The interesting part of this is not the legal-ness of the robot-car, but the breathless, and supposedly “green” aspect of robot cars. Techno-utopianism – both “green” and “high-tech” are the same thing.
I’d sure like to see exactly how the Google car works. Which model for driverless cars was it based off? The Stanford/Stanley model? the CMU AI model?
If that wasn’t enough tech-slobber for you, try this earlier explosion of tech-religion:
What I find amazing is that the writers apparently don’t know much about the long history of driverless cars (especially the DARPA Grand Challenge just a few years ago), or don’t feel it is important, now that GOOGLE is involved. Yeech.
But, as always, China is the next America – driverless cars ahoy!
The fact is, these robot cars appear to use little more than cameras and (possibly) laser rangefinders – the same as driverless cars for the last two decades. Sure, they can drive themselves with a GPS rout pointing the way – just like cars in the Grand Challenge and the Urban Challenge. But is their driving comparable to humans? In contrast to the cars, humans use lots of sensory cues ? It’s doubtful that the robo-cars are really driving as well as humans, simply because since sound is a major component of human driving, and directly related to accident avoidance. Also, while the cars are most likely registering acceleration/deceleration, it is very unlikely that they are using this data like a human driver would. In short, the driverless cars are still old-school.
They might work as a few isolated machines in a sea of humans, but that’s about it. Remember, most driving involves cooperative interaction between drivers. We don’t know, especially for city driving, how much human drivers are compensating for the robot car’s mistakes. Without factoring in the ‘help’ coming from other drivers adjusting to the robot car, we don’t know if the car is operating at human level, or, like any “student driver” being adjusted for.
The only way we’ll ever have this kind of car replacing humans is a much more orderly world. That a typical city. True to form, the wet-dream visions of artists show clean, straight city streets, with buildings and curbs that look like they were designed in a computer (they were). This is the opposite of Robots That Jump, which requires robots deal with the dirtiest of the real world. Instead, in order to get robot cars we need to make the cities – well, more robotic. Since computer-driven cars have trouble with reality, why not make reality more like the inside of a virtual world, where robot cars run and play without crashes or conflict?
In fact, why not just declare reality “broken”, and fix it by making it like a video game?
Seems hard to justify that Darwinian evolution make us able to work out the real world in games, but I suppose the Goddess Techna directed our evolution to be perfectly adapted to the Xbox, less so with dirty things like ground. The prevailing view, driven by rising time spent in the virtual world of cyberspace, is that the world fails by not being like a computer model. This is an interesting twist – models were supposed to be imperfect representations of reality, not the other way around.
IMHO, this is an extension of the ideas I head about in the 1960s. Then, we were going to hitch our cars to some sort of clamp in the freeway, and detach when we reached our destination. The modern formulation just assumes less central control (my guess is that it would still be needed).
An interesting driverless car would have the following:
- Sensors everywhere
- A variety of sensors, smell, touch, vision, hearing, gps, cellphone, radio, electric sense, whatever
- Some basic set of ‘instincts’ to behave in a certain way.
- Integration of all the sensory input via training, similar to the Stanford “stanley” car that run the 2005 Grand Challenge.
- At that point, we’d have a robot car that could jump. It would almost be as good as a horse.