It's certainly possible that it works that way, but my hunch is that it doesn't, and that what's going on with a project like a self-driving car is as much about assembling good training data and tuning the constants to get the desired behavior as it is about inventing the algorithms. If they are at the point of having actual cars driving on actual roads then the theoretical work must have been finished long ago, and what they are actually doing is not some patentable new idea, but the down-to-earth engineering work of making it useful under as broad a range of real-world conditions as possible.
It has certainly never been true, in my experience, that the actual software development is the easy part, regardless of the quality of the spec. A spec detailed enough to make the software development easy would, itself, be a piece of software; it is precisely that process of reducing the idea to all of its details and harmonizing all of its competing constraints that is the work of software development.
I have no background in machine vision, navigation, or the like, but I have done some embedded software development, and one of my projects was a wearable device which did sensor fusion, data logging, networking, etc. It was a relatively small project, and the core algorithms were basically obvious as soon as we knew what the product was supposed to do, but it still took almost two years to implement it all, test it, see what real world data looked like, tune our constants, and run more tests, etc. There were many practical tasks involving power consumption, peripheral management, network communication, debugging, manufacturing tests, and the like which contained absolutely no new ideas but still had to be implemented in order for the product to succeed.
There may have been a patentable invention in there somewhere, but who would ever license it? Anyone trying to solve the same problem would come up with something similar, and they'd still have 18-24 months of expensive development work to do getting a product designed and ready for market. If we'd patented that hypothetical idea, we'd have contributed nothing to the public sphere; the only value in such a patent would be to make it more difficult for someone else to follow in our footsteps, which is exactly the opposite of the reason the patent system was established in the first place!