id
stringlengths
36
36
source
stringclasses
15 values
formatted_source
stringclasses
13 values
text
stringlengths
2
7.55M
4b665037-1644-4081-a3a2-53e6fdbb2971
trentmkelly/LessWrong-43k
LessWrong
Goal-Direction for Simulated Agents tldr: consistent LLM failure suggests possible avenue for alignment and control epistemic status: somewhat hastily written, speculative in places, but lots of graphs of actual model probabilities Today in ‘surprisingly simple tasks that even the most powerful large language models...
2d7a572e-82e7-4aec-afe2-74eac8885b82
trentmkelly/LessWrong-43k
LessWrong
Starting Meditation [Update] this intent has been put on hold. Check this comment for details. This note was put at the top, because whenever I see a post about "I'm going to do XYZ" my most burning question is always, "Well did they?" I think since about last winter I've been convinced that there is something really...
840d1fef-b8fb-4b6a-bbd6-77cc6e2befaa
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
[AN #122]: Arguing for AGI-driven existential risk from first principles 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 **[thi...
30c99f82-928f-4205-8fdb-5f8461d78c34
trentmkelly/LessWrong-43k
LessWrong
Wireheading Done Right: Stay Positive Without Going Insane
a649e96f-4305-4229-9818-a42765f76678
trentmkelly/LessWrong-43k
LessWrong
Clipboard Filtering Here's a pattern I find pretty useful: pbpaste | some_command | pbcopy For example: * Converting spaces to tabs, for pasting into a spreadsheet program: pbpaste | tr ' ' '\t' | pbcopy * Converting tabs and newlines to html table formatting: pbpaste | sed 's/^/<tr><td>/' | sed 's/\t/<td>/g'...
d8e1265f-35ec-4449-8be3-80e0130e0c36
trentmkelly/LessWrong-43k
LessWrong
The Jackpot Jinx (or why “Superintelligence Strategy” is wrong) On March 5, 2025 Dan Hendrycks, Eric Schmidt, and Alexandr Wang published “Superintelligence Strategy”, a paper that suggests a number of policies for national security in the era of AI. Central to their recommendations is a concept they call “Mutual Assu...
e2bc1826-2345-4bf1-8af7-f55b325e43e7
trentmkelly/LessWrong-43k
LessWrong
Is the length of the Covid-19 incubation period likely to be affected by whether you are vaccinated? Covid-19 symptoms typically present 2-14 days after exposure in unvaccinated people, with an average presentation at 5 days post-exposure. Is this period likely to be affected by having been vaccinated? Amongst those ...
56d2ad50-2656-47ee-83ca-64b9f86b8893
trentmkelly/LessWrong-43k
LessWrong
St. Louis, Missouri Meetup - now happening every week! I got inspired by the NY meetup post by Cosmos, and I'd love to get a recurring meetup started in St. Louis. Where: Kayak Coffee at Skinker and Forest Park Parkway, near Washington University. When: Every Tuesday from 6-8 PM, until and including May 3, 2011 and ...
a2a36d70-5e13-44ee-bba8-b1fdf8e81aff
StampyAI/alignment-research-dataset/special_docs
Other
Challenges of Aligning Artificial Intelligence with Human Values 54Margit Sutrop Acta Baltica Historiae et Philosophiae Scientiarum Vol. 8, No. 2 (Autumn 2020) Acta Baltica Historiae et Philosophiae Scientiarum Vol. 8, No. 2 (Autumn 2020) DOI : 10.11590/abhps.2020.2.04 Challenges of Aligning Artificial Intelligen...
9b8c7507-c41d-415f-be0a-5edcc95c1644
StampyAI/alignment-research-dataset/lesswrong
LessWrong
The issue of meaning in large language models (LLMs) *Cross posted* [*from New Savanna*](https://new-savanna.blogspot.com/2023/03/the-issue-of-meaning-in-large-language.html)*.* Noam Chomsky was trotted out to write a dubious op-ed in *The New York Times* about large language models and Scott Aaronson registered his...
bfc7303c-1784-481a-b6fa-a4d8f125475b
trentmkelly/LessWrong-43k
LessWrong
Spreadsheet for 200 Concrete Problems In Interpretability Thanks to the Open Source Mechanistic Interpretability Slack for their feedback. When going through Neel's excellent 200 Concrete Problems In Interpretability sequence, I found myself thinking "It sure would be great to have a single document I could go throug...
c3783189-f422-4fe3-a177-9b7d5c6ef608
trentmkelly/LessWrong-43k
LessWrong
Book Summary: Willpower by Baumeister, Tierney I recently read this book. I've tried to summarize the main points below -- you can read my notes here (MSWord doc). You might also find Derek Sivers' notes useful, which can be found here. NOTE: The general model of willpower (as a finite resource consumed with use) use...
4e926ba9-ab40-4126-863f-f8aae83d0397
trentmkelly/LessWrong-43k
LessWrong
Contra Scott on Abolishing the FDA Scott Alexander’s recent post on the FDA raises the average level of discourse on the subject. He starts from the premise that the FDA deserves destruction but cautions against rash action. No political slogan can be implemented without clarification and “Abolish the FDA” is no diffe...
6808bb29-ac74-4953-a73f-44aafd0403e2
trentmkelly/LessWrong-43k
LessWrong
Setting the Zero Point Context: I have hopes of assembling a full "defense against the dark arts" sequence. This essay will not necessarily be the very first one in the sequence, but it's One Of The Basic Spells To Defend Against, and it's the one I happen to have done the most data gathering on so it's getting writte...
367d135d-0eeb-4a6c-aaf3-ecf79b0afa5d
trentmkelly/LessWrong-43k
LessWrong
The Bitter Lesson for AI Safety Research Read the associated paper "Safetywashing: Do AI Safety Benchmarks Actually Measure Safety Progress?": https://arxiv.org/abs/2407.21792 Focus on safety problems that aren’t solved with scale. Benchmarks are crucial in ML to operationalize the properties we want models to have ...
c6c826fe-55bf-4344-b149-98652b0327d2
trentmkelly/LessWrong-43k
LessWrong
Environments as a bottleneck in AGI development Given a training environment or dataset, a training algorithm, an optimiser, and a model class capable of implementing an AGI (with the right parameters), there are two interesting questions we might ask about how conducive that environment is for training an AGI. The fi...
fcb1ec5a-f9ed-483f-81d3-aa5853352884
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
Modeling the impact of safety agendas *This post, which deals with how safety research - that is, technical research agendas aiming to reduce AI existential risk - might impact risks from AI, is part 6 in our* [*sequence on Modeling Transformative AI Risk*](https://www.alignmentforum.org/s/aERZoriyHfCqvWkzg)*. In this...
4af8c191-3c76-4f13-ae87-6d1d80b4aaef
trentmkelly/LessWrong-43k
LessWrong
Relationship Outcomes Are Not Particularly Sensitive to Small Variations in Verbal Ability After a friendship-ending fight, you feel an impulse to push through the pain to do an exhaustive postmortem of everything you did wrong in that last, fatal argument—you could have phrased that more eloquently, could have antici...
8a539c2e-f5d8-4665-8642-5efb49fd5051
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
Utility uncertainty vs. expected information gain It is a relatively intuitive thought that if a Bayesian agent is uncertain about its utility function, it will act more conservatively until it has a better handle on what its true utility function is. This might be deeply flawed in a way that I'm not aware of, but I'...
d9658a21-00c7-4823-bbe4-ecab3a5d00f4
trentmkelly/LessWrong-43k
LessWrong
Debating with More Persuasive LLMs Leads to More Truthful Answers We've just completed a bunch of empirical work on LLM debate, and we're excited to share the results. If the title of this post is at all interesting to you, we recommend heading straight to the paper. There are a lot of interesting results that are har...
f9e0858c-de8e-45d2-9d15-f7fbf278ff06
trentmkelly/LessWrong-43k
LessWrong
New AI risks research institute at Oxford University The Oxford Martin Programme on the Impacts of Future Technology (aka FutureTech) is a new research department at Oxford University, roughly a spin-off of FHI, but focusing on AI and nanotech risks and differential technological development. Like FHI, this department...
3d5b5539-3d99-453d-981c-687799a66254
trentmkelly/LessWrong-43k
LessWrong
The Regularizing-Reducing Model The linked document is my (last-minute) submission for the AI alignment prize. Feedback is appreciated, even though it cannot change my submission entry.
654ced21-1ec6-42cc-9f06-9123c1a2519e
StampyAI/alignment-research-dataset/aisafety.info
AI Safety Info
What is meta-RL? You can learn about this computer science concept [here](https://www.youtube.com/playlist?list=PLoROMvodv4rNjRoawgt72BBNwL2V7doGI).
e4a16214-2c0e-4bc2-bb90-45e137b5f6af
trentmkelly/LessWrong-43k
LessWrong
Distillation of "How Likely Is Deceptive Alignment?" Summary Deceptive alignment seems to be the likely outcome of training a sufficiently intelligent AI using gradient descent.  An ML model with a long term goal and an understanding of its training process would probably pretend to be aligned during training, as this...
d08853ca-0f31-4098-a8a1-109136e08458
trentmkelly/LessWrong-43k
LessWrong
Episode 5 of Tsuyoku Naritai (the 'becoming stronger' podcast): The Stance Transcript in show notes, as always! https://anchor.fm/tsuyokunaritai/episodes/Episode-5---The-Stance-e434pj
3d0154cb-af64-4851-8370-d78378b9bf25
trentmkelly/LessWrong-43k
LessWrong
Your inner Google I just heard a comment by Braddock of Lovesystems that was brilliant:  All that your brain does when you ask it a question is hit "search" and return the first hit it finds.  So be careful how you phrase your question. Say you just arrived at work, and realized you once again left your security pass...
0466349e-4c5b-4e18-9bda-9d496a25eab8
trentmkelly/LessWrong-43k
LessWrong
Ureshiku Naritai This is a supplement to the luminosity sequence.  In this comment, I mentioned that I have raised my happiness set point (among other things), and this declaration was met with some interest.  Some of the details are lost to memory, but below, I reconstruct for your analysis what I can of the process....
03236393-82ef-4d57-8325-c33b75bb8460
trentmkelly/LessWrong-43k
LessWrong
Meta wants to use AI to write Wikipedia articles; I am Nervous™ Meta is experimenting with using AI to write Wikipedia articles: https://ai.facebook.com/research/publications/generating-full-length-wikipedia-biographies-the-impact-of-gender-bias-on-the-retrieval-based-generation-of-women-biographies  I personally hav...
f39978cc-7a69-40de-bb5b-a24ad2464676
trentmkelly/LessWrong-43k
LessWrong
AI #86: Just Think of the Potential Dario Amodei is thinking about the potential. The result is a mostly good essay called Machines of Loving Grace, outlining what can be done with ‘powerful AI’ if we had years of what was otherwise relative normality to exploit it in several key domains, and we avoided negative outco...
c40e46eb-adbe-43da-a6bd-a9eea0bc87c7
trentmkelly/LessWrong-43k
LessWrong
Walk while you talk: don't balk at "no chalk" Some days ago, after a long conversation with a friend, I took retrospective notes, recalling what we discussed. (I wrote over 50 items. This is not a small sample.) For parts of the time, we walked; for other parts, we stayed in roughly the same place. Writing the notes, ...
f9d74a89-c2e0-4acd-b186-e822c9d53e38
trentmkelly/LessWrong-43k
LessWrong
Heritability, Behaviorism, and Within-Lifetime RL I’m a firm subscriber to both: * (A) The theory that people’s personalities are significantly predictable from their genes, and mostly independent of how their parents raised them (at least within the typical distribution, i.e. leaving aside cases of flagrant abuse a...
c305fdae-239d-43af-ad84-05e2bfaa63d5
trentmkelly/LessWrong-43k
LessWrong
[AN #172] Sorry for the long hiatus! Listen to this newsletter on The Alignment Newsletter Podcast. Alignment Newsletter is a publication with recent content relevant to AI alignment. Find all Alignment Newsletter resources here. In particular, you can look through this spreadsheet of all summaries that have ever bee...
e549bb8c-4b75-4b53-818e-63e430c31ca4
trentmkelly/LessWrong-43k
LessWrong
Examples of Mitigating Assumption Risk One of the subskills mentioned in Eliezer's Security Mindset post is mitigating assumption risk–that is, the risk of losing utility because some of your assumptions are wrong. There are two main ways to do this: 1. Gain more information about whether your assumptions hold 2. M...
effd36ec-73b4-4a9e-9026-18df30454d9f
trentmkelly/LessWrong-43k
LessWrong
TOMORROW: the largest AI Safety protest ever! Tomorrow, PauseAI and collaborators are putting on the largest AI Safety protest to date, across 7 locations in 6 countries. All are eagerly welcomed!  Your presence at this protest is a rare impact opportunity when in-person volunteering is not fungible with money or int...
564f4e66-fb78-427f-9e6b-2305a6c2dec3
StampyAI/alignment-research-dataset/blogs
Blogs
Another Critique of Effective Altruism I’ve decided to branch out a bit from technical discussions and engage in, as Scott Aaronson would call it, some [metaphysical spouting](http://www.scottaaronson.com/blog/?cat=12). The topic of today is the effective altruism movement. I’m about to be relentlessly critical of it,...
fd4523f5-fc12-4c42-9fae-3fa2774bcfe3
trentmkelly/LessWrong-43k
LessWrong
Simpson's Paradox This is my first attempt at an elementary statistics post, which I hope is suitable for Less Wrong. I am going to present a discussion of a statistical phenomenon known as Simpson's Paradox. This isn't a paradox, and it wasn't actually discovered by Simpson, but that's the name everybody uses for it,...
de7dd82e-2fc4-4ab1-8b2b-e192dbd9fc7a
trentmkelly/LessWrong-43k
LessWrong
Asgardia - The Space Kingdom
c9e2eb69-30b9-4a44-8dfe-9a2c1b981276
trentmkelly/LessWrong-43k
LessWrong
My MATS Summer 2023 experience This post may be interest people who - are interested in getting into AI alignment / the MATS program - are interested in the soft skills that I've found valuable in developing when working on a research project Background In 2023 I was working as a machine learning engineer. I wanted ...
f36fc467-d58a-48e8-9153-86f670639515
trentmkelly/LessWrong-43k
LessWrong
Religion = Cult + Culture [copied in full -- request to develop community knowledge/practices?] Cults are not necessarily bad. Cults provide value. People join them to get things they need which aren’t provided elsewhere. Every cult is a spiritual start-up, doing its best to serve a neglected segment of the populatio...
3440581e-3b6f-49f4-adc3-3909380e9ece
trentmkelly/LessWrong-43k
LessWrong
Vote in the LessWrong review! (LW 2022 Review voting phase) Click here to begin voting. Click here for a general overview of the 2022 review. The final phase of the 2022 review is upon us! We’ve spent 2 weeks nominating posts, and a month reviewing them. And today, we begin the final vote! The LessWrong Review has t...
364c8790-1598-4a4c-ae61-998215ee0e48
StampyAI/alignment-research-dataset/eaforum
Effective Altruism Forum
AI Risk US Presidental Candidate With AI risk getting significant media attention lately (FLI open letter, Eliezer's letter), I think it's plausible that the Overton window has expanded enough where a presidential candidate talking about it won't get immediately laughed off stage. In addition, there seems to be incre...
b9873e0c-5d7f-4121-be47-a735f960b5ad
trentmkelly/LessWrong-43k
LessWrong
Crossing the Rubicon. I heard someone say that 2022 was the year that AI crossed the Rubicon and I thought that was an apt description. To be sure, the actual date when the Rubicon was crossed was before the release of ChatGPT, but for most humans on the planet Earth that was the day the world took notice, if only fo...
397de659-2ccb-4791-926a-c5bc3e504ce7
trentmkelly/LessWrong-43k
LessWrong
Stuff I might do if I had covid In case anyone wants a rough and likely inaccurate guide to what I might do if I had covid to mitigate it, I looked into this a bit recently and wrote notes. It’s probably better than if one’s plan was to do less than a few hours of research, but is likely flawed all over the place and ...
d496734b-7ac2-4824-8bdb-6b4064708e24
trentmkelly/LessWrong-43k
LessWrong
Pomodoros worked during a day: an analysis of Alex’s data Regular readers know that I am always on the lookout for software engineering data.  One search technique is to feed a 'magic' phrase into a search engine, this can locate data hiding in plain sight.  This week the magic phrase: "record of pomodoros" returned p...
09ef1785-e421-4d15-9f97-963bd59628eb
trentmkelly/LessWrong-43k
LessWrong
Some mind experiments This post’s aim is not to prove or disprove the morality of utilitarianism. I only want to raise some provocative questions through mind experiments that shows extreme consequences of some forms of utilitarianism. Let’s assume that Omega has an evil twin: Psi. Psi really likes to play with hu...
8b433b3b-3ea3-4ebb-b5fb-e35f964b9b95
trentmkelly/LessWrong-43k
LessWrong
[Intro to brain-like-AGI safety] 15. Conclusion: Open problems, how to help, AMA (Last revised: July 2024. See changelog at the bottom.) 15.1 Post summary / Table of contents This is the final post of the “Intro to brain-like-AGI safety” post series! Thanks for reading this far! * In Section 15.2, I’ll list seven ...
f5e0ee39-8e34-44f5-9bce-34a7656b63b6
StampyAI/alignment-research-dataset/eaforum
Effective Altruism Forum
Resilience Via Fragmented Power This post explores a big-picture idea - that our civilization is more resilient when power is fragmented, in contrast with concentrated power.  A world with fragmented power cannot be dominated or destroyed by a small group, and so is more resilient.  Concentrated power is what enables ...
6f6e1ee0-ae86-4bcc-aa6b-8e93d4da02b6
trentmkelly/LessWrong-43k
LessWrong
Orthogonality Thesis in layman's terms. I've created a video as an intuition pump for the Orthogonality Thesis and I'd like feedback from the community. Here is the transcript:    Now, if you ask: Why would something so clever want something so stupid, that would lead to death or hell for its creator? you are missin...
57ab999b-3b4e-4153-93da-a0d87c2d8a9b
trentmkelly/LessWrong-43k
LessWrong
A Hypothetical Takeover Scenario Twitter Poll I ran an experimental poll sequence on Twitter a few weeks back, on the subject of what would happen if a hostile ASI (Artificial Superintelligence) tried to take over the world and kill all humans, using only ordinary known techs. The hope was that by exploring this, it ...
ff66c710-7432-4b4c-b777-06aa23e493cb
trentmkelly/LessWrong-43k
LessWrong
Posts with comments crash (bug's probably in comment rendering). The comment rss works. I'm just putting this here in case it helps someone fix the problem, mods; feel free to delete this once things are fixed. It seems posts with comments crash (so the bug's probably in the comment rendering code or something). Post...
a20d7706-954d-4bec-b857-321e955c6f13
trentmkelly/LessWrong-43k
LessWrong
On the nature of help - a framework for helping It is important to define ‘help’ before delving further into the topic. I would like to avoid using dictionary definitions when trying to be concrete about the essence of a word. Albeit sometimes useful, dictionaries strive to provide a short and concise explanation of a...
f4c69c3a-11ff-4b4a-b687-c11377f6735f
awestover/filtering-for-misalignment
Redwood Research: Alek's Filtering Results
id: post123 This is a personal post and does not necessarily reflect the opinion of other members of Apollo Research. If we want to argue that the risk of harm from scheming in an AI system is low, we could, among others, make the following arguments: Detection: If our AI system is scheming, we have good reasons to be...
4d921914-f39f-415f-bb22-f8329e69d4a2
StampyAI/alignment-research-dataset/arbital
Arbital
Group presentation summary(Technical): A presentation $\langle X \mid R \rangle$ of a [group](https://arbital.com/p/3gd) $G$ is a set $X$ of *generators* and a set $R$ of *relators* which are words on $X \cup X^{-1}$, such that $G \cong F(X) / \llangle R \rrangle^{F(X)}$ the [https://arbital.com/p/-normal_closure](htt...
ab4abe8d-03f9-49bd-be31-60c67d70adb5
StampyAI/alignment-research-dataset/arxiv
Arxiv
Integration of Imitation Learning using GAIL and Reinforcement Learning using Task-achievement Rewards via Probabilistic Graphical Model 1 Introduction --------------- The integration of reinforcement learning (RL) and imitation learning (IL) is an important problem that has been studied for a long time in the fiel...
52a4e1f7-6bd6-4ff7-b52a-28699d656ce4
trentmkelly/LessWrong-43k
LessWrong
[SEQ RERUN] Sympathetic Minds Today's post, Sympathetic Minds was originally published on 19 January 2009. A summary (taken from the LW wiki):   > Mirror neurons are neurons that fire both when performing an action oneself, and watching someone else perform the same action - for example, a neuron that fires when you...
c1ba06af-3804-4584-bb65-bc218cc6b2d4
trentmkelly/LessWrong-43k
LessWrong
Meetup : Melbourne social meetup Discussion article for the meetup : Melbourne social meetup WHEN: 16 March 2012 07:00:00PM (+1100) WHERE: see mailing list; Carlton, VIC Melbourne's next social meetup is on the 16th of March - as usual, at Ben's house. See the mailing list for the address - or you can call or text ...
88c41432-d033-49df-ba16-e7c6d4e3e865
StampyAI/alignment-research-dataset/lesswrong
LessWrong
Human beats SOTA Go AI by learning an adversarial policy *See also [article in Financial Times](https://www.ft.com/content/175e5314-a7f7-4741-a786-273219f433a1)* Apparently, a human (Kellin Pelrine, a solid player but not even a Go professional) was able to beat some state-of-the-art Go AIs (KataGo and Leela Zero) b...
2be71727-de41-438e-b959-f585bebaaaed
trentmkelly/LessWrong-43k
LessWrong
Sparse MLP Distillation This is a research report about my attempt to extract interpretable features from a transformer MLP by distilling it into a larger student MLP, while encouraging sparsity by applying an L1 penalty to the activations, as depicted in Figure 1. I investigate the features learned by the distilled M...
9162f139-4880-4ae0-9de6-48d3e3bb9e50
trentmkelly/LessWrong-43k
LessWrong
Training Rationality through Deckbuilders Deckbuilding games are a great way to train rationality. They're usually composed of high-dimensional cost/benefit analyses under uncertainty: you don't know precisely what future enemies you might encounter, or the particular hand you'll be dealt. All you can do is shift prob...
1e024b7d-42bd-4ac7-800d-b967f008ca7f
trentmkelly/LessWrong-43k
LessWrong
How to determine if my sympathetic or my parasympathetic nervous system is currently dominant? I've been reading CFAR Handbook's chapter about Againstness. The chapter's idea is that when your sympathetic nervous system (SNS) is dominant over parasympathetic nervous system (PSNS), your introspection is impaired and yo...
ec1ef24f-3278-4b0f-b043-173c591891cf
trentmkelly/LessWrong-43k
LessWrong
JumpReLU SAEs + Early Access to Gemma 2 SAEs New paper from the Google DeepMind mechanistic interpretability team, led by Sen Rajamanoharan! We introduce JumpReLU SAEs, a new SAE architecture that replaces the standard ReLUs with discontinuous JumpReLU activations, and seems to be (narrowly) state of the art over exi...
73ad1d77-ea2d-488b-b0e9-85951f1fe665
StampyAI/alignment-research-dataset/eaforum
Effective Altruism Forum
AGI x-risk timelines: 10% chance (by year X) estimates should be the headline, not 50%. Artificial General Intelligence (AGI) poses an [existential risk](https://en.wikipedia.org/wiki/Existential_risk_from_artificial_general_intelligence) (x-risk) to all known sentient life. Given the stakes involved -- the whole wor...
16aaebaa-14e4-4dc1-8eb1-16f3e29fc0db
trentmkelly/LessWrong-43k
LessWrong
Operationalizing Newcomb's Problem The standard formulation of Newcomb's problem has always bothered me, because it seemed like a weird hypothetical designed to make people give the wrong answer. When I first saw it, my immediate response was that I would two-box, because really, I just don't believe in this "perfect ...
3a2dbeb9-f4ce-460e-b537-b0c74cee4894
trentmkelly/LessWrong-43k
LessWrong
Portable Chargers are Great Portable chargers are great and underrated. For $20 you can get a little box that can charge a phone ~5x or laptop ~1x. [1] You can pay a bit more for one with built-in wall-charging and cables. It's great throwing one in my backpack and knowing I don't have to ration or worry, especially w...
0a8fd769-6f46-4295-a6c6-b9a7baa1a416
trentmkelly/LessWrong-43k
LessWrong
How dangerous is human-level AI? I don't expect anything here to be original; this is just me thinking things through, and perhaps this post will be useful as a relatively stand-alone argument. Typically, you can envision many many more future states than there are future states you could bring about. You can imagine...
9630b74c-f516-479c-8572-c03e163bbbfc
trentmkelly/LessWrong-43k
LessWrong
Technology path dependence and evaluating expertise Muireall I often wonder if I'd be able to tell if a technology were uncompetitive for fundamental viability reasons or for reasons of path dependence—if the ultimate performance would be a winner once we'd worked out a few trillion dollars of kinks, but we never go d...
f73a7742-3979-491a-82b6-23cb6b257fef
trentmkelly/LessWrong-43k
LessWrong
Instead of "I'm anxious," try "I feel threatened" cw: teaching to learn I have a long history with anxiety, and I’m pretty good at noticing when it’s happening. The problem is that I’m always anxious. Noticing anxiety doesn’t snap me out of anxiety– in fact, it often produces meta-anxiety, anxiety about feeling anxio...
5a9d6101-1657-4828-9ca0-f4b4dcb46b23
awestover/filtering-for-misalignment
Redwood Research: Alek's Filtering Results
id: post369 In this post, I’ll talk about trustworthiness , a concept similar to scheming and deceptive alignment. I prefer this concept, as I find it A) is directly important in practice, B) more readily distinguishes between capability for scheming and in fact scheming , and C) is clearer how to evaluate for. I’ll a...
d30e4ab7-0113-4140-ac10-ef795059b496
trentmkelly/LessWrong-43k
LessWrong
Grading on Word Count One of the most common forms of schoolwork is the writing assignment. The idea is that, if a student understands the subject, they should be able to demonstrate that by writing about the subject. These kinds of assignments are widely considered to improve comprehension and writing skill in additi...
35c4e9ed-e165-4b9a-b995-818670733e3d
trentmkelly/LessWrong-43k
LessWrong
Summary: "How to Read Books and Write Essays" by OSP's Blue This post is part of my "research" series. Blue. How to Read Books and Write Essays. Overly Sarcastic Productions, 2019. https://www.youtube.com/watch?v=E3v8-j1znXM. Why This got recommended at the end of Red's video on how to do research and can be consid...
bedfe035-1843-422d-b550-11d2ccc72f8f
trentmkelly/LessWrong-43k
LessWrong
"Other people are wrong" vs "I am right" I’ve recently been spending some time thinking about the rationality mistakes I’ve made in the past. Here’s an interesting one: I think I have historically been too hasty to go from “other people seem very wrong on this topic” to “I am right on this topic”. Throughout my life,...
9b492d9d-10b3-4ed9-900b-f71fe99152f0
trentmkelly/LessWrong-43k
LessWrong
ARC paper: Formalizing the presumption of independence (I did not have anything to do with this paper and these are just my own takes.) The Alignment Research Center recently published their second report, Formalizing the presumption of independence. While it's not explicitly about AI alignment, it's probably still i...
07dc0cdf-0d00-4f98-bfa8-e639a6352e05
StampyAI/alignment-research-dataset/youtube
Youtube Transcripts
Concrete Open Problems in Mechanistic Interpretability: Neel Nanda at SERI MATS kill yeah that'd be too um all right so for the people I've not met before um hi I'm Neil I used to work at anthropic doing mechanistic interpretability of Transformers I have spent the last couple of months as an independent researcher do...
a7c0c57a-2a6b-4543-8cfa-55167a99bd79
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
AI Services as a Research Paradigm Introduction ============ I would hereby like to advertise a document [Systems of Services as a Paradigm for AI Alignment](https://docs.google.com/document/d/1Jk2GgJnF9pLIQqg9hgX0Tvdukq6MR2qmXW3FyvqiiUg/edit?usp=sharing). I hope that it can serve as a starting point for investigati...
9e154aae-fb5b-42fb-b187-2f20d6917304
trentmkelly/LessWrong-43k
LessWrong
Division of cognitive labour in accordance with researchers' ability The variance in productivity between people is much greater in some areas than in others. The most productive cleaner is probably less than twice as productive as the average cleaner, whereas the most productive entrepreneur arguably is many times mo...
9880b18f-a5e6-4bf5-b7ec-602939f57902
StampyAI/alignment-research-dataset/lesswrong
LessWrong
Optimization and the Intelligence Explosion Among the topics I haven’t delved into here is the notion of an optimization process. Roughly, this is the idea that your power as a mind is your ability to hit small targets in a large search space—this can be either the space of possible futures (planning) or the space of ...
5ae2c4ca-2ec2-4b68-84e2-9fb7aaa4d49e
trentmkelly/LessWrong-43k
LessWrong
Bayesian Flame There once lived a great man named E.T. Jaynes. He knew that Bayesian inference is the only way to do statistics logically and consistently, standing on the shoulders of misunderstood giants Laplace and Gibbs. On numerous occasions he vanquished traditional "frequentist" statisticians with his superior ...
266fa1c3-4f32-4442-8ee1-a7e37ebb6734
trentmkelly/LessWrong-43k
LessWrong
Against utility functions I think we should stop talking about utility functions. In the context of ethics for humans, anyway. In practice I find utility functions to be, at best, an occasionally useful metaphor for discussions about ethics but, at worst, an idea that some people start taking too seriously and which ...
4908a10e-2748-4d2b-a82e-6b81e19c27ce
trentmkelly/LessWrong-43k
LessWrong
Fiction: A Setting Justifying the Epistemic Aggressiveness Of A Religion Stand-in I've just been thinking (a dangerous pastime, I know) and remembering my religious days, and one of the things that made me so anxious back then, and made adopting atheism so gratifying, what the sheer drought of information. I had so ma...
35bb6681-a9f8-4d8d-bbcf-8632fefa6318
trentmkelly/LessWrong-43k
LessWrong
Oracles: reject all deals - break superrationality, with superrationality tl;dr Acausal trade between Oracles is a problem even if our Oracle is a causal decision theorist. To avoid this, we need an AFDT Oracle - an Oracle that reasons partially like a functional decision theorist, partially like a causal decision th...
f5f7c009-6ecf-4fde-b82b-8ce894a319fe
trentmkelly/LessWrong-43k
LessWrong
Justifying (Improper) Priors EDIT: I tested this idea in a simulation and it (un)fortunately failed. ---------------------------------------- This post is an attempt to combine some parts of frequentism with some parts of Bayesianism. Unfortunately, there is a good deal of probability theory. I assume that "impr...
904bc7e2-5a98-4356-98ea-9bb02a19d575
trentmkelly/LessWrong-43k
LessWrong
An Invitation to the Guild of the Rose — Alpha (formerly the Guild of Servants) Salutatioиs, You are cordіally invited to apply to the Guild of the Rose (edit: formerly the Guild of Servants), an organization dedicated to more effective tһinkіng and ӏiving, and providing support to our members and communities in this...
9676976a-9a5f-4774-9440-0e44ddbcaa16
trentmkelly/LessWrong-43k
LessWrong
What are information hazards? This post was written for Convergence Analysis. Cross-posted to the EA Forum. The concept of information hazards is highly relevant to many efforts to do good in the world (particularly, but not only, from the perspective of reducing existential risks). I’m thus glad that many effective ...
5ed7a52c-854c-4436-b075-81b67f429352
trentmkelly/LessWrong-43k
LessWrong
Meetup : LessWrong Scotland Discussion article for the meetup : LessWrong Scotland WHEN: 12 April 2015 02:00:00PM (+0000) WHERE: Biblos, 1 Chambers Street, Edinburgh EH1 1HR We're going to meet in Edinburgh this time! I don't know the layout of the venue, but we'll be recognisable by books, stuffed owls, signs with...
573a5245-e02b-4234-9257-e0613c0f193f
trentmkelly/LessWrong-43k
LessWrong
Announcing Data Secrets Lox: An Alternative to Slate Star Codex Open Threads In light of the shutdown of Slate Star Codex, and in particular, the lack of new SSC Open Threads, much of the SSC community has moved to a new forum: Data Secrets Lox. Data Secrets Lox is a sort of "permanent open thread" for the SSC readers...
780bc200-538a-4477-ac38-653655ff1840
trentmkelly/LessWrong-43k
LessWrong
Ideas for rationalist meetup topics (From Zvi Mowshowitz, leader of the New York group, and based on his experience) Category 1: Discussions - The Base A: Less Wrong topics. Usually recent posts. Often big hits. B: Rationalist group planning and organization. Meta-topics. Dealing with group issues on occasion. C: Sp...
2be24a35-2bdd-4774-9b05-85eab350b5a1
trentmkelly/LessWrong-43k
LessWrong
Vague concepts, family resemblance and cluster properties My definition of a "vague concept" (A): * It creates a group of things that share family resemblance or "family difference". * It doesn't have any meaning outside of context. But in a specific context it obtains specific meaning. * It's very hard to unders...
2e24daf6-116d-41ca-ad24-2d2cd1e850d4
trentmkelly/LessWrong-43k
LessWrong
Risk Map of AI Systems Joint work with Jan Kulveit. Many thanks to the various people who gave me feedback. This text describes a model of the space of possible AI systems, as we try to navigate it safely towards AGI.[1] Its primary goal is to provide a simple mental model that people with disagreeing opinions about ...
5cee260e-fae7-4796-8844-4fc9e8a0beaf
trentmkelly/LessWrong-43k
LessWrong
What a 20-year-lead in military tech might look like I've spent way too much time speculating about near-future military technology. Here is a list of technologies I think are likely to be important to a conventional war in, say, 2040. If you think this list shouldn't be taken seriously because I don't have official ...
6e3fc89a-27e7-4742-a03a-08f586ef9bd6
trentmkelly/LessWrong-43k
LessWrong
AI, Intellectual Property, and the Techno-Optimist Revolution The emergence of generative Artificial Intelligence (“AI”) is putting computers at odds with creators and raising novel questions about the future of Intellectual Property (“IP”) law. Though scholars are scrambling to anticipate the legal consequences of AI...
eafb2849-7f15-4adf-aac6-810e104d868f
StampyAI/alignment-research-dataset/eaforum
Effective Altruism Forum
My current take on existential AI risk [FB post] *Cross-posted from something I* [*wrote on Facebook*](https://www.facebook.com/permalink.php?story_fbid=pfbid02KA372m29bwMPdiPPEYmhjDEfLP2ayyZPMJrgJxoBMPW5iyZFVVchbCbs53dTpMo6l&id=100010608396052)*. I doubt there's really anything new or important here for an EA audienc...
6fc8f7ef-40a3-4282-b967-380fe3359d5d
StampyAI/alignment-research-dataset/arxiv
Arxiv
Humans learn too: Better Human-AI Interaction using Optimized Human Inputs Introduction ------------ In an industrial setting, production processes involving humans and robots are constantly improved to reduce errors and improving efficiency. Improvements target measures towards humans and machines alike in a holis...
a9318986-880d-498c-a3f1-c597ceb4319f
trentmkelly/LessWrong-43k
LessWrong
How likely are malign priors over objectives? [aborted WIP] This was written as part of an entry to the FTX AI competition, but I presume that's off now and I can't justify spending more time on it. Neither my thinking nor my writing is polished here. I believe some of the arguments are original. In many (though not ...
e79035e5-511d-4ab9-a7fb-971ac0aed22b
trentmkelly/LessWrong-43k
LessWrong
The biological intelligence explosion Summary:  Human genetic engineering could lead to intelligence enhancement that leads to genetic engineers who are better at genetic engineering (and research on pathways to improving intelligence).  Which leads to a continuing process of greater and greater intelligence. This ite...
ceb4299a-287b-4e7a-81ea-4c2c2b8339e1
trentmkelly/LessWrong-43k
LessWrong
Learning to make better decisions None
3413cc84-dc36-44d9-b27a-b18e5909ed70
trentmkelly/LessWrong-43k
LessWrong
Defusing Hate: A Strategic Communication Guide to Counteract Dangerous Speech
701b8d93-5558-42b2-92ce-7e98e96c060e
StampyAI/alignment-research-dataset/alignmentforum
Alignment Forum
Mech Interp Project Advising Call: Memorisation in GPT-2 Small I've recently been having advising calls with REMIX teams (Redwood's interpretability sprint) trying to give advice & feedback on projects. As an experiment, I've published a recording of one advising call (with Tessa Barton & Kushal Jain on memorisation i...
2ba684ee-d828-4441-8b5b-c607b3741e5b
trentmkelly/LessWrong-43k
LessWrong
Rational Feed Note: I am trying out a weekly feed.  ===Highly Recommended Articles: Superintelligence Risk Project: Conclusion by Jeff Kaufman - "I'm not convinced that AI risk should be highly prioritized, but I'm also not convinced that it shouldn't. Highly qualified researchers in a position to have a good sense ...
f9bd05ad-27d4-4247-91e8-68a6e13f79fb
StampyAI/alignment-research-dataset/special_docs
Other
Small Data’s Big AI Potential September 2021 Small Data’s Big AI Potential CSET Issue Brief AUTHORS Husanjot Chahal Helen Toner Ilya Rahkovsky Center for Security and Emerging Technology | 1 Executive Summary This issue brief provides an introduction to and overview of “small data” artificial intelligence approa...
093c248f-6af7-456d-aa7c-3a883e1194c4
trentmkelly/LessWrong-43k
LessWrong
How Many of Me Are There? This post is not about many worlds. It is somewhat disjointed, but builds to a single point. If an AI was asked today how many human individuals populate this planet, it may not return a number the several-billions range. In fact I’d be willing to bet it’d return a number in the tens of thou...