Charlie Loyd @vruba@everything.happens.horse

Same general idea and source data, but this time it’s just z-score (https://en.wikipedia.org/wiki/Standard_score), with red positive and green negative. Hard to say that this makes it easier to spot the interesting anomalies, but it seemed like the common sense thing to try. You can see the the OceanGate incident, the earthquake in Turkey and Syria, probably some festivals, etc.

Woops, the tag didn’t work in the original post. Credit to @pnorman for the idea.

The answers to your questions are:

1. Yacht and rally car races.
2. Yes, well noticed! We call it “Null Island”.
3. No, but it would be nice if it were practical.
4. People scraping tiles.
5. Yes, because the distribution of traffic follows Zipf’s law.
6. If I have time.

Inspired by @pnorman’s animation, here’s a year of daily OpenStreetMap tile traffic. One frame per UTC day, from the start of September 2022 through end of August 2023. It’s from the public log aggregates, so some low-traffic tiles are left out for (I assume) efficiency and privacy.

The amount of high-quality open data in these releases is amazing, and it’s been consistently timely.

If I were doing disaster response work at the moment, I would be building heavily on this data.

Hard to think of another company giving out this much top-tier product, under CC-BY-NC and with no sign-up, just to help with disasters.

https://fosstodon.org/@marcpfister/111053289241223580

I wish to break my several-year streak of not complaining about software redesigns to say: The new Slack is in fact bad.

[clears throat] [looks around] The reinvention of potato.

[slightly louder:] The reinvention of potato?

https://www.nature.com/articles/s41422-021-00542-5

(Via @alexismadrigal’s newsletter.)

Could get half a dozen of these for about $200.

Where can I order an enamel pin that says “The LLM discourse is really tangled up and unproductive in ways that are too annoying and complex to even outline on this pin – there are a lot of different aspects to the situation that it would be unfair to leave out of even a short summary”?

This is, to be clear, me subtweeting people getting mad at ChatGPT for using a modest amount of water compared to … almost anything else you could name? Every single almond, like literally one lone almond, is about 3,000 ml of virtual water. A cup of coffee is 100,000+ ml; pee it out into a high-efficiency toilet and that’s a minimum 4,000 ml flush. May I suggest that the rogue billionaires explicitly aiming to stop paying for labor are the problem to focus on here.

Opinions my grandparent had about cars, c. 2000 CE:

1. Electric cars are wildly dangerous because you can’t hear them coming; they will kill dozens of times as many people as gas cars do.

2. It’s hard to get across town these days what with all the traffic caused by big busses.

3. The New Beetle should be banned from public roads since it’s front/back symmetrical: you can’t tell which way it’s going.

Cars are in fact bad but not really for any of those reasons. I remember this sometimes.

And I should say: the way this actually works in practice is you take the timeseries of each pixel in the video, in oklab or your preferred color space, and do an interquartile (or similarly trimmed) mean on it.

Anyway, the trick is, you take a low discrepancy sequence on a sphere (I used R[2] and a Lambert azimuthal projection) and use those angles as lines to sample depth – a measure of centrality – along; you map these angle samples over your points and reduce with min(), and you get an approximation of Tukey’s multivariate depth, which is a generalization of the median (or quantile). I may explain this better eventually but then again I may not.

A wonderful thing about image processing is you’re always discovering completely new and fascinating ways to make pictures look worse.

Made a little thing to take fast approximate multivariate medians, at no small cost to my sanity, and it turns out that when I apply it to pixels in California Landscapes time lapse videos (e.g., https://onewilshire.la/@CALandscapeBot/110699293237082262) it’s mostly just Compression Artifact Finder 2023 Pro Edition.

Partly I think the Anti faction seems to be doing a reasonable job of shifting from a “say scary and sweeping things at random to see what sticks” to “let’s try to articulate what actually concerns us here”, and that’s good. I am still not a card-carrying member of that camp, but I think it’s being a lot more honest now than it was in March, and I respect that.

I still think the discourse around LLMs right now is bad. By which I mean: most of the stuff I see on my feeds, boosted by people I like and whose ideals I tend to share, seems intellectually lazy, politically shortsighted, rhetorically misdirected, etc., in ways worse than we accept on other topics.

And yet! The average quality has gone way up over the last few months. And that gives me some hope.