text stringlengths 300 320k | source stringlengths 52 154 |
|---|---|
# The Fragility of Life Hypothesis and the Evolution of Cooperation
> *This part 2 in a 3-part sequence summarizes my book,* [*The Darwinian Trap*](https://www.amazon.com/Darwinian-Trap-Evolutionary-Explain-Threaten/dp/0593594053)*, (see* [*part 1*](https://www.lesswrong.com/posts/q3YmKemEzyrcphAeP/darwinian-traps-and... | https://www.lesswrong.com/posts/kBG6bNckhZeamQJsQ/the-fragility-of-life-hypothesis-and-the-evolution-of |
# Against Explosive Growth
*Epistemic status: hot take, the core of which was written in ~10mn.*
Assumptions/Definitions
-----------------------
I'm conditioning on AGI sometime within next 3~40 yrs.
"Explosive growth" ≝ the kind of crazy 1yr or 6 month GDP (or GWP) doubling times Shulman talks about [here](https:/... | https://www.lesswrong.com/posts/4cjxj5yEFfYnBH7sy/against-explosive-growth |
# Executable philosophy as a failed totalizing meta-worldview
*(this is an expanded, edited version of an [x.com post](https://twitter.com/jessi_cata/status/1831152278004498814))*
It is easy to interpret Eliezer Yudkowsky's main goal as creating a friendly AGI. Clearly, he has failed at this goal and has little hope ... | https://www.lesswrong.com/posts/Bi4yt7onyttmroRZb/executable-philosophy-as-a-failed-totalizing-meta-worldview |
# on Science Beakers and DDT
## tech trees
There's a series of strategy games called [Civilization](https://www.youtube.com/results?search_query=let%27s+play+civilization). In those games, the player controls a country which grows and develops over thousands of years, and science is one of the main types of progress.... | https://www.lesswrong.com/posts/MXpiDr63FJaSeEwFg/on-science-beakers-and-ddt |
# How to Fake Decryption
*{Epistemic Status: mostly just writing up an idea I came up with a while ago. I’ve done non-trivial coursework in cryptography, so I think I’ve avoided obvious errors, though there may be non-obvious ones. Consult an expert before using.}*
Suppose you knew that someone was spying on your enc... | https://www.lesswrong.com/posts/nAgEBsy9bPoKfjExx/how-to-fake-decryption |
# What program structures enable efficient induction?
*previously:* [*My decomposition of the alignment problem*](https://www.lesswrong.com/posts/yLpWBKBnRfSFbWMct/my-decomposition-of-the-alignment-problem)
A simple model of meta/continual learning
=========================================
In the framework of solomo... | https://www.lesswrong.com/posts/nyooPYkvTcCKCNxbL/what-program-structures-enable-efficient-induction |
# Why Swiss watches and Taylor Swift are AGI-proof
The post [What Other Lines of Work are Safe from AI Automation?](https://www.lesswrong.com/posts/tgrdfvN8f3WzvZGcr/what-other-lines-of-work-are-safe-from-ai-automation) from Roger Dearnaley examined candidate job categories "for which being an actual real human is a p... | https://www.lesswrong.com/posts/oiymn5qNCLc9SEGs2/why-swiss-watches-and-taylor-swift-are-agi-proof |
# Why Reflective Stability is Important
Imagine you have the optimal AGI source code $O$. If you run $O$ the resulting AGI process will not self-modify.[^42imbsrhqz4] You get a non-self-modifying AI for free.
This would be good because a non-self-modifying AI is much easier to reason about.
Imagine some non-optimal... | https://www.lesswrong.com/posts/iWfde3aSDnyMnbBxT/why-reflective-stability-is-important |
# A bet for Samo Burja
I'm listening to Samo Burja talk on the Cognitive Revolution podcast with Nathan Labenz.
Samo said that he would bet that AGI is coming perhaps in the next 20-50 years, but not in the next 5.
I will take that bet. I can't afford to make an impressively large bet because my counterfactual income... | https://www.lesswrong.com/posts/48ZNJxbGZ5w7gEuvi/a-bet-for-samo-burja |
# Conflating value alignment and intent alignment is causing confusion
*Epistemic status: I think something like this confusion is happening often. I'm not saying these are the only differences in what people mean by "AGI alignment".*
### **Summary:**
Value alignment is better but probably harder to achieve than *p... | https://www.lesswrong.com/posts/83TbrDxvQwkLuiuxk/conflating-value-alignment-and-intent-alignment-is-causing-1 |
# instruction tuning and autoregressive distribution shift
*\[Note: this began life as a "Quick Takes" comment, but it got pretty long, so I figured I might as well convert it to a regular post.\]*
In LM **training**, every token provides new information about "the world beyond the LM" that can be used/"learned" in-c... | https://www.lesswrong.com/posts/JeuAk53QfWauaGWfS/instruction-tuning-and-autoregressive-distribution-shift |
# What is SB 1047 *for*?
Emmett Shear [asked on twitter](https://x.com/Raemon777/status/1831748181392257126):
> I think SB 1047 has gotten much better from where it started. It no longer appears actively bad. But can someone who is pro-SB 1047 explain the specific chain of causal events where they think this bill bec... | https://www.lesswrong.com/posts/ehYuLsnoFch5QcaWD/what-is-sb-1047-for |
# AI x Human Flourishing: Introducing the Cosmos Institute
[**Watch the Cosmos Mission Video**](https://www.youtube.com/watch?v=2Zbg5iuDxOg&list=PLVaBJ6bgxgRZYwxCOOjvrCQ2Fqioj5p_b)
[Earlier this year](h... | https://www.lesswrong.com/posts/ts4dNmiHX3Xjvs2rW/ai-x-human-flourishing-introducing-the-cosmos-institute |
# Distinguish worst-case analysis from instrumental training-gaming
*"I" refers to Olli. Thanks to Buck, Ryan Greenblatt, Fabien Roger and Zach Stein-Perlman for feedback.*
The word “schemer” is used to refer to multiple things, and in particular to the following two concepts:
Schemer as a **power-motivated instrume... | https://www.lesswrong.com/posts/h2CKojGfWfB44ryT2/distinguish-worst-case-analysis-from-instrumental-training-2 |
# Investigating Sensitive Directions in GPT-2: An Improved Baseline and Comparative Analysis of SAEs
*Experiments and write-up by Daniel, with advice from Stefan. Github repo for the work can be found* [*here*](https://github.com/dajale423/error_pathology)*.*
*Update (October 31, 2024): Our* [*paper*](https://arxiv.o... | https://www.lesswrong.com/posts/dS5dSgwaDQRoWdTuu/investigating-sensitive-directions-in-gpt-2-an-improved |
# Backdoors as an analogy for deceptive alignment
ARC has released a paper on [**Backdoor defense, learnability and obfuscation**](https://arxiv.org/abs/2409.03077) in which we study a formal notion of backdoors in ML models. Part of our motivation for this is an analogy between backdoors and _deceptive alignment_, th... | https://www.lesswrong.com/posts/efwcZ35LwS6HgFcN8/backdoors-as-an-analogy-for-deceptive-alignment |
# Adam Optimizer Causes Privileged Basis in Transformer LM Residual Stream
Diego Caples ([diego@activated-ai.com](mailto:diego@activated-ai.com))
Rob Neuhaus ([rob@activated-ai.com](mailto:rob@activated-ai.com))
Introduction
============
In principle, neuron activations in a transformer-based language model residua... | https://www.lesswrong.com/posts/yrhu6MeFddnGRSLtQ/adam-optimizer-causes-privileged-basis-in-transformer-lm |
# Is this voting system strategy proof?
My voting system works like this. Each voter expresses their preferences for all candidates on a real numbered utility scale.
Then a Maximal lottery takes place over all lotteries over candidates. [https://en.wikipedia.org/wiki/Maximal_lotteries](https://en.wikipedia.org/wiki/... | https://www.lesswrong.com/posts/m6qoRhwyayfRQGbh5/is-this-voting-system-strategy-proof |
# Fun With CellxGene
[
](https://substackcdn.com/image/fetch/f_auto,q_auto:good,fl_progressive:s... | https://www.lesswrong.com/posts/3jXqvHej48Q3t8yjv/fun-with-cellxgene |
# Excerpts from "A Reader's Manifesto"
[“A Reader’s Manifesto”](https://archive.is/DtsmU) is a July 2001 *Atlantic* piece by B.R. Myers that I've returned to many times. He complains about the inaccessible pretension of the highbrow literary fiction of his day. The article is mostly a long list of critiques of various... | https://www.lesswrong.com/posts/NiTuqw6LRFifqzRRH/excerpts-from-a-reader-s-manifesto |
# Pay Risk Evaluators in Cash, Not Equity
Personally, I suspect the alignment problem is hard. But even if it turns out to be easy, survival may still require getting at least the absolute basics right; currently, I think we're mostly failing even at that.
Early discussion of AI risk often focused on debating the via... | https://www.lesswrong.com/posts/sMBjsfNdezWFy6Dz5/pay-risk-evaluators-in-cash-not-equity |
# Jonothan Gorard:The territory is isomorphic to an equivalence class of its maps
Jonothan Gorard is a mathematician for Wolfram Research and one of the cofounders of the wolfram physics project. I recently came across this twitter thread from him and found it particularly insightful:
> **Jonothan Gorard:** The terri... | https://www.lesswrong.com/posts/cfq9ReeAeeT3tz3MS/jonothan-gorard-the-territory-is-isomorphic-to-an |
# That Alien Message - The Animation
*Our new video is an adaptation of* [*That Alien Message*](https://www.lesswrong.com/posts/5wMcKNAwB6X4mp9og/that-alien-message)*, by* [*@Eliezer Yudkowsky*](https://www.lesswrong.com/users/eliezer_yudkowsky?mention=user)*. This time, the text has been significantly adapted, so I i... | https://www.lesswrong.com/posts/Q9omyL3qooXdjnyZn/that-alien-message-the-animation |
# I want a good multi-LLM API-powered chatbot
Chatbots are pretty useful to me, but the pro plans are expensive. I don't want to spend 20$/month here and there. However, I couldn't find a free chatbot app with at least OpenAI and Anthropic API support with a similar level of functionality as the online gpt or claude c... | https://www.lesswrong.com/posts/JTLquqW9GNtMYBrsZ/i-want-a-good-multi-llm-api-powered-chatbot |
# My Number 1 Epistemology Book Recommendation: Inventing Temperature
In [my last post](https://epistemologicalfascinations.substack.com/p/my-model-of-epistemology), I wrote that no resource out there exactly captured my model of epistemology, which is why I wanted to share a half-baked version of it.
But I do have o... | https://www.lesswrong.com/posts/TbaCa7sY3GxHBcXTd/my-number-1-epistemology-book-recommendation-inventing |
# Fictional parasites very different from our own
*Note: this is a fictional story. Heavily inspired by SSC’s similar posts on* [*fictional legal systems*](https://slatestarcodex.com/2020/03/30/legal-systems-very-different-from-ours-because-i-just-made-them-up/) *and* [*fictional banned drugs*](https://slatestarcodex.... | https://www.lesswrong.com/posts/eQhDXk57ZgE6Sa8dK/fictional-parasites-very-different-from-our-own |
# On Fables and Nuanced Charts
*Written by Spencer Greenberg & Amber Dawn Ace for* [*Asimov Press*](https://www.asimov.press/p/charts)*.*
In 1994, the U.S. Congress passed the largest crime bill in U.S. history, called the Violent Crime Control and Law Enforcement Act. The bill allocated billions of dollars to build ... | https://www.lesswrong.com/posts/dwvusp6wr9GfZYkrK/on-fables-and-nuanced-charts |
# Lighthaven Sequences Reading Group #2 (Tuesday 09/17)
Come get old-fashioned with us, and let's read the sequences at Lighthaven! We'll show up, mingle, do intros, and then split off into randomized groups for some sequences discussion. Please do the reading beforehand - it should be no more than 20 minutes of readi... | https://www.lesswrong.com/events/kFnfQwpca9MNHwGp6/lighthaven-sequences-reading-group-2-tuesday-09-17 |
# Pollsters Should Publish Question Translations
[Pew recently commissioned](https://www.pewresearch.org/short-reads/2024/08/26/many-israelis-say-social-media-content-about-the-israel-hamas-war-should-be-censored/) Gallup to run a poll in Israel on attitudes to social media censorship. They found high support for bann... | https://www.lesswrong.com/posts/28eyDuZ6FKrHpQ2FK/pollsters-should-publish-question-translations |
# Has Anyone Here Consciously Changed Their Passions?
I'm sorry if this question is a bit fuzzy, maybe my understanding of a "passion" isn't the most coherent. But, I'd like to know if anyone has made a successful effort to get themselves to be passionate about doing X where they were not passionate about X before. I ... | https://www.lesswrong.com/posts/CQjxKzyezt2autG9N/has-anyone-here-consciously-changed-their-passions |
# Building an Inexpensive, Aesthetic, Private Forum
**Reason for Writing**: I want to gather information for a friend who is a professional in the behavioral sciences.
**Epistemic Status:** Less than 2 hours googling how to build a forum, and a couple youtube videos
**Requirements:**
1. **Aesthetic**
2. **Inexpe... | https://www.lesswrong.com/posts/u5ZsKovMFGpLjPAKb/building-an-inexpensive-aesthetic-private-forum |
# My takes on SB-1047
I recently decided to sign a [letter of support](https://calltolead.org/) for SB 1047. Before deciding whether to do so, I felt it was important for me to develop an independent opinion on whether the bill was good, as opposed to deferring to the opinions of those around me, so I read through the... | https://www.lesswrong.com/posts/kBg5eoXvLxQYyxD6R/my-takes-on-sb-1047 |
# AI forecasting bots incoming
In a recent appearance on *Conversations with Tyler*, famed political forecaster Nate Silver expressed skepticism about AIs replacing human forecasters in the near future. When asked how long it might take for AIs to reach superhuman forecasting abilities, Silver replied: “15 or 20 \[yea... | https://www.lesswrong.com/posts/4kuXNhPf9FBwok7tK/ai-forecasting-bots-incoming |
# Simon DeDeo on Explore vs Exploit in Science
Good news everyone: I’m getting into podcasting!
As part of my investigation into how scientific paradigms are created, I hosted a talk by [Simon DeDeo, Professor Of Many Topics](https://sites.santafe.edu/~simon/) at Carnegie Mellon and the Santa Fe Institute, recorded ... | https://www.lesswrong.com/posts/guxhjoiDgEFRtqAb2/simon-dedeo-on-explore-vs-exploit-in-science |
# Physical Therapy Sucks (but have you tried hiding it in some peanut butter?)
The worst part about physical therapy is not knowing if it’s working. The timescale for improvement is dreadfully long and the day-to-day changes are imperceptible.
65% of people who start physical therapy [fail to adhere](https://pubmed.n... | https://www.lesswrong.com/posts/rZxGTFdnXqX282NZs/physical-therapy-sucks-but-have-you-tried-hiding-it-in-some |
# Economics Roundup #3
I am posting this now largely because it is the right place to get in discussion of unrealized capital gains taxes and other campaign proposals, but also there is always plenty of other stuff going on. As always, remember that there are plenty of really stupid proposals always coming from all si... | https://www.lesswrong.com/posts/cc9fXQLzAc52kMBHx/economics-roundup-3 |
# The Best Lay Argument is not a Simple English Yud Essay
*Epistemic status: these are my own opinions on AI risk communication, based primarily on my own instincts on the subject and discussions with people less involved with rationality than myself. Communication is highly subjective and I have not rigorously A/B te... | https://www.lesswrong.com/posts/CZQYP7BBY4r9bdxtY/the-best-lay-argument-is-not-a-simple-english-yud-essay |
# AI #80: Never Have I Ever
(This was supposed to be on Thursday but I forgot to cross-post)
Will AI ever make art? Fully do your coding? Take all the jobs? Kill all the humans?
Most of the time, the question comes down to [a general disagreement about AI capabilities](https://thezvi.substack.com/p/ai-and-the-techno... | https://www.lesswrong.com/posts/x77vDAzosxtwJoJ7e/ai-80-never-have-i-ever |
# Formalizing the Informal (event invite)
**Formalizing the Informal**
============================
One way to view MIRI's Agent Foundations research is that it saw the biggest problem in AI safety as "human preferences are informal, but we need to somehow get formal guarantees about them" -- and so, in response, it ... | https://www.lesswrong.com/posts/cd4SwAGLav4WAuxeN/formalizing-the-informal-event-invite |
# A necessary Membrane formalism feature
Summary
========
Consider a Membrane that is supposed to protect Steve. An AI wants to hurt Steve but does not want to pierce Steve's Membrane. The Membrane ensures that there is zero effect on predictions of things inside the membrane. The AI will never take any action that ... | https://www.lesswrong.com/posts/GJbejYAWutHFbPdpT/a-necessary-membrane-formalism-feature |
# Reformative Hypocrisy, and Paying Close Enough Attention to Selectively Reward It.
People often attack frontier AI labs for "hypocrisy" when the labs admit publicly that AI is an extinction threat to humanity. Often these attacks ignore the difference between various kinds of hypocrisy, some of which are good, incl... | https://www.lesswrong.com/posts/oRNqZwjgWSGfFdyND/reformative-hypocrisy-and-paying-close-enough-attention-to |
# Seeking Mechanism Designer for Research into Internalizing Catastrophic Externalities
### **TL;DR **
**I need help finding/developing a mech that can reliably elicit honest and effortful risk-estimates from frontier AI labs regarding their models, risk-estimates to be used in risk-priced premiums that labs then pay... | https://www.lesswrong.com/posts/67jgcEwrxh8B3gg2n/seeking-mechanism-designer-for-research-into-internalizing |
# Refactoring cryonics as structural brain preservation
I first learned about cryonics when I read Eliezer and Robin's posts about it on Overcoming Bias years ago. I got cryopilled. Somewhat amazingly to me, I'm now a researcher in this field. So I thought this community might be interested to know that I was one of s... | https://www.lesswrong.com/posts/PG4D4CSBHhijYDvSz/refactoring-cryonics-as-structural-brain-preservation |
# Programming Refusal with
Conditional Activation Steering
*For full content, refer to the arXiv preprint at* [*https://arxiv.org/abs/2409.05907*](https://arxiv.org/abs/2409.05907)*. This post is a lighter, 15-minute version.*
Abstract
========
* Existing activation steering methods alter LLM behavior indiscrimin... | https://www.lesswrong.com/posts/HiG479grQtkut7svb/programming-refusal-with-conditional-activation-steering |
# When can I be numerate?
I recently started working in a setting making physical products, where I discovered that I had a whole nother section of my brain purposed for physical interactions between things and when things might fall on other things or melt or fall apart or shatter or scar or any number of other thing... | https://www.lesswrong.com/posts/tgQftJRqSdCekBcvG/when-can-i-be-numerate |
# AI #81: Alpha Proteo
Following up on Alpha Fold, DeepMind has moved on to Alpha Proteo. We also got a rather simple prompt that can create a remarkably not-bad superforecaster for at least some classes of medium term events.
We did not get a new best open model, because that turned out to be a scam. And we don’t ha... | https://www.lesswrong.com/posts/YMaTA2hX6tSBJWnPr/ai-81-alpha-proteo |
# Open Problems in AIXI Agent Foundations
*Originally stand-alone, added to the AIXI agent foundations sequence, underlying research supported by the LTFF.*
I believe that the theoretical foundations of the AIXI agent and variations are a surprisingly neglected and high leverage approach to agent foundations research... | https://www.lesswrong.com/posts/MvfD4tmzyuCYFqB2f/open-problems-in-aixi-agent-foundations |
# How to Give in to Threats (without incentivizing them)
[*(available as a paper.)*](https://contact.ms/How_To_Give_In_To_Threats.pdf)
TL;DR: using a simple mixed strategy, [LDT](https://arbital.greaterwrong.com/p/ldt_intro_compsci?l=5d6) can give in to threats, ultimatums, and commitments - while incentivizing coope... | https://www.lesswrong.com/posts/TXbFFYpNWDmEmHevp/how-to-give-in-to-threats-without-incentivizing-them |
# OpenAI o1
It's more capable and better at using lots of inference-time compute via long (hidden) chain-of-thought.
**OpenAI pages: **[Learning to Reason with LLMs](https://openai.com/index/learning-to-reason-with-llms/), [o1 System Card](https://openai.com/index/openai-o1-system-card/), [o1 Hub](https://openai.com... | https://www.lesswrong.com/posts/bhY5aE4MtwpGf3LCo/openai-o1 |
# Contra papers claiming superhuman AI forecasting
*\[Conflict of interest disclaimer: We are *[*FutureSearch*](http://futuresearch.ai)*, a company working on AI-powered forecasting and other types of quantitative reasoning. If thin LLM wrappers could achieve superhuman forecasting performance, this would obsolete a l... | https://www.lesswrong.com/posts/uGkRcHqatmPkvpGLq/contra-papers-claiming-superhuman-ai-forecasting |
# My career exploration: Tools for building confidence
*Crossposting from* [*my blog*](https://lynettebye.com/blog/2024/7/29/my-career-exploration-tools-for-building-confidence)
I did a major career review during 2023. I’m sharing it now because:
1. I think it’s a good case study for [iterated depth decision-making... | https://www.lesswrong.com/posts/kcCNMk26zCvkD8yuf/my-career-exploration-tools-for-building-confidence |
# Evidence against Learned Search in a Chess-Playing Neural Network
Introduction
------------
There is a [new paper](https://arxiv.org/pdf/2406.00877) and [lesswrong post](https://www.lesswrong.com/posts/GfwdBoaLw3ef3zBqe/) about "learned look-ahead in a chess-playing neural network". This has long been a [research i... | https://www.lesswrong.com/posts/hPGw7hWYbYyvDcqYK/evidence-against-learned-search-in-a-chess-playing-neural |
# AI, centralization, and the One Ring
People thinking about the future of AI sometimes talk about a single project ‘getting there first’ — achieving AGI, and leveraging this into a decisive strategic advantage over the rest of the world.
I claim we should be worried about this scenario. That doesn’t *necessarily* me... | https://www.lesswrong.com/posts/dLDczS6Xi59GNKd3H/ai-centralization-and-the-one-ring |
# "Real AGI"
I see "AGI" used for everything from existing LLMs to superintelligence, and massive resulting confusion and illusory disagreements. I finally thought of a term I like for what I mean by AGI. It's an acronym that's also somewhat intuitive without reading the definition:
***Reasoning, Reflective Entities ... | https://www.lesswrong.com/posts/YW249knFccwATGxki/real-agi |
# The Great Data Integration Schlep
[](https://substackcdn.com/image/fetch/f_auto,q_auto:good,fl_prog... | https://www.lesswrong.com/posts/7L8ZwMJkhLXjSa7tD/the-great-data-integration-schlep |
# How difficult is AI Alignment?
*This article revisits and expands upon the AI alignment difficulty scale, a framework for understanding the increasing challenges of aligning artificial intelligence systems with human values. *
*We explore how alignment difficulties evolve from simple goal misalignment to complex sc... | https://www.lesswrong.com/posts/Wz42Ae2dQPdpYus98/how-difficult-is-ai-alignment |
# Why I'm bearish on mechanistic interpretability: the shards are not in the network
Once upon a time, the sun let out a powerful beam of light which shattered the world. The air and the liquid was split, turning into body and breath. Body and breath became fire, trees and animals. In the presence of the lightray, any... | https://www.lesswrong.com/posts/XNnPsTKgiQtsypPfg/why-i-m-bearish-on-mechanistic-interpretability-the-shards |
# Keeping it (less than) real: Against
ℶ₂ possible people or worlds
Epistemic status and trigger warnings: *Not rigorous in either math or physics. Not proof-read by any third party yet. May contain original research. Lengthy. Some sections are tongue-in-cheek. Anthrophics. Cosmology. Theism.*
TL;DR: There are good ... | https://www.lesswrong.com/posts/cEq2ZMSJ88gyTryio/keeping-it-less-than-real-against-possible-people-or-worlds |
# Can startups be impactful in AI safety?
With [Lakera's](https://lakera.ai/) strides in securing LLM APIs, [Goodfire AI's](https://goodfire.ai/) path to scaling interpretability, and 20+ [model](https://www.harmonyintelligence.com/) [evaluations](https://haizelabs.com/) [startups](https://www.vectorview.ai/) among [m... | https://www.lesswrong.com/posts/Mf7wcRDoMAzbECfCw/can-startups-be-impactful-in-ai-safety |
# If-Then Commitments for AI Risk Reduction [by Holden Karnofsky]
Holden just published this paper on the Carnegie Endowment website. I thought it was a decent reference, so I figured I would crosspost it _(included in full for convenience, but if either Carnegie Endowment or Holden has a preference for just having an... | https://www.lesswrong.com/posts/sMtS9Eof6QC6sPouB/if-then-commitments-for-ai-risk-reduction-by-holden |
# Estimating Tail Risk in Neural Networks
Machine learning systems are typically trained to maximize average-case performance. However, this method of training can fail to meaningfully control the probability of tail events that might cause significant harm. For instance, while an artificial intelligence (AI) assistan... | https://www.lesswrong.com/posts/xj5nzResmDZDqLuLo/estimating-tail-risk-in-neural-networks |
# Avoiding the Bog of Moral Hazard for AI
Imagine if you will, a map of a landscape. On this map, I will draw some vague regions. Their boundaries are uncertain, for it is a new and under-explored land. This map is drawn as a graph, but I want to emphasize that the regions are vague guesses, and the true borders could... | https://www.lesswrong.com/posts/pieSxdmjqrKwqa2tR/avoiding-the-bog-of-moral-hazard-for-ai |
# Compression Moves for Prediction
Imagine that you want to predict the behavior of some system -- the weather, a computer program, a real-world language... However you go about it, I bet you will do some compression: you'll find simple and minimal and concrete ways to think about the vast complexity of your system th... | https://www.lesswrong.com/posts/NXDofSmd5pxhrHhFB/compression-moves-for-prediction |
# OpenAI o1, Llama 4, and AlphaZero of LLMs
GPT-4 level open weights models like Llama-3-405B don't seem capable of dangerous cognition. OpenAI o1 demonstrates that a GPT-4 level model can be post-trained into producing useful long horizon reasoning traces. AlphaZero shows how capabilities can be obtained from compute... | https://www.lesswrong.com/posts/Rpc9ACd6zCy3FssTj/openai-o1-llama-4-and-alphazero-of-llms |
# How you can help pass important AI legislation with 10 minutes of effort
*Posting something about a current issue that I think many people here would be interested in. See also the related *[*EA Forum post*](https://forum.effectivealtruism.org/posts/AWYmFwpCrqkknLKdh/how-to-help-crucial-ai-safety-legislation-pass-wi... | https://www.lesswrong.com/posts/8BqovAoFbGjwwDpji/how-you-can-help-pass-important-ai-legislation-with-10 |
# AlignedCut: Visual Concepts Discovery on Brain-Guided Universal Feature Space
Authors: [Huzheng Yang](https://arxiv.org/search/cs?searchtype=author&query=Yang,+H), [James Gee](https://arxiv.org/search/cs?searchtype=author&query=Gee,+J), [Jianbo Shi](https://arxiv.org/search/cs?searchtype=author&query=Shi,+J).
Abstr... | https://www.lesswrong.com/posts/Q8qwjXmv4Hd2Dw6Bi/alignedcut-visual-concepts-discovery-on-brain-guided |
# Not every accommodation is a Curb Cut Effect: The Handicapped Parking Effect, the Clapper Effect, and more
In the fields of user experience and accessibility, everyone talks about the [curb](https://en.wikipedia.org/wiki/Curb_cut_effect) [cut effect](https://ssir.org/articles/entry/the_curb_cut_effect): Features tha... | https://www.lesswrong.com/posts/G3nh24cHJTT9z6gjC/not-every-accommodation-is-a-curb-cut-effect-the-handicapped |
# Did Christopher Hitchens change his mind about waterboarding?
There's a popular story that goes like this: Christopher Hitchens used to be in favor of the US waterboarding terrorists because he thought it wasn't bad enough to be considered torture. Then he had it tried on himself, and changed his mind, coming to bel... | https://www.lesswrong.com/posts/fNqEGTmkCy9sqZYm7/did-christopher-hitchens-change-his-mind-about-waterboarding |
# Proveably Safe Self Driving Cars [Modulo Assumptions]
I’ve seen a fair amount of skepticism about the “Provably Safe AI” paradigm, but I think detractors give it too little credit. I suspect this is largely because of [idea inoculation](https://en.wikipedia.org/wiki/Inoculation_theory) \- people have heard an undeve... | https://www.lesswrong.com/posts/3P8WBwLyfPBEkbG3c/proveably-safe-self-driving-cars-modulo-assumptions |
# Why I funded PIBBSS
I just left a [comment](https://manifund.org//projects/pibbss---affiliate-program-funding-6-months-6-affiliates-or-more?tab=comments#7aa374d7-c42a-4519-9be2-08ccc03fed62) on [PIBBSS](https://pibbss.ai/)' Manifund grant request (which I funded $25k) that people might find interesting. PIBBSS needs... | https://www.lesswrong.com/posts/Yjiw5rnu4mJsYN8Xc/why-i-funded-pibbss-1 |
# Superintelligence Can't Solve the Problem of Deciding What You'll Do
Superintelligence can solve all problems, even as it won't necessarily do so. But not quite. In the modern world, you [exist within physics](https://www.lesswrong.com/posts/NEeW7eSXThPz7o4Ne/thou-art-physics), and all you do obeys its laws. Yet it'... | https://www.lesswrong.com/posts/KKGbzDZdbTCLHZjTN/superintelligence-can-t-solve-the-problem-of-deciding-what |
# If I wanted to spend WAY more on AI, what would I spend it on?
Supposedly intelligence is some kind of superpower. And they're now selling intelligence for pennies/million tokens. Logically, it seems like I should be spending [way more](https://x.com/kellerjordan0/status/1834990147555017180) of my income than I cu... | https://www.lesswrong.com/posts/bX7q9NcoGpb5KdZzQ/if-i-wanted-to-spend-way-more-on-ai-what-would-i-spend-it-on |
# Hyperpolation
Interpolation, Extrapolation, Hyperpolation: Generalising into new dimensions
by [Toby Ord](https://en.wikipedia.org/wiki/Toby_Ord)
Abstract:
> This paper introduces the concept of hyperpolation: a way of generalising from a limited set of data points that is a peer to the more familiar concepts of... | https://www.lesswrong.com/posts/DY8PKdPsCoWonA7nQ/hyperpolation |
# What's the Deal with Logical Uncertainty?
I notice that reasoning about logical uncertainty does not appear more confusing to me than reasoning about empirical one. Am I missing something?
Consider the classical example from the [description of the tag](https://www.lesswrong.com/tag/logical-uncertainty):
> Is the ... | https://www.lesswrong.com/posts/H229aGt8nMFQsxJMq/what-s-the-deal-with-logical-uncertainty |
# Can subjunctive dependence emerge from a simplicity prior?
Suppose that an embedded agent models its environment using an approximate simplicity prior, would it acquire a [physicalist agent ontology or an algorithmic/logical agent ontology](https://www.lesswrong.com/posts/dmjvJwCjXWE2jFbRN/fdt-is-not-directly-compar... | https://www.lesswrong.com/posts/QyfescBSt7nLfkGBs/can-subjunctive-dependence-emerge-from-a-simplicity-prior |
# GPT-o1
Terrible name (with a terrible reason, [that this ‘resets the counter’ on AI capability to 1](https://x.com/btibor91/status/1834686946846597281), and ‘o’ as in OpenAI when they previously used o for Omni, very confusing). Impressive new capabilities in many ways. Less impressive in many others, at least relat... | https://www.lesswrong.com/posts/zuaaqjsN6BucbGhf5/gpt-o1 |
# Secret Collusion: Will We Know When to Unplug AI?
**TL;DR:** We introduce the **first comprehensive theoretical framework for understanding and mitigating secret collusion** among advanced AI agents, along with **CASE, a novel model evaluation framework**. CASE assesses the cryptographic and steganographic capabilit... | https://www.lesswrong.com/posts/smMdYezaC8vuiLjCf/secret-collusion-will-we-know-when-to-unplug-ai |
# Generative ML in chemistry is bottlenecked by synthesis
**Introduction**
================
Every ... | https://www.lesswrong.com/posts/K8vtpBvGzEyFEjhxK/generative-ml-in-chemistry-is-bottlenecked-by-synthesis |
# MIRI's September 2024 newsletter
MIRI updates
------------
* Aaron Scher and Joe Collman have joined the Technical Governance Team at MIRI as researchers. Aaron previously did independent research related to sycophancy in language models and mechanistic interpretability, while Joe previously did independent resea... | https://www.lesswrong.com/posts/HJwxirN8nGaHG2Rr7/miri-s-september-2024-newsletter |
# Book review: Xenosystems
I've met a few Landians over the last couple years, and they generally recommend that I start with reading Nick Land's (now defunct) Xenosystems blog, or *Xenosystems*, a [Passage Publishing book](https://passage.press/products/xenosystems) that compiles posts from the blog. While I've read ... | https://www.lesswrong.com/posts/cfw7L22oLAm7Zcmff/book-review-xenosystems |
# Does life actually locally *increase* entropy?
The usual materialist story of life I've heard is that life acts like an entropy pump, creating local reductions of entropy within the organism but increasing the entropy outside of the organism. (I think I've even seen that in The Sequences somewhere? But couldn't find... | https://www.lesswrong.com/posts/2ShXvqzA27qJTkYwq/does-life-actually-locally-increase-entropy |
# I finally got ChatGPT to sound like me
I publish lots of words on the Internet. These words are fed into LLMs such as ChatGPT. I have published enough training data that the world's most powerful AIs have a simulation of me inside of them. I have basically uploaded a copy of myself into the cloud. The AI simulation ... | https://www.lesswrong.com/posts/2d5o75nmTpLiSP4WL/i-finally-got-chatgpt-to-sound-like-me |
# Monthly Roundup #22: September 2024
It’s that time again for all the sufficiently interesting news that isn’t otherwise fit to print, also known as the Monthly Roundup.
#### Bad News
[Beware the failure mode in strategy and decisions that implicitly assumes competence](https://x.com/robertskmiles/status/1830925270... | https://www.lesswrong.com/posts/4gAqkRhCuK2kGJFQE/monthly-roundup-22-september-2024 |
# How harmful is music, really?
For a while, I [thought music was harmful](https://dkl9.net/essays/music_harm.html), due largely to pervasive and arbitrary earworms. More recently, I started to find that [earworms are ephemeral and lawful](https://dkl9.net/essays/earworm_mechanics.html). A contrarian belief held like ... | https://www.lesswrong.com/posts/dbkXFiB3JbD64W6fu/how-harmful-is-music-really |
# Where to find reliable reviews of AI products?
Being able to quickly incorporate AI tools seems important, including for working on AI risk (people who disagree: there's a thread for doing so in the comments). But there are a lot of AI products and most of them suck. Does anyone know a good source of reviews, or ev... | https://www.lesswrong.com/posts/jMn3YXCjqmJu3Zx5o/where-to-find-reliable-reviews-of-ai-products |
# Skills from a year of Purposeful Rationality Practice
A year ago, I started trying to deliberate practice skills that would "help people figure out the answers to confusing, important questions." I experimented with [Thinking Physics](https://www.lesswrong.com/posts/PiPH4gkcMuvLALymK/exercise-solve-thinking-physics)... | https://www.lesswrong.com/posts/thc4RemfLcM5AdJDa/skills-from-a-year-of-purposeful-rationality-practice |
# Is "superhuman" AI forecasting BS? Some experiments on the "539" bot from the Centre for AI Safety
*Disclaimer: I am a computational physicist’s and this investigation is outside of my immediate area of expertise. Feel free to peruse the experiments and take everything I say with appropriate levels of skepticism.*
... | https://www.lesswrong.com/posts/ZhqEShJMMbASxzfEL/is-superhuman-ai-forecasting-bs-some-experiments-on-the-539 |
# Pronouns are Annoying
This post isn’t totally about the culture war topic du jour. Not at first.
As with any other topic that soaks up angst like an ultra-absorbent sponge, I wonder how many have lost... | https://www.lesswrong.com/posts/zrHHLngm5CmQLD2zk/pronouns-are-annoying |
# The case for a negative alignment tax
TL;DR:
------
Alignment researchers have historically predicted that building safe advanced AI would necessarily incur a significant [alignment tax](https://www.alignmentforum.org/tag/alignment-tax) compared to an equally capable but unaligned counterfactual AI.
We put forwar... | https://www.lesswrong.com/posts/xhLopzaJHtdkz9siQ/the-case-for-a-negative-alignment-tax |
# Intention-to-Treat (Re: How harmful is music, really?)
I have long wanted to write about intention-to-treat because it's such a neat idea, and the recent article [How harmful is music, really?](https://www.lesswrong.com/posts/dbkXFiB3JbD64W6fu/how-harmful-is-music-really) spurred me to finally do it.
* * *
The rep... | https://www.lesswrong.com/posts/CtsqycgJFNgm7Pe3g/intention-to-treat-re-how-harmful-is-music-really |
# How to choose what to work on
So you want to advance human progress. And you’re wondering, what should you, personally, *do*? Say you have talent, ambition, and drive—how do you choose a project or career?
There are a few frameworks for making this decision. Recently, though, I’ve started to see pitfalls with some ... | https://www.lesswrong.com/posts/8qLMF25s3ctCpgx7e/how-to-choose-what-to-work-on |
# The Obliqueness Thesis
In my [*Xenosystems* review](https://www.lesswrong.com/posts/cfw7L22oLAm7Zcmff/book-review-xenosystems), I discussed the Orthogonality Thesis, concluding that it was a bad metaphor. It's a long post, though, and the comments on orthogonality build on other *Xenosystems* content. Therefore, I t... | https://www.lesswrong.com/posts/vdATCfuxdtdqM2wh9/the-obliqueness-thesis |
# Which LessWrong/Alignment topics would you like to be tutored in? [Poll]
Would you like to be tutored in [applied game theory](https://www.lesswrong.com/tag/game-theory), [natural latents](https://www.lesswrong.com/posts/mMEbfooQzMwJERAJJ/natural-latents-the-concepts), [CFAR-style rationality techniques](https://www... | https://www.lesswrong.com/posts/Hb9LGGyv4RqWE9jBq/which-lesswrong-alignment-topics-would-you-like-to-be |
# Slave Morality: A place for every man and every man in his place
Lately, there has been a discussion between [Bentham's Bulldog](https://benthams.substack.com/p/shut-up-about-slave-morality), [Scott Alexander](https://www.astralcodexten.com/p/matt-yglesias-considered-as-the-nietzschean) and others about the nature o... | https://www.lesswrong.com/posts/crBCGazM6Hedqb83J/slave-morality-a-place-for-every-man-and-every-man-in-his |
# AI #82: The Governor Ponders
The big news of the week was of course OpenAI [releasing their new model o1](https://thezvi.substack.com/p/gpt-4o1). If you read one post this week, read that one. Everything else is a relative sideshow.
Meanwhile, we await Newsom’s decision on SB 1047. The smart money was always that G... | https://www.lesswrong.com/posts/Y4nS3yMWfJdmeoLcQ/ai-82-the-governor-ponders |
# [Intuitive self-models] 1. Preliminaries
1.1 Summary & Table of Contents
===============================
This is the first of a [series of eight blog posts](https://www.lesswrong.com/s/qhdHbCJ3PYesL9dde). Here’s an overview of the whole series, and then we’ll jump right into the first post!
1.1.1 Summary & Table o... | https://www.lesswrong.com/posts/FtwMA5fenkHeomz52/intuitive-self-models-1-preliminaries |
# Laziness death spirals
I’ve claimed that [Willpower compounds](https://patrickdfarley.com/wielding-willpower/) and that small wins in the present make it easier to get bigger wins in the future. Unfortunately, procrastination and laziness compound, too.
You’re stressed out for some reason, so you take the evening o... | https://www.lesswrong.com/posts/JBR6AF9Gusv4u6Fwo/laziness-death-spirals |
# To CoT or not to CoT? Chain-of-thought helps mainly on math and symbolic reasoning
Authors: [Zayne Sprague](https://arxiv.org/search/cs?searchtype=author&query=Sprague,+Z), [Fangcong Yin](https://arxiv.org/search/cs?searchtype=author&query=Yin,+F), [Juan Diego Rodriguez](https://arxiv.org/search/cs?searchtype=author... | https://www.lesswrong.com/posts/gxweu2tD8z9y8qFD6/to-cot-or-not-to-cot-chain-of-thought-helps-mainly-on-math |
# Why good things often don’t lead to better outcomes
Many people believe that when something good happens, positive outcomes naturally follow — but I'm here to explain why that's often not the case. In complex systems, improvements or advancements frequently set off a chain of reactions that can undermine the origina... | https://www.lesswrong.com/posts/4J6n3djNcHGqQuCbu/why-good-things-often-don-t-lead-to-better-outcomes |
# AISafety.info: What are Inductive Biases?
*AISafety.info writes AI safety intro content. We'd appreciate any* [*feedback*](https://docs.google.com/document/d/1buCYKL0fYgkXU17Cfvra6_1VDk7y76RDy7Dnr8hzG7I/edit?usp=sharing)*.*
The [inductive bias](https://en.wikipedia.org/wiki/Inductive_bias) of a learning process is ... | https://www.lesswrong.com/posts/8hyrX3EdRv8iqcxoz/aisafety-info-what-are-inductive-biases |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.