Sunday, July 3, 2016

open door, open loop: self-driving cars in an age of self-interest


Glimpses of the future can sometimes be seen in Princeton, despite its preoccupation
with the rear-view mirror. A recent example is the new valet parking service, operated
by Newtown-based Open Door Valet. More info on the company and its owner here.
The town stands to collect 6% of any revenue from this license arrangement. Parking
valet service is available "in front of Heinz Plaza" which most of us know as Hinds Plaza.

In the article How Driverless Cars Can Reshape Our Cities, Professor Alain Kornhauser
says "The biggest impact is going to be on parking. We aren't going to need it, definitely
not in the places we have it now. Having parking wedded or close to where people spend
time, that's going to be a thing of the past. [...] The current shopping center with the sea
of parking around it, that's dead." [  A sea of parking ... and of curbing made of "setts" ]

Kornhauser has led an effort to make New Jersey a research hub for driverless cars, and
has published widely on the topic of shared autonomous taxi networks, or vehicle fleets.
His proposal to create a Center for Automated Road Transportation Safety (CARTS) at
Fort Monmouth, the former Army installation, may lack high-level political support, though.

The enabling technology for driverless cars has been the recent confluence of three factors:
powerful graphics processor hardware, artificial intelligence "deep learning" algorithms, and
"big data" for efficient training of the robotic systems. As explained here, the architecture of a
graphics processing unit (GPU) is a very good match for the highly parallel requirements of
so-called "biologically-inspired multi-layer perceptrons" aka convolutional neural networks.

Decades ago, groundbreaking research in this area was published by John Hopfield, who
has spent most of his career at Princeton University. Collaborative research at Bell Labs in
Holmdel launched in the 1980's eventually led to character and voice recognition products;
the legacy of this neural network research program and scientific talent is carried on at Nvidia,
a leading vendor of GPU silicon. They've installed a research team in what is nowadays known
as Bell Works, the huge/famous building formerly home to Bell Labs on Crawfords Corner Road.
Here's a link to Nvidia's paper End to End Learning for Self-Driving Cars and the related video.

A provocative article "Makers of driverless cars want cyclists and pedestrians off the roads" is here.

The sentiment is that driverless cars will improve the safety of our roads. But in a likely setback for
the technology, the The National Highway Traffic Safety Administration (NTHSA) has launched a
probe into the fatal collision of a Tesla and an 18-wheeler rig, in Florida around seven weeks ago.
In this accident, which featured poor decision-making by Tesla's "Autopilot" controller - a mode which
had been banned in Hong Kong - and by the human drivers, both of whom sported safety violations,
the Tesla became instantly and literally driverless when it failed to notice the low clearance of a trailer
across its path on the highway. Tesla's Autopilot feature depends on "public beta testing" and is
never reviewed by regulators -- "highlighting what some say is a gaping pothole on the road to
self-driving vehicles: the lack of federal rules".

Meanwhile, a recent article asked Can a 700 M.P.H. Train in a Tube Be for Real? about the Hyperloop
project launched by Tesla CEO Elon Musk. Princeton University has joined five other schools 
(Cornell, Harvey Mudd College, Memorial University, Northeastern University and the University
of Michigan) in the OpenLoop Consortium, competing against other collegiate teams at the
SpaceX test track in Hawthorne, CA. More info

NY Times' Silicon Valley correspondent and avid bicyclist [link] John Markoff, in his latest

             "People say that one day, perhaps in the not-so-distant future, they'd like 
             to be passengers in self-driving cars that are mindful machines doing their
             best for the common good. [...] A new research study, however, indicates that
             what people really want to ride in is an autonomous vehicle that puts its
             passengers first. If its machine brain has to choose between slamming into a
             wall or running someone over, well, sorry, pedestrian."

The article quotes Harvard psychologist Joshua D. Greene, who, when a
grad student at Princeton, had  legendary ethicist Peter Singer, author
of How Are We to Live? Ethics in an Age of Self-Interest as well as
the somewhat bike-related Is Doping Wrong?, on his thesis committee.

The article in The Guardian puts it as follows "... designers of future driverless 
cars find themselves in a moral maze from which there is no easy way out". Or
per the headline in Spectrum magazine's articlePeople Want Driverless Cars
with Utilitarian Ethics, Unless They're a Passenger.

The NYT article mentions that study of "the trolley problem" got its start in 
1967 with Philippa Foot, a British philosopher who also had a Princeton
connection: her mother, born in the White House, was a daughter of
Grover Cleveland. Trolleyology, "a neologism ... shorthand for a flourishing
cottage industry in a well-known philosophical dilemma: should you push a fat
man onto a track to save five innocent people from being hit by an oncoming
railway trolley?" is from a book by NYU's Kwame Anthony Appiah during his years
at Princeton U. Appiah's quote "My philosophy is that everything is more complicated
than you thought" stands in contrast to Facebook's "move fast and break things".


By way of a postscript about current events, click here to watch a 30-second
video of Brexit "Leave" proponent Boris Johnson being booed and blocked by
cyclists. Johnson, the former mayor of London, made his city more bike-friendly.
Economist/commentator Paul Krugman would normally have much to say about
Brexit's ramifications, but was away on a bike trip near the Vermont/Quebec border.












No comments:

Post a Comment