Where the Silicon Meets the Road

One of the best books about technology and its impact written in recent years is The New Division of Labor, by Frank Levy and Richard Murnane, which came out in 2004. Its second chapter is titled “Why People Still Matter.” The authors propose that we still matter, even in an era of countless powerful computers, because there are some common and important tasks that we accomplish effortlessly, but which pose daunting challenges to digital entities.

One of the main examples used is piloting a bakery truck through a left turn in traffic. Levy and Murnane write

The bakery truck driver is processing a constant stream of [visual, aural, and tactile] information from his environment… to program this behavior we could begin with a video camera and other sensors to capture the sensory input. But executing a left turn against oncoming traffic involves so many factors that it is hard to imagine discovering the set of rules that can replicate a driver’s behavior.

The authors contrast the task of vacuuming a room, which by 2004 had been competently automated by the Roomba robot, with driving the truck:

The Roomba’s software exploits the common features shared by most rooms. The environment confronting the bakery truck driver, however, is much more diverse. As the driver makes his left turn against traffic, he confronts a wall of images and sounds generated by oncoming cars, traffic lights, storefronts, billboards, trees, and a traffic policeman. Using his knowledge, he must estimate the size and position of each of these objects and the likelihood that they pose a hazard… Articulating this knowledge and embedding it in software for all but highly structured situations are at present enormously difficult tasks… Computers cannot easily substitute for humans in [jobs like truck driving]

The results of the first DARPA Grand Challenge, held in 2004, supported this conclusion. The challenge was to build a driverless vehicle that could navigate a 150-mile route through the unpopulated Mohave Desert. The ‘winning’ team made it less than 8 miles.

Later on, I saw headlines indicating that teams in subsequent Grand Challenges had done better, but I didn’t pay them much mind. The recent New York Times headline “Google Cars Drive Themselves, in Traffic,” however, got my full attention. John Markoff’s story relates how a Prius loaded up with a bunch of sensors, computers, and other technology has driven more than 1000 miles on American roads without any human involvement at all, and more than 140,000 miles with only minor inputs from the person behind the wheel (to comply with driving laws, Google felt that they had to have a person behind the steering wheel at all times).

Google’s feat, though, does not mean that Levy and Murnane were wrong in their book. Automatic driving on populated roads is an enormously difficult task, and it’s not easy to build a computer than can substitute for a human bakery truck pilot. Not easy, but possible —  this enormously difficult task has been accomplished.

The Google technologists got it done not by taking any shortcuts around the challenges listed by Levy and Murnane, but instead by meeting them head-on. They use the staggering amounts of data collected for Google Maps and Street View to provide as much information as possible about the roads they were traveling (in other words, to provide as much structure as possible to the situation). They also collect huge volumes of real-time data using video, radar, and LIDAR gear mounted on the car; these data are fed into software that take into account the rules of the road, the presence, trajectory, and likely identity of all objects in the vicinity, driving conditions, and so on. This software controls the car; I am confident it provides better awareness, vigilance, and reaction times than any human driver could. Here’s a video from the Times:

None of this is easy. But in a world of plentiful accurate data, powerful sensors, and massive storage capacity and processing power it IS possible. And in a world where the costs associated with all these factors decline over time at consistent and astonishing rates, new feats of digitization become possible all the time. This is the world we live in now. It’s one where, as I wrote earlier, “computers improve so quickly that their capabilities pass from the realm of science fiction to the realm of the mundane not over the course of a human lifetime, but rather well within the span of one professional’s career.”

It looks like I was being too conservative. It took just 6 years for autonomous real-world cars to go from being textbook examples of the near-impossible to being something you can cruise around busy American streets in.

The pace of improvement and innovation is becoming faster and faster, and there’s no slowdown in sight. It’s causing us to rethink firm convictions about what can and can’t be done (“we’ll never have good voice recognition technology;” “we’ll never get good machine translations;” “we’ll never have practical robot cars”), and forcing thoughtful people to revisit some of their theories.

As I learn more and more about what the alpha geeks are accomplishing these days I keep getting reminded of Arthur C. Clarke’s great insight that “any sufficiently advanced technology is indistinguishable from magic.”

What do you think will be the next digital magic act? What’s the next astonishing feat that computers will pull off? Will it be beating a human champion in Jeopardy! Passing a Turing Test? Where will science fiction next become reality? Leave a comment, please, and let us know what you think.