Data is a substitute for doing engineering by hand. If you do the latter, you require less data.
So that's wrong and untrue, as you recognize in this latest comment by saying you can partially define F. To make a concrete example, the convolution operator, the number of layers, their connectivity--all of that partially defines F. Finding a structure is what allows you to learn from the x/y pairs instead of just storing them in a look-up table.
It makes sense, then, that finding a better structure may allow you to learn from fewer x/y pairs. This is Apple's purported approach.
It also means that if you have a huge number of x/y pairs, you might benefit from (or get away with) using a simpler structure. This is Google's approach in the domains where they have a lot of data.
It is true that an important lesson has been "when designers build structure into machine learning systems, they often build in the wrong things, and they end up making commitments that the data cannot undo". But there's still a lot of structure that we put into machine learning system, and we're still learning about what good structures are for different applications, and different regimes of data.
That said, a dynamical system may exhibit chaos for one setting of the parameters, and may not exhibit chaos for other settings of the parameters. It may be true that the equations governing traffic flow have the potential to exhibit chaos, and do exhibit chaos when 1. humans drive, or 2. the drivers have only local sensing. But that doesn't mean it would still be true for driverless cars.
But again, that's not the issue at play. We wouldn't fault an driverless car for getting stuck in traffic. It's more like it has to work when it's snowing, when you can't see the lane markers, when a kid in a trashbag costume jumps into the street, etc. Those are the edge cases that we care about right now.
But in the more general case, the optimal solution for humans and the optimal solution for machines is not necessarily the same. So in driving, for instance, it may be the case that operating a vehicle in one way is safest for a human - yet not necessarily so for a machine. It's not hard to come up with examples. We are awesome at intuition and awful at microsecond reactions. Computers are awful at intuition and awesome at microsecond reactions. The optimal style of 'play' is likely going to differ. Yet if the machine is trained based on human data it would end up trying to shoehorn itself into replicating human behavior which may be suboptimal for it.
All that said, this sounds more like marketing talk. My guess is that Apple simply decided that self driving is extremely difficult and is going to come with some major growing pains which could negatively affected their crafted brand image, making it generally a poor investment of time and resources.
Er.... that's not how it works. As a Valve employee put it (a talk about ML recognising cheaters), machine learning is just an equation:
y = f(x)
where you can actually derive the FUNCTION f if you have enough x/y pairs. Without those pairs then most of the successful techniques in machine learning fail to function.
Yes you can generate your own data but generating realistic stuff is a huge can of worms.