[LINK] Robot cars and the fear gap

[Jim Birch]

BRD wrote:

So what will the court's reaction be if nobody can understand it?
>

No one understands the neural networks of humans in absolute terms, yet
courts can deal with their actions. :)  The questions that arise with self
driving cars are akin to a captain putting a ship or plane on autopilot.
Was this appropriate use of the system?  Was the system properly
maintained?  Did the manufacturer follow best practice in design and
construction?  Was it adequately tested?

This type of process applies everywhere.  No one knows how most drugs
work.  Their use entails risks because biochemistry is seriously
incomplete.  A drug manufacturer is not culpable on a drug failure.  They
are culpable if they withheld information on adverse reactions or made
false claims for efficacy.

Tesla label their robot driving software as "beta".   They say that their
software has been shown to be significantly safer ikn is usage domain than
the average driver and this label is to warn drivers that it is not
perfectly reliable.  The say they will remove the label when the software
is ten times safer than the average driver.


> A second question is: will the software be licenced or will the vendor own
> it?
>

Not sure what the implication is.  I would expect, as a practical matter of
management and safety, that the software would need to be continuously
updated.  In this case, the vendor is either running a taxi service (which
I think will end up the most common usage scenario) or selling a vehicle
with the best current software.  In the latter case the owner would be
culpable if they stopped the update process - if this were possible - or
messed with the software or sensors - again, if this were possible.  These
actions would presumably brick your car.  The vendor would be culpable if
he didn't collect usage information, produce updates when required, test
updates, and ensure that all "nodes" receive critical updates.

Jim