Is the current NoA helpful or harmful?

[Nocturnal]

Harmful? That's ridiculous. If you are using it the way it's intended then it does it's thing while you are still there watching it. It's only dangerous if you take a nap or watch your phone etc.

My only complaint is that sometimes it doesn't want to get over even though it has space.

 

[jdcollins5]

After CR’s report giving Cadillac’s AP a better score than Tesla’s because Cadillac limits its use to pre-defined highways, it is no surprise at this report. It is evident that CR expects close to perfection and limited use before an AP product is released. This is contrary to Tesla’s approach.

I think NoA does a good job in light to moderate traffic. It needs improvement in the heavier traffic situations. With Tesla’s fleet experience this will come quickly.

 

[Nocturnal]

After CR’s report giving Cadillac’s AP a better score than Tesla’s because Cadillac limits its use to pre-defined highways, it is no surprise at this report. It is evident that CR expects close to perfection and limited use before an AP product is released. This is contrary to Tesla’s approach.

I think NoA does a good job in light to moderate traffic. It needs improvement in the heavier traffic situations. With Tesla’s fleet experience this will come quickly.

It's pointless if a system works better but you can never actually use it.

 

[jfrieden]

Respectfully, I haven't been able to use NOA successfully in the DC Area on the major highways. I think it probably does a great job in some areas, but it is harmful and puts me in harms way in this area. I appreciate everything regular AutoPilot can do and does, but NOA is not ready for primetime (everywhere at least).

 

[Big Earl]

It’s good that you realize its current limitations. It is essentially useless in the DC area in its current form. That will certainly improve with time.

 

[GolanB]

You can see where NoA is headed and in the right situation it’s magic, but I think they gave the option for it being on automatically at every Autopilot engagement too soon. It needed more time in the oven.

My experience has led me to drive without NoA by default when using AP/AS. Occasionally I will reenable it to track its progress.

My primary impression, like yours, is that it's not quite ready. I have often felt that auditing correctly adds a burden which is more taxing than driving alone - with some exceptions like when driving in back-to-back traffic conditions. A primary contributing reason for this is because highway driving has become almost a subconscious extension of myself — whereas auditing requires more rigorous analysis, including being hyper-conscious and vigilant.

There is also a lingering fear that something could go wrong that I don't have when I'm merely driving on my own. That fear is not without merit, as I regularly need to make corrections, but also because I've had to intervene in a smaller number of cases that would likely have resulted in a collision.

To make matters more complicated, the freeway system here in the NY/NJ area is one of the most complex, is often under construction, and supports some of the most aggressive and unpredictable drivers I've encountered in my 30 years of driving. I would characterize myself as a Tesla enthusiast, and regularly make use of its automation. At the same time, I would not want my teenage son, who is learning to drive now to use it.

 

[mtndrew1]

My experience has led me to drive without NoA by default when using AP/AS. Occasionally I will reenable it to track its progress.

My primary impression, like yours, is that it's not quite ready. I have often felt that auditing correctly adds a burden which is more taxing than driving alone - with some exceptions like when driving in back-to-back traffic conditions. A primary contributing reason for this is because highway driving has become almost a subconscious extension of myself — whereas auditing requires more rigorous analysis, including being hyper-conscious and vigilant.

There is also a lingering fear that something could go wrong that I don't have when I'm merely driving on my own. That fear is not without merit, as I regularly need to make corrections, but also because I've had to intervene in a smaller number of cases that would likely have resulted in a collision.

To make matters more complicated, the freeway system here in the NY/NJ area is one of the most complex, is often under construction, and supports some of the most aggressive and unpredictable drivers I've encountered in my 30 years of driving. I would characterize myself as a Tesla enthusiast, and regularly make use of its automation. At the same time, I would not want my teenage son, who is learning to drive now to use it.

Agreed. I find regular Autopilot to dramatically reduce my fatigue and catch unsafe situations that I might miss, making driving easier and more relaxing.

NoA has the opposite effect for me right now; there’s a constant overhead of monitoring it that wears me out. It can be remarkable and fun to watch but it’s mostly not useful yet. If used properly I don’t think it’s dangerous though.

 

[ucmndd]

Autopilot is great and I use it every single day as a driver aid / stress reducer.

NoA requires significantly more oversight and babysitting right now - dare I say more than normal manual driving. As such it’s not useful to me outside of the occasional parlor trick. CR is absolutely right IMO.

 

[Enginerd]

So at this point, there probably isn't much statistical relevance in comparing.

I think the best thing we can do is to answer the question based on our own personal experience with NoA rather than rely on any stats.

I'm taking on the statistical performance part. I'll be providing regular performance updates based on about 600 miles per week.
Navigate On Autopilot: automatic lane change results

It's all about context. Does it require intervention now and then? Yes, Absolutely.

Does it make my commute in traffic safer because it is more attentive than me? Yes, Absolutely.

I'm in this to watch the improvement over time, and experience this once-in-a-lifetime transition to automated driving. As an engineer (unrelated to auto or self driving), I find this whole thing fascinating.

This. I enjoy being a part of the sausage as it's being made.

 

[Spalding]

On this web site most will disagree with Consumer Reports' hand-wringing conclusions because we have direct experience with the product and know when NOA is beneficial and when it isn't. The article reads like they expect NOA to be full Level 5 autonomy, bootstrapping the sometimes hyperbolic descriptions given to it by our favorite CEO. It's not Level 5, of course, and it isn't marketed that way.

Regardless, what we say to each other on this website is basically meaningless. Most people will read the CR article, and the million web sites that will link to it, and conclude that all the naysayers are right. Useful comments like the ones here need to be posted on the reply sections of the CR web site and others.

 

[dfwatt]

I ran across this article about NoA today. Curious to know what others think about this.
Tesla's Navigate on Autopilot Requires Significant Driver Intervention — Consumer Reports

It's another questionable CR report. They clearly don't know what to make of the model 3, and it was shortly after the model 3 came back as the highest rated ever car by owners that they pulled their recommendation of it. This was due to open quotes reliability close quotes issues but when you drill down on that, it's things like cracked glass, poor paint and screen reboots - hardly the stuff of reliability nightmares or anything that might produce stranded drivers. When you ask CR for their data sets to look at the incidence of problems, they refuse to to hand it over for independent review. Bottom line, I've lost all respect for CR.

As Kirby64 indicated, there are for sure problems that we're all aware of with autopilot, but on balance, it makes a huge difference in terms of reducing driver fatigue which is frankly one of the largest risk factors for highway accidents. Once again CR fails to mention this, more evidence that they've lost the forest for the trees.

 

[afadeev]

Human + Machine > Human or Machine.

Not if it's an erratic and a poorly programmed machine!
Automation needs to deliver certain (high) success rate where supervision becomes a relief, and not a distraction.
AP is getting there. NOA is just at the start of the journey.

Agreed. I find regular Autopilot to dramatically reduce my fatigue and catch unsafe situations that I might miss, making driving easier and more relaxing.

NoA has the opposite effect for me right now; there’s a constant overhead of monitoring it that wears me out. It can be remarkable and fun to watch but it’s mostly not useful yet. If used properly I don’t think it’s dangerous though.

This.

On this web site most will disagree with Consumer Reports' hand-wringing conclusions because we have direct experience with the product and know when NOA is beneficial and when it isn't. [...] Regardless, what we say to each other on this website is basically meaningless. Most people will read the CR article, and the million web sites that will link to it, and conclude that all the naysayers are right.

1. Some folks will disagree with everything that is less than 110% veneration of Tesla. These folks are the vocal minority on this forum, but they are highly vocal. They loved CR when it gave MS 100% review score, and now gladly tear into it when it found and documented issues with NoA that we all know are there.
2. Some will take a pragmatic approach to the new features that Tesla is rolling out, and explore them with a healthy grain of salt. Both groups own Tesla.
3. The third, and the largest group, has zero or little first-hand experience with Tesla's, will read CR, and accept its conclusions as gospel. They are the majority of the US population, and for that reason, CR's reviews are extremely powerful. Usually, they are also objective, but somewhat shallow.

Like it or not, this forum consists primarily of groups 1 & 2.
Group 3 is the one Tesla needs the most to increase sales and return to profitability.
As of today, that job has just become a little harder.

a

 

[Eno Deb]

On this web site most will disagree with Consumer Reports' hand-wringing conclusions because we have direct experience with the product and know when NOA is beneficial and when it isn't.

What makes you think the CR guys don't? They have had a permanent Model S and Model 3 for a couple of years now and have continuously reported their findings, including e.g. uncovering a problem with the Model 3 braking system. They have also reviewed NoA when it first came out.

 

[chickensworth]

I use EAP and NOA everyday... and Every single day, i need to intervene multiple times because it wants to do something stupid.

So... yes, NOA requires significant intervention... BUT the system is currently designed to require human supervision... and if used correctly can be a significant driver aide... as an extra set of eyes, as a way to reduce fatigue, as a way to carry 90% of the stress of driving.

i think the problem is Elon's overselling and overly optimistic timeline of full autonomy... it creates an expectation that the system is far better and more advanced than it really is. I joke, but my car can't go straight into my garage 3/5 times...so i certainly don' t trust my car to drive me and my family without keeping a close eye on what it's doing.

 

[notAnExpert]

At the current level of maturity of NoA, the thing that rings true for me is the article's concluding paragraph, quoted below.

Monitoring the system is much harder than just changing lanes yourself. Using the system is like monitoring a kid behind the wheel for the very first time. As any parent knows, it’s far more convenient and less stressful to simply drive yourself.​

 

[chronopc]

Not if it's an erratic and a poorly programmed machine!
Automation needs to deliver certain (high) success rate where supervision becomes a relief, and not a distraction.
AP is getting there. NOA is just at the start of the journey.

Yeah...because human drivers are never erratic or distracted.

 

[astrowunder]

Autopilot is awesome. NOA is an early beta. I don't mind using it occasionally to test it out, but I can't stand the way it works right now- much more stressful for me than not using it. I think CR is spot on except they should have mentioned it's in beta and the 'standard' autopilot is far better than anything else currently available.

 

[Magellan55]

For context, CR serves mostly conservative (as in not risk takers) people who aren't car or tech enthusiasts, who just want a reliable appliance to get them safely from A to B. So, yeah for their readership NoA is not ready, and certainly doesn't meet the marketing / tweeting hype. Agree CR should have stated it's Beta (though their readers have no clue what that is) and added that non-NoA AP is great.

I don't use it much, but bought it to watch history being made (and be part of it), like some others on the thread. Watching it slowly navigate a clover leaf last week, then be hesitant to merge onto the next highway, I too felt like it was my 15 year old behind the wheel first time on a highway. And my wife was definitely not impressed with it. Absolutely takes more effort than to just drive myself, but I engage it when i'm alone and can spare the mental cycles to help "teach" the neural network.

The only thing I'm worried about in terms of investing in FSD is whether we have good enough sensors for it. We have 8 cameras but they are mono-vision, and my car has a lot of trouble judging where other cars are when first merging onto a highway. I know the neural network makes use of relative motion to make up for lack of stereo vision, but that doesn't work when the other cars are moving too so there's not enough relative movement for it to do it's magic. I'm predicting there will be an upgrade to the cameras, or maybe additional ultrasonic sensors like the "cross traffic alert" sensors many cars have in the rear quarter panels now, added to new cars at some point now that there's a fast enough CPU to handle it. The question I have, is will us early adopters get them too?

 

[dfwatt]

After CR’s report giving Cadillac’s AP a better score than Tesla’s because Cadillac limits its use to pre-defined highways, it is no surprise at this report. It is evident that CR expects close to perfection and limited use before an AP product is released. This is contrary to Tesla’s approach.

I think NoA does a good job in light to moderate traffic. It needs improvement in the heavier traffic situations. With Tesla’s fleet experience this will come quickly.

Has anyone here actually used Cadillacs product? It's way inferior. It ping-pongs you back between lane borders in a way that is really disconcerting and it can't seem to find a stable centering on the lane.

 

[GolanB]

For context, CR serves mostly conservative (as in not risk takers) people who aren't car or tech enthusiasts, who just want a reliable appliance to get them safely from A to B. So, yeah for their readership NoA is not ready, and certainly doesn't meet the marketing / tweeting hype. Agree CR should have stated it's Beta (though their readers have no clue what that is) and added that non-NoA AP is great.

I don't use it much, but bought it to watch history being made (and be part of it), like some others on the thread. Watching it slowly navigate a clover leaf last week, then be hesitant to merge onto the next highway, I too felt like it was my 15 year old behind the wheel first time on a highway. And my wife was definitely not impressed with it. Absolutely takes more effort than to just drive myself, but I engage it when i'm alone and can spare the mental cycles to help "teach" the neural network.

The only thing I'm worried about in terms of investing in FSD is whether we have good enough sensors for it. We have 8 cameras but they are mono-vision, and my car has a lot of trouble judging where other cars are when first merging onto a highway. I know the neural network makes use of relative motion to make up for lack of stereo vision, but that doesn't work when the other cars are moving too so there's not enough relative movement for it to do it's magic. I'm predicting there will be an upgrade to the cameras, or maybe additional ultrasonic sensors like the "cross traffic alert" sensors many cars have in the rear quarter panels now, added to new cars at some point now that there's a fast enough CPU to handle it. The question I have, is will us early adopters get them too?

Musk has stated on numerous occasions that he firmly believes the sensor hardware available in current models is all that is needed. Some have questioned this logic, for various reasons - including stereoscopic vision, as well as radar sensor redundancy, but I will state that there is an area of study which reproduces stereo vision with a single camera operating multiple frames over time and distance. There have been studies, videos, and even educational coursework on this subject.

Here is one example of some prior work: Driving Computer Vision with Deep Learning