The author specifically calls ML non-human computing and I believe is talking about the advances in ML from a theoretical standpoint. I don’t believe it is possible to lump the human written software used to train models, generate adversarial examples, and perform other tasks on GPUs, as purely non human computing. In fact, a lot of the advances in computing recently are specifically to enable this type of computing to be more efficient: easily parallelize training, analyze bigger datasets, graph databases, and other “big data” technologies.

In any case, software notebooks definitely existed before Jupyter like you say (I’m not familiar with Wolf but Mathematica has a similar idea), but I don’t believe they date back to 1996. I might be mistaken. I don’t see the author mention them in the text.

Mathematica is mentioned and was released in 1988.

Also, I don’t see anything particularly new in how ML computing is organized.

We’ve had vectorization, and massively parallel processors since the 80s.

All that’s new is that those techniques are being used on desktop machines now because of Moore’s law.

I’m not saying there are no new discoveries in ML, just that the engineering isn’t new.