Fair enough. I only got into this to call attention to the insane bullshit posted by @CarlK. But then y’all kept piling up on me anyway.
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Autonomous Car Progress
- Thread starter stopcrazypp
- Start date
-
- Tags
- Autonomous Vehicles
Fair enough. I only got into this to call attention to the insane bullshit posted by @CarlK. But then y’all kept piling up on me anyway.
It was the logical extreme of the other one so that's why I added it.
It was also important to distinguish it between those who believe that Waymo doesn't have real time remote control capability. This I did see being argued, and I do see merits in this argument.
Last edited:
So here is how I see the possible options:
Remote command execution only -> the remote operator can only issue high level commands to what it wants the vehicle to do. Things like the next pickup address or re-routing around a traffic situation.
Pseudo real time remote control -> Basically they could take over the driving, but the car would assist the driving to help reduce the impact of latency. An example of this kind of remote driving is what Double Robotics does with their new Tele-presence robot. Where it will avoid obstacles while the human is doing the driving. This normally wouldn't be used, but would be a backup in case something happened.
Real time capable -> Remote safety drivers watching over the car from some control room. Where they have some kind of setup with reduced latency. I believe they've stated that they don't have this capability.
Remote command execution only -> the remote operator can only issue high level commands to what it wants the vehicle to do. Things like the next pickup address or re-routing around a traffic situation.
Pseudo real time remote control -> Basically they could take over the driving, but the car would assist the driving to help reduce the impact of latency. An example of this kind of remote driving is what Double Robotics does with their new Tele-presence robot. Where it will avoid obstacles while the human is doing the driving. This normally wouldn't be used, but would be a backup in case something happened.
Real time capable -> Remote safety drivers watching over the car from some control room. Where they have some kind of setup with reduced latency. I believe they've stated that they don't have this capability.
CarlK
Active Member
I don't think anyone has said they remotely "control" the car 1:1. We were only talking about remotely "monitor" the car like they do with in car "safety engineer" now. They will only take over when in an emergency and when the car is confused.
Everything discussed was for remote "monitoring" the car (like in car safety engineers do). The argument somehow turned into remotely "controlling" the car only when that @SandiaGrunt clown found out he looked so stupid and tried to divert the object.
Last edited:
CarlK
Active Member
Lol no wonder. You really need to have your head examined for its mental state. Only remote monitoring, the same as how in car safety engineers do, was said by me about Waymo. Remote controlling of course can be done too but no one has accused of Waymo of doing that.
diplomat33
Average guy who loves autonomous vehicles
"Niedermeyer reports that the trip involved an unprotected left-hand turn, "busy city streets," and speeds as high as 45 miles per hour. Niedermeyer says that his 10-minute ride from a park to a coffee shop was uneventful—which makes sense given how carefully Waymo must have prepared for it.
Still, this represents yet another baby step in Waymo's journey toward full autonomy. And despite Waymo's slow pace of progress relative to expectations it previously set for itself, the company still appears to be the industry leader. No other company is testing fully driverless rides on the scale or at the speeds Waymo is.
Moving forward, the big question remains how soon Waymo can start offering fully driverless rides to the general public at a non-trivial scale. Waymo has been tight-lipped about the scale of its current driverless testing or how soon the technology might become commercially available."
Waymo finally let a reporter ride in a fully driverless car
Still, this represents yet another baby step in Waymo's journey toward full autonomy. And despite Waymo's slow pace of progress relative to expectations it previously set for itself, the company still appears to be the industry leader. No other company is testing fully driverless rides on the scale or at the speeds Waymo is.
Moving forward, the big question remains how soon Waymo can start offering fully driverless rides to the general public at a non-trivial scale. Waymo has been tight-lipped about the scale of its current driverless testing or how soon the technology might become commercially available."
Waymo finally let a reporter ride in a fully driverless car
Waymo has a lot more autonomous miles on the clock than Tesla. According to Tesla they did zero in 2018.
Note how Tesla has only ever demonstrated their "full self driving" in good conditions too, never in the rain.
And Tesla have only demonstrated Level 2, with a driver ready to take over at all times, where as Waymo has Level 4 cars on the public roads right now.
Indeed but I am genuinely interested in Waymo's progress in various weather conditions!
I can only assume you are trying to use the California self reporting ststem as a basis for that statement. As anyone who reads the requirements know, the use of nags means the driving is not required to be reported. Otherwise, all EAP/ NoA/ Autosteer events would be required to be reported, and CA is definately aware of those features.
Unless your position is that NoA/ AP is not autonomous in which case what are you referring to by the zero in 2018? (And I'm sure you are aware other states/ cities have looser or no reporting requirement).
Other reports:
1 billion miles of autopilot in 2018: Tesla owners have driven 1 billion miles with Autopilot activated - Electrek
Tesla reports: Tesla Vehicle Safety Report
Q4 2018:
Q3 2018:
WannabeOwner
Well-Known Member
Nah, that was 2018 also to align with the claim of Tesla reporting zero autonomous miles in 2018.
Q3 2019 was:
DanCar
Active Member
There is no amount of data that can cover all edge cases. What is possible that the number of edge cases are small as to make the software driven vehicle safer than the average driver.
CarlK
Active Member
True. You could still label them as billion miles edge cases or trillion mile edge cases and not to worry about those statistically would almost never occur. Even Tesla probably will need to collect another order of magnitude data, I believe it has about 2 billion miles of data right now, to make it to work for general autonomy. Those mickey mouse things others do with a few hundreds or even a few thousand test cars will never get them there.
No it's just the opposite. See my comment above. Every Tesla car made since 2016 is an autonomous car running in the "shadow mode" and collecting data. They even learn from driver interventions to know when they need to pay more attention. Tesla is the only company that employees, and could employee deep learning to develop the algorithm. This video may help you to understand the difference between the two approaches.
Last edited:
Then why don’t they do anything autonomously, even now, almost four years later?
Literally every AV company uses deep learning. For ****’s sake, the Udacity Self-Driving Car nanodegree focuses on deep learning! Do you understand how common something has to be for it to get baked into an introductory course on the subject?
From the course description: “Deep learning has become the most important frontier in both machine learning and autonomous vehicle development. Experts from NVIDIA will teach you to build deep neural networks and train them with data from the real world and from the Udacity simulator. You’ll train convolutional neural networks to classify traffic signs, and then train a neural network to drive a vehicle in the simulator!”
Self Driving Car Engineer Nanodegree | Udacity
Last edited:
CarlK
Active Member
They do not do FSD yet but they do perform many of the functions under autopilot. FSD can be, and will be, turned on when the system matures through continuous deep learning.
Many people say that because it's fashionable. No one is capable of actually doing it though. The basic requirement for deep learning is you need to acquire HUGE amount of data to train the NN (watch the Fridman video). No one except Tesla has the large fleet to collected needed data. No one likely can have one in the near future either.
Simulated data does not work either. Those people are just fooling themselves, or to keep their job securities at Waymo or NVidia. When you know how to simulate all edge cases you have already solved the self driving problem. That smart man Elon Musk first pointed this out. He knew years ago what is needed to make it to work.
Last edited:
This is a misconception. A deep net has a capacity that is a function of the number of weights.
The number of weights is constrained by the amount of compute available on the car. Today’s deep nets have ~100M weights. Larger nets take too long to run, and cars need quick reaction times.
A net with 100M weights cannot learn everything there is to learn from a dataset with billions of examples. It will necessarily forget most of what you teach it.
Similar threads
- Article
