Beyond the views, what’s been happening with self-driving
vehicles over the past several months?
In “A Lesson of Tesla Crashes? Computer Vision Can’t Do It All Yet” (The New York Times, September 19), Steve
Lohr recapped May’s fatal Florida accident, correctly judging that “the man
placed too much confidence in Tesla’s self-driving system.” The issue is that naming it something like
“autopilot,” when it is designed only to spell operators for problem-free and
probably short times, is asking for trouble, and now, what with the
well-publicized state of advancement in the field, is unnecessary. When more progress is made with technologies
such as the one Lohr described, Stanford and Princeton Universities’ ImageNet,
and problems such as the connection between Tesla’s driving and braking systems
which caused that crash are fixed, drivers will be more able to disengage.
On that same date, in Business
Insider’s “Uber’s self-driving cars are impressive – but there’s still a
lot we can’t do,” Danielle Muoio described her experience riding in one of
Uber’s Pittsburgh test cars, summarizing that they had real problems “simply
having to deal with other human drivers on the road.” As she photographed in action, during her
ride the operator needed to take over control when an 18-wheel truck turned
right, from the left lane, in front of them.
The Uber test, per Muoio, is also uncovering issues these vehicles have: with bridges, given the lack of buildings
next to them; with snow, since it covers up lane markings and other orientation
markers; and even with trees, which look
different enough in summer and winter for the systems to identify specific ones. On the other hand, these tests are running up
miles, and Google’s has, as of October 5th (Wall Street Journal), reached 2,000,000 of them.
Sunday Review in The
New York Times is a good place for issues we should be thinking about, even
without solid conclusions, and that perfectly describes Azim Shariff, Iyad
Rahwan, and Jean-Francois Bonnefon’s November 3rd “Whose Life Should
Your Car Save?” This was not the first
time I or others have mentioned the issue of how driverless cars should be
programmed when faced with a choice “between risks to its passengers and risks
to a potentially greater number of pedestrians,” but it didn’t need to be. We learned here that when Science magazine presented the results
of people being surveyed on life-or-death scenarios, “a large majority… agreed
that cars that impartially minimized overall casualties were more ethical, and
were they type they would like to see on the road,” but also that “most people”
showed “a strong preference for buying the self-protective one” instead. That looks like a crime of the 2030s – people
illegally modifying their driverless vehicle software to protect themselves
first.
In that same newspaper the next day, Henry Fountain’s “A
Slow Ride Toward the Future of Public Transportation” raised some less commonly
presented issues. He mentioned the
appearance of driverless buses around Helsinki campuses and factories, and
suggested, correctly for once, that self-driving cars would cut auto sales “in
cities” – not in the country. Those,
along with a first experimental bus in public territory, use a combination of
environment-detecting sensors and an approach described 64 years ago for
automating barbers in Kurt Vonnegut’s Player
Piano, capturing and replicating the motions of human drivers. The piece also mentioned a local computer
application, Whim, which Finns can now use to not only plan but book
transportation, the forms to be determined by the system, by keying in starting
point and destination.
On December 13 in Salon,
we found that Google’s driverless effort is now named Waymo, and that the
company, ever optimistic, officially expects their vehicles to be “commonplace”
in four years or less. That same day, Cecilia
Kang in the Times (Cars Talking to
One Another? They Could Under Proposed
Safety Rules) suggested something I should have thought of myself, that
driverless vehicles routinely broadcast, air-traffic-control style, their
location, speed, and direction to each other.
This plan has the backing and involvement of the National Highway Traffic
Safety Administration, and, when technical and bureaucratic issues with radio
frequencies are resolved, seems a huge favorite to be implemented.
From December 14th to December 21st,
covered by the New York Times and the
Wall Street Journal, we saw the rise
and fall of Uber’s San Francisco self-driving taxi service. On that first date they announced that it,
though only provided by five cars, would be available throughout that entire
city. They tried to sidestep local
regulations by claiming they were not autonomous, and indeed each would have
two people, a driver and an engineer, in front, but the imprecision of the laws
ended up working against them, and, after the California Department of Motor
Vehicles cancelled the cars’ registrations, the service was discontinued. It will come back, though, at another place,
most likely within months.
On December 22nd in Salon, Angelo Young told us that Google’s first Waymo vehicle is
now planned to be a minivan. It’s as
least as good as any, since, as modern-day station wagons, minivans accommodate
a variety of occupants and cargo, and the expected much greater safety of
self-driving cars would appeal to the same demographic.
Fox Business peeked
at the stock investment merits of two major driverless players, including
Google’s parent company (“How Apple and Alphabet are De-risking their
Self-Driving Car Efforts,” December 23rd). Commentators Dylan Lewis and Daniel Sparks
reached the same conclusion that I did months ago, that producing such things
is too much for any single firm. They
added properly that, unlike what these companies are accustomed to in the
smartphone market, they will not be able to achieve “high-30%” profit margins,
and that both would like to minimize the manufacturing they do themselves. Yet somebody must make them – could they get
away with pawning off this tangible product bearing their nameplate on someone
else? We will see.
Finally, two days ago, Brent Snavely mentioned a name not as
commonly mentioned in the driverless world as its competitors. In “Brain in the Trunk: Ford to unveil next self-driving car” (Detroit Free Press), he says that
Henry’s old company is, along with the others, “investing heavily” in the
technology, and is focusing farther down the road, “to develop a fully
autonomous car that does not require the driver to operate the vehicle.” I’m not sure this is a good place to try to
leapfrog competitors, but what do I know?
They are hoping to do that without any special roadway technology, by
reading where they are going through LIDAR and visual sensors.
So how about jobs?
Hang on to your hat – 2017 will be quite a year. Expect an updated look at how many truck
driving, cabdriving, auto sales, and other positions might be lost by when,
sometime in the middle of it.
No comments:
Post a Comment