Incoming links and notes
March 5, 2020 — February 26, 2025
Assumed audience:
Me
Things that I think should be noted and filed in an orderly fashion, but which I lack time to address right now. Content will change incessantly.
1 To subscribe to
- Michael Levin—The Future of Intelligence: Synthbiosis — Artificiality: Minds Meeting Machines — Overcast
- https://experiencemachines.substack.com/
- https://www.elysian.press/
- https://englishinprogress.substack.com/
- The Roots of Progress | Jason Crawford
- Existential Crunch | Florian U. Jehn
- Governing with AI | Justin Bullock
2 To post to team chat
glama “All-in-one AI workspace”
Is it safe to travel to the United States with your phone right now? | The Verge
Susan McKinnon Foundation - Strengthening Australia’s Democracy
[2502.21098] Re-evaluating Theory of Mind evaluation in large language models
A Bear Case: My Predictions Regarding AI Progress — LessWrong
Can Transformer Models Generalize Via In-Context Learning Beyond Pretraining Data? | OpenReview
Machine Learning as an Experimental Science | Machine Learning
Preparing for the Intelligence Explosion | Forethought
In fact, there are ways we can prepare for these challenges, today. For example, we can:
- Prevent extreme and hard-to-reverse concentration of power, by establishing institutions and policies now. For instance, we can ensure that data centers and essential components of the semiconductor supply chain are distributed across democratic countries and that access to frontier AI continues to be available to many parties, both within and across countries.
- Empower responsible actors: Increase the chance that those actors who wield the most power over the development of superintelligence (such as politicians and AI company CEOs) are responsible, competent, and accountable.
- Build AI tools to improve collective decision-making. We could begin building, testing, and integrating AI tools which we want to be in wide use by the time the intelligence explosion is underway, across epistemics, deal-making, and decision-making advice.
- Remove obstacles to applying superintelligent AI to downstream challenges, for example by unblocking bureaucracy which makes it unnecessarily difficult to use advanced AI tools in government.
- Get started early on institutional design for new areas of governance, including on the rights of digital beings and the legal framework for property claims on offworld resources.
- Raise awareness and improve our understanding of the intelligence explosion and each of the challenges that follow from it, so that it takes less time to get up to speed at the crucial moments when decisions do need to be made.
From Design doc to code: the Groundhog AI coding assistant (and new Cursor meta)
“Torrenting from a corporate laptop doesn’t feel right”: Meta emails unsealed - Ars Technica
The Unpredictability of Reasoning Isn’t a Flaw but a Feature | by Carlos E. Perez
[2412.01786] Hard Constraint Guided Flow Matching for Gradient-Free Generation of PDE Solutions
The Best Tacit Knowledge Videos on Every Subject — LessWrong
-
What does it mean to give a model a capability? And while we’re on the subject, how do you give a model a capability?
I’m talking about AI progress in the past year again. Contrary to the expectations that GPT-4 set in 2023, research labs emphasized the “capabilities” of the models they released this year over their scale.
Moon – Bartosz Ciechanowski — incredible explorable
The Math Academy Way: Using the Power of Science to Supercharge Student Learning
Adam Mastroianni, Underrated ways to change the world
Gordon Brander, The Zombocom Problem
Make better documents. - Anil Dash
Even very smart, capable communicators routinely send important documents that distract from, or even undermine, their goals.
Experiment metascience grant
3 Links
BWNEO Resource List on communications patterns that minimise cancellation dynamics and maximise bridge building
shot-scraper: A command-line utility for taking automated screenshots of websites
Susan McKinnon Foundation - Strengthening Australia’s Democracy
Dan Simpson on GPs
The Rationalist-Etcetera Diaspora: A SPREADSHEET!! — LessWrong
Clio: Privacy-preserving insights into real-world AI use Anthropic
Artificial Analysis: Comparison of AI Models across Quality, Performance, Price
Nvidia unveils $3,000 desktop AI computer for home researchers - Ars Technica
-
Mathematics was designed for the human brain.
-
Janet is a functional and imperative programming language. It runs on Windows, Linux, macOS, BSDs, and should run on other systems with some porting. The entire language (core library, interpreter, compiler, assembler, PEG) is less than 1MB. You can also add Janet scripting to an application by embedding a single C source file and a single header.
Seems to work on webassembly too? e.g. Bauble
Multistakeholder Engagement for Safe and Prosperous AI - Future of Life Institute
Infrastructure alignment problem: How Madrid built its metro cheaply - Works in Progress
The Rising Sea is Dan Murfett’s blog
Beyond Message Passing: a Physics-Inspired Paradigm for Graph Neural Networks
Project Sid: Many-agent simulations toward AI civilization / altera-al/project-sid
An Overview of “Obvious” Approaches to Training Wise AI Advisors – AI Impacts
Alexithymia, also called emotional blindness, is a neuropsychological phenomenon characterized by significant challenges in recognizing, expressing, feeling, sourcing, and describing one’s emotions.
Free Anagram Sentence Generator for English, German, French, and Spanish
Reflections on 2020 as an independent researcher | Andy Matuschak
cloud functions
Kate Wagner’s piece Behind F1’s Velvet Curtain was taken down shortly after publication
The Shift from Models to Compound AI Systems – The Berkeley Artificial Intelligence Research Blog
Interesting synth: Vital
Ask a Neoliberal: An Interview with J. Bradford DeLong - Dissent Magazine
GitHub - ml-explore/mlx: MLX: An array framework for Apple silicon
-
Argues that for a given limited event budget, closed captioning is more inclusive than hiring a signer to interpret, since many deaf people do not sign, plus captions assist autistic and non-native English speakers.
Science communicators need to stop telling everybody the universe is a meaningless void
Accelerating Natural Gradient with Higher-Order Invariance | Yang Song
How to Find the Taylor Series of an Inverse Function - Randorithms
Remix - Build Better Websites — distributed, rather than static, site rendering. backed by Shopify.
f-DM: A Multi-stage Diffusion Model via Progressive Signal Transformation | OpenReview
[2309.10068] A Unifying Perspective on Non-Stationary Kernels for Deeper Gaussian Processes
Monitoring, Streamlining and Reorganising Work with Digital Technology
[2302.02947] GPS++: Reviving the Art of Message Passing for Molecular Property Prediction
CodaLab Competitions: An Open Source Platform to Organize Scientific Challenges
Pushover: Simple Notifications for Android, iPhone, iPad, and Desktop
Half-Truths (at Best) about Calculus of Variations and Optimal Control
“VC qanon” and the radicalization of the tech tycoons - Anil Dash
Smart Countdown Timer is a good timer.
Smart Countdown Timer allows you to use natural language to set, modify and start a countdown on your Mac. Our simple and easy to use UI just requires you to enter your countdown time using plain English, such as ‘1 hour and 35 mins’ or ‘add 25 mins’.
-
Agenda’s unique approach of organising notes into a timeline helps to drive your projects forward. While other apps focus specifically on the past, present, or future, Agenda is the only note-taking app that tracks them all at once, giving you the complete picture.
The first AI model based on Yann LeCun’s vision for more human-like AI / [2301.08243] Self-Supervised Learning from Images with a Joint-Embedding Predictive Architecture
BIMLOGIQ is where Amir Dezfouli went to work.
SPIGM @ ICML ICML 2023 Workshop on Structured Probabilistic Inference & Generative Modeling
Flyte: An Open Source Orchestrator for ML/AI Workflows - The New Stack / Build production-grade data and ML workflows, hassle-free with Flyte
Sam Kriss, All the nerds are dead, conflates geeks and nerds but is funny anyway.
The reasonable(?) effectiveness of data analysis
Why is it that we can be thrown into the work of other people, in a field we have zero experience in, and have any expectation of making any useful impact at all? When stated objectively, it sounds utterly ridiculous. But in my experience, a data team can find something to improve, even if the impact is sometimes small.
Mastroianni’s favourite Underrated ideas in psychology
I esp like
Annie Lowrey, We Haven’t Been Measuring How the Economy Really Works
TIL Apophenia vs Pareidolia
Matthew Feeney, Markets in fact-checking
Jason Collins, We don’t have a hundred biases, we have the wrong model
factorization_machine Something something kernels, something regression something interaction effects. 🚧TODO🚧 clarify
Making Friends with Machine Learning is Cassie Kozyrkov’s lecture series.
Prof Steve Keen | Creating realistic economics for the post-crash world
-
Kolibri is an open-source educational platform specially designed to provide offline access to a wide range of quality, openly licensed educational resources in low-resource contexts like rural schools, refugee camps, orphanages, and non-formal school programs.
Cult Classic ‘Fight Club’ Gets a Very Different Ending in China
In Which Long-Time Netizen & Programmer-at-Arms Dave Winer Records a Podcast for Me, Personally
Nicholas Gruen, Democracy: forking the project
Ethan Epperly, Low-Rank Approximation Toolbox: Nyström Approximation
The Australian academic STEMM workplace post-COVID: a picture of disarray
Merve Emre, Has Academia Ruined Literary Criticism?
Tom Stafford, Microarguments and macrodecisions
Kevin Munger, Why I am (Still) a Conservative (For Now)
Randy Au, in Data science has a tool obsession, talks about Gear Acquisition Syndrome for data scientists.
danah boyd, What if failure is the plan?
Adam Mastroianni, The great myths of political hatred
Big correlations and big interactions ([2105.13445] The piranha problem: Large effects swimming in a small pond)
How to keep cakes moist and cause the greatest tragedies of the 20th century
Microsoft CSR’s Law Enforcement Request Report is disconcerting transparency
Marc ten Bosch, Let’s remove Quaternions from every 3D Engine (An Interactive Introduction to Rotors from Geometric Algebra)
Ti John’s Publications
Zoomers Co-Working Community (co-working for accountability)
Oshan Jarow, Markets Underinvest In Vitality
Darren Wilkinson’s Bayesian inference for a logistic regression model 1, 2, 3, 4, 5
https://www.patreon.com/posts/taylor-modeling-70131808
Book Review: Public Choice Theory And The Illusion Of Grand Strategy
https://downforeveryoneorjustme.com/
Census is a tool that links all the weird different data storage systems and CRM stuff.
Nemanja Rakicevic, NeurIPS Conference: Historical Data Analysis
Samuel Moore, Why open science is primarily a labour issue
Machine Learning Trick of the Day (1): Replica Trick— Shakir Mohammed
What’s the difference between a tutorial and how-to guide? - Diátaxis
https://cbergmeir.com/
http://i.giwebb.com/
Francis Bach Going beyond least-squares – II : Self-concordant analysis for logistic regression
On the Generalization Ability of Online Strongly Convex Programming Algorithms
4 Illustrations
5 Homeless links
Bookmarked but where will they ever go?
Dispel your justification-monkey with a “HWA!” - Malcolm Ocean
Roger’s Bacon, Living and Dying with a Mad God
Washable & breathable flexiOH cast adapts to the patient’s skin
-
Today I took a desk lamp whose Halogen light had burned out, whose crappy transformer always made those bulbs sputter, and whose mildly art-deco appearance I’d always liked, and swapped it out to run an LED bulb off USB power. It took about an hour’s work to replace the light with an LED, the switch with a nice heavy clicky one, and now the whole thing runs off USB-C instead of wall voltage. It emits no appreciable heat, and if these calculations are to be believed, will run for decades for a few cents per year, assuming I leave it on all the time.
I hadn’t really appreciated how big a deal USB-PD voltage negotiation was until I found out that the little chips that handle that negotiation are about the size of the end of a pencil, and that if you include the USB-C port, you can replace basically any low-voltage transformer with something smaller than a quarter.
The magic search string, if you want to try this yourself, is “usb-pd trigger module.”
vscode-paste-image/README.md at master · mushanshitiancai/vscode-paste-image
mhoye/awesome-falsehood: 😱 Falsehoods Programmers Believe in
Communications’ digital initiative and its first digital event
Playable Half Earth Socialism simulator
flatmax/vector-synth: Old 2002 era vector synth code based on XFig
Nick Chater, Would you Stand Up to An Oppressive Regime
Lambda School’s Job Placement Rate May Be Far Worse Than Advertised
State Power and the Power Law, State Power and the Power Law 2
Liquid Information Flow Control, a confidential computing DSL
Kostas Kiriakakis, A Day at the Park
-
pyribs is the official implementation of Covariance Matrix Adaptation MAP-Elites (CMA-ME) and other quality diversity optimisation algorithms.[…]
Quality diversity (QD) optimisation is a subfield of optimisation where solutions generated cover every point in a measure space while simultaneously maximising (or minimising) a single objective. QD algorithms within the MAP-Elites family of QD algorithms produce heatmaps (archives) as output where each cell contains the best discovered representative of a region in measure space.
Is Pandemic Stress to Blame for the Rise in Traffic Deaths? Nope, apparently it is decreased congestion making drivers drive faster on shit roads.
Black Americans are pessimistic about their position in U.S. society
A study of lights at night suggests dictators lie about economic growth
Paint With Music — Google Arts & Culture turns brushstrokes into melodies.
Chalk is a good calculator for macOS, including useful things like matrices and bitwise ops.
Clive Thompson, The Power of Indulging Your Weird, Offbeat Obsessions
Victoria Falconer is incredible to watch.
-
River is a visual connection engine. Clear your mind and surf laterally through image space. May contain NSFW content. Right-click any image to open it on Are.na for the source, full resolution, and human-curated connections.
SoSha | Social Media Engagement for Communities, Employees, and Partners
Combining embeddable sharing buttons, advanced analytics, and AI-powered content creation to provide the highest-performing organic social media amplification platform.
He Fought for Freedom. Then He Chose Prison.; on Jimmy Lai
Indivisible: A Practical Guide to Democracy on the Brink - Google Docs
The Artist Who Trained Rats to Trade in Foreign-Exchange Markets
Specific Suggestions: Simple Sabotage for the 21st Century (Posting this link is… not advice)
A ‘Second Tree of Life’ Could Wreak Havoc, Scientists Warn - The New York Times
-
A project called CETI — Cetacean Translation Initiative — was launched to help scientists understand the language of sperm whales.
ClearerThinking.org’s courses, e.g.
- Introduction to Decision Academy: The Science of Better Decisions
- Rhetorical Fallacies: Dodging Argument Traps
- Learning from Mistakes: A Systematic Approach
- Probabilistic Fallacies: Gauging the Strength of Evidence
- Explanation Freeze: Interpreting Uncertain Events
- The Sunk Cost Fallacy: Focusing on the Future
- When to Stop Exploring | The explore-exploit trade-off
2023’s best global tech stories we wish we’d written - Rest of World
Christian Lawson-Perfect’s Interesting Esoterica is a collection of weird papers in math.
Office for the Preservation of Normalcy - I feel confident enough to post these now. Cosmic horror warning posters.
How the Kayin BGF’s business interests put Myanmar at risk of COVID-19. The ethnic border militias of Myanmar sound fascinating.
I would like to read the diaries of Usama ibn Munqidh