We all have an idea of how the future of driving looks however far-fetched and fantastical the source of that imagining might be. While there is no sign of flying cars, our fantasies are undoubtedly coming to fruition where automated driving is concerned. And given the history of technological innovation, it is quite likely that these fantastic imaginings played a part in their development. However much room there is for fantasy here, things become much more sober when it comes to distinguishing between levels and kinds of drive automation. While we have yet to achieve all the possibilities forecast in this field, two subtly different forms of drive automation represent the pinnacle of technological achievement so far – autonomous and self-driving vehicles. This post hopes to help distinguish between these two limit-possibilities of drive automation and given some timely examples of both.

Autonomous vs. Self-Driving

First, a disclaimer is necessary: the lingo surrounding vehicle automation is a matter of intense debate! This post merely represents an attempt to discern between two terms which are often used with the intention to distinguish, however obscurely, between two things. Undoubtedly, there are use cases where the two terms are indistinguishable or interchangeable.

Full Self-Driving is most prominently associated with Tesla's autopilot beta which is regularly called by that name. This suggests that consumers associate full self-driving with privately owned vehicles – something which might help us begin to draw distinctions here. Full self-driving indicates that the is vehicles is capable of fully managing all the logistics and mechanics of travel, but is still governed by a pilot or front seat passenger who has established the destination, and has the choice and power to intervene at any given time.

This appears naturally distinct from an autonomous vehicle, which suggests that in some regard the activity of the vehicle unfolds without the governance of its passengers. That would be analogous to any situation where we consider a person to be autonomous – making their own decisions, not just on the way to a particular goal, but in the establishment of a goal itself. Of course, what I am describing doesn't need to be interpreted so radically as to depict some kind of Terminator-esque dystopia. A far more domestic real-world example fits the bill too: imagine a car which has been equipped with an AI that allows it to intelligently function as a taxicab. Beyond just getting customers from destination to destination, it also has the capacity to decide which customer to pick up, on the basis of proximity, ride distance, road conditions and etc. – perhaps even coordinating these decisions with other human and automated taxi drivers as well!

If this description has you thinking that autonomous vehicles are clearly more advanced than self-driving technology, you would certainly be right – at least in some key regards. In terms of sheer processing capacity, an autonomous vehicle requires more sophisticated technology. This is reflected in the Society of Automotive Engineers (SAE) intelligent vehicle ranking system. What we are here calling full self-driving, has been argued to correspond with SAE level four – that is, in fact, what Tesla's self-driving suite is considered to be. This leaves a final SAE level – presumably that occupied by autonomous vehicles - level five, 'steering wheel optional'.

There is, actually, a caveat – even those autonomous vehicles like the ones described above, which have been implemented in limited real-world circumstances, only technically occupy level four, too! This is a matter of safety, policy compliance, and the technicalities of SAE ranking. The requisite of SAE level five is full automation in all circumstances. Many autonomous vehicles which have the capacity for complete autonomy are implemented without the real option to do so. Companies such as Waymo, a small scale American 'robotaxi' company, would not want to simply entrust the outcome of all of their fleet's maneuvers' to computer intelligence! The legal complexity of such a decision is uncharted territory, and the consequence of accidents which unfold in light of such freedom would be a moral and public relations disaster. There is the additional issue of unions, too – taxi unions are well established and the removal of a human component, and thus human employment, from the transit of peoples, is staunchly opposed by such unions.

This all serves to illustrate why 'autonomous' is such a precarious term, and why a truly autonomous car may never be seen or implemented. At least, this much is true when the sense of autonomous is interpreted strictly, or in accord with SAE ranking – at least in its present state. SAE rankings are an ever-changing quantity, though the basis of this change isn't entirely transparent. Lately it has been occurring in accord with requests for greater clarity and granularity. Given the pragmatic and semantic difficulties of reaching true autonomy in vehicles, the time will undoubtedly come when level five must be reinterpreted or stipulated to recognize the incoherence or undesirability of a truly autonomous car. For now, it is perhaps still tied too much to the fantasy that bore all these amazing vehicles.


The difference between self-driving vehicles and autonomous ones is difficult to iron out, but here I have attempted to make a practical distinction. Using SAE rankings or a strict semantic understanding of autonomy leaves us with an inaccessible and undesirable imagining of future technology – this clearly will not do, as the term is used to describe things already existing today. Some people have taken the approach of criticizing those who 'inappropriately' use the term when describing intelligent vehicles, but there is room for an understanding of autonomy which makes it usefully distinct from those with full self-driving technology. The understanding is thus: full self-driving vehicles are those, piloted by a private owner, which have the capacity to carry out the entirety of the of a trip, no intervention and little to no supervision necessary on the part of the driver. Autonomous vehicles meanwhile refer to those where, on the ground and in the vehicle, little to no decision making happens on the part of a passenger, who may be a customer or guest of some kind – if there is any human element whatsoever, and not just a load of some sort. Chief examples of this sense of the autonomous would be 'robotaxis' like Waymo, the shuttles used in the Olympic village, automated cargo trucks, or Google Maps vehicles which independently plot out the world's roads.