The Dangerous Dazzle Of The Use Case

When the use case drives a technology project rather than the pace of technological development, you get trouble.

The driverless car thing got sold to governments because it would:

Let cars drive closer together so reducing the need to build so many roads;
It would reduce accidents so reducing medical and clean-up costs;
It would give governments control over car drivers.

In making their case to governments, companies over-promised on the tech.

Having over-promised on the tech, companies are trying to push driverless technology further and quicker than they should.

Uber is said to have reduced the Lidar units in its driverless car from seven to one before killing the lady in Arizona.

The remaining single Lidar was mounted on the roof and its Velodyne sensors, while effective for 360 deg lateral vision, have limited vertical vision which is said to have created blind spots around the perimeter of the lower car body.

It was known, it was warned against, but Uber persisted with it, dazzled no doubt, by the prospect of driverless taxis.

Uber makes a $1 billion a quarter loss, it is said to only recoup 40% of the cost of a ride, Uber’s take per trip is estimated to be 40% of the fare.

Without the driver, voilà, the company breaks even.

That’s a heady prospect for Uber but not worth a human life


Comments

8 comments

  1. Ha! Good point Mike. Wonder why Elon ever said it then.

    • Same reason as he thinks he can go to Mars – he forgets there are rules in place to restrict activities hazardous to the human race.

  2. Very interesting Ian, many thanks for that.

  3. Reducing the number of Lidars from seven to one is a red herring being used to avoid the big problem. The close-in blind spots of a single Lidar are irrelevant at the distances involved in the collision and anyway are filled in by other sensors, one Lidar is 100% capable of spotting a pedestrian with a bike at this distance — as are radars and cameras, all of which the car was fitted with.

    The biq problem for Uber is that it looks like the software was either not using the sensors, ignoring the result, decided to take no action, or was partly inactivated — in any case the system should have been perfectly capable of detecting the pedestrian and take avoiding action (this isn’t a difficult “edge case”, it’s a perfectly common one which should have been simulated many times), but didn’t.

    Nvidia showed that their system successfully detected the pedestrian based only on the camera video released by the police, without the benefit of Lidar or any of the other sensors. Uber screwed up big time with their system/software/AI, and trying to blame a single Lidar is just distraction tactics.

  4. Personally I’d say that reducing the Lidars from 7 to 1 was pushing the technology to do more than it can usefully do, DontAgree, and, of course, naturally I say I’m a realist about driverless cars not a nay-sayer. I see Musk promised us an LA – New York driverless run by the end of 2017. Dibn’t happen. Why not, DontAgree?

    • Well among other reasons, there’s a large number of states in between CA and NYC who don’t allow the testing of driverless cars. Sometimes the MidWest can be quite sensible !

  5. Sorry but fail to see the relation between “reduced the Lidar units in its driverless car from seven to one” and “trying to push driverless technology further and quicker”.

    Reducing the number of Lidar units is a cost saving matter only, nothing to do with pushing technology faster.

    Then again such jumping to the wrong conclusions doesn’t surprise me from someone long time on record as a nay-sayer w.r.t. driver less cars …

Leave a Reply

Your email address will not be published. Required fields are marked *

*