but he is right you know.
the car mimics a car driven by a human. but the data set that controls it is not understood by the people who made the data set using training.
perhaps they want money to teach objectives to it? or priorisations? I mean, if you just ran a dataset to teach it to drive, presumably footage of people driving and value logs of inputs happening while they are doing so, that for example in a curve it would try to keep the car on the road by turning right/left, you would still need to teach it the things you need to know to be fit to have a drivers license - but since they don't understand their own generated set they CANNOT MODIFY IT directly. they cannot teach it directly a new traffic sign by adding to a list of traffic signs - instead all of that is somewhere there, inside the mysterious dataset they have gotten out of the box they put a bunch of inputs in. did they make that mystery box? yes? no?
or to research how they could use this(possibly off the shelf mind you, possibly) software / training system. to use it you don't need to understand it, which is really the big problem with a LOT OF SOFTWARE and not just trained matching.
basically what they have is a dog. YOU WOULDN'T WANT A TRAINED DOG driving a car. it can be done though, but neither the dog understands what it is doing or the trainer exactly understands what is going on fully in the dogs head or if it is safe to assume that it will not driver over little children - as such they have a research project they can't do anything with because presumably.. well. if they got that far, how about just attaching a debugger and trying to figure it out in a simulator?
Anyway, you certainly wouldn't call it a real AI, but for investor purposes sure why the fuck not, give me all your AI money.
those folks are in a luck though because the less they understand the more likely it is they can with a straight face speak to the vc's "sure, yeah, this is first real artificial intelligence, it's name is TOM and we want to learn about it, see it's intelligent so thats why we don't already know how it works!"