Co-pilot does autonomous lane changes with NO assistance and handles highway interchanges. You are the one who said copilot was just openpilot. You are the one who embarrassed openpilot.
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Anthony Levandowski beats Tesla's Elon Musk to first Auto Cross country (3k Miles, 0 disengagement)
- Thread starter Bladerskb
- Start date
Co-pilot does autonomous lane changes with NO assistance and handles highway interchanges. You are the one who said copilot was just openpilot. You are the one who embarrassed openpilot.
morrisdl
Member
diplomat33
Average guy who loves autonomous vehicles
My final post on this. By the video's own admission, Co-Pilot is a L2 ADAS. Levandowski replicated Autopilot and did a coast to coast trip, all highway driving, with no driver disengagements except for gas and sleep breaks, which he used to great marketing affect. Fine. It is certainly impressive what Levandowski was able to do given his limited resources. But he did not achieve full self-driving. He just created AP that works extremely well on highways. To argue that Co-Pilot is anything more than L2 is just factually incorrect. Of course, @Bladerskb takes things a step further and wants to argue that the video is proof that Tesla is "far behind" and wants to bash Tesla fans for it. But that's entirely subjective. I don't agree that Tesla is "far behind" and I've given my reasons why.
EAP and NOA aim to be something more than Co-Pilot. It is trying to push beyond L2 and eventually give us true self-driving not just on highways but on all roads. Is it there yet? Of course not. And yes, NOA has some problems but that's going to be true in any development of a new product, especially one that is trying to push the envelope so far. NOA will get better. Tesla is making progress one step at a time. Naysayers like @Bladerskb love to point to Tesla not meeting goals on a certain timetable. "NOA was supposed to be L3 and it's not! See Tesla failed again!" But I for one, prefer to focus on Tesla's progress. And I prefer to enjoy the good that my car provides to me NOW and look forward to what my car will do in the future.
EAP and NOA aim to be something more than Co-Pilot. It is trying to push beyond L2 and eventually give us true self-driving not just on highways but on all roads. Is it there yet? Of course not. And yes, NOA has some problems but that's going to be true in any development of a new product, especially one that is trying to push the envelope so far. NOA will get better. Tesla is making progress one step at a time. Naysayers like @Bladerskb love to point to Tesla not meeting goals on a certain timetable. "NOA was supposed to be L3 and it's not! See Tesla failed again!" But I for one, prefer to focus on Tesla's progress. And I prefer to enjoy the good that my car provides to me NOW and look forward to what my car will do in the future.
Last edited:
Huh? I'm sorry but keep contradicting yourself. Co-Pilot is based on Anthony's own system that is aimed for full self driving. His internal system works on city streets. Co-Pilot is just a limited highway version that he wants to release instead of waiting for when its L4 equivalent is ready. Because there are benefits you can attain now.
Yet you call Co-pilot simply lane keeping and EAP/NOA as aiming to be something more. Yet Tesla's EAP description is that its an "ADAS" and that the "driver is to remain in control at all times".
Co-Pilot is simply a result of Anthony's progress just as NOA is a result of Tesla's progress.
If Co-Pilot is 100x better than NOA. It means Anthony is ahead of Tesla in Self driving, period.
You avoided my question. When Tesla announces they did FSD trip next year and you still have to take over. Will you come back here and admit that you were wrong? This is the problem you make up criteria and goal posts but you never hold Tesla to it. So you make it up as you go. Take responsibility for the statements you make for once.
diplomat33
Average guy who loves autonomous vehicles
If that happens, yes, of course I would admit I was wrong. But I doubt that Tesla would do a FSD coast to coast demo where the driver has to take over multiple times. That would defeat the whole point of a FSD demo. The whole reason why Tesla is taking so long in doing a FSD demo coast to coast is precisely because they are waiting until they can do it without disengagements and without "cheating".
I replied again because I did not want to avoid answering your question.
"The car, a modified Toyota Prius, used only video cameras, computers and basic digital maps to make the cross-country trip.'
How does the computer turn the wheels, apply brakes etc?
Bolt-on hardware?
How did Waymo do it?
These are critical components that seem to never be part of media coverage.
What vehicles can adapt these systems, will they need this or that from the factory etc.
How does the computer turn the wheels, apply brakes etc?
Bolt-on hardware?
How did Waymo do it?
These are critical components that seem to never be part of media coverage.
What vehicles can adapt these systems, will they need this or that from the factory etc.
If that happens, yes, of course I would admit I was wrong. But I doubt that Tesla would do a FSD coast to coast demo where the driver has to take over multiple times. That would defeat the whole point of a FSD demo. The whole reason why Tesla is taking so long in doing a FSD demo coast to coast is precisely because they are waiting until they can do it without disengagements and without "cheating".
I replied again because I did not want to avoid answering your question.
You misunderstood my question or you are conflating it on purpose.
I'm saying after Tesla does its FSD demo trip and they release the update. Remeber Its supposed to be done with version 10 alpha. After they release it and you as a driver still have to take over on freeway and city streets. Will you come back here and admit you are wrong?
diplomat33
Average guy who loves autonomous vehicles
Sorry, I misunderstood your question because it was not very clear. Well yes, if I get the FSD option on my car and I download the V10 Alpha version that claims to be FSD but I still need to take over, of course I will happily admit that I was wrong. That's not a big deal to me. I am going to be honest when I am wrong. Of course, I am hopeful that Tesla is successful and is able to release FSD with V10 as they promised.
My question to you is: do you want Tesla to fail? I ask because, no offense, but it just seems like your entire purpose here is to point to missed promises and shout "see I told you so!". Even now, you seem like you can't wait for V10 to be released and for someone to post that it is not FSD so that you can rub it in our faces that we were wrong again.
This is wrong by the way, irrespective of where one believes Tesla is on the autonomous ladder. EAP (which includes NOA) is an ADAS by description. Driver will remain in charge. It will not require HW3 either as further proof it will always remain a driver’s aid.
FSD in separate and Tesla has not released or shown anything FSD like since the 2016 video.
diplomat33
Average guy who loves autonomous vehicles
Not sure about this. I used to believe that FSD was a completely separate system from EAP. But now I am not so sure. It is very possible that FSD will be the new chip and a software update that upgrades EAP to FSD capabilities, rather than a completely new and separate software. I don't think we know for sure how Tesla will implement FSD until they do it.
But all that being said, NOA is still aiming to be more than your basic L2 ADAS. NOA seeks to add features like auto lane change without driver input and highway transition and auto exit taking, which your average L2 ADAS can't do.
@diplomat33 FSD may well evolve from the same codebase but they are clearly separate products in all communications thusfar. The HW3 separation makes it even more distinct. The product called EAP (which NOA is part of) has always been called and said to remain a driver’s aid so it would take a clear sea change to change this and retrofitting HW3 into EAP cars as well.
diplomat33
Average guy who loves autonomous vehicles
Yes, I agree. When I said that EAP aims to eventually be self-driving, I was referring to the software. Yes, "EAP" the product, will always be a L2 ADAS. But when we get the FSD option enabled on our cars, EAP will change into FSD. At that moment, it won't be "EAP" anymore.
EAP on HW 2.x must be a lesser NN than FSD on HW3.0. The back end system, training data, and methodology can be shared, but due to resource constraints, Tesla must trade FSD functionality for HW 2.x processing capacity. The was alluded to in a Karpathy SW 2.0 talk regarding recompiling/ retraining your NN based on target NN size.
EAP on HW3 will (most likely) be different than EAP on HW2.x. I suspect 3.0 EAP will be the FSD code (when it exists) with abilities disabled whereas on 2.x, it would not have those abilities at all.
diplomat33
Average guy who loves autonomous vehicles
If true, does it imply that the new AP3 chip will indeed improve EAP? I know Musk has said that the AP3 chip will not be required for EAP but I do suspect it will still improve EAP performance.
Ya know, I could be wrong on my previous statement.
It is possible that they use the 2.x compatible EAP NN on 3.0 HW. That would assure feature parity car to car and also not require tweaking FSD to support EAP. Of course, the 3.0 HW would be bored...
Or, Tesla is not adverse to different experiences on different cars and makes each version custom to HW.
What I don't see them doing is removing any potential safety functions from the system. HW 3.0 must be able to handle more than 2.x, and it would be a waste to not utilize that capability. One thought ratting around in my brain due to that is 3.0 EAP will do anything FSD can, but force you to have a driver in the car.
So, maybe???
Tam
Well-Known Member
Comma.ai has proven that it can be done even for a classical Model S that has no Autopilot.
So, I don't think that's an issue.
Most cars have lane keeping assist and automatic cruise control these days (at least as an option but often standard). Toyota has is as well.
Right. You have to retrofit hardware for autonomy to cars that don't have it by default, like Anthony Levandowski had to do with his demonstration of semi-autonomy here.
Any AP2.5 Tesla owner can do the same thing today. Root your car, disable nags (or wedge nag defeating fruit onto steering wheel) and load the beta NoA that didn't require confirmation of lane changes. That's not a coast to coast autonomous drive.
Any AP2.5 Tesla owner can do the same thing today. Root your car, disable nags (or wedge nag defeating fruit onto steering wheel) and load the beta NoA that didn't require confirmation of lane changes. That's not a coast to coast autonomous drive.
Last edited:
