Pre-April-fools linkage

As always, you can see these earlier and more spread out by following my Mastodon account. These linkage posts are just a redundant copy of what I post and boost there to make it easier for later me (and maybe others) to find the same content without having to scroll through miles of other posts to reach it. The “\(\mathbb{M}\)” links go to the Mastodon posts, where there may be more discussion.

• Wackiest use of SAT solvers I’ve seen so far (but I haven’t looked very hard): compressing CSS files by determining which selectors can safely be combined (\(\mathbb{M}\), via). It’s not clear to me whether it actually finds a new combination of selectors that is as small as possible (superoptimization), or whether it merely does some heuristic compaction using the SAT solver to ensure that it preserves semantics. For full programming languages, finding the shortest program with a given output is undecidable, far beyond the capabilities of SAT solvers, but maybe static CSS is simple enough to avoid this issue? There’s some discussion of CSS being Turing complete at stackoverflow but that involves user interaction. Searching for superoptimization and CSS did not find anything promising. Maybe there’s something to do in that direction?

• How to download freely-usable art from the Getty (\(\mathbb{M}\)). I’ve seen stories elsewhere saying that the Getty has suddenly made 88k new images free to download, but that may be a misunderstanding: they started building this free resource over ten years ago, and have been gradually building it up. It reached 100k images in 2016, and 135k in 2019; the linked article says there are now over 160k. The 88k number is what you get now when you search their database for open content; I’m not sure what the discrepancy means.

• I’m currently kind of worried about Google Scholar pushing a new online pdf reader (\(\mathbb{M}\), via), presumably to capture and commercialize your scientific-article reading behavior. Leave Google Scholar alone. It is very useful for full-text literature searches and any efforts to move it in a different direction endanger that usefulness. As for online pdf readers: just give me the pdf as a file and let me read it offline. It’s annoying when journal publishers hide the download link behind their online reader and doing the same thing in Google Scholar isn’t any better. The comment in the via link suggesting that Google is testing the removal of the link for finding other papers that cite a given one is even more worrisome.

• Unusual mathematical unicode characters make it a fun exercise to imagine the motivations of their designers. Examples: ⫡, perpendicular with s; ⫸, triple-nested greater than; ⨒, line integral with rectangular path around pole; ⫓, subset above superset; ⨸, tensor division.

• Combinatorics of the nilpotent cone (\(\mathbb{M}\)), Antoine Chambert-Loir. How many nilpotent matrices are there over a given finite field? The explanation involves partitions of integers.

• My latest Wikipedia Good Article: Descartes’ theorem (\(\mathbb{M}\)). This is the one on the equation for the curvatures of four pairwise-tangent circles famously summarized by Soddy’s poem The Kiss Precise. In honor of the article’s new status, I committed a new stanza summarizing a proof of the theorem (yes, I know the determinant is not actually a norm):

  Four kissing circles’ centers spread
  as sums of inverse bends;
  squared distance into square array
  now Cayley-Menger sends.
  The volume of their convex hull,
  as Bradford saw, must equal null,
  and calculating matrix norms,
  the simplified result then forms:
  The sum of the squares of all four bends
  is half the square of their sum!

• Pappus of Alexandria’s Mathematical Collection, Book III and Book VIII (\(\mathbb{M}\)) is available in a recent open-access English translation by John B. Little of the College of the Holy Cross. Book III is the one that tells us about Pandrosion of Alexandria, the first female mathematician for whom we have evidence. Book VIII “may have originally been a free-standing introduction to mechanics” and this may be its first complete English translation.

• Eur. Math. Soc. Magazine no. 131 (\(\mathbb{M}\)) includes articles on women mathematicians in film, the gender gap in mathematics, and zbMATH beginning to index arXiv preprints, among other topics.

• New preprint by Andrew Gray estimates some 60k academic articles “contaminated” by the use of AI-generated text (\(\mathbb{M}\)), by observing the sudden new prevalence of AI-loved words like “commendable”, “meticulous”, and “intricate”.

• Popular Mechanics on Kevin Buzzard’s proposed project to formalize the proof of Fermat’s Last Theorem (\(\mathbb{M}\)).

• Flat-folded origami meets Arabic tiling patterns: Zellij Origami Tiling inspired by Bourgouin Plate 48 (\(\mathbb{M}\)), Tung Ken Lam.

• My colleague Sandy Irani is giving the Richard M. Karp Distinguished Lecture at the Simons Institute in Berkeley next Tuesday afternoon (April 2), on “Quantum Constraint Satisfaction” (\(\mathbb{M}\)). A livestream is available on a registration basis. A talk recording will also be made available, eventually, but if you don’t want to wait you need to register.

• Millions of Wikipedia readers still seeing broken pages as “temporary” disabling of graph extension nears its second year (\(\mathbb{M}\)). Before it broke, this was used for rendering chess game boards, chemical formulas, timelines and charts, etc. Meanwhile, Wikipedia’s svg-to-bitmap conversion (maybe needed to avoid security issues with displaying images as direct svg files) is broken in many cases because they have been stuck for years using an obsolete version of librsvg. Meanwhile, the less said the better about Wikipedia’s perpetual insistence on converting mathematics formulas to badly-server-rendered images, or maybe begrudgingly to badly-browser-rendered mathml, instead of even providing the hooks that would let users swap in better-rendered and long-working methods like mathjax and katex. Which kind of raises the question: As a commenter on the first link writes, Wikimedia is swimming in money. Why is it so difficult to apply some of it to the basic infrastructure of their flagship product?

• Boston University forced to walk back suggestion of dean of arts and sciences Stan Sclaroff, a respected AI/ML researcher, that generative AI could be used as a scab to replace striking grad students (\(\mathbb{M}\), archived link).