On 3/20/2018 7:08 PM, Char Jackson wrote:
On Tue, 20 Mar 2018 15:12:19 -0600, Jeff Barnett wrote:

Char Jackson wrote on 3/20/2018 10:07 AM:
On Tue, 20 Mar 2018 19:53:13 +0800, "Mr. Man-wai Chang"
wrote:

Self-driving Uber kills Arizona woman in first fatal crash involving
pedestrian

https://www.theguardian.com/technology/2018/mar/19/uber-self-driving-car-kills-woman-arizona-tempe

[multiple groups removed]

AFAIK, the facts are still being gathered, but I'd be curious to know
how many pedestrians were killed by human drivers over the same period.
If it's more than 1, which I'm assuming is the case, then I'm not
alarmed by this incident other than having sympathy for the deceased and
her family. Initial reports said she was crossing the street, but not in
or near a crosswalk, so I wonder if it would have made a positive
difference if a human had been driving. I hope testing doesn't get
curtailed by this incident.

I believe the number last year was 6000. The real question is how do
auto driven car accident statistics compare with human drivers.

The Uber accident was not necessarily the car/drivers fault. A woman was
walking a bicycle and started to cross the street, not in the crosswalk,
and was hit just as she went into the street. The car was traveling
around 40-45mph. In other words, it was HIGHLY likely she didn't look
before crossing. The municipal police are still investigating and
certainly have not assigned any blame yet. In fact they have speculated
that this might be one of those where no primary blame is asserted.

As someone else said: You should be wary and afraid of those automated
vehicles. But you should be god-awful more afraid of all those idiots
out there jacking off with their smart phones while driving.

My personnel bet is that 5 years from now we will see self drive cars
doing spectacularly better than human drive cars - better safety, better
millage, faster trips - and still a bunch of idiots (the same ones who
opposed autopilots and computer assisted landings for planes) bitching
about the supremacy of human drivers, vinyl records, doctors reading
x-rays, etc, etc, etc.

I'm with you 100%. From everything I've read, the technology is coming
along much faster than I would have ever thought.

There are two major hurdles that I see. The first, of course, is the
technology itself. We already have anti-lock brakes, lane departure
warnings, adaptive cruise control, blind spot monitors, and automatic
parallel parking, oh and 360-degree virtual overhead view on the
dashboard stitched together from multiple exterior cameras, on virtually
all new vehicles. ICBW, but I think all of those things are mandated by
2020. With that much automation already in place, it's a logical (but
difficult) next step to stitch it all together and make it work without
significant human intervention.

The second hurdle is the transition period, where semi-autonomous
vehicles are forced to share the world with us humans. We're the weakest
link by far, so the sooner we can get the humans out of the picture the
better off we'll be. If people insist on playing with Facebook while
they drive, let them play on Facebook while the car drives itself.

Uber Trucking has a good initial approach. A human drops off a semi
trailer at a hub near the edge of a city, then an Uber truck is hooked
up. The Uber truck takes the trailer to the next city or across the
country, where it's once again dropped off at a trucking hub and a human
takes it into the city. Out on the highway, there's still a human in the
truck, but he or she is there just in case, not as a primary driver.

I'm guessing you missed today's Shannon Luminary Lecture Series at Nokia
Bell Labs
by Vint Cerf, Google's Internet Evangelist ( also widely known as a
“Father of the Internet”) where he talked about many of the pitfalls of
AI being integrated in to the real world.
~
Bottom line: Debugging AI algorithms ain't so simple.
~~
[ Just sayin' ]
--
==
Later...
Ron C
--