text stringlengths 300 320k | source stringlengths 52 154 |
|---|---|
# User research as a barometer of software design
([Cross-posted](https://adamzerner.bearblog.dev/user-research-as-a-barometer-of-software-design/) on my personal blog.)
In UX design, they do something called [user](https://blog.codinghorror.com/low-fi-usability-testing/) [research](https://sensible.com/rocket-surger... | https://www.lesswrong.com/posts/BLcpCqNFnbCasAKTe/user-research-as-a-barometer-of-software-design |
# Being a donor for Fecal Microbiota Transplants (FMT): Do good & earn easy money (up to 180k/y)
TLDR
====
Help to find super donors for Fecal Microbiota Transplants (FMT)! By donating stool for FMT, exceptionally healthy people can earn $180,000/yr with little effort. Simultaneously they can help chronically ill peo... | https://www.lesswrong.com/posts/i48nw33pW9kuXsFBw/being-a-donor-for-fecal-microbiota-transplants-fmt-do-good |
# What New Desktop Should I Buy?
It's time for me to buy a new desktop, as my old one isn't working so well anymore.
It also happens to be 'black Friday in July' which opens up the opportunity to perhaps get a good sale price somewhere.
Time is valuable, so I'm happy to spend a reasonably large amount if it actuall... | https://www.lesswrong.com/posts/ktcigmup94jeeZhpu/what-new-desktop-should-i-buy |
# Research Notes: What are we aligning for?
As part of learning the field and maximizing on new ideas, I've been trying to figure out what the goal of AI alignment is. So far I've found out what outer alignment is as a [concept](https://www.lesswrong.com/tag/outer-alignment), but not what it should be as an instantiat... | https://www.lesswrong.com/posts/zpz929rJvJ8xCBiGZ/research-notes-what-are-we-aligning-for |
# MATS Models
I’ve been using the summer 2022 [SERI MATS program](https://www.serimats.org/) as an opportunity to test out my current best guesses at how to produce strong researchers. This post is an overview of the methods I’ve been testing, and the models behind them.
The Team Model
==============
My MATS partici... | https://www.lesswrong.com/posts/nvP28s5oydv8RjF9E/mats-models |
# Changes in Community Dynamics: A Follow-Up to 'The Berkeley Community & the Rest of Us'
*I've received a few notifications in the last couple weeks that* [*this post*](https://www.lesswrong.com/posts/zAqoj79A7QuhJKKvi/the-berkeley-community-and-the-rest-of-us-a-response-to-zvi) *I wrote in 2018 about the relationshi... | https://www.lesswrong.com/posts/wLSo6os9etg2BPDtc/changes-in-community-dynamics-a-follow-up-to-the-berkeley |
# I’ve become a medical mystery and I don’t know how to effectively get help
To put it very mildly, I’m in a really bad way emotionally because of this and I’m running out of ideas. Any ideas are welcome—specific to symptoms, or just generally how to deal with having a problem and not being able tot get diagnosis or t... | https://www.lesswrong.com/posts/7uZ9aQYukQ5AjMjHE/i-ve-become-a-medical-mystery-and-i-don-t-know-how-to |
# Ars D&D.sci: Mysteries of Mana
*This is an entry in the 'Dungeons & Data Science' series, a set of puzzles where players are given a dataset to analyze and an objective to pursue using information from that dataset. *
**STORY (skippable)**
=====================
For thousands of years the Order of Athena has preser... | https://www.lesswrong.com/posts/9P4yLsfhj2Jjm4yqg/ars-d-and-d-sci-mysteries-of-mana |
# Making it harder for an AGI to "trick" us, with STVs
| Summary / Preamble ================== AI Alignment has various sub-areas. The area I focus on here is ways we might use a superintelligent AGI-system to help with creating an aligned AGI-system, even if the AGI we start out with isn’t fully aligned. Imagine a su... | https://www.lesswrong.com/posts/xERh9dkBkHLHp7Lg6/making-it-harder-for-an-agi-to-trick-us-with-stvs |
# Using Ngram to estimate depression prevalence over time
Historical language records reveal a surge of cognitive distortions in recent decades
-------------------------------------------------------------------------------------
My summary: People diagnosed with depression tend to exhibit characteristic patterns of ... | https://www.lesswrong.com/posts/Sbd3AHrKtJRXhkwMQ/using-ngram-to-estimate-depression-prevalence-over-time |
# Train first VS prune first in neural networks.
This post aims to answer a simple question about neural nets, at least on a small toy dataset. Does it matter if you train a network, and then prune some nodes, or if you prune the network, and then train the smaller net.
What exactly is pruning.
---------------------... | https://www.lesswrong.com/posts/PLqopCagHKo2EK5cE/train-first-vs-prune-first-in-neural-networks |
# Grouped Loss may disfavor discontinuous capabilities
*Thanks to Evan Hubinger and Beth Barnes for comments on these ideas.*
Language models exhibit [scaling laws](https://arxiv.org/abs/2001.08361), where the loss is a power-law in model size. This offers a lot of predictive power, and seems like a useful thing to k... | https://www.lesswrong.com/posts/PmhTzHHFEem5hX79R/grouped-loss-may-disfavor-discontinuous-capabilities |
# 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 i... | https://www.lesswrong.com/posts/KBjRoc7WccbMehnbN/report-from-a-civilizational-observer-on-earth |
# Comment on "Propositions Concerning Digital Minds and Society"
> _I will do my best to teach them
> About life and what it's worth
> I just hope that I can keep them
> From destroying the Earth_
>
> —Jonathan Coulton, "The Future Soon"
In a recent paper, Nick Bostrom and Carl Shulman present ["Propositions ... | https://www.lesswrong.com/posts/nqXcgsqQBw2doAnXu/comment-on-propositions-concerning-digital-minds-and-society |
# Taste & Shaping
*Author's note: while it does not belong in the handbook proper, Duncan Sabien's essay* [*Goodhart's Imperius*](https://www.lesswrong.com/posts/hBKnkx8xvmhGZPdmg/goodhart-s-imperius-1) *is very close to (and partially derived from) the content in this chapter, and is a good extension or follow-up.*
... | https://www.lesswrong.com/posts/wanmjz6wkTCNtyKLX/taste-and-shaping |
# Hessian and Basin volume
*Thanks to Thomas Kwa for the question which prompted this post.*
**Note:** This is mostly a primer / introductory reference, not a research post. However, the details should be interesting even to those familiar with the area.
When discussing “broad basins” in the loss landscape of a DN... | https://www.lesswrong.com/posts/QPqztHpToij2nx7ET/hessian-and-basin-volume |
# Why Portland
_Meta: This is a personal post. Almost like a journal entry, but also loosely intended to be enjoyable and/or useful to other people._
([Cross-posted](https://adamzerner.bearblog.dev/where-portland/) on my personal blog.)
Why do people live where they live? Everyone's got a different answer.
I lived ... | https://www.lesswrong.com/posts/5usKtyM6QvpN8jf5u/why-portland |
# My Opportunity Costs
When I was younger, I had time and energy to spare on weird side-projects, and a comparative lack of ideas worth spending them on. Now I have some ideas, but no time and energy to spare. I share these ideas below, in case anyone is in the same situation I was a decade ago[^d8zcs177yu4].
Editing... | https://www.lesswrong.com/posts/wtGqr9AhseajtHmDY/my-opportunity-costs |
# Avoid the abbreviation "FLOPs" – use "FLOP" or "FLOP/s" instead
Especially in discussions about AI, the abbreviation "FLOPs" is being used for both "floating point operations per second" (a measure of computational power) and "floating point operations" (a measure of total computations, and equivalent to the previou... | https://www.lesswrong.com/posts/XiKidK9kNvJHX9Yte/avoid-the-abbreviation-flops-use-flop-or-flop-s-instead |
# Immanuel Kant and the Decision Theory App Store
*\[Epistemic status: About as silly as it sounds.\]*
Prepare to be astounded by this rationalist reconstruction of Kant, drawn out of an unbelievably tiny parcel of Kant literature![^9lk7ht49za]
Kant argues that all rational agents will:
* “Act only according to t... | https://www.lesswrong.com/posts/byMNKEXBn4RTcaaa6/immanuel-kant-and-the-decision-theory-app-store |
# The Alignment Problem
Last month Eliezer Yudkowsky wrote ["a poorly organized list of individual rants"](https://www.lesswrong.com/posts/uMQ3cqWDPHhjtiesc/agi-ruin-a-list-of-lethalities) about how AI is going to kill us all. In this post, I attempt to summarize the rants in my own words.
These are not my personal o... | https://www.lesswrong.com/posts/G6nnufmiTwTaXAbKW/the-alignment-problem |
# Checksum Sensor Alignment
# Problem
Suppose you have a powerful AGI protecting a diamond in a vault[^1]. You have various sensors in the room, all monitoring the diamond. There is a camera, watching the diamond. There is a pressure plate under the diamond confirming it is still in its stand. There are a total of 10... | https://www.lesswrong.com/posts/9sqL83o7emTxnEikh/checksum-sensor-alignment |
# How do AI timelines affect how you live your life?
This question is more about personal decision-making rather than for example deciding to work on AI safety for altruistic reasons. If I were thinking about this from a purely selfish perspective, it seems pretty likely that if I expect transformative AI to arrive i... | https://www.lesswrong.com/posts/xupJnpdRjbExwdb8J/how-do-ai-timelines-affect-how-you-live-your-life |
# Systemization
**Epistemic status**: Anecdotally strong
*Many of the flawed heuristics and biases that the Systemization unit seeks to address are well-known and well-researched (such as the planning fallacy and failure to account for switching costs). The underlying theory of attention draws on a combination of Dan... | https://www.lesswrong.com/posts/QKpmPBpyCdNbBKAdR/systemization |
# Marriage, the Giving What We Can Pledge, and the damage caused by vague public commitments
I believe honesty is very important. I think most people agree that honesty is like, pretty important, but I think it’s a lot more important than that. I basically think that people will be dishonest in ways that hurt them and... | https://www.lesswrong.com/posts/XXFiZ4swvxN9ogKqz/marriage-the-giving-what-we-can-pledge-and-the-damage-caused |
# Book Review: Neal Stephenson’s “Termination Shock”
*Cross posted from the Words&Dirt blog. I'm a longtime reader and fan of lesswrong, but not a regular poster. My best friend submitted this review of Neal Stephenson's 2021 book 'Termination Shock' into the ACX book review contest this year. It didn't make the final... | https://www.lesswrong.com/posts/fAAzDqjcPxtPK7AoJ/book-review-neal-stephenson-s-termination-shock |
# Rainmaking
There is an old TV show called 'The Wire', and in it a con artist fleeces a mark.
The mark's friend tries to wake him up, whereupon the mark insists that the con artist has helped him. The mark's friend tells him to 'wake up, he rain-made you!'
"What?"
"A guy says if you pay him, he can make it rain. ... | https://www.lesswrong.com/posts/bouGmyCPtH5E8LTyg/rainmaking |
# On how various plans miss the hard bits of the alignment challenge
*This post has been recorded as part of the LessWrong Curated Podcast, and can be listened to on* [*Spotify*](https://open.spotify.com/episode/6cDwctrmpWW6UNEKTT3Vvf?si=7T7SaHNCQ9axzL4D-crPQw)*,* [*Apple Podcasts*](https://podcasts.apple.com/us/podca... | https://www.lesswrong.com/posts/3pinFH3jerMzAvmza/on-how-various-plans-miss-the-hard-bits-of-the-alignment |
# How do you concisely communicate & navigate the politics / culture at your job working at a large corporation or institution?
Previously I've worked at small companies and state government or educational institutions. Now I work at a ~30,000 person private corporation and need to better adapt to that cultural situat... | https://www.lesswrong.com/posts/W3kykr5ycpm3HGiCN/how-do-you-concisely-communicate-and-navigate-the-politics |
# Mosaic and Palimpsests: Two Shapes of Research
*(Minor update to change Steve's labelling following* [*this comment*](https://www.alignmentforum.org/posts/4BpeHPXMjRzopgAZd/mosaic-and-palimpsests-two-shapes-of-research?commentId=poz8BXhgR9HJojs3Q)*, and also because I realized that I never added the footnotes...)*
... | https://www.lesswrong.com/posts/4BpeHPXMjRzopgAZd/mosaic-and-palimpsests-two-shapes-of-research |
# Response to Blake Richards: AGI, generality, alignment, & loss functions
Blake Richards is a neuroscientist / AI researcher with appointments at [McGill](https://www.mcgill.ca/neuro/blake-richards-phd) & [MiLA](https://mila.quebec/en/person/blake-richards/). Much of his recent work has involved making connections be... | https://www.lesswrong.com/posts/rgPxEKFBLpLqJpMBM/response-to-blake-richards-agi-generality-alignment-and-loss |
# Progress links and tweets, 2022-07-12
**Links**
---------
* [Patrick Collison on universal covid vaccines: we could have them very soon, if it weren’t for bureaucratic roadblocks](https://www.slowboring.com/p/we-could-have-universal-covid-vaccines)
* [“Current funding organizations cluster heavily around a few ... | https://www.lesswrong.com/posts/fdhyp8rfQSw4mdMwP/progress-links-and-tweets-2022-07-12 |
# Acceptability Verification: A Research Agenda
[This Google doc^](https://docs.google.com/document/d/199Lkh78UA2uI9ljLEy_aWR8RBetQLRO6Kqo3_Omi1e4/view) is a halted, formerly work-in-progress writeup of [Evan Hubinger’s](https://www.lesswrong.com/users/evhub) AI alignment research agenda, authored by Evan. It dates ba... | https://www.lesswrong.com/posts/GeabLEXYP7oBMivmF/acceptability-verification-a-research-agenda |
# Alien Message Contest: Solution
Thanks to everyone who participated in the [Alien Message contest](https://www.lesswrong.com/posts/dFFdAdwnoKmHGGksW/contest-an-alien-message), and congratulations on successfully being able to decode the message. I gave a time limit of 2 weeks at the start of the contest, but you com... | https://www.lesswrong.com/posts/Rs2fnNCgwpggEEpxj/alien-message-contest-solution |
# Straw-Steelmanning
I've noticed that when people are asked to "Steelman" a position, they sometimes instead do what I would call "Straw-Steelmanning". Someone can also straw-steelman without having been asked to steelman or having said that they would do so.
What is straw-steelmanning? Assume someone makes an argu... | https://www.lesswrong.com/posts/Fph8Z4BoMtZkxxdB5/straw-steelmanning |
# Everyone is an Imposter
**I.**
I am generally considered to be a pretty smart guy.
Now to say this directly is to invite doubt and disaster. I know that. I feel it deep in my bones. It’s been drilled into me that, outside of certain domains (like rap and sports), it’s impolite to talk directly about one’s abiliti... | https://www.lesswrong.com/posts/XYhwst2Yh4dLRa6ek/everyone-is-an-imposter |
# John von Neumann on how to safely progress with technology
From a 1955 article written by von Neumann titled, "[Can We Survive Technology?](http://geosci.uchicago.edu/~kite/doc/von_Neumann_1955.pdf)"
> What safeguard remains? Apparently only day-to-day — or perhaps year-to-year — opportunistic measures, along seque... | https://www.lesswrong.com/posts/4f3FvHK5qTKrGqte4/john-von-neumann-on-how-to-safely-progress-with-technology |
# Passing Up Pay
I recently switched from [earning money to donate](https://www.jefftk.com/p/earning-to-give-transcript) to [doing something directly useful](https://www.jefftk.com/p/leaving-google-joining-the-nucleic-acid-observatory). I think this work is very important (or I wouldn't have switched!), and if I were ... | https://www.lesswrong.com/posts/uLR2ghKSXbu8fKzoz/passing-up-pay |
# MIRI Conversations: Technology Forecasting & Gradualism (Distillation)
*This was submitted as part of UC Berkeley's* [*distillation competition*](https://eaberkeley.com/aims-distillation)*. Many thanks to everyone who gave feedback on this post – in particular Hannah Erlebach, Jamie Bernardi, Dewi Erwan, Arthur Conm... | https://www.lesswrong.com/posts/DQMD5XZgBXaegRDQv/miri-conversations-technology-forecasting-and-gradualism |
# App idea to help with reading STEM textbooks (feedback request)
**Problem:** STEM textbooks often reference figures and equations from earlier in the textbook. They usually do this with statements like "the shearing stress τ... may be obtained from the shearing-stress-strain diagram of Fig. 3.30." But Fig. 3.30 migh... | https://www.lesswrong.com/posts/5MiQ2u9f7Fh4fgXoX/app-idea-to-help-with-reading-stem-textbooks-feedback |
# Making decisions using multiple worldviews
*Tl;dr: the problem of how to make decisions using multiple (potentially incompatible) worldviews (which I'll call the problem of meta-rationality) comes up in a range of contexts, such as epistemic deference. Applying a policy-oriented approach to meta-rationality, and eva... | https://www.lesswrong.com/posts/6QPFKHRsuY63cuJwh/making-decisions-using-multiple-worldviews |
# Goal Alignment Is Robust To the Sharp Left Turn
[A central AI Alignment problem](https://www.lesswrong.com/s/v55BhXbpJuaExkpcD/p/GNhMPAWcfBCASy8e6) is the "sharp left turn" — a point in AI training under the SGD analogous to the development of human civilization under evolution, past which the AI's capabilities woul... | https://www.lesswrong.com/posts/vix3K4grcHottqpEm/goal-alignment-is-robust-to-the-sharp-left-turn |
# Any tips for eliciting one's own latent knowledge?
I struggle to write. I cannot figure out how to access my own thoughts and ideas. They come in non-linguistic forms - vague feelings, 3D images, animations - with a verbal narrator on top, but my very bad memory, and the fact that my best thinking is done in the mid... | https://www.lesswrong.com/posts/EN6NczhybtGFM6ZsG/any-tips-for-eliciting-one-s-own-latent-knowledge |
# Artificial Sandwiching: When can we test scalable alignment protocols without humans?
*Epistemic status: Not a fleshed-out proposal. Brainstorming/eliciting ideas.*
*Thanks to Ben Mann, Pablo Moreno, and Jared Kaplan for feedback on early drafts.*
Overview
--------
* I’m convinced [sandwiching](https://www.ali... | https://www.lesswrong.com/posts/nekLYqbCEBDEfbLzF/artificial-sandwiching-when-can-we-test-scalable-alignment |
# Deep learning curriculum for large language model alignment
This is a deep learning curriculum with a focus on topics relevant to large language model alignment. It is centered around papers and exercises, and is biased towards my own tastes.
It is targeted at newcomers to deep learning who are already familiar wit... | https://www.lesswrong.com/posts/5uNfgjaDAwkhcLJca/deep-learning-curriculum-for-large-language-model-alignment |
# Notes on Love
This post examines **love** as a virtue, as part of a sequence of posts about virtues. I mostly explore what other people have learned about love, rather than sharing my own research or opinions about it, though I’ve been selective about what I found interesting or credible, and my research has been sc... | https://www.lesswrong.com/posts/KjScjtBpGboH6uELz/notes-on-love |
# How to impress students with recent advances in ML?
I am preparing resources for an equivalent of the [mlab](https://www.lesswrong.com/posts/3ouxBRRzjxarTukMW/apply-to-the-second-iteration-of-the-ml-for-alignment) camp, organized by [EffiSciences](https://www.effisciences.org/), that will be held in France in a mon... | https://www.lesswrong.com/posts/Wj97sy5eQTTjLKg7o/how-to-impress-students-with-recent-advances-in-ml |
# Humans provide an untapped wealth of evidence about alignment
*This post has been recorded as part of the LessWrong Curated Podcast, and can be listened to on* [*Spotify*](https://open.spotify.com/episode/0jpI7LLNzKsn6lwrsoDCc9)*,* [*Apple Podcasts*](https://podcasts.apple.com/us/podcast/humans-provide-an-untapped-w... | https://www.lesswrong.com/posts/CjFZeDD6iCnNubDoS/humans-provide-an-untapped-wealth-of-evidence-about |
# Criticism of EA Criticism Contest
Back when it was announced, I [toyed with the idea](https://twitter.com/TheZvi/status/1532140957470269441) of criticizing [the EA criticism and red teaming contest](https://forum.effectivealtruism.org/posts/8hvmvrgcxJJ2pYR4X/announcing-a-contest-ea-criticism-and-red-teaming) as an e... | https://www.lesswrong.com/posts/pohTfSGsNQZYbGpCy/criticism-of-ea-criticism-contest |
# Covid 7/14/22: BA.2.75 Plus Tax
That’s the new variant for this week. I am not worried about it, but there is little other news on the Covid front and I’m still attempting not to reuse titles. As time goes by the events are still new but they tend to rhyme. This update is inessential.
In addition to various non-Cov... | https://www.lesswrong.com/posts/fpyxnnwTJfS9gpHGs/covid-7-14-22-ba-2-75-plus-tax |
# Circumventing interpretability: How to defeat mind-readers
*(Post now available as a pdf:* [*https://arxiv.org/abs/2212.11415*](https://arxiv.org/abs/2212.11415) *)*
| TL;DR: Unaligned AI will have a convergent instrumental incentive to make its thoughts difficult for us to interpret. In this article, I discuss man... | https://www.lesswrong.com/posts/EhAbh2pQoAXkm9yor/circumventing-interpretability-how-to-defeat-mind-readers |
# Better Quiddler
[Quiddler](https://boardgamegeek.com/boardgame/339/quiddler) is a card game, a cross between rummy and scrabble. I've been playing it on vacation with in-laws, and it suffers from the common drawback of games where your turn begins with drawing and ends with discarding: you begin your turn with new i... | https://www.lesswrong.com/posts/jGNXFAvLmYstBGM4k/better-quiddler |
# Potato diet: A post mortem and an answer to SMTM's article
Slime Mold Time Mold (SMTM) has just published the [results of their studies on the potato diet](https://slimemoldtimemold.com/2022/07/12/lose-10-6-pounds-in-four-weeks-with-this-one-weird-trick-discovered-by-local-slime-hive-mind-doctors-grudgingly-respect-... | https://www.lesswrong.com/posts/oh3YWqobjHAjNvNqd/potato-diet-a-post-mortem-and-an-answer-to-smtm-s-article |
# Inward and outward steelmanning
The usual steelmanning is when you try to find the best possible arguments for a position you don't like. Or the hardest to refute version of an idea you don't like. Let's call this "inward steelmanning". You're mostly fixing internal cogs of arguments and positions.
I want to propos... | https://www.lesswrong.com/posts/ZamwhJaHwuNALJjwq/inward-and-outward-steelmanning |
# A note about differential technological development
Quick note: I occasionally run into arguments of the form "my research advances capabilities, but it advances alignment more than it advances capabilities, so it's good on net". I do not buy this argument, and think that in most such cases, this sort of research do... | https://www.lesswrong.com/posts/vQNJrJqebXEWjJfnz/a-note-about-differential-technological-development |
# Upcoming heatwave: advice
There's a heatwave coming (or already arrived) in the UK and western Europe. Many of these places are not equipped for dealing with high temperatures and have large at risk populations* - not simply those with preexisting health conditions, but those living in accomodation grossly unsuited ... | https://www.lesswrong.com/posts/bjPMgb2NYPqztK5dF/upcoming-heatwave-advice |
# Don't use 'infohazard' for collectively destructive info
Suppose there were a cheap way to make nuclear weapons out of common household materials. Knowing this information is not directly harmful to the median individual like yourself. On average, it may even be beneficial to the average person, to know that secret ... | https://www.lesswrong.com/posts/yET7wbjjJZtpz6NF3/don-t-use-infohazard-for-collectively-destructive-info |
# Peter Singer's first published piece on AI
Hi! I am Tse Yip Fai. This is a linkpost to Peter Singer's first published piece on AI, it's also my first [published paper](https://link.springer.com/article/10.1007/s43681-022-00187-z) (second author). Since the paper is open access, I won't be doing any copy-and-paste h... | https://www.lesswrong.com/posts/TdfFYpsHrkuFm8nqX/peter-singer-s-first-published-piece-on-ai |
# A review of Nate Hilger's The Parent Trap
I enjoyed this a lot. It’s in t... | https://www.lesswrong.com/posts/BSypz7PrxftCTqyp7/a-review-of-nate-hilger-s-the-parent-trap |
# Review of The Engines of Cognition
Introduction[^3viitbggdgj]
--------------------------
For those who read my review of the first Lesswrong collection of essays, A *Map that Reflects the Territory* (see [here](https://www.lesswrong.com/posts/JXTEDFCC5r4dW2tta/review-of-a-map-that-reflects-the-territory) for my rev... | https://www.lesswrong.com/posts/4DncC47Q9RcqsyvxW/review-of-the-engines-of-cognition |
# Notes on Learning the Prior
*This post was written to fulfil requirements of the SERI MATS Training Program.*
One of the goals in writing this post is to arrive at an explanation of a particular set of issues raised by Irving in the [Learning the smooth prior](https://www.alignmentforum.org/posts/9x5mtYjHYfr4T7KLj/... | https://www.lesswrong.com/posts/ukidKsEio8hfB9uHT/notes-on-learning-the-prior |
# Highlights from the memoirs of Vannevar Bush
Vannevar Bush—head of military research during WW2, author of “As We May Think” and “Science, the Endless Frontier”—wrote a memoir late in life, *Pieces of the Action.* It was out of print and hard to obtain for a long time, but Stripe Press has brought it back in [a new ... | https://www.lesswrong.com/posts/oJk6cb9umgjNJsxGA/highlights-from-the-memoirs-of-vannevar-bush |
# A time-invariant version of Laplace's rule
What is the probability that the sun will rise tomorrow? What are the chances of a pandemic happening next year? What are the odds of survival of a new surgery that has been successfully executed only once?
These and many other questions can be answered appealing to a gene... | https://www.lesswrong.com/posts/wE7SK8w8AixqknArs/a-time-invariant-version-of-laplace-s-rule |
# Comfort Zone Exploration
**Epistemic status**: Mixed
*The concepts which inspired the CoZE unit (such as exposure therapy and the explore/exploit problem in probability) are generally well-researched and under- stood. However, our combination of these concepts with an outlook of curiosity and epistemic uncertainty ... | https://www.lesswrong.com/posts/ifa45nkAZiF5Msubi/comfort-zone-exploration |
# Safety Implications of LeCun's path to machine intelligence
Yann LeCun recently posted [A Path Towards Autonomous Machine Intelligence](https://openreview.net/forum?id=BZ5a1r-kVsf), a high-level description of the architecture he considers most promising to advance AI capabilities.
This post summarizes the archite... | https://www.lesswrong.com/posts/GrbeyZzp6NwzSWpds/safety-implications-of-lecun-s-path-to-machine-intelligence |
# A summary of every "Highlights from the Sequences" post
1
=
I recently finished reading [Highlights from the Sequences](https://www.lesswrong.com/highlights), 49 essays from [The Sequences](https://www.lesswrong.com/rationality) that were compiled by the LessWrong team.
Since moving to Berkeley several months ago... | https://www.lesswrong.com/posts/pmRxRzgZWHezk7GdR/a-summary-of-every-highlights-from-the-sequences-post |
# QNR Prospects
Approximately a book review: Eric Drexler\'s [QNR
paper](https://www.lesswrong.com/posts/FKE6cAzQxEK4QH9fC/qnr-prospects-are-important-for-ai-alignment-research).
\[Epistemic status: very much pushing the limits of my understanding.
I\'ve likely made several times as many mistakes as in my average... | https://www.lesswrong.com/posts/3FRixme9q6MFNCjfb/qnr-prospects-1 |
# All AGI safety questions welcome (especially basic ones) [July 2022]
**tl;dr: Ask questions about AGI Safety as comments on this post, including ones you might otherwise worry seem dumb!**
Asking beginner-level questions can be intimidating, but everyone starts out not knowing anything. If we want more people in th... | https://www.lesswrong.com/posts/auPkxnLb3R9vXjEzo/all-agi-safety-questions-welcome-especially-basic-ones-july |
# Alignment as Game Design
*This essay is a continuation of the* [*Naive Hypothesis approach*](https://www.lesswrong.com/posts/ubdp8qAL8Gfki2pYo/naive-hypotheses-on-ai-alignment) *wherein I attempt to write up and share thoughts on AI alignment before having absorbed much of the existing literature. The goal is to pre... | https://www.lesswrong.com/posts/cEQyKsreistXxFEeF/alignment-as-game-design |
# Resolve Cycles
**Epistemic status**: Anecdotally strong
*This technique was largely developed by Kenzi Amodei in the context of after-workshop followups and pair debugging. It has been refined and iterated, and has proven highly useful to our alumni, but all theorizing is post-hoc and untested, and direct research ... | https://www.lesswrong.com/posts/jiJquD34sa9Lyo5wc/resolve-cycles |
# Why I Think Abrupt AI Takeoff
*(note: Quickly written. I've attempted to number my arguments "formally", but I have no training in this format. Edits/suggestions welcome.)*
1. Ability to generalize is "lumpy": unpredictable -- sometimes lots of inputs lead to little progress, sometimes small inputs lead to lots of... | https://www.lesswrong.com/posts/wv8cKEyaRqLfHqHzZ/why-i-think-abrupt-ai-takeoff |
# Four questions I ask AI safety researchers
Over the last few months, I’ve been trying to develop a stronger inside view on AI safety research agendas.
As part of this quest, I’ve been having conversations with AI safety researchers. I notice myself often asking the following questions:
1. What are you working on... | https://www.lesswrong.com/posts/B5ARfCcwYacZkeSQm/four-questions-i-ask-ai-safety-researchers |
# Examples of AI Increasing AI Progress
Recursive self-improvement is already here.
This point is far from original. It’s been described before, for instance [here](https://www.lesswrong.com/posts/aqTAd7KzsYmHWYdei/why-copilot-accelerates-timelines), in Drexler’s [Reframing Superintelligence](https://www.fhi.ox.ac.uk... | https://www.lesswrong.com/posts/W3tZacTRt4koHyxbr/examples-of-ai-increasing-ai-progress |
# Why you might expect homogeneous take-off: evidence from ML research
*This write-up was produced as part of the SERI MATS programme under Evan Hubinger’s mentorship. It is also my first post on LW, so feedback is very welcome!*
Introduction
============
This article aims to draw a connection between recent ML rese... | https://www.lesswrong.com/posts/RQn45KzN5cojLLb3L/why-you-might-expect-homogeneous-take-off-evidence-from-ml |
# How Interpretability can be Impactful
*This post was written as part of the* [*Stanford Existential Risks Initiative ML Alignment Theory Scholars (MATS) program*](https://www.alignmentforum.org/posts/FpokmCnbP3CEZ5h4t/ml-alignment-theory-program-under-evan-hubinger). *thanks to Evan Hubinger for insightful discussio... | https://www.lesswrong.com/posts/Cj4hWE2xBf7t8nKkk/how-interpretability-can-be-impactful |
# Deception?! I ain’t got time for that!
*Or ... How penalizing computation used during training disfavors deception*.
*This post was written under Evan Hubinger’s direct guidance and mentorship, as a part of the *[*Stanford Existential Risks Institute ML Alignment Theory Scholars (MATS) program*](https://www.lesswro... | https://www.lesswrong.com/posts/C8XTFtiA5xtje6957/deception-i-ain-t-got-time-for-that |
# What should you change in response to an "emergency"? And AI risk
*This post has been recorded as part of the LessWrong Curated Podcast, and can be listened to on* [*Spotify*](https://open.spotify.com/episode/50rRKLATgHhZeb0jsUp162?si=EO0KSvVdT2iatHEDMQu7DQ)*,* [*Apple Podcasts*](https://podcasts.apple.com/us/podcas... | https://www.lesswrong.com/posts/mmHctwkKjpvaQdC3c/what-should-you-change-in-response-to-an-emergency-and-ai |
# Forecasting ML Benchmarks in 2023
_Thanks to Collin Burns, Ruiqi Zhong, Cassidy Laidlaw, Jean-Stanislas Denain, and Erik Jones, who generated most of the considerations discussed in this post._
[Previously](https://bounded-regret.ghost.io/ai-forecasting-one-year-in/), I evaluated the accuracy of forecasts about per... | https://www.lesswrong.com/posts/arveXgFbJwascKtQC/forecasting-ml-benchmarks-in-2023 |
# A distillation of Evan Hubinger's training stories (for SERI MATS)
*This post is a* [*distillation*](https://www.lesswrong.com/posts/zo9zKcz47JxDErFzQ/call-for-distillers) *of Evan Hubinger's post "*[*how do we become confident in the safety of a machine learning system?*](https://www.alignmentforum.org/posts/FDJnZt... | https://www.lesswrong.com/posts/wPLeBqsLgJyFyuTr7/a-distillation-of-evan-hubinger-s-training-stories-for-seri |
# Training goals for large language models
*This post was written under Evan Hubinger’s mentorship, as part of the *[*Stanford Existential Risks Initiative ML Alignment Theory Scholars (SERI MATS) program*](https://www.lesswrong.com/posts/8vLvpxzpc6ntfBWNo/seri-ml-alignment-theory-scholars-program-2022)*. Many of the ... | https://www.lesswrong.com/posts/dWJNFHnC4bkdbovug/training-goals-for-large-language-models |
# Conditioning Generative Models for Alignment
*This post was written under Evan Hubinger’s direct guidance and mentorship, as a part of the *[*Stanford Existential Risks Institute ML Alignment Theory Scholars (MATS) program*](https://www.lesswrong.com/posts/8vLvpxzpc6ntfBWNo/seri-ml-alignment-theory-scholars-program-... | https://www.lesswrong.com/posts/JqnkeqaPseTgxLgEL/conditioning-generative-models-for-alignment |
# Quantilizers and Generative Models
*Thanks to Evan Hubinger for discussions about quantilizers, and to James Lucassen for discussions about conditioned generative models. Many of these ideas are discussed in Jessica Taylor's* [Quantilizers: A Safer Alternative to Maximizers for Limited Optimization](https://intellig... | https://www.lesswrong.com/posts/tz3hoCs2efHjzNYm5/quantilizers-and-generative-models |
# Machine Learning Model Sizes and the Parameter Gap [abridged]
**Summary**: The model size of notable Machine Learning systems has grown ten times faster than before since 2018. After 2020 growth has not been entirely continuous: there was a jump of one order of magnitude which persists until today. This is relevant ... | https://www.lesswrong.com/posts/p62uihoFGH3HwGg4P/machine-learning-model-sizes-and-the-parameter-gap-abridged |
# Launching a new progress institute, seeking a CEO
**Summary:** The Roots of Progress is planning a major expansion of our activities, and we are seeking a Chief Executive Officer to lead the new organization in partnership with me (I will remain Founder & President). We’re taking this step because we see an opportun... | https://www.lesswrong.com/posts/zEBPn5ThEB27XjEja/launching-a-new-progress-institute-seeking-a-ceo |
# Addendum: A non-magical explanation of Jeffrey Epstein
Around seven months ago I wrote the post "[a non-magical explanation of Jeffrey Epstein"](https://www.lesswrong.com/posts/hurF9uFGkJYXzpHEE/a-non-magical-explanation-of-jeffrey-epstein). In it, I make the argument that Jeffrey Epstein did actually kill himself a... | https://www.lesswrong.com/posts/DZoGEHzZNRsMjfpfE/addendum-a-non-magical-explanation-of-jeffrey-epstein |
# Turning Some Inconsistent Preferences into Consistent Ones
*cross-posted from [niplav.github.io](https://niplav.github.io)*
### Epistemic Status
This is still a draft that I [was told](https://schelling.pt/@niplav/108403526373636608) to already post here, which includes working (but very slow) code for one special... | https://www.lesswrong.com/posts/H2SizXNd99FcwLsCC/turning-some-inconsistent-preferences-into-consistent-ones |
# Without specific countermeasures, the easiest path to transformative AI likely leads to AI takeover
I think that [in the coming 15-30 years](https://www.cold-takes.com/where-ai-forecasting-stands-today/), the world could plausibly develop “transformative AI”: AI powerful enough to bring us into a new, qualitatively ... | https://www.lesswrong.com/posts/pRkFkzwKZ2zfa3R6H/without-specific-countermeasures-the-easiest-path-to |
# Are Intelligence and Generality Orthogonal?
A common presupposition seems to be that intelligent systems can be classified on two axes:
* Intelligence (low to high)
* Generality (narrow to general)
For example, AlphaGo is presumably fairly intelligent, but quite narrow, while humans are both quite intelligent... | https://www.lesswrong.com/posts/ijKSLz54NspYe5grg/are-intelligence-and-generality-orthogonal |
# Modelling Deception
*This post was made as part of John Wentworth's* [*SERI MATS*](https://www.serimats.org/) *skill-up phase. I'd like to thank Chu Chen, Stephen Fowler, John Wentworth, and LessWrong's review services for help reviewing.*
When I was a kid, I used to play poker[^a7j4dt36vi] with my brothers. I prid... | https://www.lesswrong.com/posts/PYS2nvffCAyBCTmn2/modelling-deception |
# Marburg Virus Pandemic Prediction Checklist
**Current score: 4/14**
**Last updated: 18 July 2022**
**Summary:**
* [Two Marburg virus infections have been detected in Ghana.](https://www.reuters.com/world/africa/ghana-confirms-two-cases-highly-infectious-marburg-virus-2022-07-17/)
* **Prediction:** Since 1967,... | https://www.lesswrong.com/posts/yYozpuBqzJtKLGJuo/marburg-virus-pandemic-prediction-checklist |
# Ars D&D.Sci: Mysteries of Mana Evaluation & Ruleset
This is a follow-up to [](https://www.lesswrong.com/posts/cesr2yGoZ2y327qKQ/d-and-d-sci-divination-nine-black-doves) [last week's D&D.Sci scenario](https://www.lesswrong.com/posts/9P4yLsfhj2Jjm4yqg/ars-d-and-d-sci-mysteries-of-mana): if you intend to play that, and... | https://www.lesswrong.com/posts/dTxikwQDjBuKkaNvF/ars-d-and-d-sci-mysteries-of-mana-evaluation-and-ruleset |
# Help ARC evaluate capabilities of current language models (still need people)
\*\*Still looking for people as of Sep 9th 2022\*\*
I (Beth) am looking for 3-10 new contractors to work with me and my existing team of 3-4 contractors/interns on probing model capabilities to evaluate how close models are to being able ... | https://www.lesswrong.com/posts/e3j7h4mPHvkRynbco/help-arc-evaluate-capabilities-of-current-language-models |
# Bounded complexity of solving ELK and its implications
This post was written for the SERI MATS program. I thank Evan Hubinger and Leo Gao for their mentorship in the program. Further thanks go to Simon Marshall and Leo Gao (again) for specific comments regarding the content of this post.
The Eliciting Latent Knowle... | https://www.lesswrong.com/posts/8cr9fJnay97GEYPt3/bounded-complexity-of-solving-elk-and-its-implications |
# Abram Demski's ELK thoughts and proposal - distillation
This post was written for the SERI MATS program. I thank Evan Hubinger and Leo Gao for their mentorship in the program. Further thanks go to Evan Hubinger (again), Simon Marshall, and Johannes Treutlein for specific comments regarding the content of this post.
... | https://www.lesswrong.com/posts/kF74mHH6SujRoEEFA/abram-demski-s-elk-thoughts-and-proposal-distillation |
# Spending Update 2022
Every few years ([2020](https://www.jefftk.com/p/spending-update-2020), [2018](https://www.jefftk.com/p/spending-update-2018), [2016](https://www.jefftk.com/p/spending-update), [2014](https://www.jefftk.com/p/spending-over-time)) I like to look over our spending and write something up publicly. ... | https://www.lesswrong.com/posts/GbjMyccyhHFEXcCH7/spending-update-2022 |
# LessWrong Meetup - Gendlin's Focusing
Focusing has long been part of the rationalist toolbox, since it can be seen as a technique to bridge "gut feelings" (system I) and "thinking" (system II), a way to more clearly access knowledge we already have, and possibly fine tune it.
The basics are quite simple, and in thi... | https://www.lesswrong.com/events/rawvyqfaCHzCgjBCZ/lesswrong-meetup-gendlin-s-focusing |
# Sexual Abuse attitudes might be infohazardous
*Content warning: sexual abuse, rape, arguably trivialising thereof.*
---
When I was 11 an older boy used to pull me behind a door in one of the school corridors, trap me there, shove his hand down my pants, and touch my penis.
This happened about once a week for a co... | https://www.lesswrong.com/posts/NuueGqPZdotjMQKLu/sexual-abuse-attitudes-might-be-infohazardous |
# Applications are open for CFAR workshops in Prague this fall!
tldr: **CFAR’s running updated versions of our mainline rationality workshop this fall in the Prague area. **[**Apply here**](https://docs.google.com/forms/d/e/1FAIpQLScnHIPwdDGp1aNyOncfEJAIgl2uWxQvqiI4QOPiIy7_mI-4Gw/viewform)**. **
===
CFAR workshops a... | https://www.lesswrong.com/posts/fJb36xZxqFtpoyRv7/applications-are-open-for-cfar-workshops-in-prague-this-fall |
# Progress links and tweets, 2022-07-19
**Links**
---------
* [The first images from the JWST look amazing](https://www.nasa.gov/webbfirstimages). See also [this comparison with Hubble](https://twitter.com/ianlauerastro/status/1546625603641790465)
* [Helion Energy is hiring. “Clean, continuous energy for 1 cent p... | https://www.lesswrong.com/posts/BSiY6ueMidvcJAini/progress-links-and-tweets-2022-07-19 |
# A daily routine I do for my AI safety research work
I’m a Ph.D student at MIT working in AI safety--mostly interpretability in deep networks. I thought I would very briefly share a list of links that I have bookmarked and go through every weekday. It usually takes me <15 minutes. Since I started doing this, I’ve bee... | https://www.lesswrong.com/posts/NENrmKWoEzXXmQjhr/a-daily-routine-i-do-for-my-ai-safety-research-work |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.