Reply to post:

I for one welcome our robot overlords.

Not quite so fast! Where AI is going is for proprietary AI systems to be sold to the world as a service. The world of machine learning is developing nicely, but the developers of these systems are being routinely bought out by big tech giants. In most cases Google excepted) the tech giants don't have the data to run AI against (by definition, if the problem's big enough to benefit from AI/ML, then it needs vast and multiple data sets to run against). And not just one data set - if all you've got is a load of customer history data, then algorithmic approaches work just fine, no matter how big that data set is.

Now, the experts are of the view that open data and open research into AI is the optimal way forward (as Amanfrommars1 suggests in his own unique style below), but in practice that's not going to match business needs. Some aspects may be patented, but I think it is more probable that companies will evolve their AI systems to as near general purpose problem solvers as is feasible, but keep the method and underlying designs resolutely confidential, and then rent them out as a black box. This also fits with the fact that this isn't cloud computing, it is massively parallel, fabulously low latency recursive computing on expensive, complex and dedicated machines. The customer company sends the their data to the tech company, who feed it into the system and then help the client interpret the models and results.

Because each situation is different, the output models will be unique to each situation, and that doesn't lend itself to traditional IP protection frameworks.