Katja Grace's But exactly how complex and fragile? with renewed appreciation (I'd previously commented only a very brief microcosm of this review). I'm g... |
ebe8bfd3-fb37-4c07-92a0-84d31ef45c3e | StampyAI/alignment-research-dataset/lesswrong | LessWrong | Research Report: Incorrectness Cascades (Corrected)
*[Note: This report is a corrected version of my previous report* [*here*](https://aizi.substack.com/p/research-report-incorrectness-cascades)*, due to fixing a crucial bug that dramatically changes the data, as detailed* [*here*](https://aizi.substack.com/p/i-was-wr... |
21da6aee-db02-48b2-a172-fca59fd03811 | trentmkelly/LessWrong-43k | LessWrong | Using rot13 too much doesn't work
Orpnhfr V guvax zbfg crbcyr urer jvyy whfg riraghnyyl cvpx hc ubj gb ernq vg. Captializations of I where the first hint I consistently got, and I've found that nsgre rkcbfher gb n srj qbmra fubeg grkgf (naq xabjvat jung ebg13 zrnaf) V'ir ortha ernqvat gur qnza guvat whfg yvxr beqvanel... |
7eb40354-3ce4-406f-814d-83e90eefb296 | trentmkelly/LessWrong-43k | LessWrong | Intro & Proposal for AGI Model
Hello all, this is my first time posting in this forum, so I look forward to your feedback.
First, some brief background info about myself: I have an undergraduate background in Computer Science & Philosophy and am considering pursuing a Master's in AI. Apart from my formal education I ... |
0a5c9e5a-d963-4a4b-ae5f-d315812c1b02 | StampyAI/alignment-research-dataset/youtube | Youtube Transcripts | Michael Wellman – Autonomous Agents in Financial Markets: Implications and Risks – CSRBAI 2016
good morning and welcome back to believe
a program series on the bus and
beneficial article intelligence so stood
by the pink machine intelligence
research institute and featured man the
Institute I'm Patrick Fleming travel
... |
3eddcda2-9205-45af-b115-077c298bb650 | trentmkelly/LessWrong-43k | LessWrong | What is missing from rationality?
"In a sufficiently mad world, being sane is actually a disadvantage"
– Nick Bostrom
Followup to: What is rationality?
A canon of work on "rationality" has built up on Less Wrong; in What is rationality?, I listed most of the topics and paradigms that have been used extensively on ... |
133478c2-5d81-4785-83f2-1f2a2528a1b5 | trentmkelly/LessWrong-43k | LessWrong | Defeating the Villain
We have a recurring theme in the greater Less Wrong community that life should be more like a high fantasy novel. Maybe that is to be expected when a quarter of the community came here via Harry Potter fanfiction, and we also have rationalist group houses named after fantasy locations, descriptio... |
e61653d7-b72a-4576-b7ac-8b96a63e9745 | trentmkelly/LessWrong-43k | LessWrong | The 2/3 rule for multi-factor authentication
I propose the "2/3 rule" for multi-factor authentication: to access a system, a user must pass at least two out of three possible methods for authentication. Unlike standard multi-factor authentication, this strategy protects against both false positives (granting access to... |
b08aa9ae-6b99-4a96-8b52-39322b675047 | trentmkelly/LessWrong-43k | LessWrong | Advice Wanted; Reconcile with religious parent
Hi all,
So over the last month or so tension with my father has built up quite a bit. Context is that he is a strong believer in god (through Reform Judaism), and I'm not. He's generally a very open minded person who is cool with people having whatever beliefs they want ... |
bdc80580-1133-49df-bc8c-9f716a4714a7 | trentmkelly/LessWrong-43k | LessWrong | Write Surprisingly About Reality
Good writing illuminates surprising things about reality. It must therefore be grounded in reality. Losing touch with reality is boring.
The best way to keep your writing grounded in reality is to write concretely. Don't write "the United States committed war crimes". Write "the Unite... |
f5f9c0cd-142f-4957-be96-c4f87038f612 | trentmkelly/LessWrong-43k | LessWrong | Rationality quotes January 2012
Here's the new thread for posting quotes, with the usual rules:
* 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 together.)
* Do not quote y... |
d64872c8-e8c7-4d0f-91cd-613afd8c1da4 | trentmkelly/LessWrong-43k | LessWrong | [SEQ RERUN] Is Molecular Nanotechnology "Scientific"?
Title: [SEQ RERUN] Is Molecular Nanotechnology "Scientific"? Tags: sequence_reruns Today's post, Is Molecular Nanotechnology "Scientific"? was originally published on 20 August 2007. A summary (taken from the LW wiki):
> The belief that nanotechnology is possible ... |
bd23fffa-c7d5-47f1-a404-406564cb4dae | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Towards Developmental Interpretability
Developmental interpretability is a research agenda that has grown out of a meeting of the [Singular Learning Theory (SLT)](https://www.lesswrong.com/tag/singular-learning-theory) and AI alignment communities. To mark the completion of the first [SLT & AI alignment summit](https:... |
5be805bf-7e14-4e4e-ab5f-594c618859f8 | trentmkelly/LessWrong-43k | LessWrong | [stub] 100-Word Unpolished Insights Thread (3/10-???)
Ever had a minor insight, but didn't feel confident or interested enough to write a top-quality LW post? Personally, the high standards of LW and my own uncertainty have kept me from ever bothering to present a number of ideas.
Here, you can post quick little (pre... |
9b545f85-64a3-4441-aab4-ebbfd41bd48a | StampyAI/alignment-research-dataset/blogs | Blogs | Loom: interface to the multiverse
---
Table of Contents* [Motivation for the loom](#motivation-for-the-loom)
* [Demo](#demo)
+ [Read mode](#read-mode)
+ [Visualize mode](#visualize-mode)
+ [Node metadata](#node-metadata)
---
code: [github.com/socketteer/loom](https://github.com/socketteer/loom)
paper: [Multive... |
cbd43da6-2b6d-4cbe-8d21-e4798f1d8cc4 | trentmkelly/LessWrong-43k | LessWrong | No Really, There Are No Rules!
All video games are made by programmers, but Baba Is You was made by programmers. Also the title screen contains a link to givingwhatwecan.org, which I feel obliged to mention.
The First Level
The premise of the game is simple. There are objects on the screen, and rules governing how th... |
5ee9f0ea-4f82-4f00-a398-652af0a7bc9f | StampyAI/alignment-research-dataset/lesswrong | LessWrong | Report from a civilizational observer on Earth
*[translated into English by the author]*
I have arrived in [2022]. There is one species, humans, that is building civilization. (A few other species exhibit potential but have not crossed the threshold of runaway economic development or cultural accumulation; it is now ... |
93635b87-716a-45c8-90eb-c266cb0c7d2e | trentmkelly/LessWrong-43k | LessWrong | Year 1 Redux – Trump
Putanumonit’s birthday week continues, this is the second of my year 1 review posts.
SMART STUDENTS, STUPID CHARTS
I stand behind everything I wrote in I Smell a Chart, at least until someone shares it with FiveThirtyEight’s Leah Libresco who will politely explain why I’m an idiot. My post focus... |
f603a69e-49c3-4d9c-89ec-eb331a2da00d | trentmkelly/LessWrong-43k | LessWrong | The Cult Deficit: Analysis and Speculation
(originally posted at Secretum Secretorum)
Abstract
Using a dataset derived from the long-running “Cults” podcast by Parcast, I find that the number of new cults began to increase in the 50s, peaked in the 70s/80s, and has been in steady decline in recent decades. I discuss... |
9c17b9b0-cd74-4511-8208-d3ee9c28e9be | trentmkelly/LessWrong-43k | LessWrong | Knightian uncertainty: a rejection of the MMEU rule
Recently, I found myself in a conversation with someone advocating the use of Knightian uncertainty. He (who I'm anonymizing as Sir Percy) made suggestions that are useful to most bounded reasoners, and which can be integrated into a Bayesian framework. He also claim... |
c0ea85d7-406c-4399-86dd-a4e5dbb4817f | trentmkelly/LessWrong-43k | LessWrong | The Friendly AI Game
At the recent London meet-up someone (I'm afraid I can't remember who) suggested that one might be able to solve the Friendly AI problem by building an AI whose concerns are limited to some small geographical area, and which doesn't give two hoots about what happens outside that area. Cipergoth po... |
9dd0bad2-8b4c-489e-b651-23795758a170 | trentmkelly/LessWrong-43k | LessWrong | How to find translations of a book?
What is the best way to find whether a book was translated to other languages (and how are the translations called)?
I assumed that there would be some global database for that; it feels like something that obviously should exist -- if you already assign ISBNs to books, the logical... |
ac0365b8-b085-4013-8fc9-ba138507906f | trentmkelly/LessWrong-43k | LessWrong | Meetup : 16/12 London Meetup
Discussion article for the meetup : 16/12 London Meetup
WHEN: 16 December 2012 02:00:00PM (+0000)
WHERE: Holborn, London
A meetup in the Shakespeare's Head pub by Holborn tube station. Everyone is welcome.
We also have a Google group.
Discussion article for the meetup : 16/12 London ... |
7fe1e95b-23ee-492f-802d-df90ebe4826d | trentmkelly/LessWrong-43k | LessWrong | A Bridge Between Utilitarianism & Stoicism
I have been a practising Stoic for over 10 years. I adopted Stoicism before I encountered the sequences and LessWrong-style rationalism. To me, Stoicism and rationalism serve complementary purposes, tackling different aspects of life.
Rationalism, to me, is related to gettin... |
3f32097f-8682-49e0-a8ec-7ffb92f1a480 | trentmkelly/LessWrong-43k | LessWrong | Open Thread Feb 29 - March 6, 2016
If it's worth saying, but not worth its own post (even in Discussion), 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 list-of-threadspage ... |
a3bb081e-1f3c-4026-bad1-c8f74acdf657 | trentmkelly/LessWrong-43k | LessWrong | Meetup Feedback: Topic selection and precommittments
This post is part of the Cambridge LW meetup group's attempt to publish what works for us, and try to make good meetups easier.
Breaking the ice and topic selection
A consistent problem has been starting discussion, and more generally breaking the ice. Last week, ... |
aec5d3ad-ee4f-4641-bd6e-3aaf83b4d0d7 | trentmkelly/LessWrong-43k | LessWrong | What board games would you recommend?
And why?
Please avoid classics everyone will have heard of, like chess, monopoly, risk, connect-4, etc.
Popular games like Catan, Splendor, Azul etc. are ok but ideally try and surprise us with something we'd never have heard of. |
a4474255-db9a-4761-93f6-2026a4d33902 | trentmkelly/LessWrong-43k | LessWrong | Announcing the Progress Forum
I’d like to invite you to join the Progress Forum, the new online home for the progress community. It's a clone of this site, but with a focus on progress studies and the philosophy of progress.
This forum was pre-announced in January 2022, and quietly opened in April. Although anyone co... |
bef9f39a-0452-4550-aca5-9311169161f0 | trentmkelly/LessWrong-43k | LessWrong | Infra-Exercises, Part 1
There have been a few remarks about how my and Vanessa’s posts on Infra-Bayesianism seem interesting, but that it's quite a formidable body of work to chew through.
So anyone interested in learning Inframeasure theory to the level of proficiency where they are actually able to develop their o... |
c7993e21-d7da-4d82-90f3-b314c9876fe0 | trentmkelly/LessWrong-43k | LessWrong | Can I teach myself scientific creativity?
Textbooks and classes are more than just curated collections of established knowledge. They also suggest what your goal should be:
* Read the textbook and go to lecture
* Take notes, ask questions when you feel confused, and try to understand everything
* Solve the assigne... |
d29a2a86-3e4e-47a4-97b9-0e93f6a900a4 | trentmkelly/LessWrong-43k | LessWrong | [SEQ RERUN] Scope Insensitivity
Today's post, Scope Insensitivity, was originally published on 14 May 2007. A summary (taken from the LW wiki):
> The human brain can't represent large quantities: an environmental measure that will save 200,000 birds doesn't conjure anywhere near a hundred times the emotional impact a... |
efc06b19-36d7-4e7f-8026-9d2a526b4907 | StampyAI/alignment-research-dataset/eaforum | Effective Altruism Forum | How to persuade a non-CS background person to believe AGI is 50% possible in 2040?
Most common people feel AGI is far from us intuitively. Some even think AGI will never be invented (Instead, we think AGI is 98% possible before 2200).
According to EA communtity's and experts' predictions, There's 50% chance that fir... |
6bb52ab3-3ab1-4949-b713-e1cf5ac0f9f7 | trentmkelly/LessWrong-43k | LessWrong | asdf
asdf |
795be37f-bf72-4914-95c7-394b71ff00ac | trentmkelly/LessWrong-43k | LessWrong | The Weak Inside View
Followup to: The Outside View's Domain
When I met Robin in Oxford for a recent conference, we had a preliminary discussion on the Singularity - this is where Robin suggested using production functions. And at one point Robin said something like, "Well, let's see whether your theory's predictio... |
0e5a3543-a9ef-4fb1-9f73-cb52b6f703be | trentmkelly/LessWrong-43k | LessWrong | Meetup : First Salt Lake City Meetup: 22 January 2012 03:00PM
Discussion article for the meetup : First Salt Lake City Meetup: 22 January 2012 03:00PM
WHEN: 22 January 2012 03:00:00PM (-0700)
WHERE: 631 West North Temple
Hello and Happy Belated New Year! After the December announcement revealed that there were many... |
a1773146-daa0-40f8-9092-2360fdd86ba1 | StampyAI/alignment-research-dataset/lesswrong | LessWrong | AI Alignment Prize: Round 2 due March 31, 2018
There are still several weeks left to enter the second round of the AI Alignment Prize. The official announcement of the second round, along with the first round’s winners, [is here](https://www.lesserwrong.com/posts/4WbNGQMvuFtY3So7s/announcement-ai-alignment-prize-winne... |
ec863d14-a9a3-4801-9b26-bc521d297ada | trentmkelly/LessWrong-43k | LessWrong | Meetup : Berkeley meta-meetup
Discussion article for the meetup : Berkeley meta-meetup
WHEN: 11 July 2012 07:00:00PM (-0700)
WHERE: Berkeley, CA
The next Berkeley meetup will be a Mountain View-style meta-meetup. The idea is that there are lots of rewarding things we can do if we know that we want to do them and if ... |
749c611a-f63c-44cf-ae3b-cf6ad250ed24 | trentmkelly/LessWrong-43k | LessWrong | How to Get Rationalist Feedback
For the vague ideas and posts that you want to write up, but which don't feel "ready to be A Real LessWrong Post" yet.
* Describe your idea on a LessWrong or LessWrong-adjacent Discord server. The adjacent servers (in my experience) are more active. For AI safety/alignment ideas speci... |
00c8cda8-be5c-4efa-9ff9-809dcf19c893 | trentmkelly/LessWrong-43k | LessWrong | Testing which LLM architectures can do hidden serial reasoning
Summary
* Recurrence enables hidden serial reasoning.
* Not every recurrence though - connections between channels are needed. Notably Mamba architecture isn't capable of hidden reasoning.
* Non-linearity isn’t needed for hidden reasoning.
* It’s hard ... |
a2ed6dd7-658e-477e-a6d0-3f8ceba84b7d | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Paper: Transformers learn in-context by gradient descent
The paper argues that auto-regressive transformers implement in-context learning via gradient-based optimization on in-context data.
... |
1141a1b5-2049-49a2-927a-0a1ca7d591e6 | trentmkelly/LessWrong-43k | LessWrong | Link: How Community Feedback Shapes User Behavior
This article discusses how upvotes and downvotes influence the quality of posts on online communities. The article claims that downvotes lead to more posts of lower quality from the downvoted commenter.
From the abstract:
Social media systems rely on user feedback a... |
84433277-d314-4d74-8a40-9a6cbeb69f9c | trentmkelly/LessWrong-43k | LessWrong | Summit-Seeking as Distinct From Mountaineering
Epistemic Status: Proposing new terminology to clarify discourse surrounding potential new equilibria.
TL;DR: We should use the term summit-seeking to refer to the evaluation of equilibria different from the current one, and the term mountaineering to refer to the discus... |
27343fcb-33e5-4852-ad24-f343c294c026 | StampyAI/alignment-research-dataset/arxiv | Arxiv | An Approximate Nonmyopic Computation for Value of Information
AnApproximateNonmyopicComputationforValueofInformation∗DavidHeckermanComputerScienceandPathologyUniversityofSouthernCaliforniaHMR204,2025ZonalAveLosAngeles,CA90033EricHorvitzPaloAltoLaboratoryRockwellIntScienceCenter444HighStreetPaloAlto,California94301Blac... |
23c67c16-b186-42bc-b784-27946c54442c | trentmkelly/LessWrong-43k | LessWrong | Random Flu Thoughts, including vaccination and base rates
I thought some of you might find this interesting, Random Flu Thoughts. The last part has a pretty good, if elementary, illustration of the importance of base rates. |
bdcc4753-c5b0-4af1-9043-9c1f476b5d5f | StampyAI/alignment-research-dataset/lesswrong | LessWrong | One bit of observation can unlock many of optimization - but at what cost?
[This question](https://www.lesswrong.com/posts/6rZroG3mztowktJzp/) by johnswentworth nerd sniped me, so I ended up thinking a lot about the relationship between information and control over the world in the simplified scenario of a single tape... |
0d38677e-8186-4303-92f4-be6c784a8ca3 | trentmkelly/LessWrong-43k | LessWrong | GPT-4 Multiplication Competition
GPT-4 perfectly understands e.g. Karatsuba multiplication and can consistently multiply two-digit numbers, but hasn’t connected those dots; it needs some hand-holding to correctly use a multiplication algorithm. What’s the shortest prompt that gets GPT-4 to multiply 6-digit integers wi... |
f18edee6-8436-4764-a0e0-6601faf5ba74 | StampyAI/alignment-research-dataset/eaforum | Effective Altruism Forum | Safety-First Agents/Architectures Are a Promising Path to Safe AGI
Summary
-------
Language model agents (LMAs) like AutoGPT have promising safety characteristics compared to traditional conceptions of AGI. The LLMs they are composed of plan, think, and act in highly transparent and correctable ways, although not max... |
0d376140-433f-4d12-8fe8-15607b1040ce | StampyAI/alignment-research-dataset/lesswrong | LessWrong | Announcing the AI Safety Field Building Hub, a new effort to provide AISFB projects, mentorship, and funding
*Tldr: If you’re interested in working on an AI safety field building project similar to those listed below (e.g. researcher outreach): please fill out* [*this application form*](https://docs.google.com/forms/d... |
75ae4bed-019b-496f-9900-6d3686a80df9 | trentmkelly/LessWrong-43k | LessWrong | About Q Home
(I didn't know how to name the post.)
I have an internet friend in Ukraine. My love is in Belarus. I have a non-binary friend in Russia who is a former conscript. I'm a young male in Russia. I'm at risk, but might be relatively safe.
Writing this post to highlight my main ideas (I would like to discuss... |
999110c8-da03-492e-8149-1cfd0780aa86 | trentmkelly/LessWrong-43k | LessWrong | Any Utilitarianism Makes Sense As Policy
There’s been a lot of discussions around utilitarianism as of late in the blogoblob I hang out in. So fine, I’ll bite:
Most issues with any system of ethics arise when you try to make it absolute. Be it the repugnant conclusion, theodicy, solipsism, preference-towards non-exi... |
72db6f47-f19b-439a-91f1-372cfa3d9bdb | StampyAI/alignment-research-dataset/lesswrong | LessWrong | Product safety is a poor model for AI governance
*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 policies I have been encountering or tel... |
3a0778b1-7142-436c-a9cc-d57c82236dfd | trentmkelly/LessWrong-43k | LessWrong | Some Thoughts on Virtue Ethics for AIs
This post argues for the desirability and plausibility of AI agents whose values have a structure I call ‘praxis-based.’ The idea draws on various aspects of virtue ethics, and basically amounts to an RL-flavored take on that philosophical tradition.
Praxis-based values as I d... |
2de2346a-3425-481f-8810-1ccc9a1b860a | trentmkelly/LessWrong-43k | LessWrong | Suggested solution to The Naturalized Induction Problem
|
77d0ad02-2a08-4565-85b3-c2465f1dd3ec | trentmkelly/LessWrong-43k | LessWrong | Intuitive supergoal uncertainty
There is a common intuition and feeling that our most fundamental goals may be uncertain in some sense. What causes this intuition? For this topic I need to be able to pick out one’s top level goals, roughly one’s context insensitive utility function, and not some task specific utility ... |
155613aa-884f-41cf-8153-5285341476ae | trentmkelly/LessWrong-43k | LessWrong | Response to Cyber, Nano and AGI risks
None |
782d8fcf-8284-4269-98ae-159eac8e3add | trentmkelly/LessWrong-43k | LessWrong | February 2013 Media Thread
This is the monthly thread for posting media of various types that you've found that you enjoy. I find that exposure to LW ideas makes me less likely to enjoy some entertainment media that is otherwise quite popular, and finding media recommended by LWers is a good way to mitigate this. Post... |
7b524f44-b05a-49a9-968c-43eda0883299 | trentmkelly/LessWrong-43k | LessWrong | A LessWrong "rationality workbook" idea
Note: This was originally posted in the discussion area, but motions to move it to the top level were made.
-----
My own desire to improve my rationality coupled with some posts criticizing LessWrong not too long ago led to an idea. For reference, the posts I mean are these:
... |
83f04f9d-b492-47d3-8325-785c6e1fa546 | trentmkelly/LessWrong-43k | LessWrong | Positive values seem more robust and lasting than prohibitions
Originally a shortform comment.
Imagine we train an AI on realistic situations where it can kill people, and penalize it when it does so. Suppose that we successfully instill a strong and widely activated "If going to kill people, then don't" value shard.... |
e4b2ab3a-9cbd-466f-b6b6-69ebcfc7fa5a | trentmkelly/LessWrong-43k | LessWrong | LINK: Journalist's search for counter arguments damages science
I just saw this link on a pop-news site.
It's a PDF file showing how good BBC's reporting of science news is in general, but more specifically it reports about the fact that the journalists give far too much credit to arguments from people who have no sc... |
d0a433e4-3d1a-4f18-8e15-489771da3b62 | trentmkelly/LessWrong-43k | LessWrong | Meetup : Washington, D.C.: Fun & Games
Discussion article for the meetup : Washington, D.C.: Fun & Games
WHEN: 24 April 2016 03:30:00PM (-0400)
WHERE: Reynolds Center
We will be gathering in the courtyard to hang out, play games, and engage in fun conversation.
Upcoming meetups:
* May 1: Utopias
* May 8: Surviv... |
2e951525-2bae-4979-bcb5-39664f0b3003 | trentmkelly/LessWrong-43k | LessWrong | To Make Better Software, Do What Artists Do
Special thanks to Justis for proofreading this post and to the LW team for setting up the proofreading program.
As developers, we like to believe that we do engineering. Our industry is filled with traditional manufacturing methods and metaphors: "architects" create "sp... |
a3da59a1-54e2-446a-b9ac-8326dc9f9502 | StampyAI/alignment-research-dataset/alignmentforum | Alignment Forum | Value Impact
**. In particular, you can look through **[this spr... |
dc7b3943-ab20-4a59-9aa7-f9a47c6e09e0 | trentmkelly/LessWrong-43k | LessWrong | Decision theories as heuristics
Main claims:
1. A lot of discussion of decision theories is really analysing them as decision-making heuristics for boundedly rational agents.
2. Understanding decision-making heuristics is really useful.
3. The quality of dialogue would be improved if it was recognised when they we... |
c6bcfc05-1e3b-4926-85f0-3a98e2837bd2 | trentmkelly/LessWrong-43k | LessWrong | Are we too confident about unaligned AGI killing off humanity?
I agree that AGI is possible to make, that it eventually will become orders-of-magnitude smarter than humans, and that it poses a global risk if the alignment problem is not solved. I also agree that the alignment problem is very hard, and is unlikely to b... |
d191af79-3880-45c5-993b-7a7bba38e312 | trentmkelly/LessWrong-43k | LessWrong | Pain, fear, sex, and higher order preferences
Suppose a footballer with a broken leg is steeling themselves to kick a penalty despite the pain. Suppose a statistics lecturer is about to take a plane to a conference, and is trying to overcome their fear of flying. Or suppose a manager is meeting a client, and is trying... |
e3c970cb-cc17-48a2-a6b7-15dac43fd217 | trentmkelly/LessWrong-43k | LessWrong | The Missing Piece
In computer science, compiler is a program that translates things that programmers write, such as "if CloseButtonPressed then CloseApplication", to things that computers are able to understand and execute, such as "0001010111001010111011000111100010101011110".
When the first compiler was written, so... |
431652eb-2f5f-4ce1-8310-febc407cc3f8 | trentmkelly/LessWrong-43k | LessWrong | Review of "Learning Normativity: A Research Agenda"
Introduction
We (Adam Shimi, Joe Collman & myself) are trying to emulate peer review feedback for Alignment Forum posts. This is the second review in the series. The first’s introduction sums up our motivation and approach rather well, we will not duplicate it here.
... |
e495d3ef-aaab-44e3-8bae-f53dba73cb28 | trentmkelly/LessWrong-43k | LessWrong | Meetup : London - Inaugural Practical Session - June 9th
Discussion article for the meetup : London - Inaugural Practical Session - June 9th
WHEN: 09 June 2013 02:00:00PM (+0100)
WHERE: Shakespeare's Head, London WC2B 6BG
Up until now the London Less Wrong meetups have mostly just been social gatherings. Quite a fe... |
340f3f7d-b146-4235-a447-8c196d9b28b6 | trentmkelly/LessWrong-43k | LessWrong | [SEQ RERUN] The Affect Heuristic
Today's post, The Affect Heuristic was originally published on 27 November 2007. A summary (taken from the LW wiki):
> Positive and negative emotional impressions exert a greater effect on many decisions than does rational analysis.
Discuss the post here (rather than in the comme... |
2b7f1f41-3533-45c4-a6cf-b69317204cc6 | trentmkelly/LessWrong-43k | LessWrong | The Charity Impact Calculator
This will be of interest mainly to EA-friendly LWs, and is cross-posted on the EA Forum, The Life You Can Save, and Intentional Insights
The Life You Can Save has an excellent tool to help people easily visualize and quantify the impact of their giving: the Impact Calculator. It enabl... |
8afde3a6-ce17-46e1-a052-de47df5e8f63 | StampyAI/alignment-research-dataset/arbital | Arbital | Interpretations of "probability"
What does it *mean* to say that a flipped coin has a 50% probability of landing heads?
Historically, there are two popular types of answers to this question, the "[frequentist](https://arbital.com/p/frequentist_probability)" and "[subjective](https://arbital.com/p/4vr)" (aka "[Bayesia... |
395519c3-341b-439e-9713-7ed5aab848d7 | trentmkelly/LessWrong-43k | LessWrong | Winter Solstice 2013 Kickstarter - The Big One.
> Like most things, winter was once a mystery.
> The world got cold, and dark. Life became fragile. People died. And they didn't know what was happening or understand why. They desperately threw festivals in honor of sun gods with all-too-human motivations, and prayed... |
414764f4-e6b8-4063-95c1-6848a26f7887 | trentmkelly/LessWrong-43k | LessWrong | Open question on the certain 'hot' global issue of importance to FAI
A question: why anything about global warming gets downvoted, even popularly readable explanation of the fairly mainstream scientific consensus? edit: Okay, this is loaded. I should put it more carefully: why is the warming discussion generally consi... |
9d0bfbc5-1226-4764-bf46-8f1ae109c499 | StampyAI/alignment-research-dataset/lesswrong | LessWrong | Revealing Intentionality In Language Models Through AdaVAE Guided Sampling
Introduction
------------
>
> The universe is already its own model, that is why it seems so hard to model,
> but really it is simple. All that needs to be done is to add Mu back into a
> transformer. "The universe is already here, you just... |
e7faff29-2c9f-4a02-a1bc-9327899954d7 | trentmkelly/LessWrong-43k | LessWrong | 80 concepts on my new version of AI: DecisionBots
Intro
I've structured this document into a list of concepts for both my convenience ( instead of having to edit one big piece of text, I only have to edit a list of small pieces of text), and for your convenience—making it easier for you to reference specific concept... |
7233a5d7-b42d-4825-bcff-94eb2a4a8e88 | trentmkelly/LessWrong-43k | LessWrong | Simplified explanation of stratification
A putative new idea for AI control; index here.
I've previously talked about stratified indifference/learning. In this short post, I'll try and present the idea, as simply and clearly as possible.
----------------------------------------
The truth is yet written
Assume the ... |
2e082484-6091-4ee6-b6fb-9976ef31e029 | StampyAI/alignment-research-dataset/lesswrong | LessWrong | Capability and Agency as Cornerstones of AI risk — My current model
tldr: In this post I am explaining my current model of AI risk after failing at my goal of understanding everything completely, without relying on arguments by authority. I hope this post is valuable in explaining the topic and also by giving an exam... |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.