[LINK] Robot cars and the fear gap

[Michael]

On 14 July 2016 at 16:22, Karl Auer & Brendan wrote:
<SNIP>

Talk of the trolley problem seems kind of irrelevant to what is likely to
happen in the real world.
How do autopilot systems in a jetliner handle such a quandary?
They do not, in any meaningful sense.
Robot vehicles will likely drive slower in poor conditions than humans, and
will likely drive faster in tight traffic on motorways when 100% of the
surrounding vehicles are also signalling that they are robots.
They will likely be programmed to apply the brakes as best they can and
divert their course if they can do so safely, should a soft bodied human
unexpectedly jump off an overpass in front of them.
Would we say the autopilot was unsafe if somebody deliberately steered a
hang glider into the jetliner's path? What about a normal car with cruise
control engaged?
My expectation is that should an emergency like the over-pass jumper occur,
all vehicles in the vicinity will immediately reduce speed, signalling the
emergency to each other so they can behave in a co-ordinated way.

A lot of the discussion here seems hung up on transition issues (how will
robots handle a drunk in another car!?!) and other interfaces between the
robot and the world.
I think the most likely scenario is we will begin making the world more
accommodating to robots where it makes sense (robot only lanes like bus
lanes in certain high frequency routes, for example) while the robots
improve year after year to adapt to the messy world around them.

During the transition decades the robot function may only operate
autonomously on motorways, perhaps. But that would still be a great benefit
for drivers with a long commute.
Later, we might have cars that operate autonomously all the time, but only
at 40km/h on minor or remote roads that are poorly mapped or don't have
hard shoulders or similar. City drivers might happily put up with a journey
that takes twice as long if they can relax and let a robot drive on the
occasional times they encounter such a road, while the farmer at the end of
the 60km gravel track from town might always choose manual driving to get
him home quicker. And the beacon on the farmer's car signals his approach
well in advance, so the robot car in front slows by 5% 2km ahead of the
overtaking lane so it times it's arrival at the dual lane just in time to
ensure it doesn't hold up the impatient human coming up behind.

And I expect the statistics will show these cars make fewer errors, are
routinely travelling slower than humans when they do face uncertain
situations, so they will suffer fewer catastrophic accidents.
And after a few more years of this, more and more drivers will see that a
relaxing journey with a robot doing the driving beats a stressful trip, so
more trips will be robot driven and more sales of robot vehicles will occur.
And the greater numbers of robot vehicles will benefit from a network
effect as they have to adjust for human drivers less frequently,
accelerating greater safety.

We are right at the start of this period. The existing autonomous systems
are very limited, and with no environmental accommodation for robots yet,
but I can imagine this changing rapidly, and in 10 years, when half the
cars on the road will be newer than now, I can foresee much more robot
involvement. Another 10 years and I wouldn't be surprised to see humans in
control for a minority of journeys.

Regards,
Michael Skeggs