Forbes ran a ten year study and found that Teslas autopilot was just as good as a human 94% of the time. 3% it was slightly better 3% of the time it was slightly worse in performance. I still can't believe that Tesla never engineered a way to wash the cameras Similar to how Volvo and Audi had headlamo washers
FSD/Tesla driving software
- Thread starter Torrid
- Start date
The windshield washer cleans the camera. It’s in the glass. That’s not the problem. The problem is it can’t cope with fog or extremely heavy rain or snow.
Human driving is a pretty low bar... Ha!
OMG, so true.
Well, real drivers fail mostly because of in-attention, excessive speed, or not predicting how a situation could lead to collision, and taking some precautions.
Autonomous systems probably still have almost no capability to predict a potential situation developing, and have relatively short fields of awareness as well. Lots of people drive that way too, but they are the ones getting into accidents....
Self driving is better than unsafe drivers, but if you pay attention and use your large brain to make decisions that lower the risk of accidents, I think you are better off doing that.
I'll take my chances and skip the FSD. As it stands now I'll rely on my ability to drive.
I'm close to taking delivery. I opted for the full self-driving option.
100%. As far as I see it if FSD is slightly better than an inattentive driver and it’s programmed to always be processing, what’s its shortcoming? It doesn’t fall asleep or get tired.
I think any alert driver is a much more reliable option. Even better an alert driver with a driver aid that operates as another layer of safety. I think it’s terrible so many are making excuses for autonomous driving solutions being better than an inattentive driver.
True, but you can also argue that a 16 year old would be 94% as good as a 20 year experience long haul truck driver, but you probably wouldn't want to hire a 16 year old with new license to drive a long haul truck.
The fundamental of a lidar based self driving system will always be better than a camera based one, all else being equal.
Definitely. That's why I always point out that humans have multiple senses, not just vision. Musk claimed this week that the only reason that Tesla doesn't have lidar is that "humans don't shoot lasers from our eyes" And he wonders why people are starting to not want his products.
I saw this. I’m not sure what the difference is that this is the reason. While I haven’t tried to blast through any fake walls, I still find HW4 in my Y to make the same mistakes my HW3 3 makes in the same conditions. Light glare in certain spots causes the car to stand hard on the brakes. Infuriating is too mild a word for it.
It’s very interesting technology. I just recommend taking its abilities with a grain of salt. It does some unexpected things at times. I find at least for me it’s just more relaxing to drive myself. It’s not terrible on open multi lane highways, but it hesitates at lights around town and people will honk at you. It also doesn’t have good etiquette on 2 lane 55mph blacktop. It won’t hug the shoulder when opposing traffic is passing.
The problem is that since I only want to use it on highways, Autopilot with Autosteer can handle that. It just won’t change lanes on its own.
exactly, I'll take my chances with an automated system vs the humans I encounter every single day.
elon is just cheap.
One of the largest issues with camera only systems is that as far as I'm aware that can't calculate oncoming traffic speeds nor oncoming objects. There have been numerous times that Tesla drivers have either hit something or were close as lidar and radar sends out beams and can calculate the distance and speed of vehicles. Cameras can't.
In many cases. Tesla skipped a $10 rain sensor and decided to use its cameras. All fine and well except the cameras are already processing a large amount of data and images. This causes numerous issues.
My complaints get labeled as Tesla hate by the the typical Tesla crowd even considering I own two of them.
Beyond that the camera just doesn’t work as a rain sensor even when driving yourself. Wrong tool for the job.
A camera can estimate oncoming speed by analyzing the size change of 2 consecutive images, and what portion of the FOV it takes up. Precision depends on the resolution of the camera and how well it can define the blob. It should definitely be able to determine the difference within +/- 5mph in decent conditions I would guess? The issue is getting a decent image and blob analysis to begin with. ie if your looking at a white object with the sun to your back - good luck.
I wonder how well it works considering the cameras are so close together. It definitely has trouble with distance a lot.
Last edited: