Sunday, January 31, 2016

"Don’t worry Uber/Lyft drivers: self-driving cars won’t take your job for at least a decade"

From Medium:
Truth is people like me have overhyped self-driving cars. Now that I’ve had a great look at the state of the art I’m quite convinced that your jobs are safe for at least a decade. After that you are pretty screwed, though, so you better take that decade and invest in your education and learn something else.
Here’s the inside scoop, from some of the pioneers in the industry.
I visited Virginia’s Tech Transportation Institute and spent an hour with Jared Bryson, director of its smart road testing factility. I live video broadcast that on Facebook, here, so you can see what we discussed. Afterward I visited Michael Fleming, CEO of TORC Robotics which is just across the field from the smart road. His team at Virginia Tech came in third at the DARPA Challenge back in 2007 (Stanford and Carnegie Mellon teams joined Google to form its autonomous driving teams).
In his building he has military vehicles, all sorts of earth moving equipment, and consumer cars that are outfitted with LIDARs, cameras, and other sensors. TORC is working with many manufacturers to bring self-driving capabilities to all of us.
Finally, I’ve spent quite a bit of time with Mercedes Benz’ autonomous driving team, and I’ve met many others who are working on such technology at Google, Alibaba, and other places. (I interviewed the Stanford team back in 2007 and had the first video of the Google car driving around in Silicon Valley way back in 2009).
Mercedes gave me a lengthy ride around the Nevada desert in their latest E Class, which has autonomous driving features. You can see all that video here. (You see the head of the Mercedes team, Michael Hafner, in front of that car on this post).
Just so you can see that I’m probably more up to date on the topic than other people.
Which gets me to the point of this article.
When I think of self driving cars I am not thinking of mere driver assistance features like pulling out of your garage, radar-assisted cruise control (all three of my Toyotas have that, by the way), keep in lane features, or parking features.
No, I am talking about the full monty: where a car can drive around town without any human inside AND without a steering wheel.
That’s what it will take to threaten the Uber driver’s livelihood. We are a LONG way from being able to buy that kind of car for less than $40,000, which is what it would take too. Some of the insiders say we’re 15 years away from such a car. Some say even more.
There are HUGE challenges to having such a car driving around that simply haven’t been solved yet. Here’s just some:
  1. Humans don’t like programming their cars. Heck, most humans don’t even like using Waze, which helps make your drive much nicer (by routing you around traffic accidents, warning you of objects in road, potholes, and stuff like that). My wife won’t even use cruise control and we have really amazing cruise control systems in our Toyotas.
  2. We have thousands of social rules when it comes to cars. Think about it, when you are in a crosswalk you look at the eyes of the driver coming toward you to know whether or not she/he will stop. That’s just one rule. There are many, many, other rules and we don’t yet have a car that communicates with other drivers, or pedestrians, in any real way.
  3. We don’t have sensors that work in all situations. Last February I drove in a heavy snowstorm where I could hardly discern where the lane was. Yes, I know, that was stupid of me to do, but sensors simply aren’t good enough to withstand those kinds of situations yet.
  4.  Our current compute power isn’t good enough. In China recently my driver went through a crowd of hundreds of people. Our current systems can’t handle that kind of complexity right up close to the car. How did our driver handle it? By relying on social rules. Going slow, honking, and watching for people who weren’t paying attention. The people working on these technologies say that’s just too complex for our current systems.
  5. Maps aren’t good enough. Next month I’ll be headed to Seattle to see Here Maps (formerly Nokia/Navtek/Microsoft map team). They are building new maps specifically for self driving cars. Why aren’t current maps good enough? They aren’t accurate enough, they don’t have all the signage on them. They haven’t yet figured out where all the potholes are and aren’t updatable easily enough by sensors being driven around....MORE
That last is why there was a big money bidding war last summer:
Uber Bids for Nokia Maps Service to Lessen Google Reliance
Why Three German Carmakers Will Pay $3 Billion for Nokia's Mapping Service

Uber ended up buying Microsoft's mapping technology and talent after the Germans took the Nokia business. They also bought the company that powers GM's OnStar called deCarta.

See also:
Winning The Driverless Car Wars
Mercedes-Benz Mulling On-Demand Autonomous Limo Service (mit umlauts)
Night of the Long Knives: "Google Vs. Uber in the Rush To Drive You Around, Driverless" (GOOG)
"Hidden Obstacles for Google’s Self-Driving Cars" (GOOG)