Don’t we all use this to suggest new stuff to buy? Don’t we all use this to give you a better “feed” of whatever we’re pushing out as “content”? Don’t we use this to know in advance where you’ll be or what movie to watch? Don’t we use this to try and forecast finances?
The problem with supervised learning is the data sets. The problem is twofold:
- where do you get all that data?
- do you know what you want?
The use of the ML platforms of today I think, is more of the way alpha versions or developer preview versions work, to test the stuff and see what can be done with it, because we either know what to do with ML but it can’t do it yet, or we don’t know what ML can do for us.
Then there is also the business problem. For small companies ML in its current capability range doesn’t give much of a return. Small dents in some metrics, maybe. It works wonders for some giants but not for small and medium players.
Then, there is also the farther future, who will be the winning platform? Because ML and AI (not SAI) will be a service, you will not get it as a stand alone system because as the tech is now it is at the stage of the original IBM room sized computers. So when performance and optimization and the combination of algorithms will be good enough to be used by the mass markets some of the platforms you enumerated will get the crown jewels: subscriptions by the millions, some will get the next best thing, enterprise data input and some will get the stick and get wound down.