Consumer Reports: Tesla Autopilot a 'Distant Second' To GM Super Cruise (arstechnica.com) 131
An anonymous reader quotes a report from Ars Technica: Cadillac Super Cruise has retained its title as the best driver assistance system on the market, Consumer Reports declared in a new ranking. Super Cruise also won CR's last ranking in 2018. While Super Cruise started out as a Cadillac-only feature, GM is planning to bring it to 22 vehicles by 2023. Tesla's Autopilot came in second place -- a "distant second" according to Consumer Reports. The group says it saw "minor improvements in lane keeping performance" from Tesla's offering since the system was last evaluated in 2018.
Those minor improvements were enough for Autopilot to get the top spot in the "lane keeping and performance" category of CR's report. CR ranked Autopilot 9/10 for performance, while Super Cruise scored 8/10. Tesla also got top marks for Autopilot's ease of use. But Cadillac got a much better score -- seven points versus three for Tesla -- for its driver monitoring system. Cadillac Super Cruise uses a driver-facing camera to verify that the driver has their eyes on the road. The system uses lights and sounds to aggressively alert the driver if they stop paying attention. By contrast, Tesla uses a steering wheel torque sensor to determine whether the driver is holding the wheel. As CR points out, having hands on the wheel "does not necessarily mean the driver is actually looking at the road ahead."
CR also gave Cadillac credit for limiting the use of Super Cruise to pre-mapped highways. GM has made detailed maps of thousands of miles of highway across the US. The maps help Super Cruise understand the road and allows the vehicle to alert drivers ahead of time if they need to take the wheel. Tesla got dinged for allowing the use of Autopilot in residential areas -- areas where drivers need to pay especially close attention to avoid hitting pedestrians or other obstacles. Finally, CR found that Super Cruise did the best job of safely bringing the vehicle to a stop if the driver became unresponsive -- for example because she fell asleep or had a heart attack.
Those minor improvements were enough for Autopilot to get the top spot in the "lane keeping and performance" category of CR's report. CR ranked Autopilot 9/10 for performance, while Super Cruise scored 8/10. Tesla also got top marks for Autopilot's ease of use. But Cadillac got a much better score -- seven points versus three for Tesla -- for its driver monitoring system. Cadillac Super Cruise uses a driver-facing camera to verify that the driver has their eyes on the road. The system uses lights and sounds to aggressively alert the driver if they stop paying attention. By contrast, Tesla uses a steering wheel torque sensor to determine whether the driver is holding the wheel. As CR points out, having hands on the wheel "does not necessarily mean the driver is actually looking at the road ahead."
CR also gave Cadillac credit for limiting the use of Super Cruise to pre-mapped highways. GM has made detailed maps of thousands of miles of highway across the US. The maps help Super Cruise understand the road and allows the vehicle to alert drivers ahead of time if they need to take the wheel. Tesla got dinged for allowing the use of Autopilot in residential areas -- areas where drivers need to pay especially close attention to avoid hitting pedestrians or other obstacles. Finally, CR found that Super Cruise did the best job of safely bringing the vehicle to a stop if the driver became unresponsive -- for example because she fell asleep or had a heart attack.
That's a feature? (Score:3, Insightful)
Cadillac Super Cruise uses a driver-facing camera to verify that the driver has their eyes on the road. The system uses lights and sounds to aggressively alert the driver if they stop paying attention. By contrast, Tesla uses a steering wheel torque sensor to determine whether the driver is holding the wheel. As CR points out, having hands on the wheel "does not necessarily mean the driver is actually looking at the road ahead."
I'm not sure most consumers would consider that a positive feature. Like my riding mower that I have to push and hold a button to use reverse while the blades are engaged. Totally pointless on multiple levels, yet the kind of thing I'm sure Consumer Reports would give a *better* score for.
When it comes to the true performance of the system - IE how well it maintains the lanes, which is truly the important part, Telsa actually beats out GM. Yet they scored worse overall. Maybe if your system does a better job at the task at hand, it's not as important to have a human monitoring it continuously?
Re:That's a feature? (Score:4, Insightful)
The people who died in autopilot crashes after they decided to to not monitor it and take a nap or watch netflix might disagree.
Like, if they were still alive I mean.
Re: (Score:2)
I don't feel bad for those people. You have to acknowledge several warnings beforehand. You know simple shit like you may wreck if you stop paying attention.
Re: (Score:2)
That 'nanny' will be less likely to drive your car straight into a barrier at 60mph, or the back of a fire truck.
Re: (Score:2)
Yes, but that is because the nanny is less likely to be driving in the first place.
Re: (Score:2)
Tesla doesn't care if you drive into a wall, GM does.
Wow. (Score:5, Insightful)
Seriously. You are not a good human.
Re: (Score:2)
Re: (Score:2)
Right. Tesla a "distant second" because the car is a nanny? Get. The. Fuck. Out. -a-. Here. Having driven with both systems, its not even close. Not by miles. The same fuckers who think the product should control the user are the same ass-hats who think government should put people in cages or rob them of their earnings for refusing to cover their faces with a mask. Fascists.
Knowing and adhering to the limits of the system is absolutely crucial. "Your freedom stops where mine begins" - and a car is a very dangerous item. You can easily hurt someone other than yourself with it if the system let itself be used in situations and areas it can't handle.
Re: (Score:2)
If the only risk was the car plowing into a bridge abutment and taking the willing participants out, I would agree. But if it plows into a school yard, that's a bunch of innocent 3rd parties that all suffer worse than the negligent "driver" snoozing in the back seat with a fishing weight tied to the wheel.
Re:That's a feature? (Score:5, Insightful)
Porsche 911: "Fantastic car, love the handling, great acceleration off the line. But it doesn't have a breathalyzer built in and some people have crashed while drunk. 6/10"
Ford Fiesta: "Horrible handling, slow, felt like driving a limp pool noodle with wheels. Did include breathalyzer that locks out drunk drivers. 8/10"
"In the end we had to go with the Ford Fiesta. While it's handling was lousy, it prevented me from driving while drunk. And since I can't be trusted to operate a motor vehicle legally, that's a good thing. The Ford Fiesta is clearly the superior automobile."
Re: (Score:2)
Re: (Score:2)
To be honest, with cadillacs it's probably important to have heart attack and narcolepsy monitoring.
Re: (Score:2)
It doesn't sound like they were reviewing the cars, only the driver assist system.
A proper analogy would be the Ford getting 8/10 on it's breathalyzer, and the Porsche not receiving a rating, since it doesn't have a brethalyzer.
Re:That's a feature? (Score:4, Interesting)
It doesn't sound like they were reviewing the cars, only the driver assist system.
A proper analogy would be the Ford getting 8/10 on it's breathalyzer, and the Porsche not receiving a rating, since it doesn't have a brethalyzer.
Not being able to drive on many roads that haven't been loaded into a database does not sound like it drives on roads better to me. If they are testing the driving feature then Testla wins. If they are testing the driver nanny feature, then Cadalack wins and Testla gets no rating, since it doesn't have a driver nanny feature.
Re: (Score:2)
The difference being that these driver aids are basically designed to lull the driver into a false sense of security by being as good as possible yet needing constant monitoring to be safe. It's well known that humans are terrible at monitoring things that work fine 99.9% of the time and only fail occasionally.
The ones with over-the-air updates like Tesla are even worse because the behaviour of the system changes with little warning. It might have been fine yesterday but today it will kill you.
Re:That's a feature? (Score:5, Interesting)
The people who died in autopilot crashes...
Yep, there were 38,800 deaths in the US. Oh wait, those were with human drivers not using Autopilot. My bad!
Source: National Safety Council -- https://www.nsc.org/road-safet... [nsc.org]
Re:That's a feature? (Score:4, Interesting)
You can also defeat eye tracking. You just need to have your phone / tablet forward, in the direction of the windscreen. Trying to stop idiots from idioting is a futile effort. Thankfully, the vast majority of people are not idiots, and the software is good enough to deal with idiots the vast majority of the time - which is why the rate of accidents and fatalities with AP on is so low.
CR's system is pretty ridiculous. AP was the best performing system in their testing. Period. And that's saying something given that Super Cruise only works on pre-mapped roads, so it's beating systems that are basically cheating on the test. If you use CR's methodology, my living room couch would be a better Driver Assist. Heck, let me write a Driver Assist program ("ReiPilot") for you right now and use CR's methodology on it.
bool Attempt_To_Activate_ReiPilot()
{
alert_user("This driver assist program cannot be run under any circumstances.");
return false;
}
How would CR rank it?
* Capability and Performance: 0/10 (does nothing)
* Ease of Use: 10/10 (couldn't be easier!)
* Clear When Safe To Use: 10/10 (will never engage!)
* Keeping the Driver Engaged: 10/10 (has to drive self!)
ReiPilot takes the top spot among driver assist systems according to CR's insane methodology.
The only reasonable safety measure (ignoring measures of user experience) is simple: (HowItReducesYourLikelyhoodOfAnAccident) x (HowOftenIsTheUserLikelyToHaveItEngaged). Indeed, in regards to that, the more annoying, restrictive, and onerous the system, the worse it should rank (unless the system is so bad in general as to increase your odds of an accident, wherein your top safety rating comes when it's active 0% of the time).
Re:That's a feature? (Score:5, Informative)
CR's criteria were bogus.
Keeping the car in the center of the lane should not be a simple criterion. What happens when you have an 18-wheeler in the lane on the right, occupying all of the lane and nothing on the left. You don't actually want your car in the center of the lane in this case.
What about poor road conditions? Can it deal with stretches of road where the lane markings are indistinct, or even non-existent?
Can it avoid potholes? How well does it recognize and deal with objects (such as a ladder, or shovel -- all things I have encountered) on the road?
How well does it respond to other vehicles that might collide with the vehicle?
Re: (Score:2)
Keeping the car in the center of the lane should not be a simple criterion. What happens when you have an 18-wheeler in the lane on the right, occupying all of the lane and nothing on the left. You don't actually want your car in the center of the lane in this case.
What about poor road conditions? Can it deal with stretches of road where the lane markings are indistinct, or even non-existent?
Remember that these are only driver aids. These are not self driving cars. If that situation comes up the driver is required to take over control of the vehicle.
That is true even of Tesla's so-called "full self driving", it's merely a level 2 assistance feature like cruise control.
Re: (Score:3)
Actually, Teslas are very good in these situations.
When you start with an assertion like that and don
Re: (Score:2)
It says very clearly that the driver must be paying attention at all times and ready to take the wheel even before the car warns them to. Particularly with the full self driving beta it warns that the car may do unexpected and extremely dangerous things.
It's very much level 2.
Re: (Score:2)
Arguing about how to classify the system isn't useful. Discussing what it actually does and does not do is useful.
Re: (Score:2)
At least if you have your phone held up high enough to fool eye tracking the road will be in your field of view.
Super Cruise also has better driver alerting. With Tesla you get some beeps, with Super Cruise there is a huge green/red light on the steering wheel, audio feedback and eventually it starts jolting you with the brakes. Nissan ProPilot does that too, seems pretty likely to wake up a sleeping driver.
Note that officially Autopilot should only be used on certain roads, it's not supposed to work everyw
Re: (Score:3)
Hey, guess what? Tesla Autopilot will do the following things if you aren't paying attention and don't move the steering wheel or turn signal stalk when prompted:
1. prompt you to nudge the wheel
2. blue light flashes on the display to get your attention
3. audio signal telling you to take over
4. car actually pulls over and stops, disabling your ability to enable autopilot until you put the car in park, lock the doors, and walk away for some interval of time.
So basically they all do what you're saying, and yo
Re: (Score:2)
Oh, does it pull over now? And you have to lock the doors?
Last I checked it just stops. In the middle of the road. And you can reset it by going into park and then back into drive. Good that they improved it, must have had too many people dozing off and then trying to continue their journey.
To be fair most of the others are just as bad. At least with some jolts from hard braking there is a better chance you will wake up. Nissan does that, and turns the AC on max to blow cold air in your face. Not sure about
Re: (Score:2)
The AC chilling is pretty clever. Yanking the seat belt would be interesting, but I'm not sure what mechanism it would have for doing that, unless it's something specific they engineered to do that, or it has those shitty automatic seat belts from the early 90s that are supposed to run along the roofline of the door and always break - most seat belt systems have an inertial lock that engages at some particular negative-g deceleration, so perhaps it mashing the brake on you engages that?
It's interesting to
Re: (Score:2)
The seatbelt has a pre tensioning system. When the front radar sees that the car is about to crash it tightens the belt so you don't get thrown forward into it. Reduces injuries.
Re: (Score:3)
Oh, so the people that disregarded every warning and agreement they were given in order to use that feature, would all of a sudden care about the exact thing they were warned about, and agreed to all of a sudden?
Remember when personal responsibility was a thing? Do you fault Ford when someone drives into someone else with cruise control engaged? Or do you think that maybe the operator of the vehicle who wasn't paying attention might have something to do with it?
Re: (Score:2)
The people who died in autopilot crashes after they decided to to not monitor it and take a nap or watch netflix might disagree.
Like, if they were still alive I mean.
What separates me from accidents like the above is money. Having a fancy car does not make me a better person. If I need to use a car to impress someone, I will rent it for a day. My day-to-day car is 10 years old, with only 80,000 km driven.
Re: (Score:2)
Re: (Score:2)
Re:That's a feature? (Score:5, Funny)
They also give Super Cruise a better score because it works in fewer places and is therefore safer.
Apparently, it would get an even better score if it worked nowhere, with no risks of any accidents at all.
Re: (Score:2)
Re: (Score:2)
Re: (Score:3)
Nobody takes a nap with old-fashioned cruise control enabled, because that wouldn't come close to working. I would think that the better a driver-assist program stays in the lane, the more likely some drivers will be reading their email, posting in Slashdot, whatever.
Given that better lane following means more drivers not paying attention, that creates a very important threshold -
Is the driver assist system so good that it's no longer a driver assist, but full autonomy, so good that the driver doesn't NEED
Re:That's a feature? (Score:4, Funny)
Nobody takes a nap with old-fashioned cruise control enabled, because that wouldn't come close to working.
Talking about self-driving cars reminds me of that old joke about how my grandpa died in his sleep.
When I die, I hope it's in my sleep, like my grandpa. Not screaming, like the passengers in his car.
Re: (Score:2)
When I die, I hope it's in my sleep, like my grandpa. Not screaming, like the passengers in his car.
Emo Philips. One of the comedy greats.
Re: (Score:2)
Re: (Score:2)
Beg to differ. Especially out on the prairie where steering isn't all that engaging an activity, people take naps with old-fashioned cruise control enabled. You're right though, it doesn't work. A friend of mine tried it actually. The rehab worked well. You can't even tell today.
Re: (Score:2)
I don't know but to me, a camera that is checking if I am paying attention is much better than a wheel torque sensor.
In fact, having my car warn me in case I am dozing off is a feature I want as a driver. Having a self-steering car require me to put the hands on the wheel feels more like nagging.
I mean, if I am paying attention, grabbing the wheel takes only a fraction of a second, but having the hands on the wheel without paying attention is pointless.
Another useful "annoying" feature of Supercruise, accor
Re: (Score:2)
"I mean, if I am paying attention, grabbing the wheel takes only a fraction of a second, but having the hands on the wheel without paying attention is pointless."
A fraction of a second is all it takes to make the difference between life and death for you, your passengers, or the adults and children you might hit.
Re: (Score:2)
That's a Feature? Part 2 (Score:2)
yet the kind of thing I'm sure Consumer Reports would give a *better* score for.
Just to follow on your original thought that what consumer reports considers a feature seems pretty messed up, there is also this:
CR also gave Cadillac credit for limiting the use of Super Cruise to pre-mapped highways.
How does that not DOCK points? What good is a system that is limited to a handful of roads?
GM, I award you no points and may God have mercy on your soul.
Re: (Score:2)
It's worse than that. SuperCruise can detect construction zones and refuse to work in them.
In 2017 SuperCruise performed much better than AutoPilot where it worked. That isn't the case anymore. To me that is the big story.
Thinking same thing (Score:2)
In 2017 SuperCruise performed much better than AutoPilot where it worked. That isn't the case anymore. To me that is the big story.
I was thinking the same thing some time after I posted!
Even on roads where SuperCruise has heavily mapped the environment, Tesla STILL performs better without any specific knowledge of the location!
It seems like that indicates Tesla has a massive lead, unlike being behind as Consumer Reports seems to indicate at first glance.
Re: That's a feature? (Score:2)
Telsa actually beats out GM
Of course they do. I had a piece-of-shit '85 Audi wagon with far-better-functioning cruise control than my 2015 Suburban has.
Tesla is electronics.
Re: (Score:2)
"Telsa actually beats out GM. Yet they scored worse overall. "
It's the usual game. They use the latest 2021 GM software and a 2018 Tesla to test.
It's not the first time they cheat that way.
Re: (Score:3)
... and the 2018 Tesla is still far better in any given situation. And works in far more places. And is a system that people actually use on a daily basis. And will be further improved by software updates that are coming before the end of the year (allegedly) making it even better than the existing SuperCruise. And will receive the one thing that Consumer Reports says makes SuperCruise better - driver attention tracking via camera - Model 3 and Model Y have an interior camera specifically for this, and
What the headline giveth the summary taketh away (Score:3, Informative)
Re: (Score:2)
The only way that a "refusal to work in unmapped areas" would be superior is if it was a true level 3 system where in predefined areas you could actually shift your attention to something else and watch Netflix or read a book until you approached a disengagement zone and needed sufficient preparation time to resume driving.
I think Tesla should map that and if every driver in that lane at that GPS waypoint disengage, it should give you a warning that it's about to encounter a known bug.
I suspect whatever aut
Re:What the headline giveth the summary taketh awa (Score:5, Insightful)
It's a safer system, in that regard it's a better system.
Of course Tesla drivers don't want a safe system, they want a system which tells them "watch the road, nudge nudge wink wink".
Re: (Score:2)
The fact that this is a viable business is pretty scary.
https://www.autopilotbuddy.com... [autopilotbuddy.com]
Anyone who bought one of those is an idiot and a menace to other road users. Go ahead and Darwin yourself but don't take my and my family with you.
Re: (Score:2)
Different Goals (Score:5, Interesting)
I think the difference here is that GM and Tesla have different goals. Tesla is aiming for the driver not needed to be there at all. On any road, under any conditions. They're not there yet, but that's the goal.
GM are aiming for a driver assist feature and are absolutely relying on the fact that there is a driver ready to take over if they system gets confused.
Change the goal of the test, and the outcome changes... whowodathunkit!
Re: (Score:3, Insightful)
Re: (Score:3)
GM has been in the vehicle business for a long, long time.
Tesla is one of those Silicon Valley 'break things' experiments.
This is true... but it doesn't mean GM is right and Tesla is wrong. Nor does it mean the reverse. This is a new area and the expectations are still being defined, so experience isn't the only relevant factor.
Given the outcomes of the AutoPilot crashes and lawsuits so far, it seems like Tesla's approach is workable, and probably better. Tesla may arguably even be overcautious from a legal and liability perspecitve, given the lack of consequences Uber faced when killing a woman in circumstances that should
Re: (Score:2)
Re: (Score:2)
It's really sad that Tesla's solution is legal, because that makes Tesla's customers look like absolute fools.
How so?
Re: (Score:2)
Re: (Score:2)
When you use autopilot, you are risking diverting your eyes for a second and having to answer for all its failures.
AutoPilot doesn't require you to take your eyes off the road; no more than using the speedometer, etc. does.
Re: (Score:2)
Re: (Score:2)
But people do.
So people who do are fools... as are people who don't have AutoPilot and take their eyes off the road. I fail to see how this makes Tesla customers fools in general.
Re: (Score:2)
Re: (Score:2)
You're pretending like people have control over this. Some do some don't. Regardless, you are taking a chance because you don't really know if and when you will do it.
Of course people have control over it. You're being ridiculous.
Re: Different Goals (Score:2)
Re: (Score:2, Interesting)
How many headlines have you read about a GM vehicle crashing into something at freeway speed while the driver was asleep at the wheel? How many have you read about for Tesla? I'm guessing you've read about a lot more for Tesla than you have for GM. From a public safety standpoint GM wins this round.
Both GM and Tesla have fairly good systems, the difference is in the design philosophy. Tesla wants to let you be lazy. GM wants to be a drivers assistant.
Tesla got greedy and reached for too much when their syst
Re: (Score:2)
How many headlines have you read about a GM vehicle crashing into something at freeway speed while the driver was asleep at the wheel? How many have you read about for Tesla?
But is this because there are more crashes per vehicle mile, or because Tesla is the cool new company and gets the press? Or is it because GM crashes are so common they aren't newsworthy, or at least not worthy of being reported nationwide? Keep in mind that "reported by the news" means "not a common occurrence", by definition.
There's a serious risk of falling victim to availability bias [wikipedia.org] here.
I'm not saying your conclusion is wrong. I don't know if it is or not. I'm just pointing out that your method fo
Re: (Score:2)
All things considered, I'm inclined to concede to your point.
I certainly don't know the rates of accidents between the two. I do know that Tesla's generally do quite well in accidents. Unfortunately you don't benefit from this in terms of better insurance rates due to their vehicles almost always being totaled. As a result their insurance rates are through the roof due to their corporate policies of refusing to work with third parties for repairs. Both companies do some things well and some not well in both
Re: (Score:2)
I was looking at things from the perspective of risk, which is much more in my wheelhouse. Best practice is that the risk to human life is always supposed to be the highest consideration.
I agree, but this is a little tricky with self-driving systems, because the baseline risk (human driver) is very bad. So if the automated system is better than the baseline, then it reduces risk even if it's not great.
GM has taken that for their advanced cruise control features and Tesla has not. GM's feature won't work without the backup feature (the driver), whereas the Tesla feature will work without the backup feature.
This statement presumes that the backup feature is better than the system. That's not a given in all circumstances. In some situations the driver is better, because the driver has broader awareness. In some situations the system is better because it has faster reactions, constant 360-degree vi
Re: (Score:2)
Re: (Score:2)
I think it boils down to whether or not you have a backup system or not.
I think you have to consider the quality of the backup system as well.
Other thought is that while humans are generally not very good drivers, I'm not very confident in AI driving in winter conditions either. Can AI do winter driving conditions in a safe manner?
At present, I don't think so. Waymo is also testing in Michigan and has been for a few years. Odds are they're the furthest along, but they haven't brought a product to market. Tesla's system refuses to operate in bad weather of any sort, and I'm sure SuperCruise does, too.
Re: (Score:3)
GM are aiming for a Level 5 fully self driving vehicle too. The difference is that Tesla is beta testing their software on the public, with untrained drivers.
It's quite clever really. No need to get permits or pay safety drivers, just let customers to the beta testing for you, using their own insurance. If they die in a wreck well it's their own fault, after all it does say beta!
Needs weighting to the scoring (Score:4, Insightful)
This is like a consumer reports review going like this:
TV Product A:
Picture Quality: 10/10
Remote Layout: 2/10
Sound Quality: 3/10
Total Score: 15/30 F
TV Product B:
Picture Quality: 2/10
Remote Layout: 10/10
Sound Quality: 10/10
Total Score: 22/30 B+
Ok, yes TV A scored way lower than TV B. But when you are reviewing televisions, sound quality isn't really all that important. And you can replace the sound and remote with much better solutions like stand alone surround sound speakers and universal remotes.
This is the problem inevitably with using a spreadsheet based rating that's "fair" and "empirical". How you weight the scores inevitably drives who wins.
Do you stack the categories to favor autopilot:
Quality of Lane Changes: Autopilot 8, Super Cruise 0
Quality of Clover leafs: Autopilot 6, Super Cruise 0
Quality of Exits: Autopilot 9, Super Cruise 0
Quality of lane position: Autopilot 5, Super Cruise 9
Total: Autopilot 28, Super Cruise 9
Or do you rate it based on the amount of time the driver is being assisted by the feature.
Quality of Lane Changes: Autopilot 20, Super Cruise 0
Quality of Clover leafs: Autopilot 2, Super Cruise 0
Quality of Exits: Autopilot 3, Super Cruise 0
Quality of lane position: Autopilot 50, Super Cruise 90
Total: Autopilot 75, Super Cruise 90
Consumer Reports shouldn't have gone off of a scoring system and then added up esoteric features to try and appear "fair and balanced" because that doesn't help consumers. They should have done a purely subjective test where they asked how much each consumer reviewer to just give a subjective quality rating. How pleasant was each technology?
Re: Needs weighting to the scoring (Score:2)
This reminds me of a very relevant xkcd comic that illustrates this problem perfectly:
https://imgs.xkcd.com/comics/t... [xkcd.com]
It's for a hypothetical app to warn you about approaching tornadoes with 4 out of 5 stars... several 5-star ratings for things like having a clean UI, not crashing, etc... and a single 1-star rating because it failed to warn someone about an actual tornado.
Re:Needs weighting to the scoring (Score:5, Interesting)
In addition, the part of the test that scores actual self driving performance includes only a few basic scenarios. The Tesla scored top marks there, but if the test had included more complex scenarios involving serious driver error from other road users, it would have scored even better compared to Mercedes and BMW, which for instance (IIRC) make no attempt to avoid being sideswiped.
But the most outrageous reason given for the poor score in the Driver Engagement category is that it includes consumer information, for which Tesla got penalized heavily because of the "misleading nature of the term Autopilot". So: you built a very good driver assist system but we don't like the name.
Re: (Score:2)
Those are all valid criticisms. Super Cruise has a great UI, there is a huge light on the steering wheel that is either green for engaged or red to warn when you must take over, or off when the system is turned off. You can't miss it, and there are audio cues as well.
Mode confusion, as it's called, is a not uncommon cause of accidents. Being in reverse when you want to go forward, having normal cruise control engaged when you thought you were on manual etc.
I like how the Honda system works for take-overs. I
Re: (Score:2)
Clearly it depends what you want. If you want ultra reliable and safe cruising on highways the Super Cruise is the best system.
If you want beta quality handling of various other scenarios then Autopilot may be your thing.
Obviously Consumer Reports ranks "not dying" pretty highly in their overall scoring system, but your preferences may vary.
Re: (Score:2)
Obsolete System (Score:2)
Well, they reported on Tesla's system just after it became obsolete. They're starting to roll out a complete re-write with a massive increase in functionality. Of course, it appears that Consumer Reports is already ignoring a number of features like lane changes and taking exits. And if you're one of the lucky few to already have the new software, it does stops and turns on local streets.
Interesting (Score:2, Insightful)
I'd never heard of GMs offereing before and I was a little gleeful to read the headline and imagine a poke in the eye to the Tesla cock-gobblers.
But reading further, it sounds to me like Tesla has a better or more capable autopilot than GM does. But the limitations of the GM system seem to appeal more to Luddites as Consumer Reports.
'GM less scary, so better'?
I can't wait to put my life in GM's hands! (Score:3)
Yes, I for one cannot wait to put my life in GM's hands. I don't care about their abysmal history of safety flaws, cost cutting, lemons and lying to the public...
What is the purpose of these systems? (Score:4, Informative)
I have driven with the Audi and Tesla "driver assist" systems this year and they both left me wondering why they were there.
Maybe it's Toronto/Ontario roads, but neither one was what I would call competent in lane keeping - both tried to steer the car out of the lane (many of the times I felt it was clearly marked) and they weren't well integrated with the GPS systems (shouldn't they be guiding you into the lanes needed for turning?) The only feature I found really useful is the radar assisted cruise control (because sometimes you look away from the car in front of you) although there were a number of times when I could see that there was a problem and I would naturally brake earlier than the system to keep things from being exciting for the passengers.
It's good that the CR rating includes "ease of use" as going through the options can be a bit daunting and, when you get right down to it, you should probably just be going with a basic default and not give a "driver" the ability to tune the system for their style of driving.
Personally, either the car drives itself or a person does, there shouldn't be anything in the middle which is why I'm asking what is the purpose of the system?
Re:What is the purpose of these systems? (Score:4, Informative)
I had that happen once in my Tesla. It didn't take the correct when there was an exit on a left bend in the road. That happened two years ago and it hasn't happened since. Autopilot on Tesla is very good, even handling short stretches of road with poor or non-existent lane markers.
That's what NOA does: it will change lane for you, take exit ramps and make other maneuvers to follow the navigation directions.
Much reduced stress when driving down a freeway.
I suspect you haven't driven a Tesla recently.
Volvo in the UK was pretty good (Score:2)
At lane following. But it did not remind me to take the (small) freeway exit!
Re: (Score:3)
Tesla's self driving is in two separate parts.
Lane assist and adaptive cruise control are free. This mode also includes a (fairly anemic) summon feature and an option to change lanes on the highway to avoid traffic. (It's called Mad Max mode.)
Full Self Driving (FSD) is what costs extra $$$ ($8k now, $10k starting next week, I think).
In full self driving mode, the car will currently take exit ramps to get off the highway if that's what the GPS tells it to do. Right now, that's the only extra thing FSD does
Re: (Score:2)
There are some minor errors in your description, at least as compared to how my car (2020 Model S, w/FSD) works.
Lane assist and adaptive cruise control are free. This mode also includes a (fairly anemic) summon feature and an option to change lanes on the highway to avoid traffic. (It's called Mad Max mode.)
Automatic lane changing is part of FSD [tesla.com]. Fully-automatic lane changing isn't enabled unless you activate Navigate on Autopilot (NOA). There's a setting that allows you to specify how aggressive it should be -- meaning how much difference between the speed of the car you're following and your target speed will trigger the car to look for a better lane . "Mad Max" is the most aggressive setting. I k
Re: (Score:2)
I test drove a 2020 Model X in August of this year from the dealer for two days.
Don't know what to say - the off ramps came up on the centre console but the vehicle didn't do anything other than notify me that they were coming up (which I would get in any other GPS system).
Re: (Score:2)
Either the car didn't have or you didn't enable the NoA features.
Re: (Score:2)
there shouldn't be anything in the middle which is why I'm asking what is the purpose of the system?
Absolutely absurd. There are many ways we take control away from the driver in the name of safety and all of them have two things in common:
1) On their introduction people thought they were better, thought they were a waste of time and insisted they wouldn't be safe and the driver should be in charge.
2) People who assumed 1 were proven wrong over and over again.
You aren't in control of your car without a lot of systems in the middle. Unless your car was built in the 60s in which case we kindly ask you to ge
Clear Perspective (Score:4, Funny)
TL;DR Tesla was the clear winner in Capabilities and Performance along with Ease of Use. However, Tesla severely lagged in the nag features.
Re: (Score:3)
Re: (Score:2)
or to put it another way, Tesla performed better in function but sucked at safety.
You're making an unjustified leap there, assuming that the nag features actually improve safety. Maybe they do, maybe they don't. Data is required to make that call. Also, it may be that SuperCruise is safer than AutoPilot when active, but that both are safer than the unassisted driver (note that this is possible even if the driver assist system occasionally causes accidents, as long as it prevents more than it causes). If that's the case, then SuperCruise's more limited scope of usability actually reduces
Re: (Score:2)
Which is a huge change from 2017 where SuperCruise was flat out better in terms of performance in the limited areas that it worked. Autopilot has come a long ways. And they were comparing with Auto Pilot not Enhanced Auto Pilot or Full Self Driving.
Tesla's crippleware is more featured than GMs $5k option
Consumer Reports stopped mattering years ago (Score:5, Interesting)
Consumer Reports claims to represent consumers, but there are numerous examples where they've failed to do that.
One of the most notable that sticks out in my mind was a few years back when they encountered a battery issue while testing the MacBook. They were barely getting half or a third of the advertised life, even though previous models had matched the advertised battery life. Rather than questioning whether the sudden and unexpectedchange that no one else seemed to be reporting might be indicative of a problem with their "real world" testing, or at the very least point to the presence of a bug that affected their specific methodology, they went to print with a sensationalist headline that slammed the MacBook for its poor battery life, saying that it didn't get anywhere close to the advertised life in "real world" testing.
But then no one could reproduce the problem for days. The laptop was getting its advertised life in actual real world use. Consumer Reports was the only one having battery issues, and they were even able to reproduce the issue on multiple laptops, even though no one else was seeing it. Rather than recognizing that there had to be something particularly to their methodology that was leading to the issue, they defended their methodology as reflective of real world usage and slammed the MacBook in a second, sensationalist headline.
Eventually it came out that their "real world" testing was nothing of the sort: they had turned on Safari's Developer mode (which was a hidden feature back then), enabled a seldom-used setting that was disabled by default, set it to a non-default option, and apparently were the first to discover a performance bug with that setting for that feature that had been introduced in the immediately prior software update. Any typical developer with those same, unusual settings would lose maybe a few seconds of battery life in a day, but Consumer Reports' automated workflow hit the bad path tens of thousands of times in quick succession, hence their woes.
That could have been the end of the story, with Consumer Reports issuing a followup report to explain how they had worked with Apple to isolate the bug and would be modifying their methodology to more closely resemble real world use cases. Instead, Consumer Reports wrote another hit piece, defending their methodology and slamming the laptop. This, despite the problem being quickly identified once Consumer Reports divulged their unusual testing methodology to Apple, despite the bug being isolated to the exact version of Safari they were using, despite a fix being promised by the next week (which was delivered), despite a beta version that was suitable for testing being available with the fix, and despite the issue being specific to the automated use of a seldom-used, developer mode-only feature, none of which was typical of real world use.
And then, once the fix was released, Consumer Reports went silent. Whereas they had been issuing multiple reports every week to drag out the issue and draw attention to themselves, once the problem was addressed they dragged their heels for several weeks before updating the score to reflect the reality that everyone else already knew.
I had issues with them prior to that event, but they lost all credibility in my book with that series of stunts. It was self-serving, clickbait garbage.
But even without that one situation, I've found their reports rather lacking in detailed analysis when compared to the Wirecutter. Wirecutter explains what you should be looking for in the product category, the rough rubric they use for grading, who's doing the grading, why they're qualified, what the top picks are, why those were the picks, a list of alternatives with explanations for why they weren't picked, and a list of competing products/features on the horizon that are likely to change things up. It's also clear when updates are made, what the most recent updates were, and there's a comments section at the bottom with plenty of users calling out question
Alternate Headline: (Score:2)
GM Supercruise superior to Autopilot due to all the many ways it refuses to work.
It's a great system if you stay on the Interstate (Score:2)
Otherwise, it's not. Sorry, but that's not full self driving. Unless you can drive ANYWHERE the system doesn't work. As usual, Consumer Reports testing cars is what is deficient here.
There was a cartoon is C&D or R&T several years ago. On a distant hill several men in glasses and white coats carrying clipboards were making notes as one car after another as pushed off a cliff. In the foreground one guy says to the other, "Oh, that's just Consumer Reports testing cars again."
Consumer Reports "distant" from reality (Score:2)
CR seem to focus on arbitrary things that sometimes don't make sense. We don't need them anymore since the public can publish less biased reviews based on real user experience.
Re: (Score:2)
Do you think CR's approach of giving much higher scores specifically for working less often is a good one?
Re: (Score:2)
It's the opposite. If safety is a high criterion, you want autopilot to be enabled as much as possible to actually see how safe it is in the real world.
Limiting the systems to only very safe situations doesn't really tell you anything, and is totally at odds with how people would use the systems in the real world.