1925: Self-Driving Car Milestones
|Self-Driving Car Milestones|
Title text: I'm working on a car capable of evaluating arbitrarily complex boolean expressions on "honk if [...]" bumper stickers and responding accordingly.
| This explanation may be incomplete or incorrect: Can somebody proofread this? Other than that, it seems like we're finished.Do NOT delete this tag too soon. |
If you can address this issue, please edit the page! Thanks.
With the creation of self driving cars, many new milestones are being found and / or solved thanks to them. Some are good, and some are downright weird. This comic lists some that have already been achieved, some that that are being worked on, and some that are facetious "milestones".
|Automatic emergency brakes||This is another reference to how hard it can be to program human-obvious stuff (as in 1425: Tasks). A self driving car has to be able to distinguish a danger (cliff, person on foot/cycle/etc., other cars coming the wrong way/doing weird stuff) from the side of the road, the background, the other cars or even a light pole safely standing on the side of the road. Then the car also has to decide the optimal response, taking into account weather conditions, road type and traffic - whether to turn aside, just slow down (as danger is not imminent), or actually do the strong brake. There are big potential advantages for self-driving cars, if this problem can be solved: computers don't tend to panic as much as humans, would have faster reaction times, and would have more reliable judgment.|
|Highway lane-keeping||Sometimes, especially on highways where road delimitations might be faint or absent, or when lane markings could have faded away, a self-driving car programmed to pilot based on road markings would have issues holding to the correct side of the road. This is a bigger problem on highways than in cities, as cars move faster on highways, so the danger detection mentioned above might not manage to detect danger in time, while braking or avoiding the obstacle needs to be anticipated much more.|
|Self-parking||Already implemented in recent normal cars, this feature is important to remove the car from the road while not in use, and is sometimes considered a difficult maneuver for drivers to master, as it requires a good "feeling" of the car dimensions, as well as of distances and maneuverability of the car, and information about surrounding barriers. The latter parameters, being easy to sense with radar and back-camera aide, are made more reliable with computers.|
|Full highway autonomy||The ability for a car to drive itself on a highway. As of 2017, there are plans under consideration to set highway lanes aside for self-driving cars, but this milestone would require a car to be able to operate on a highway that also has human-driven cars, as well as wildlife, pedestrians, debris, and other obstacles, should they enter the highway.|
|First sex in a self-driving car||This is not a milestone for the cars themselves, but just the age-old practice of having sex in cars, performed in a car that happens to be self-driving. Whether or not this would happen while the car is in motion (other than that induced by the passengers) or on a public road is not specified, though both are implied. Given the nature of human sexuality, it is probable this has already happened, but there has not been a public documentation of this milestone.|
|Full trips with no input from driver||The main point of self-driving cars, allowing all humans within to act as passengers. As of 2017, self-driving cars require a human to be able to take over just in case, but any such trip where the human never actually took control would qualify for this milestone. However, there could be an additional joke here that the car is driving without human input including the destination. In this case, the car itself is choosing where to go, leaving the humans helpless.|
|Full trips by empty cars||A more complete version of the above, since with no humans present, no human can take control. This could be considered fulfilled by the DARPA Grand Challenge entrants, as the challenges are racing competitions of autonomous cars with no humans on board.|
|Self-refueling of empty cars||This would require either: a robotic fuel station, able to refuel cars with humans inside as well; an ordinary full-service fuel station (that is, one where the station's employee performs the refueling of the car) that happens to service a self-driving car with no humans aboard (which could be arranged as a publicity stunt); a specially designed fuel station that would allow self-driving cars to refuel by docking to it (likely to require fine control of the docking procedure that would render it unsuitable for more fallible human-driven cars); or, perhaps least likely, a robotic arm attachment on the car that would allow it to use a normal self-service fuel station. Currently Tesla's robotic charging station is the closest thing to this accomplishment.|
|An empty car wandering the highways for months or years until someone notices the credit card fuel charges||The first completely facetious milestone of the list (since "first sex", despite having little to do with self-driving cars, has probably happened). Cars are expensive enough that, were one to drive itself off and wander, some effort would be made to track it down. As this would require the self-refueling milestone, local fuel stations could be alerted to look for the "rogue" car—and in any case, whatever payment method is used to pay for the fuel would be traced.|
|Cars that read other cars' bumper stickers before deciding whether to cut them off||Another facetious milestone, implying self-driving cars might obtain the capacity to hold and act upon opinions that might override safety and efficiency of transit. This would be generally considered undesirable, so this seems unlikely to actually happen, except perhaps as an unintended consequence of runaway self-learning.|
|Autonomous engine revving at red lights||Mimicking the human practice. This is often done by human drivers who wish to draw attention to their car and then speed off as quickly as possible once the light turns green, but is regarded by most as being a nuisance. As such, this is an unlikely goal for self-driving cars to achieve.|
|Self-loathing cars||This would require cars to become sentient enough to understand, and have negative opinions about, themselves. Depending on one's definition, though, self-diagnostic software might qualify, as they would be running on a car's computer and could express a negative opinion about the car (albeit normally limited to the context of the car needing maintenance).|
|Autonomous canyon jumping||Although it seems unlikely that a navigation routine would ever decide that jumping a canyon is part of an optimal route, a car could be programmed to jump a canyon as part of a stunt or show, with no human driver (or any other human aboard) at the time of the jump. It is questionable how "autonomous" such a car would be, though. Could also be a reference to the next point, with another popular setting in below mentioned discussions: "should a self-driving car leave the road and drive into a canyon, which will kill the driver (and passengers), or stay on the road and kill others?". Possibly a reference to when a Tesla was driven off a cliff and the driver and his passenger survived without injury. The car was not on autopilot at the time. Could also be a reference to the previous point where the car develops enough self-loathing to want to commit suicide. Or it may be a reference to certain Knight Rider episodes.|
|Cars capable of arguing about the trolley problem on Facebook||The Trolley problem is a well-known thought experiment in ethics, in which a person must choose between passively allowing several people to die, or actively causing a single person to die. With the increasing likelihood of fully autonomous vehicles, there's been a flurry of interest in this problem, centered around what a vehicle should be programmed to do in such a case (for example, if avoiding a high-speed collision required running over a pedestrian). Munroe seems to mock this debate by arguing that the true milestone would not be when the vehicle can make such a decision, but when it can argue about it on Facebook. This may refer to the idea that humans aren't capable of agreeing on a resolution to the problem, so expecting a vehicle to resolve it would be less reasonable than expecting it to be able to debate. On the day this comic was released the Youtube channel Vsauce posted a video, The Greater Good - Mind Field S2 (Ep 1), where they for real tested peoples reactions to the trolley problem in a (fake) situation where the subjects really believed they where in a situation where they where choosing between saving five from an oncoming train by killing one on another track! It is almost too much a coincidence to not have some relation.|
|Evaluating arbitrarily complex Boolean expressions on "honk if [...]" bumper stickers and responding accordingly (title text)||As with the cut-off milestone, this implies development of artificial intelligence unrelated to the basic functions of a car, though still imitating human drivers' behavior. This joke is a reference to a previous comic about honking and formal logic.|
- Upcoming and recently-achieved
- Self-driving car milestones
- Automatic emergency braking
- Highway lane-keeping
- Full highway autonomy
- First sex in a self-driving car
- Full trips with no input from driver
- Full trips by empty cars
- An empty car wandering the highways for months or years until someone notices the credit card fuel charges
- Cars that read other cars' bumper stickers before deciding whether to cut them off
- Autonomous engine revving at red lights
- Self-loathing cars
- Autonomous canyon jumping
- Cars capable of arguing about the trolley problem on Facebook
- The Trolley problem became part of the joke a month after this comic in 1938: Meltdown and Spectre. And earlier, in 1455: Trolley Problem, it is even the entire subject.
add a comment! ⋅ add a topic (use sparingly)! ⋅ refresh comments!