I think both Tesla, Google and others won't get to self driving cars at all, because they work from the faulty assumption that 'self driving' means that the car needs to do all the work.
Whereas 'self driving' only needs to mean that the human driver no longer has to do the work.
The easiest and cheapest way of ensuring the car has all the right information is to ensure it runs on tracks of data collected and provided by the environment mostly. Certainly not collected solely by the car.
The car has by definition a limited sensor range, the environment too but isn't on the move so can always be completely aware within its own radius.
That blind spot ensures that the self driving time horizon is moving backwards as fast as the projects progress, and has been for at least a decade.
To do get there one needs to fully take on board that:
The car is not the sole locus of sensing, everything else is a way more relevant locus of sensing
The car is not the sole source of data, it's more likely the smallest source of data, relating only to its current intentions in order to broadcast that to everything else.
The car is not the sole unit of decision making, it's more likely it needs to be the recipient of mostly outside instructions.
As I blogged in frustration in 2015 https://www.zylstra.org/blog/2015/10/why-false-dilemmas-must-be-killed-to-program-self-driving-cars/
I think both Tesla, Google and others won't get to self driving cars at all, because they work from the faulty assumption that 'self driving' means that the car needs to do all the work.
Whereas 'self driving' only needs to mean that the human driver no longer has to do the work.
The easiest and cheapest way of ensuring the car has all the right information is to ensure it runs on tracks of data collected and provided by the environment mostly. Certainly not collected solely by the car.
The car has by definition a limited sensor range, the environment too but isn't on the move so can always be completely aware within its own radius.
That blind spot ensures that the self driving time horizon is moving backwards as fast as the projects progress, and has been for at least a decade.
To do get there one needs to fully take on board that:
The car is not the sole locus of sensing, everything else is a way more relevant locus of sensing
The car is not the sole source of data, it's more likely the smallest source of data, relating only to its current intentions in order to broadcast that to everything else.
The car is not the sole unit of decision making, it's more likely it needs to be the recipient of mostly outside instructions.
As I blogged in frustration in 2015 https://www.zylstra.org/blog/2015/10/why-false-dilemmas-must-be-killed-to-program-self-driving-cars/