>This thing has 8 motors. It’s easy to imagine that it could have 8 independent batteries and could be able to perform a controlled descent if two fail. Similarly, it could have redundant sensors and use commodity “majority rules” logic circuits where all calculations are run three times. On top of that, they can already cut over to ground control if the computer can’t proceed.

Most of these are a total nightmare to implement. There is no such thing as a "controlled descent" into a dense urban center. Even keeping maneuverability with a single defective motor is a hard task, you need to develop and test for this.

>On top of that, they can already cut over to ground control if the computer can’t proceed.

So ground control can do what? The only reasonable way these could ever safely operate is with a high degree of sensor fusion, information about nearby crafts, their trajectories, obstacles, etc. combined together. You can not put a human into that system and have him make split-second decisions.

>With all that, it’s “just” a software problem.

Passenger plane automation is also "just" a software problem. It is just an enormously difficult software problem with exceedingly high stakes. The problem is not that it is "impossible", but that it is extremely difficult and not worth the risk getting it wrong.

Again, what is the acceptable failure rate for these things crashing into sky scrapers? If it is zero, we are very far away from them being able to take flight.