Photo: HandWashingRooster / Twitter
The study was conducted on drivers using Tesla's Autopilot, but to be perfectly fair toward the EV maker, the numbers could most likely apply to any other advanced driver assist system out there that includes auto-steering and traffic-aware cruise control (TACC).
However, there is a reason why the authors of this study chose Tesla's Autopilot: according to the paper's introduction, it's because it is "considered to be one of the most capable systems commercially available". Well, "one of the most capable" is definitely not something a lot of people will argue against.
The premises of the study are something a lot of people have argued before, only to fall on deaf ears when it came to Tesla and the company's most ardent followers. It's been pointed out that a partially automated driving system runs the almost guaranteed risk of lowering the driver's attention levels. This will lull them into a sense of security that is only true for as long as no unexpected situation (or "edge case", as Tesla likes to call them) comes up that the AI can't handle.
As we've seen so far, these can vary from an emergency vehicle stopped on the side of the road to a semi-truck performing a U-turn, and who knows how many other scenarios the system hasn't been exposed to yet. In these situations, the driver would be required to resume full control of the vehicle in an instant. However, if their attention levels won't be at the same level as if they had been driving the whole time, the reaction might not be entirely satisfactory.
This has been argued over and over again, and yet Tesla apologists seem to ignore it. Yes, other manufacturers offer similar Level 2 systems, but the difference here is that they don't claim their systems to be any more than they are. GM isn't saying you can use Super Cruise to drive from one coast to another with no interventions, is it?
According to the study's findings, the duration for both on- and off-road glances featured an average difference of 0.3 seconds between manual driving and using Autopilot (AP). The difference favored manual driving in both cases (longer on-road and shorter off-road glances), but the one aspect where AP use stood out was in the proportion of off-road glances at the center-stack (infotainment system) that exceeded two seconds - here, we're looking at 4% in manual mode versus 22% for AP.
Tesla says AP relieves drivers from having to think about certain tasks and enables them to focus on the environment, essentially making them more attentive to their surroundings and thus, better, safer drivers. Well, we all know that's a load of crap, and it's not even Tesla's fault, just human nature. Give us the apparent choice between doing something we don't necessarily like and any form of entertainment, and we'll go for the latter.
The idea that a more relaxed driving experience makes for more focused drivers is also completely wrong, and you don't need any ADAS to notice it. All you need is a few hundred miles and a highway. You are much more likely to become bored and be taken by surprise by a car coming from behind than if the highway is extremely busy. An increased number of stimuli may make a driver tired more quickly, but it will also keep them engaged and on their toes. Freeing them of responsibility will do the exact opposite.
More and more companies are realizing Level 3 autonomy is tricky, not so much from a technical point of view, but rather a legal stance. No, actually it's not even about legal liability - Tesla is getting away with it right now just on the basis of a disclaimer. It's actually the moral side of things that gets in the way. Some people in the industry believe that Level 3 autonomy (vehicles capable of driving themselves in virtually all circumstances but require human supervision at all times) - something Tesla, or any other manufacturer for that matter, don't offer yet - should be skipped altogether until Level 4 is doable precisely because of these very conflicting signals it sends for the driver.
On the one hand, they're telling them to relax and enjoy the ride, but on the other, they're supposed to be ready to jump in at any moment as if they'd been driving the car the whole time, something everyone knows is not only unlikely, but virtually impossible.
There are plenty of ways in which this study can be labeled as irrelevant, and a lot of Tesla fans are already doing that. However, it shouldn't be interpreted as a dig at their favorite EV maker, but rather a warning to everyone that Level 3 is a dangerous game to play. Either you take full responsibility as a manufacturer for what your system is doing, or you stick to Level 2 and limit its functions to the highway. It may sound retrograde, but retrograde isn't that bad when the alternative is sharing the road with imperfect AIs and oblivious drivers.
The premises of the study are something a lot of people have argued before, only to fall on deaf ears when it came to Tesla and the company's most ardent followers. It's been pointed out that a partially automated driving system runs the almost guaranteed risk of lowering the driver's attention levels. This will lull them into a sense of security that is only true for as long as no unexpected situation (or "edge case", as Tesla likes to call them) comes up that the AI can't handle.
As we've seen so far, these can vary from an emergency vehicle stopped on the side of the road to a semi-truck performing a U-turn, and who knows how many other scenarios the system hasn't been exposed to yet. In these situations, the driver would be required to resume full control of the vehicle in an instant. However, if their attention levels won't be at the same level as if they had been driving the whole time, the reaction might not be entirely satisfactory.
This has been argued over and over again, and yet Tesla apologists seem to ignore it. Yes, other manufacturers offer similar Level 2 systems, but the difference here is that they don't claim their systems to be any more than they are. GM isn't saying you can use Super Cruise to drive from one coast to another with no interventions, is it?
According to the study's findings, the duration for both on- and off-road glances featured an average difference of 0.3 seconds between manual driving and using Autopilot (AP). The difference favored manual driving in both cases (longer on-road and shorter off-road glances), but the one aspect where AP use stood out was in the proportion of off-road glances at the center-stack (infotainment system) that exceeded two seconds - here, we're looking at 4% in manual mode versus 22% for AP.
Tesla says AP relieves drivers from having to think about certain tasks and enables them to focus on the environment, essentially making them more attentive to their surroundings and thus, better, safer drivers. Well, we all know that's a load of crap, and it's not even Tesla's fault, just human nature. Give us the apparent choice between doing something we don't necessarily like and any form of entertainment, and we'll go for the latter.
The idea that a more relaxed driving experience makes for more focused drivers is also completely wrong, and you don't need any ADAS to notice it. All you need is a few hundred miles and a highway. You are much more likely to become bored and be taken by surprise by a car coming from behind than if the highway is extremely busy. An increased number of stimuli may make a driver tired more quickly, but it will also keep them engaged and on their toes. Freeing them of responsibility will do the exact opposite.
More and more companies are realizing Level 3 autonomy is tricky, not so much from a technical point of view, but rather a legal stance. No, actually it's not even about legal liability - Tesla is getting away with it right now just on the basis of a disclaimer. It's actually the moral side of things that gets in the way. Some people in the industry believe that Level 3 autonomy (vehicles capable of driving themselves in virtually all circumstances but require human supervision at all times) - something Tesla, or any other manufacturer for that matter, don't offer yet - should be skipped altogether until Level 4 is doable precisely because of these very conflicting signals it sends for the driver.
On the one hand, they're telling them to relax and enjoy the ride, but on the other, they're supposed to be ready to jump in at any moment as if they'd been driving the car the whole time, something everyone knows is not only unlikely, but virtually impossible.
There are plenty of ways in which this study can be labeled as irrelevant, and a lot of Tesla fans are already doing that. However, it shouldn't be interpreted as a dig at their favorite EV maker, but rather a warning to everyone that Level 3 is a dangerous game to play. Either you take full responsibility as a manufacturer for what your system is doing, or you stick to Level 2 and limit its functions to the highway. It may sound retrograde, but retrograde isn't that bad when the alternative is sharing the road with imperfect AIs and oblivious drivers.
Google News