Testing “Self-Driving” Cars – The Buck Stops Where?!

22 Jan

Hello D.O.T.,

I am very concerned regarding the U.S. Department of Transportation’s recently implied ownership of the “Self-Driving Car” testing process.

In truth, the testing of “self-driving” cars is another major software testing project (since cars do not, and never will, make decisions themselves).  This will necessitate the verification of acceptable vehicle behaviors under a huge number of real world scenarios.  All kinds of computer coding decisions will be made regarding the legal and safety “choices” exhibited by these cars. Perhaps these decisions will even be made (on occasion) by young urban computer whizzes who have not yet gotten around to obtaining their own driver’s licenses.  In addition, these scenarios, errors, and fixes will need to be documented, addressed and verified not just once, but for all of the competing proprietary systems utilized by the various car companies.  These separate “self-reported” real world “beta tests” will be occurring at the public’s risk.

Ideally, of course, this would all be combined and addressed by the D.O.T. as one overall beta test.  And by definition, this beta test would include the establishment of an appropriate project tracking system in which the public is encouraged to “report bugs”, in which issues are “logged” using a unique identifier, in which “accountability” or the “present ownership” of a particular bug is immediately identifiable at any given moment in time, and which would include the establishment of a clear decision making hierarchy regarding legalities, project priorities, and the postponing of less critical issues.

My concerns are as follows:

From my own experience – in which the D.O.T. has failed to assume ownership of any of the numerous road dangers I have documented over the past few years – it seems the D.O.T. does not currently have a public oriented, problem reporting system similar to the one described above.  Incidentally, there are certainly a number of “Best Practices” principles to be gleaned from my previous correspondences (more on that later).

On two separate occasions I was told by your employees that the US D.O.T. is prohibited from direct involvement in relation to state specific legislation.  I have also learned through experience that state legislators are not likely to admit that one of the safety “improvements” they previously voted for turned out to be dangerous.  On top of this, future dangers occurring as a result of interactions between “self-driving” cars and people will likely be a bit tricky to describe.  This is because many of society’s most logical thinkers (the aforementioned computer programmers) will have already taken into account the handling of many of the more commonly encountered driving scenarios.  These new “state specific dangers” will require much more involved, mind numbing explanations leading to even less likelihood that politicians will expend their political capital on such “I was wrong” campaigns.  And given the “recent trend” nature of the dangers I myself have been warning about, I am concerned that the D.O.T. does not understand the scope, magnitude, or irreversibility of the problems created should it let the genie out of the bottle prematurely!

One of the hardest hitting fruits the D.O.T. will be wielding, as overseers of this project, is the (largely non-binding) “Best Practices” report it will assemble in relation to self-driving cars.  I am not knowledgeable as to the inner workings of the D.O.T. or what it has in mind.  I will say, however, that nearly every one of my previous “dangerous scenarios” involved a tricky real life situation that should prove even trickier for driverless cars – I mean “computer programmers”.  When the D.O.T. discusses plans to issue this report, I can only wonder nervously – “Based on what?” “Coming from who?”  And prior to issuing these recommendations – should there not at least be a few tentative demands made?  Perhaps in relation to an end goal of “complete compatibility” between the now competing and proprietary computerized systems under development by the various car companies?  After all, aren’t these cars going to be “talking” to each other?  Wouldn’t it be nice if in the future, every time we need to correct their grammar, or slap them on the wrist, these adjustments could be made in just one location without the need to duplicate, triplicate, or quadruplicate our efforts?  Will the current climate surrounding these unseen computer codes (or “competitive advantages”) be addressed such that, if one particular company designs code that is far superior at handling a particular road danger, this life saving knowledge will be immediately disseminated to all involved before my dog “Buffy” is run over by a competitor’s car?  Wouldn’t it be nice if the D.O.T. – before handing over the keys – laid down some ground rules for these young cars who have just received their learner’s permits?

Some things in life – like “blindingly bright” modern headlight technology (also not yet addressed nationally) – can be seen from miles away.  In the case of future accidents caused by these cars, I mean “computer programs”, there will be one thing we can say for sure – “victims” and “owners” alike will be looking to put the blame on the auto manufacturers.  And since we already know “there ain’t a snowball’s chance in global warming” that any of these cars are going to end up on the roads without some degree of formal or informal immunity granted upfront to these auto companies, the question then becomes – “Who will Kenneth Feinberg be working for in these quandaries?”  …..  GM?  Takata?  …..  Hakuna  ….. Matata?  Comprende?  …..(DeNada).  It seems a bizarre possibility that those working within the federal D.O.T. – who, remember, are not the actual programmers or project managers of this computer code, who are not directly performing the testing, who do not have a complete and comprehensive setup for overseeing this testing (my deduction), and who tell me that they are prohibited by law from getting involved with state legislative decisions – will have it in their self-interests (personally and organizationally) to maintain a certain distance during this testing process.  They may actually be headed towards a bizarre alliance with the D.O.T.’s usual arch nemesis – “Plausible Deniability”.  Wow!  What a web!  And not a simple “Charlotte’s Web”, but a “Jack Webb”! (Extra ‘b’ in there)

I seriously wonder if the buck is going to stop anywhere on this project.  If the buck ends up stopping in the middle of the road and is then hit by one of these “self-driving” cars, I hope the D.O.T. will at least properly record the “Cause of Accident” under the newly added description – “Computer Crash”.