Machine Learning: The Great Stagnation

Not sure that it's as much stagnation as commoditization.

Won't they be much fewer in number, though?

A number of industrial domains have largely benefitted from this tool and outsiders can even set up their own virtual r&d lab at almost no upfront investment these days. As usual, a tool is (still...) only a tool, and it (still, just...) needs domain knowledge to work in full and produce noticeable ROI?

Your qualification of value destroying/value-creating makes little sense to me. If I need a news article or some other webpage translated I can do it thanks to Google Translate. If Google Translate was not here, I would simply not get it translated and lose the information, there is no way I would contact a translation firm for that kind of stuff. To me it is firmly value creating.

At any rate here are way more than 3 other uses of DL today off the top of my head:

* Autocompletion (be it in search engines toolbars or in Gmail/Word/...)

* Superresolution GAN, the most interesting example to me being NVIDIA DLSS, you render a game at ~720p or less and then upscale it to the target resolution of 1080p or 4k, allowing to get quality that the machine would not have been able to support at the target resolution directly.

* Image recognition/tagging: Most of this is used in the security domain, but there is also a lot of stuff around inventory management, safety etc.

* Semantic search

* Protein folding (AlphaFold)

* In astrophysics: detection of supernovaes, FRBs and probably a bunch of other stuff I'm not aware of

* Self-driving cars: Even assuming self-driving technology does not evolve anymore from now on, the current state of the art is still a selling point.

* Predictive maintenance: Used for plane engines and other things

This seems fair to me. The executive view of ML is "can you do me a magic?" And as this article's "Graduate Student Descent" bit makes clear, the worker response is often to semi-randomly perturb code, show some graphs, and say, "Is this a magic?"

For me, most software development is about finding something boring and laborious. We get a computer to do the work so humans can level up and work on something requiring actual thought. That requires getting a deep understanding of the actual work.

Some of that definitely happens in well-run ML projects. But there's a bunch of Silver Bullet Syndrome stuff going on, where ML's shiny results and magazine articles lead to inflated expectations and inflated claims of success. A fellow nerd says, "I did an algorithm!" Some turns that into an impressive presentation with claims of X% gains in the Key Business Metric, hallowed be its name. In reality, it's more plus or minus X% when you account for externalities, natural variation, and actor adaptation. But that's ok, because by the time anybody finds out, attention is elsewhere.

That's not to really blame ML for that. For a period years ago, I kept getting asked, "Can we use a wiki for that?" I would start an explanation of what it actually takes to make a wiki work (hint: it's not the software). Their eyes would glaze over in short order, because they realized that it would take actual work. So many people want the silver bullet, the magic pill. Especially people in the managerial caste, as the reigning dogma there is that management is a universal skill. Details are for the little people.