Self driving mirage

 

Everybody is trying to build a self driving car today. Google has been testing their solution for the past ten years or so, Tesla just announced they'd be putting the "self driving hardware" onto their newly manufactured cars, Uber has a big effort with Volvo in Pittsburgh, comma.ai is trying to ship a box for outfitting certain cars with a self driving mode etc. Obviously the car manufacturers are following with Ford making announcements recently, BMW working silently and so on and so on. Some of these efforts are explicitly cautious on whet they promise (driver assist technology rather than full autonomy such as e.g. Toyota), but many voices, particularly the VC's from the Bay area are hyperactive announcing how the life will be great and how the self driving car (in the sense of full autonomy) is a done deal.

Well I would not be a sceptic if I did not put all those hyper-optimistic statements to doubt. Let me go through a few claims about self driving cars one by one and put my sceptical comment next to each statement. To be frank: I'm not against the technology, I'm against the hype.

  1. Self driving cars will be safer than human drivers. Given all of the fancy sensor technology on those vehicles, they will be safer in the sense of not bumping into any static obstacles. But even though the absolutely basic level of obstacle perception can be solved with technology, the more advanced reasoning about the situation and broader context remains elusive. The car knows there are other cars on the road, can even detect turn signals, but cannot  read the intention of those on the road, neither can it deal with unusual situations. The autonomous cars cannot be well communicated with verbally (such as drivers can) therefore cannot be easily instructed to follow a particular pattern in a given unusual situation and so on. In this broader sense the self driving cars can be actually more dangerous then an alert human driver.
  2. Self driving cars will hugely improve safety on the road. Maybe. But this stuff is tricky to measure. Statements like: self driving car is safer than an average human in metric "X" can be misleading. First of all, metric "X" can be irrelevant. But even if it is relevant,  the distribution of readouts of "X" for human drivers may not have a very well defined mean (e.g. if that distribution is a long tail). Consequently the mean can be substantially affected by a few outliers (e.g. a group of very unsafe drivers such as drunk teenage) and a great majority of drivers may be way safer than an "average driver". Now sending as fleet of autonomous cars safer than the "average human" may in fact drastically decrease safety on the road.
  3. Cars sit on the lot for a greater part of the day, autonomy will allow for better usage. To some degree yes. But the unevenness of usage of cars during the day stems not only from the fact that they need to be driven, but purely from the fact that the demand for transportation varies enormously during the day. That is, if we build enough autonomous cars to cover the peak traffic hour, then majority of these cars will be sitting on the lot for the rest of the day much like traditional cars do. Just because there will be no demand for transportation off the peak.
  4. People want self driving cars. As much as the dream is romantic, majority of the random people I've talked to actually don't want a self driving car. They like to drive themselves and are cautious about technology. They would welcome drive assist features but having a car without a steering wheel is a whole different story. Since the product is not even really available yet, it is not clear to me at all if the demand is there. And even if that demand is currently there, it can easily get demolished by a few tragic accidents with a horror story of somebody inside an autonomous car being killed without any ability to act (particularly if the situation that lead to the crash was actually easy to avoid by a human).
  5. Self driving cars may not be smart enough to handle other humans on the road but soon there will only be self driving cars on the road and therefore it will be much safer. This is really not clear since we are entering the area of nonlinear effects. Even if a single self driving cars is "safer" by some measure than a human driven car, the freeway full of self driving cars may not be any safer than a freeway full of regular cars. There are numerous scenarios, notably: somebody hacks the self driving system leading to a major disruption (best case) or a massive series of horrible accidents (worst case). Or several autonomous cars interacting together lead to horrible mistakes: e.g. cars rely heavily on other cars for lane detection. Some cars begin to drift dragging more cars behind them into a dangerous situation and so on.
  6. Self driving cars will be expensive but it does not matter because they will be shared. Well, ride sharing is certainly a thing, car sharing cannot really take off. People are actually quite used to their car, they have their own custom mess in there. Families have car seats, young males have their fancy audio systems and tuned engines and so on. Self driving/sharing culture is all against that. It may find its niche, but does not seem like it can easily replace our current life style.

I think before we start blowing the hype about the self driving car technology we should seriously start discussing the issues such as above. A lot of the problems with todays traffic congestion/pollution can be fixed by building better public transportation, which in the end might be cheaper, safer and easier to put together, particularly utilising some of the recent technology. The self driving car seems a little bit like a romantic dream from the Jetsons show. Even though seemingly great, deeper down may be a mirage. Not really a solution to a real problem but a costly, semi functional, hype driven caprice.

Update:

On Oct 28 2016 comma.ai decided to cancel their product comma one which was supposed to outfit some cars with a "self driving mode". The cancellation came after a very reasonable letter from NHTSA which expressed concerns over product safety.

In addition here are a few videos of Tesla running autopilot crashing into other cars:

The footage that actually cause a fatality of Tesla slamming into a white truck is not available. There is however a picture of the wreck visible here:

I think enthusiasts of self driving car should watch those movies carefully.

Comments

comments