Wednesday, May 11, 2011

Google is putting the "auto" into automobile.

If the last big revolution was replacing muscle power with machines, the next one is automating and augmenting more mental tasks. Henry Ford and compatriots replaced the horse, now Google is working to replace the driver.

According to John Markoff in the New York Times, their self-driving cars have now logged over 140,000 miles on California roads, including highway 1 between L.A. and San Francisco. They are now lobbying Nevada to allow these cars on public roads.

The project leader, Dr. Sebastian Thrun has argued that robotic vehicles would increase energy efficiency, reduce road injuries and deaths, and cut the number of cars needed in the United States in half.

“What if I could take out my phone and say, ‘Zipcar, come here,’ ” he asked an industry conference last year, “and a moment later the Zipcar came around the

I suspect the biggest barrier to the adoption of self-driving cars is not technological -- these videos show how the systems are rapidly progressing. Instead, the impediments will be regulatory and cultural. There are about 40,000 deaths on the America's roads each year with our human drivers. But suppose the robotic cars were 100 times safer. That would still be 400 deaths per year. Can you imagine the public outcry, no to mention the legal judgments, that would follow the first time a human was killed due to an error by a robotic car? Will they have to be 100% perfect before the are adopted?


  1. I would like to take a step back from the current hypothesis and try to put things in to perspective while I derive certain learning points:

    The "auto" has been very prevalent in the airline industry and the "auto pilot" mode is increasingly being used for more and more critical aspects of flying an aircraft such as, take off and landing, specially landing during extremely bad weather (or zero vision scenarios). Not to mention high altitude cruising.

    Based on the above premise, we can derive certain learning points:
    1. The auto pilot mode or artificial intelligence did not totally eliminate the need of a human intervention in the form of pilot.
    2. The auto pilot mode can easily be used during mundane aspects of flying such as cruising (in car terminology: long drives on expressways)
    3. The auto pilot mode is increasingly used during critical aspects of flying (as summarized above) and the auto car mode could use it for driving during bad weather or to take over driving when the driver's ability to drive is compromised.

    In my opinion, we are far from a driver-less situation, but we are very much into an "auto-assisted" era in which the robotic car doesn't need to be 100% perfect before adoption.

  2. If it is going to cut the number of cars needed in the US in half, the biggest barrier to the adoption of self-driving cars is neither regulatory nor cultural. The automobile manufactures would lobby their hardest not to adopt this, perhaps.

  3. Robotic cars will come incrementally, feature by feature, starting with high end cars, and sliding down to low end. cruise control, ABS brakes, electronic stability control, hands free parallel parking, automatic radar braking, lane keeping, vehicle-to-vehicle awareness preventing collisions, etc. It's happening. Complete automation in slow, congested freeway driving will be a noticeable feature very shortly. Avoid rear-end collisions at 5 mph because you are texting that you will be late to the meeting.

  4. The American workers' compensation system should provide a workable template for limiting the litigation downside. The bigger hurdle is social acceptance.