Baby you can drive my car

Posted on Friday, November 23, 2018

My last thoughts (for now at least) on the 5G business case: I do not expect to see fully autonomous passenger-car driving in my lifetime. Sorry.

I have another 20 or 30 years on the planet. Maybe. Hopefully, I suppose. I do not expect to travel in an autonomous vehicle except perhaps on a closed track – like getting from an airport parking lot to the main terminal or at a theme park like Disney World (which, come to think of it, without grandchildren, I do not really expect to ever see again either). Maybe you will if you are younger (statistically most of my readers should be younger) but I do not expect to be traveling in my own or someone else’s autonomous family car on the open road. Even in North America, Europe or developed Asia.

Yes, there are already autonomous vehicles in mines and large factories or warehouses. These operate in what I referred to a ‘closed track’ where the complexity is far lower and the consequences far fewer.

But I believe we are years if not decades away from having the same technology work on the open road with dozens of vehicles operating at speed within a few car-lengths of each other, thousands of cars sharing a highway (and with trucks and buses), perhaps millions of cars in a city.

Too many legal problems: Who pays for the accident? Too many regulatory issues: Who certifies the software? Too many moral issues: No time to brake and no place to swerve: Hit the dog? Hit the bus? Hit the baby carriage? Hit the old man? Drive off the cliff and maybe kill the driver? Too many infrastructure issues: Are even the roads in Korea or Japan or Singapore or Dubai ready for autonomous driving? Too many implementation issues: How many things have to work absolutely perfectly and in harmony? This will take a very long time to become widespread.

Drivers today deal with these issues but they are individually liable for their decisions. Of course, there is insurance but in some cases the liabilities go beyond insurance and there are moral or even penal consequences. Who is liable if something goes wrong in an autonomous car? How long will it take us to be comfortable with a solution and develop the legal frameworks to implement it?

And how do we jump from today’s mix of vehicles of varying ages and technologies to a perfectly homogenous fleet of cars all with the latest software load so they can communicate with each other? Or will we build dedicated roads only for autonomous vehicles? And who will pay for that? (And since I always approach these questions from an Emerging Markets, especially a Latin American viewpoint, can anyone out there with knowledge of the road systems in these countries put their hands on their hearts and swear that we will see autonomous driving by, say 2030? Especially in hybrid situations where there are still Latin American drivers operating non-autonomous vehicles?)

Do not yell at me about all the trials that are going on. Those are – by definition, trials – and they do not represent ‘normal driving conditions’. The open road trials going on have drivers behind the wheel, attentive drivers who are (presumably) not reading the newspaper or watching a movie or (as in the latest idea) lying down in a bed, asleep while the autonomous car takes them to their destination. (Anyone noticed that these trials all seem to be done in the US Southwest where there is minimal rainfall and no snow? Or black ice? When will we see a trial in Eastern Ontario in November or February?)

The controlled conditions that govern the trials actually prove my point: that it will be many years before the controls will be removed and this works, on a broad scale, on the open road, with human beings of various levels of moral and even scientific awareness. The driver who died thinking that his Tesla was fully autonomous when, in fact, it only had various level of driver assistance and only in certain circumstances, illustrates my concern.

Yes, I know that autonomous cars are theoretically safer and that the trials have reinforced the statistical conclusions. The few accidents that have occurred represent far fewer incidents per kilometer-driven than normal driving.

But the events of the last few years should have shown us that the average voter does not make decisions like an engineer. Since all news, especially statistics, are presumed to be lies, slanted one way or another for political gain, people have taken to reacting viscerally. A few, unfortunate, but inevitable, high-profile accidents will overwhelm the statistical evidence that there are far fewer such events than if humans are ‘running the show’.

We are also far from believing that machines make better decisions than humans, mostly because machines are at their most effective when they make choices that a human would not have made. Thus even though their decisions may be better (according to the criteria with which they are programmed), the strangeness of such decisions is perceived to be somehow ‘wrong’.

Machines also are perceived, rightly perhaps today, to be unable to make moral decisions like those in the crash avoidance exercise I used previously. A human driver would likely not be judged for the consequences (“Poor sod! He had a split-second to decide.”) and in any event, his decision would be ‘moral’ because he had made it based on the best information that he had at that split second. An autonomous car that made the same decision would be judged on a different time-scale: the machine’s ability to process more information and think faster than the human in the split second plus the programmer’s ability to have calmly and rationally worked out all the possibilities when the software was constructed.

Thus, perhaps perversely, we feel better about a human sniper taking out a target (perhaps incorrectly) or a human drone operator assassinating someone (perhaps incorrectly) than an autonomous drone making its own decisions about which target to attack. The public has (at least so far) been more comfortable with moral decisions where a human being was ultimately responsible for making the final choice, rather than a ‘blind’ machine.

And then there are the lawyers, who, by the way, happen to dominate the political ranks where these decisions will be made. They are likely to make low risk decisions and ensure there is a human being clearly in the liability chain.

Bottom-Line: So yes, I believe there will be autonomous vehicles in industrial settings especially mines, airports or large parking lots. We may have autonomous trains (although union issues will probably scuttle that idea). We may even have ‘trains’ of autonomous buses or autonomous trucks, moving at high speeds but with minimal separation (so no idiot non-autonomous driver can slip into the middle).

There will be increasing levels of driver assistance in passenger vehicles but even if full autonomy is achieved technologically, the driver will remain responsible for anything that happens. No one should be lowering their seat back to a fully reclined position and ‘catching a few winks’. That is essentially the situation today and I cannot see it changing in my lifetime.

I can accept that maybe current High-Occupancy Vehicle lanes could become dedicated to (nearly) autonomous cars. The lanes will have more physical separation (like walls so no idiot non-autonomous driver can slip into the lanes).  All cars will drive at the same speed and will have isolated on/off ramps so if some idiot is sleeping or reading a book when the vehicle gets to the designated exit, their car is not hurtling driverless into traffic with other non-autonomous vehicles.

I still do not know who will pay for the not-inconsiderable conversion cost of doing this. Taxpayers? Including those who do not have or want to have autonomous vehicles? Car manufacturers? Tolls?

For this reason, I have trouble conceiving of dedicated highways where autonomous cars can whiz around each other at different speeds, while their owners sleep, play Fortnite or catch up on the latest series on Netflix. It is not that I do not believe that this is technically possible. It likely is. I just cannot understand who would pay for it. The World’s Most Advanced Economy – the US – has serious infrastructure issues and it cannot convince taxpayers to do basic repairs. These same taxpayers are going to agree to build dedicated highways for autonomous vehicles? If car manufacturers picked up the tab, how much would they have to charge as a monthly subscription fee to pay for infrastructure normally estimated at more than US$1M per kilometer?

To repeat, I do not doubt that we will develop the technological capability to have fully autonomous passenger cars. However, I doubt that these would be able to deal with a hybrid environment (autonomous and non-autonomous cars sharing the same road) or even with existing road infrastructure (especially in countries like Emerging Markets or perhaps even the US which have serious infrastructure deficits).

I do not doubt that, if possible, fully autonomous passenger cars would be safer, use less fuel, be better for the environment etc. I do not doubt that, using these benefits, a business case could be made for autonomous vehicles.

I seriously doubt that fully autonomous vehicles where the driver is not expected to be alert, aware, behind the wheel at all times and fully liable for any thing that happens are politically / legally feasible, at least within my remaining lifespan.

Title Reference: The context is obvious, I think, and maybe I don’t need to explain where this comes from. But I will anyway: Drive My Car is from the Beatles’ 1965 Rubber Soul album in the UK and from the Yesterday and Today compilation released in 1966 in the US.  Beep beep’m beep beep yeah


No Comments »

Leave a Reply