3 ms·

I'd like to see the data that the NHTSA used to conclude that the risk of a malfunction is higher than the risk of human error. Millions of crashes each year re

by thesash 13y ago

I'd like to see the data that the NHTSA used to conclude that the risk of a malfunction is higher than the risk of human error. Millions of crashes each year result in tens of thousands of deaths [1]. Humans drive while tired, drunk, or simply distracted, and of course, tend to have slower reflexes than computers. Even if the technology isn't ready for mainstream adoption today, I bet we'll look back in horror at the period in our history during which humans careened unaided down highways at breakneck speeds, certified in their driving ability by nothing more than a test easily passed by your average teenager. [1] http://en.wikipedia.org/wiki/List_of_motor_vehicle_deaths_in_U.S._by_year http://en.wikipedia.org/wiki/List_of_motor_vehicle_deaths_in...

BorgHunter 13y ago
I don't believe the report[1] actually said that the risk of a malfunction is higher than the risk of human error. In fact, it's quite optimistic about the ability of automated systems to save lives, for precisely the reasons you described. They do, however, correctly point out that such systems can malfunction, and will malfunction, and that safety impacts of such malfunctions should be minimized. Because malfunctions in automated transportation can lead to, say, June 2009 on the DC Metrorail.[2] Such incidents are rare, and through hard work and foresight will be rare in the future, but that doesn't happen by itself. You have to make a concerted effort to ensure such systems fail gracefully. [1] http://www.nhtsa.gov/staticfiles/rulemaking/pdf/Automated_Vehicles_Policy.pdf http://www.nhtsa.gov/staticfiles/rulemaking/pdf/Automated_Ve... [2] https://en.wikipedia.org/wiki/June_2009_Washington_Metro_train_collision https://en.wikipedia.org/wiki/June_2009_Washington_Metro_tra...
acchow 13y ago
Certification? Hah. My friend got a G1 "learner's license" in Ontario (which requires only a multiple-choice test). When he moved to New York, he "traded in" the G1 for a full license. I guess that's the exchange rate.
DougWebb 13y ago
It's rare for all of the human drivers on a road to make the same mistake at the same time under the same conditions, but that's a likely failure condition for automated driving software if the software and sensors are standardized. This could lead to much worse accidents then you'd get with human drivers, because all of the cars will do the same thing. An all-human example: driving too fast in foggy/whiteout conditions and not having time to stop when the road is blocked. When this happens today, you get 100 car pileups because everyone is doing the same thing: driving along until they suddenly come upon the pileup, and smashing into it because they don't have time to stop. What if a software error causes this same kind of behavior in more common conditions? Google's testing wouldn't expose this kind of failure mode, because (afaik) they haven't done extensive tests driving fleets of automated cars together. All of their cars have been surrounded by human drivers, and for all we know the varied reactions of the human drivers have prevented them from hitting the Google car when it does something odd.

by thesash 13y ago

BorgHunter 13y ago

I don't believe the report[1] actually said that the risk of a malfunction is higher than the risk of human error. In fact, it's quite optimistic about the ability of automated systems to save lives, for precisely the reasons you described. They do, however, correctly point out that such systems can malfunction, and will malfunction, and that safety impacts of such malfunctions should be minimized. Because malfunctions in automated transportation can lead to, say, June 2009 on the DC Metrorail.[2] Such incidents are rare, and through hard work and foresight will be rare in the future, but that doesn't happen by itself. You have to make a concerted effort to ensure such systems fail gracefully. [1] http://www.nhtsa.gov/staticfiles/rulemaking/pdf/Automated_Vehicles_Policy.pdf http://www.nhtsa.gov/staticfiles/rulemaking/pdf/Automated_Ve... [2] https://en.wikipedia.org/wiki/June_2009_Washington_Metro_train_collision https://en.wikipedia.org/wiki/June_2009_Washington_Metro_tra...

acchow 13y ago

Certification? Hah. My friend got a G1 "learner's license" in Ontario (which requires only a multiple-choice test). When he moved to New York, he "traded in" the G1 for a full license. I guess that's the exchange rate.

DougWebb 13y ago

It's rare for all of the human drivers on a road to make the same mistake at the same time under the same conditions, but that's a likely failure condition for automated driving software if the software and sensors are standardized. This could lead to much worse accidents then you'd get with human drivers, because all of the cars will do the same thing. An all-human example: driving too fast in foggy/whiteout conditions and not having time to stop when the road is blocked. When this happens today, you get 100 car pileups because everyone is doing the same thing: driving along until they suddenly come upon the pileup, and smashing into it because they don't have time to stop. What if a software error causes this same kind of behavior in more common conditions? Google's testing wouldn't expose this kind of failure mode, because (afaik) they haven't done extensive tests driving fleets of automated cars together. All of their cars have been surrounded by human drivers, and for all we know the varied reactions of the human drivers have prevented them from hitting the Google car when it does something odd.