There has been a fair amount of speculation about how the advent, Real Soon Now, of the self-driving Google Car will change life as we know it forever.
Because the temper of the times is flowing toward urbanization and away from suburbanization, lots of people have assumed that having a robot car would be like having a chauffeur-driven limousine, which is a really nice way to live in Manhattan. Thus, in the future, everybody will live in high rises and ride around in Google Cars, and you won’t have to go through all the hassle of parking as you visit another high rise. Your Google Car will just go somewhere while you are inside, and then be waiting for you as you step off the elevator as you leave.
Right now, if you are a Master of the Universe, Senior Grade, your limousine drops you off at the door of the building you’re visiting, then goes away somewhere, then picks you up again at the door. You’re not exactly sure where it parks itself while you are inside, but that’s not your problem.
Of course, this is immensely expensive: in Bonfire of the Vanities, for example, Sherman McCoy, junior grade MotU, reflects bitterly upon how much it costs him to rent a limo a la carte to take him and his wife to a party a mile away and then pick them up five hours later.
But what if the technology of the Google Car doesn’t evolve to deal well with crowded, narrow urban streets and parking garages? What if the Google Car evolves to deal best with, say, freeways retrofitted with electronic signals?
The autonomous Google car may never actually happen.
By Lee Gomes
In summary, so far the Google Car doesn’t represent what we imagine to be artificial intelligence: it’s strength isn’t in responding brilliantly to the ever changing outside world, it’s strength is in having an incredibly detailed inch-by-inch map of the streets around Google’s Mountain View headquarters downloaded into its memory.
It’s like how Google Translate doesn’t actually understand what that French website is saying and translates it into English for you, it just looks up phrases in bilingual documents published by the government of Canada and the like and lets you, the human being, make sense out of what it comes up with. If you have enough data, Google don’t have to be terribly smart. But Google Translate presupposes that an intelligent human being will be able to make sense of what it dishes up.
But, because the point of Google Car is to take away the human intelligence at the end, a lot of human-hours have been put in earlier in the process into interpreting those super-maps of the Mountain View area to make automated driving safe.
My guess is that mapping can work well on controlled roadways like freeways, but the streets of Manhattan are constantly changing with transient obstacles, such as pedestrians and the remarkable number of holes in the streets being dug at all hours by workmen.
Google may have the resources to someday monitor Manhattan’s streets second by second, but it seems unlikely, contrary to much speculation, that the Google Car would lead to the rest of America being Manhattanized.
For one thing, parking is a nightmare for Google right now. Parking doesn’t map well because the obstacles keep changing. And parking is one of the key tasks people who have the Limo Model in mind want Google Car to do: drop you off downtown and automatically drive off somewhere and park itself where parking doesn’t cost $20 per day.
Instead, Google Cars might lead to a revival of the exurbanization trend that died with the increase in the price of gasoline in the first half of 2008. With the price of gasoline seemingly moderating, it’s worth thinking about how computers might make long distance commuting more attractive.
Say Google worked with governments to have freeways and major highways retrofitted with electronic sensors and the like that would interface very well with the Google Car. So, your Google Car could drive on the freeway for you, but you’d have to drive the surface streets and park yourself. For example, say your daily commute looks like:
Driveway to freeway onramp: 5 minutes
Freeway: 60 minutes
Freeway offramp to office parking spot: 5 minutes
Office to freeway onramp: 5 minutes
Freeway: 60 minutes
Offramp to driveway: 5 minutes
Currently, if you commute five days a week, that’s 11.6 hours per week you need to have at least one hand on the steering wheel and shouldn’t be reading.
In the scenario I’ve outlined, your Google Car could let you do reading/typing work ten additional hours per week while you automatically ride the freeway. But you have to drive the surface streets yourself for 1.6 hours per week.
This is much like taking commuter rail to work, which is a pretty nice way to live. The Chicago metropolitan area has a lot of commuter rail lines and houses near stops, such as in Lake Forest, are at a premium. (So, you could model how much people would be willing to pay for a car that drives the freeways by itself by looking at home prices near and away from commuter rail stops.)
In the Chicago area, commuter rail differs from the crowded inner El in that everybody who gets on in Lake Forest gets a seat. It’s an extremely civilized Mad Men way to live. But, it’s very hard to build more commuter rail lines these days. Light rail that has to stop at red lights doesn’t cut it. Heavy rail that rips along without stopping is wonderful, but just a gigantic problem to retrofit into a developed metropolis.
Moreover, most of the United States isn’t Chicago where job concentrations grew up around rail lines. Sunbelt cities grew up around freeways. Automated freeway driving opens up the possibility of the convenience of working while moving to places that aren’t going to get commuter rail.
Retrofitting freeways with transponders or whatever will take decades, but all freeway lanes have to be torn up and repaved every so many decades anyway, so embedding electronics in the pavement isn’t asking too much: look how those shiny bumps between lanes got embedded over the years.
So, a Google Car that automatically drives the freeways but not the surface streets would kind of like be extending commuter rail networks.
(Competing with this, of course, are voice recognition technologies that some people could use while driving themselves safely.)
But all this suggests that the 2020s or 2030s might be an era not of Manhattanization, but of Lake Forestization.