Autonomous cars and some semi-autonomous cars will be able to potentially do facial recognition on us while we are walking around. Getty
Recent headlines have made well-known the ban that San Francisco has enacted on facial recognition within its borders, becoming the first major city to take such action. Some assert this exemplar will not be the last, even potentially opening the regulatory floodgates toward a mounting wave of similar directives .
But despite some breathless headlines, the ban is not quite what it seems on the surface. The scope is limited to the use of such technology by the city departments of San Francisco, plus there are a number of carveouts and exceptions allowed in the new regulation.
For example, this city-bold proclamation does not encompass say federal agencies, it doesn’t encompass private companies by-and-large, and it doesn’t “ban” the city departments per se since it also allows for exigent circumstances for the city departments to possibly make use of facial recognition technology (meanwhile, the regulation tries to mitigate this exclusionary allowance by requiring justification and some transparency in reporting on what and why the presumed intrusive act was undertaken).
What is especially striking is that among the nationwide heated debates about the societal value of facial recognition, with strong cases being made on either side of whether to ban or not ban, in a sense the city of San Francisco has made their decision and laid down the gauntlet on the matter, namely this is what the proclamation indicates : “The propensity for facial recognition technology to endanger civil rights and civil
liberties substantially outweighs its purported benefits, and the technology will exacerbate
racial injustice and threaten our ability to live free of continuous government monitoring.” Notice that there is an unequivocal calculus proclaimed that the societal costs of facial recognition outweigh the societal benefits. No confusion there.
For some, this is a weighing of the scale that does not quite fit their assessment of the tradeoffs involved in utilizing facial recognition. There are those that would be quick to point out that when there is a widespread act producing horrific carnage, and if facial recognition could be used to find and catch the culprits, the benefits of facial recognition are well-justified as to potentially preventing further carnage and stopping the criminals cold.
In fact, perhaps surprising to some, the San Francisco regulation tends to recognize this exigency, by providing this carveout : “Exigent circumstances means an emergency involving imminent danger of death or serious
physical injury to any person that requires the immediate use of Surveillance Technology or the
information it provides.” And so this exception clause would seem to allow for the prevention of facial recognition, predominantly, whilst also sensibly making use of it when a dire situation merits doing so.
But there’s a crucial factor to be considered about this idea of being able to suddenly turn on a switch and make use of facial recognition, namely that if the overall ban has pretty much precluded putting such technology into place, how can you all of a sudden on the spur of the moment make it magically appear and leverage it?
This would seem like a rather incredible feat of pulling a rabbit out of a hat, for which there wasn’t a rabbit put into the hat to begin with (oops, sorry, spoiler alert about the rabbit in the hat magical trick). If the city departments know they aren’t supposed to generally be using facial recognition, presumably they aren’t going to invest in buying, deploying, and keeping it active, merely to await the day or hour that it might urgently be needed. Thus, the exception clause almost seems like a false carveout due to not being especially practical given the overall tone and demeanor of the proclamation.
Here’s a question for you, would the advent of autonomous cars impact such a ban and if so in what ways?
I’m glad that you are willing to ponder the question since I’ve got some thoughts about it that you might find of interest.
Autonomous Cars And Recognition Technology
For true self-driving driverless autonomous cars , those of the Level 5 and Level 4, and to some degree the semi-autonomous cars as well at Level 2 and Level 3 , there is usually a slew of sensory devices included into the car for purposes of sensing the surroundings of the car, including the use of cameras, radar units, ultrasonic devices, LIDAR , thermal detection subsystems, etc.
Cameras are pointed outward to capture the visual aspects of what is nearby to a driverless car, collecting streaming video images that are then processed by the AI on-board of the vehicle. This allows the AI to try and figure out that there is a bicyclist over there, a pedestrian up on the curb and another car is ahead and moving at a good clip. Vision processing is akin to the “eyes” of the self-driving car , enabling the AI to scan for objects that are stationary and are in motion, along with classifying the objects as to what they might be and what they portend for the driverless car.
Currently, the vision processing capabilities of autonomous cars are normally only concentrating on objects as a kind of blob. For example, the human that’s standing at the street corner waiting for the light to change and might cross the street, well, it is a human being that is likely an adult since their height is six feet or so, but otherwise the human is not further classified or visually dichotomized.
There’s no particular reason that the visual analysis couldn’t go further.
Assuming the camera is good enough to capture a high-quality image, and assuming that you’ve got sufficient computing capabilities packed into the driverless car, the AI system could try to do some more in-depth processing and figure out that the human being is likely a male, favoring their right leg, carrying a briefcase, and wears glasses. This is somewhat easy to visually calculate by merely inspecting the picture images being collected.
Guess what, this could also include doing facial recognition.
I’d like to have you soak in that point. I’ll wait.
Why is this a point worthy of some in-depth contemplation?
Because once there is widespread use of autonomous cars, it implies that there is a possibility of having those roaming and continually roving driverless cars acting as a facial recognizer that could potentially track lots and lots of people as they proceed to walk around during their daily lives.
Driverless Cars Everywhere As Rolling Video Recorders At Every Turn
Think about the number of cars that go past you as you walk in any downtown city, making your way from your office to the local pub for an after-work respite. Now, imagine that many or say even all of those cars had video recorders, plus those cars were analyzing the video in real-time and doing facial recognition.
If we could stitch together the findings of those cars, admittedly a bit of a difficulty having to get the data from those disparate cars, but possible , we could likely tell you, or anyone else, where you walked, how long it took, whether you were looking up or down or toward a billboard, whether you were happy or sad, whether you were talking and possibly even do some lip reading analysis via the onboard computer to know what you said, and so on.
Plus, since most autonomous cars are going to be electronically transmitting their onboard data up to the cloud , this data about you and the facial recognizing analysis could be done on a much larger scale , tying you to every act of being outside, such as in the morning capturing the fact that you came out of your home to get the daily newspaper on your driveway and then got into your car, and then you during work in the afternoon went outside to get your lunch, and then at the end of the workday you walked over to the pub.
I realize that some will say I am trying to suggest the sky is falling, which might seem that way, but it is simply the reality of what might be possible down-the-road, once we have a prevalence of autonomous cars on our public roadways.
All of this is technologically feasible, already, and it is more a matter of whether the automakers and tech firms would want to implement something like this, and whether society would want them to or might balk at the adoption of a seemingly intrusive form of automated recognition. As earlier mentioned, there are tradeoffs of whether this kind of Big Brother approach will make our lives worse, or whether the potential benefits would make it tenable.
I’ve so far herein focused on facial recognition, which is the presumed scope of the San Francisco ban, as per the headlines, but I think that perhaps some of the media did not go the trouble to actually read the formal regulation, which states : “Surveillance Technology means any software, electronic device, system utilizing an electronic
device, or similar device used, designed, or primarily intended to collect, retain, process, or share
audio, electronic, visual, location, thermal, biometric, olfactory or similar information specifically
associated with, or capable of being associated with, any individual or group.” Please carefully read and digest that statement from the regulation. It’s quite a bit broader by far than merely facial recognition. In that manner, you could say that likely any or all of the sensory devices of an autonomous car would be encompassed.
Which, in view of the desire to apparently prevent tracking of us humans, the regulation appears to be trying to ensure that no loopholes are allowed. Keep in mind that the facial recognition that could be undertaken via the cameras of a driverless car could be augmented by or perhaps even directly undertaken via the other kinds of sensors onboard too. It’s definitely harder to use those other sensors for quite the same reliability of tracking a person , but they could certainly help in doing so.
The final comment herein about the San Francisco ban involves this equally important point that seemed to not grab the attention of the widespread media, specifically this aspect : “Ordinance amending the Administrative Code to require that City departments
acquiring Surveillance Technology, or entering into agreements to receive information
from non-City owned Surveillance Technology…” The rub on this point is that the city departments aren’t supposed to acquire (and presumably nor put in place) these technologies, something that I think was already generally expected via the ban, and nor enter into agreements to get such data from others such as private businesses. That’s an intriguing extension, and I suppose once again fits with trying to tie off any loopholes.
I’ll return to my rabbit in the hat point, and emphasize that suppose a firm like an automaker or tech said it could provide to the city an ability to rapidly scan across all of its fleet of driverless cars to find a mass killer that has just performed an unspeakable act. In theory, this kind of prior arrangement would be precluded, thus, making it harder to try and leverage the capability, doing so for the benefit of the city, on a spur of the moment matter.
One last thought on this is whether the emergency vehicles to be used by the city are within the scope of this ban, which it certainly seems that they are, and once those police cars, ambulances, fire trucks, and the like are either semi-autonomous or autonomous, they too would by definition have the sensory devices for collecting video and other aspects, of which, the possibility of doing facial recognition (or other kinds of recognition) would be feasible. It will be interesting to see how the city wrestles with this aspect.
All told, I’m not arguing that the societal benefits outweigh the societal costs, and nor am I claiming that the societal costs outweigh the societal benefits, but instead trying to awaken us as a society to the approaching era upon […]