Cities are a central engine of the modern economy. Enormous gains come from folks interacting and specializing more in bigger cities. What limits these gains, and keeps us from all living in one mega-city, is transportation costs. While the cost of transporting goods and people once mattered similarly, today people transport costs dominate. And while hopes for mass-transit remain, cars clearly dominate human transport today. Thus the near-term future of cities, and of which cities dominate the world, comes down to how cities handle auto innovation. I see three main innovations to consider:
If a big city could coordinate to create subways, etc. on the scale and quality of New York, it could support densities like New York. The level of investment and coordination required to pull this off, however, seems well beyond what any known city can muster.
Then shouldn't mass-transit projects be funded by the federal government!
New York's mass transit is a proven model for achieving high densities, whereas it isn't clear whether auto-driving cars will produce high urban densities or suburban sprawl.
A hacker could only mass-crash cars if they were controlled by a central agent. Most robot cars are driven by an on-board computer that senses the environment around it and reacts accordingly. The only thing it needs to talk to is GPS and that is pretty hard to hack.
How does the City of New York deciding to replace the elevated railroads with subways, and then issuing bonds to fund subway construction constitution a dot com boom in private subway construction?
Was the dot com boom fueled by government borrowing to contract for buying all the new technology made by private industry?
We need the public's consent because a large fraction of the risks are imposed on them. After-the-fact legal responsibility is a poor substitute for before-the-fact precautions. Also, large corporations routinely wriggle out of legal responsibility for damage they cause. (Consider how little Union Carbide paid for Bhopal, for instance).
I think greg has an excellent point - even diligent programmers have much less skin in the game than drivers do.
In principle, I agree that if these cars wind up being safer than human drivers, that could be an advantage - if this doesn't wind up pissed away in compensating changes (such as platooning) that crank the risks back up again. In addition, there are some common mode failure risks that would be introduced that aren't there now. The simplest is just that if, as Prof Hanson advocates, this development was to generally increase the size of cities, then it makes the cities more attractive targets, both for conventional military attacks and for NGO asymmetric warfare attacks. Less conventionally: Computer security is lousy. If 10% of the cars on the road are automated, with their steering controlled by computer, the first security hole that some hacker discovers can then be used to physically crash 10% of the cars at once. There is no analogous hazard today.
"Do you really think that companies should be allowed to impose arbitrary risks on the general public without the public’s consent?"
Why do we need the publics consent? Why shouldn't the driver/company be liable for any damages caused if they were responsible? If these cars really do up end up being safer then chances are, it's going to be human drivers being the ones causing these accidents and then blaming it on the computers.
Human beings in cars also impose risks on you. Seems to me we could impose the same responsibility for accidents on owners of driverless cars as we do on operators of normal ones. Might be even more effective since you can't punish a bad driver after he's killed in a crash, but the owner of the driverless car will presumably be safely far away at the time of the accident.
Your warning proposal sounds like a low-cost workaround. Robin, do you object to that?
Another possibility is to start introducing lanes for driverless cars, and then once people get comfortable with them allowing to operate on more road.
While I love the concept of automated cars, I can't find it in me to endorse the way they are being introduced, with you, me, and the general population as non-consenting test subjects. Google may have every safety precaution in the world in place, but the people this robot interacts with are not informed of the risks they face. I feel it is unethical unless the cars wear big signs along the lines of, "Warning! I'm a driverless robot".
Information technology has largely changed the neolithic basis of civilization and additional innovations will usher in a postcivil era of much richer human choice and sustainability. Postcivil society is coming. The transition will be rough. Empty the cities now to minimize human suffering during the transition.
A possible fourth possibility is Personal Rapid Transit (PRT) -- public transit, but not mass transit. Sometimes I am optimistic that it could lead to another boom in public transportation, that it can be both far more economical than traditional services, while also providing much better service. New York transit is nice, but it's also a freakin' slow and a pain in the ass. I think it is exulted too often.
Other times I'm rather pessimistic about PRT, but I think it at least has as much potential as automated cars -- it is technologically much simpler, but more requires more infrastructure than automated cars. I am mostly biased because I personally would much rather live in a community with PRT than automated cars.
SafetySo far the linear path of safety improvements is secure according to insurance companies. The more technology you apply, the lazier the driver; one still ends up with better safety records.Liability in an accidentTake the black box data, run it against the driving rules and the offending robot is identified.SpeedRobocar technology can drive large vehicle on the road at sustained speeds of 150 MPH. When you drive up speed, transaction times go way down, less queuing.HOT lanes:Even with human drivers and no digital assist, the HOT lane transponders can signal to all drivers on the hot lanes and each driver can use electronic collision warning, with an add on dash top device.AggregationRobocar technology allows us to build buses with five to six carriages moving five hundred people.
Security from terrorist threats? A difficult but solvable problem. The black boxes need tamper proof packaging, built in hardware checksums, and more frequent inspection.
The implication of your statement is that subways, and only subways, can support New York-like density. My point is that this is not true – it only appears that way to modern observers because subways are the only vestige of the private mass transit network that's left. Like I said, the streetcars and elevated lines were far more extensive than the subway system has ever been, and could support the density just fine. In fact, many of the subways were duplications of existing elevated lines, and on net didn't add much capacity. (You can argue that elevated lines are unseemly and not suited to the modern day, but the Chicago real estate market seems to be doing alright. And for passengers, els are actually a lot nicer and are generally perceived to be safer than subways.)
My big problem with platooning automated cars (which is what I am sure Google is counting on to double the capacity of a road lane) is that automobiles are NOT virtual entities. They exist and operate in the real world where even if the computer's reaction time is essentially zero, the space needed to stop a car decidedly non-zero. If one such car has an accident, then all the cars in the platoon have an accident, a chain-reaction pileup on steroids.
Then there is the factor that when you automate something like driving, the person making the decisions no longer has "skin in the game". I have been part of software development teams for my whole career, and while most of the participants of such teams are diligent, I am not sure that they are as diligent as they would be if messing-up involved the possibility of dying (as it does in driving). Excessively punishing sellers for car accidents is the very _least_ you can do to preserve such incentives. If they *execute* the development team and management in the event of a massively fatal chain reaction pile up, then and only then would I find that sufficient incentive to get in an automated, platooned car (good luck finding someone to do such work, I wouldn't go near such a project). Automobile platooning is a techno-disaster waiting to happen.
The ranking of U.S. urbanized areas (not just their central cities) is remarkably stable at the top. These areas have advantages that matter and that are durable. These areas are also able to spawn (and extend into) new land use arrangements (often at their peripheries) that are favorable. Fixed-rail transit can never be flexible enough and is a serious option for meaningful numbers of riders only in NYC. The new fixed rail systems have not mattered much -- except on the cost side. So auto-orientation is here to stay. Congestion-pricing would be great. Robo-cars, who knows?
This is a good point. Most city roads waste 50% of the space on parked cars. If this could be eliminated, additional increases in transport efficiency would arise.
If roads didn't have parked cars, and robotic cars could pack closer, you might get 4 lanes of traffic on existing 2-lane roads. You might also build thinner robotic cars and get 6 lanes of traffic.
Suppose also that the cars go twice as fast, and you have 6 times the transport efficiency.
If a big city could coordinate to create subways, etc. on the scale and quality of New York, it could support densities like New York. The level of investment and coordination required to pull this off, however, seems well beyond what any known city can muster.
Then shouldn't mass-transit projects be funded by the federal government!
New York's mass transit is a proven model for achieving high densities, whereas it isn't clear whether auto-driving cars will produce high urban densities or suburban sprawl.
A hacker could only mass-crash cars if they were controlled by a central agent. Most robot cars are driven by an on-board computer that senses the environment around it and reacts accordingly. The only thing it needs to talk to is GPS and that is pretty hard to hack.
How does the City of New York deciding to replace the elevated railroads with subways, and then issuing bonds to fund subway construction constitution a dot com boom in private subway construction?
Was the dot com boom fueled by government borrowing to contract for buying all the new technology made by private industry?
We need the public's consent because a large fraction of the risks are imposed on them. After-the-fact legal responsibility is a poor substitute for before-the-fact precautions. Also, large corporations routinely wriggle out of legal responsibility for damage they cause. (Consider how little Union Carbide paid for Bhopal, for instance).
I think greg has an excellent point - even diligent programmers have much less skin in the game than drivers do.
In principle, I agree that if these cars wind up being safer than human drivers, that could be an advantage - if this doesn't wind up pissed away in compensating changes (such as platooning) that crank the risks back up again. In addition, there are some common mode failure risks that would be introduced that aren't there now. The simplest is just that if, as Prof Hanson advocates, this development was to generally increase the size of cities, then it makes the cities more attractive targets, both for conventional military attacks and for NGO asymmetric warfare attacks. Less conventionally: Computer security is lousy. If 10% of the cars on the road are automated, with their steering controlled by computer, the first security hole that some hacker discovers can then be used to physically crash 10% of the cars at once. There is no analogous hazard today.
I wrote about it yesterday after seeing Tyler's article about it, but I agree that it's largely going to be political:
http://dispersed-knowledge....
"Do you really think that companies should be allowed to impose arbitrary risks on the general public without the public’s consent?"
Why do we need the publics consent? Why shouldn't the driver/company be liable for any damages caused if they were responsible? If these cars really do up end up being safer then chances are, it's going to be human drivers being the ones causing these accidents and then blaming it on the computers.
Human beings in cars also impose risks on you. Seems to me we could impose the same responsibility for accidents on owners of driverless cars as we do on operators of normal ones. Might be even more effective since you can't punish a bad driver after he's killed in a crash, but the owner of the driverless car will presumably be safely far away at the time of the accident.
Your warning proposal sounds like a low-cost workaround. Robin, do you object to that?
Another possibility is to start introducing lanes for driverless cars, and then once people get comfortable with them allowing to operate on more road.
Do you really think that companies should be allowed to impose arbitrary risks on the general public without the public's consent?
Do you really think nothing should be allowed to change in the world around you without your informed consent?
While I love the concept of automated cars, I can't find it in me to endorse the way they are being introduced, with you, me, and the general population as non-consenting test subjects. Google may have every safety precaution in the world in place, but the people this robot interacts with are not informed of the risks they face. I feel it is unethical unless the cars wear big signs along the lines of, "Warning! I'm a driverless robot".
Information technology has largely changed the neolithic basis of civilization and additional innovations will usher in a postcivil era of much richer human choice and sustainability. Postcivil society is coming. The transition will be rough. Empty the cities now to minimize human suffering during the transition.
A possible fourth possibility is Personal Rapid Transit (PRT) -- public transit, but not mass transit. Sometimes I am optimistic that it could lead to another boom in public transportation, that it can be both far more economical than traditional services, while also providing much better service. New York transit is nice, but it's also a freakin' slow and a pain in the ass. I think it is exulted too often.
Other times I'm rather pessimistic about PRT, but I think it at least has as much potential as automated cars -- it is technologically much simpler, but more requires more infrastructure than automated cars. I am mostly biased because I personally would much rather live in a community with PRT than automated cars.
SafetySo far the linear path of safety improvements is secure according to insurance companies. The more technology you apply, the lazier the driver; one still ends up with better safety records.Liability in an accidentTake the black box data, run it against the driving rules and the offending robot is identified.SpeedRobocar technology can drive large vehicle on the road at sustained speeds of 150 MPH. When you drive up speed, transaction times go way down, less queuing.HOT lanes:Even with human drivers and no digital assist, the HOT lane transponders can signal to all drivers on the hot lanes and each driver can use electronic collision warning, with an add on dash top device.AggregationRobocar technology allows us to build buses with five to six carriages moving five hundred people.
Security from terrorist threats? A difficult but solvable problem. The black boxes need tamper proof packaging, built in hardware checksums, and more frequent inspection.
The implication of your statement is that subways, and only subways, can support New York-like density. My point is that this is not true – it only appears that way to modern observers because subways are the only vestige of the private mass transit network that's left. Like I said, the streetcars and elevated lines were far more extensive than the subway system has ever been, and could support the density just fine. In fact, many of the subways were duplications of existing elevated lines, and on net didn't add much capacity. (You can argue that elevated lines are unseemly and not suited to the modern day, but the Chicago real estate market seems to be doing alright. And for passengers, els are actually a lot nicer and are generally perceived to be safer than subways.)
My big problem with platooning automated cars (which is what I am sure Google is counting on to double the capacity of a road lane) is that automobiles are NOT virtual entities. They exist and operate in the real world where even if the computer's reaction time is essentially zero, the space needed to stop a car decidedly non-zero. If one such car has an accident, then all the cars in the platoon have an accident, a chain-reaction pileup on steroids.
Then there is the factor that when you automate something like driving, the person making the decisions no longer has "skin in the game". I have been part of software development teams for my whole career, and while most of the participants of such teams are diligent, I am not sure that they are as diligent as they would be if messing-up involved the possibility of dying (as it does in driving). Excessively punishing sellers for car accidents is the very _least_ you can do to preserve such incentives. If they *execute* the development team and management in the event of a massively fatal chain reaction pile up, then and only then would I find that sufficient incentive to get in an automated, platooned car (good luck finding someone to do such work, I wouldn't go near such a project). Automobile platooning is a techno-disaster waiting to happen.
The ranking of U.S. urbanized areas (not just their central cities) is remarkably stable at the top. These areas have advantages that matter and that are durable. These areas are also able to spawn (and extend into) new land use arrangements (often at their peripheries) that are favorable. Fixed-rail transit can never be flexible enough and is a serious option for meaningful numbers of riders only in NYC. The new fixed rail systems have not mattered much -- except on the cost side. So auto-orientation is here to stay. Congestion-pricing would be great. Robo-cars, who knows?
This is a good point. Most city roads waste 50% of the space on parked cars. If this could be eliminated, additional increases in transport efficiency would arise.
If roads didn't have parked cars, and robotic cars could pack closer, you might get 4 lanes of traffic on existing 2-lane roads. You might also build thinner robotic cars and get 6 lanes of traffic.
Suppose also that the cars go twice as fast, and you have 6 times the transport efficiency.