During the week, I often collect and share links I find interesting on Elder Research’s Slack. These are some links shared March 17–21, 2025.
My Thoughts on the Future of “AI”
I have very wide error bars on the potential future of large language models, and I think you should too.
I’m not a language-model or AI researcher in any meaningful way, so I can feel like I’m getting whiplash between the ‘AGI is imminent’ and ‘AI is worthless’ camps. A wide set of error bars is a mental model I find helpful. Will be reading more of Nicholas’s writing.
Defense Against Dishonest Charts
This is a guide to protect ourselves and to preserve what is good about turning data into visual things.
This is a good idea, and the interactivity provides clear demonstrations of the different principles.
State Space Models & Structural Time Series, with Jesse Grabowski
This episode was mentioned in a discussion of time-series and state-space models. The particular episode is fine, but I’ve gotten a lot from the Learning Bayesian Statistics podcast overall in the last few years. I remember particularly enjoying the episode on principled modeling with Michael Betancourt.
Stop using the elbow criterion for k-means and how to choose the number of clusters instead
I don’t do much in the way of clustering, and I figured the “elbow criterion” was the right one because it’s what everyone is doing. Here are some alternatives that don’t seem too difficult to implement.
fastplotlib: Next-gen fast plotting library
GPU-accelerated plotting that supports Metal!
How Fast the Days Are Getting Longer
A fun tour of the physics and mathematics of a simple question: “How much longer are the days getting?” Like many ‘elementary’ questions in physics, the zeroth-order answer might be pretty simple, but there are many, layers of complexity as precision becomes more important.
I’ve always had a particular fascination for the trigonometric functions that go to infinity and how, in physics, these map onto physically unrealizable states.