Statistical Literacy

In the summer of 2024 Metaculus reached out and invited me to participate in an ai Forecasting Benchmark Tournament.11 This is not about forecasting ai progress, to be clear, but about programming an ai to make forecasts. At the time, I didn’t build anything, because I had too much else going on. I did get around to it back in October, though, and I call the result the Basean Loom.

I recently published my acx book review contest 2024 submission on Savage Money. My 2023 submission last year was a review of War and Peace. The review did not nail the notes I wanted it to11 I’m still searching for that. but it turned out good anyway. Have a hot beverage, stake out some time for a long review, and Merry Christmas!

During the second world war, the Germans invented the cruise missile and sent free samples over the channel to London. The missiles appeared to hit some places in London more than others (train tracks?), and the British government was worried the Germans were somehow able to target the missiles effectively.

In the future correlations article we will learn the following transformation of the correlation coefficient33 The logit function captures the idea that a difference in correlation between 0.95 and 0.98 is more difficult to prove than e.g. a difference between 0.2 and 0.3.:

Looking through the articles I’ve started but never finished, I stumbled over a cool experiment from a few years ago: When a build in a ci environment fails, it usually does so quicker than the time required for a full, successful build. This means we ought to be able to find the fraction of failed builds from the build times alone.11 Don’t ask when this would be useful. I don’t remember.