Skip to content

Self-driving cars: are you ready for the ride?

I always thought that we would eventually get there, sooner or later. The idea of having your own personal driver, in the form of a computer system powered by cutting-edge Artificial Intelligence (AI), that you can summon just by calling its name on your smartwatch ala KITT, is after all very exciting and, let’s be honest, something that everyone would be happy to live with. However, as any technology that once belonged only in the realm of science fiction starts to become part of our lives, there is an almost unavoidable psychological factor that comes into play and that manifest itself in various forms, from fear to enthusiasm, skepticism, hysteria and anything in between. I suppose this is somewhat related to a sort of “disconnect” between fantasy and reality wherein something alien once living only in your imagination and towards which you had no explicit feelings suddenly materializes just before your eyes giving rise to an instinctive reaction that may be one of those mentioned above.

So, when i became aware that several businesses including well established car manufacturers were getting serious about this technology, by extensively investing in its development and conducting real-world tests on public roads, like in California, i had the impression that KITT was no longer just a sci-fi character but something sitting in my garage in a not so distant future. My instinctive reaction was not fear nor unbridled enthusiasm or any other kind of extreme feeling, as my mindset prevents me from fearing such technology and my irrationality level is too low to be dragged into any kind of hype or hysteria. My initial reaction was just skepticism. I was not sure whether this technological innovation (improper term maybe since the concept is older than me) would really deliver what it promises in the short time frame that many analysts predict.

The ARGO Project [Italy 1998] a milestone in autonomous cars.
The thought of a computer driving a vehicle was not really what puzzled me, as autonomous (or better automated) driving systems are already being used in some industries with good success, for example in mining. But rather the fact that an AI system could be driving a car in a complex context such as a busy city with the same safety, awareness, experience and overall skills that I or a taxi driver have. Driving a car or any other vehicle is a pretty complex cognitive process considering the wide array of possible scenarios in which the driver (whether human or automated) is required to make decisions. Most of the experiments and tests conducted in the past years with this technology have taken place in more or less controlled/selected environments, such as particular types of roads or restricted areas. This has left many wondering whether it would really be capable of making its way into more difficult situations, especially in very busy and poorly regulated cities.

Google’s self-driving car [2015]
The last months have seen a long string of announcements from major players in this industry, who have started a large campaign of tests on public roads in several cities, where self-driving cars are sharing the roads with traditional cars. Claims of incredible results, very optimistic predictions and government bodies vowing to put in place all the regulations and infrastructures needed to support this innovation made me feel like something big was going to happen. So the time has come to ask myself a few questions: Would i let a self-driving car carry me around with the same level of worry and concern that i have when i let any other random human driver do that ? Is the technology really going to be mature enough in such a short period of time to justify so much buzz ? Who is responsible if the self-driving car crashes ? Will self-driving cars improve our lives ? I’ll try to answer these questions relatively to fully autonomous self-driving cars, that is one that does not allow the intervention of a human driver at any point in the driving process. I will not consider semi-automatic self-driving cars where control may be passed on to a person anytime during the ride for reasons that i will explain later on.

1) Would i let an autonomous car drive me around with the same level of worry and concern that i have when i let any other random human driver do that ?

The answer to this question was quickly found following a series of recent events that made me drastically lean towards being a big believer in such technology. One time i was driving on a busy road when all of a sudden i felt the need to sneeze. Well, i found myself almost hitting a car running on the opposite direction during the split second in which my eyes remained shut, putting my life (and that of other motorists) in great danger. In another occasion an innocent quick glance at a billboard almost made me bump into the car in front of me. In yet another occasion my dog distracted me causing me to overlook a crossing pedestrian. Another split second from something bad. These are just a few examples of how even the most careful driver can easily find himself in dangerous situations with catastrophic consequences.

For how powerful our cognitive capabilities may be, they are also easily affected by external events, dramatically rising the risk of distraction, error, miscalculation, fatigue and altered perception. Adding to that the influence of bad habits like drinking, doing drugs, sleep deprivation, texting, chatting, speeding, etc. one can quickly realize that the biggest source of hazards on the roads are the human drivers themselves. The majority of car accidents are due to human factors and very few are caused by events unrelated to the drivers. So it’s not unreasonable to think that a car driven by an AI system would very likely be safer than one driven by a human, since it wouldn’t be affected by all the weaknesses that are typical of humans and that cause most of the disasters on the roads. I have even come to a very radical conclusion that i never imagined i ever would, and that is: humans should not drive on public roads, unless they belong to special categories. Hopefully, AI systems will eventually surpass the driving skills of the average motorist and will be safer than humans. I wouldn’t be surprised if a time will come when insurance companies won’t issue any policy for anything driven by a human.

2) Is the technology really going to be mature enough in such a short period of time to justify so much buzz ?

Despite all the buzz, honestly i do not believe self-driving cars will be replacing traditional cars en-masse in the immediate future. And that is not only because of the price tags that they will probably carry initially, which will keep the masses away from them for quite some time, but also because of the reliability and validity of the test results that will require a substantial amount of peer-reviewing and certainly not only validation from the manufacturers alone (obviously biased). I think that the adoption will be very gradual, not widespread, probably starting in the logistics industry which will serve as a test ground to refine the technology. If any use will be made in the short term to move people around, this will regard very specific uses in restricted environments, such as dedicated lanes/roads/areas where the risk for damages is minimized. One example may be the use of fully automated taxi or shuttle services for public transport in particular city areas.

Whatever approach will be used to kick-start its adoption, there is a great deal of development, data acquisition, testing and analysis to be done in order to build robust models that ultimately will form the core components of this technology. This includes real-world real-time environmental data, mapping, and other rules in place at the location where the car is running which will be used in the decision processes by the AI algorithms. Big data analytics and cloud computing technologies, which are already extensively used today, will play an important role, along with the development of robust and efficient Inter-Vehicle Communication and Vehicle-to-Infrastructure/Roadside Communication systems that will make the analog signalling currently in use and designed for human drivers become obsolete. Something so revolutionary will, without doubts, demand extensive research, planning and development in order to put in place all the required infrastructure to fully support this technology, and, regardless of what enthusiasts and players in this field are trying to convey, it’s something that’s not going to happen overnight worldwide.

Also, the realization of fully autonomous self-driving vehicles poses the big challenge of creating AI systems capable enough to implement the decision making abilities demanded by a complex process such as driving. There will be the need for powerful technology to achieve such capabilities, along with adequate computational power to enable the implementation of algorithms and models (today represented mainly by the buzzword Deep Learning) that will lead to the design of AI systems with cognitive capabilities similar (or probably better) than those of a human being. This doesn’t mean, obviously, that automated self-driving cars will be immune from errors. But it is not unreasonable to think that they will maintain a quota of failures (these are unavoidable) that are statistically irrelevant compared to that of even the most careful human driver.

Whether the current AI technology is adequate to achieve the ultimate goal is still unclear. Probably a break-through will be necessary. But AI systems are already outperforming humans in many tasks, so it is likely that they will one day be better drivers as well. The public infrastructure is lagging behind and this is unfortunate because so long as it doesn’t catch up, the full potential of this technology cannot be exploited. Although considering the amount of regulation and standardization to be done, this is understandable.

3) Who is liable in case of a car crash?

One of the issues brought up in the debates about this subject is the problem of who is to be considered liable in the event that a self-driving car crashes. There seem to be a bit of confusion in this regard mostly due to the fact that it involves legal matter, which is not exact science and may lead to different interpretations and applications in different places. Since the legal field is not my domain i can only give my opinion from a logical point of view.

Responsibility for something involves determining if an event is caused by actions that are under the control of a subject, for which effects it is then accountable. Unless the subject accepts responsibility even for something that happens out of his control. In a fully autonomous self-driving car all the decision making would be carried out by the automated driving system since the human should not be allowed to interfere with the driving process, so he/she has no control whatsoever on any driving-related event that may happen. Therefore, logic dictates that he/she cannot be held accountable. Furthermore, unlike other computerized systems usually featured on vehicles, such as ABS, EBD, ESP, Cruise Control, Driver Assistance Systems, etc. whose actions are very limited in time and integrative (that is in addition to actions taken by the driver), an automatic driving system takes full control of the car’s ability to move and run at all times during the ride, thus it should be considered a fully liable entity, and the subject upon which all liabilities fall in case of road accidents. This is, as i see it, exactly the same as having a personal driver hired (or bought?) from the manufacturer, albeit a robotic one. Who will legally represent this robotic driver, that is who is to be considered the legal subject responsible for its actions, is yet to be seen. I suppose it would be the manufacturer.

The only case where i see liability falling on the car owner is if he/she has the possibility to program the automated driving system to an extent where there is the chance to have the car perform some illegal driving activity or if the system is tampered with. These situations, however, should not be possible in that the system should have in place the necessary components and logic to prevent such scenarios. All the car owner should be allowed to do is setting a valid, legal destination and issuing few other simple commands during the ride (such as stopping for emergencies or changing destination). From there on, the automated system will make all the relevant driving decisions relieving the car owner from any liabilities. Since self-driving cars are not going to be isolated systems but extensively connected to other vehicles or infrastructures, there is also a cyber security problem to take into account. But, by definition, hacking involves exploiting intrinsic bugs, defects and weaknesses of computer systems, so it’s easy to guess where the liabilities lie in such scenario.

However, things would be more complicated if the car owner is allowed to interfere with or participate in the driving process. For example, if the automated driving system gets to a state where it cannot make decisions with sufficient degree of confidence and reaches a stale point or takes the wrong decision, the most likely solutions would be either asking the car owner for intervention while still retaining control, thus sharing liability, or handing over total control, whether automatically or on-demand, with consequent transfer of liability to the human. I am not an advocate of these kind of assisted/mixed driving systems as they would make the lines of responsibility a bit fuzzy, consequently making the determination of where the liabilities lie a bit problematic. Furthermore, and even more worrying, a self-driving car that is not totally autonomous would put too much pressure on the car owner, who must always be alert in order to promptly intervene, and considering that disasters on the road can happen in a split second this would rise serious safety concerns. This approach is an indication that the technology is not yet mature, and trying to pass it off as “autonomous driving” is a big hoax.

4) Will self-driving vehicles improve our lives?

Well, consider this scenario: no more high/drunk/sleepy/fatigued drivers, absent-minded, fast and furious, P-platers, seniors, texters, chatters, etc. behind the wheels (not controlling the car at least) is already enough to consider it a big improvement to our lives. And everyone of us can be categorised into at least one of those even if only once in a while. Then think of all the things that can be done with the time otherwise employed driving, plus the possibility for disabled people or whoever doesn’t want or can’t drive to easily move around. For businesses, this would represent a further process that is automated, with all the benefits arising from automation, that is increased efficiency and reduced costs.

Some may argue that, after all, driving is also a leisure and that autonomous cars will deprive motor enthusiasts from an enjoyable activity. While this may be true, there are recent studies and statistics showing an interesting trend, and that is the new generations becoming less inclined to get a driver license. Times they change. With new economic crisis happening more frequently, the emergence of new technologies that allow communicating with other people without requiring to move around, the on-demand/sharing economy allowing you to get goods and services straight to your doorstep at the touch of a button, it is not hard to foresee a future where being able to drive is no longer a must for the ordinary citizen. I believe there are enough benefits to say that our lives will be improved once this technology evolves to a mature state. The future of cars is self-driving.


Published inAutomation