I felt the same about that picture, the original source is less pretty but much easier to read: https://commons.wikimedia.org/wiki/File:MatrixMultComplexity...

It made me think of the progression of men's 100m dash times. I was really hoping it would line up better. I think long jump might have been a better choice for a spurious correlation graph.

https://imgur.com/a/5quyLWC

I think I cracked it: it lags about 30 years behind the long jump progression

https://imgur.com/a/YHvXq7G

I definitely agree, but thinking about it I wonder if it was simply to get a "gut reaction" to the graphic - that things are getting better and we are reaching the limit of what is possible (in the context of matrix multiplication).

It's easier to get that feeling at a glance when the graphic is increasing left-to-right. With normal conventions the graphic would be decreasing. Which of course, that's what it is doing (and the goal). But the "gut reaction" to a decreasing graph is "things are doing worse". The gut reaction to graph increasing is "things are doing better".

I could be completely off-base, but that's the only reason I can think of to eschew normal axis conventions.

So put time on x axis and exponent on y, but make the exponent decrease from bottom to top. They've already reversed the x axis here, so they could've just done that for y instead.

Yep, that would be another option and perhaps a more suitable one to achieve the same effect -- assuming my guess as to "why" is even correct in the first place.

>What is with that banner image?

You beat me to it.