This site (QuantaMagazine) had really excellent quality content when it first launched,like the one at the OP, now it has become sort of IFLScience in terms of editorializing.
It is a theoretically interesting result about the PPAD complexity class itself. Not sure if it should be STOC best paper though, if it was nominated as best paper because of the weak link to machine learning I'd be a bit disappointed.
The recent ones haven’t remained in the same pathway unfortunately.
> ... elevation of the land is equal to the value of the function (the “profit”) at that particular spot. Gradient descent searches for the function’s local minimum ...