They need to hurry up and catch up because they are still years behind in camera based neural network development and deployment.
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Coast to coast drive happening this year for all FSD Teslas!
- Thread starter Daniel in SD
- Start date
They need to hurry up and catch up because they are still years behind in camera based neural network development and deployment.
diplomat33
Average guy who loves autonomous vehicles
They need to hurry up and catch up because they are still years behind in camera based neural network development and deployment.
Comparing it to AP2 is not an apples-to-apples comparison. You should compare EyeQ4 to AP3 since AP3 is Tesla's FSD computer, comparable with EyeQ4.
We know Tesla with AP3, has done a lot more than the list on the left. From Autonomy Investor Day, they have confirmed to have done traffic lights, road edges, traffic signs, path delimiters and maybe more.
Nice try though!
Last edited:
No I'm comparing deployed production system versus deployed production system.
AP2+ firmware on Feb 2019 versus EyeQ4 firmware in Q4 2017.
Even if you extend the comparison to AP3 and the newer firmware from this month the AP3 that's been deployed and in production still only runs those five neural networks according to verygreen. The only difference is a traffic light neural network that isn't useful at all for self driving because of its absurd amount of false positive and false negatives. It doesn't also qualify as a traffic light Neural network capable to be used in a self driving system as it doesn't recognize different types of traffic lights nor does it do traffic light relevancies, things you need in a traffic-light neural network.
Lastly there's no evidence they have networks that can do road edges, path delimiters even in development. But we are not comparing development here. We are comparing production systems. So its still 13+ versus 5, 13+ versus 6 if you want to compare to the latest May firmware and count that horrible traffic light NN.
diplomat33
Average guy who loves autonomous vehicles
I get the comparison but it is still flawed. You are taking the comprehensive list of everything that Mobileye has developed for the EyeQ4 and which has not all been deployed yet to mass market cars and comparing it to a few "beta" features that Tesla has deployed on old hardware. You mention traffic light detection but the traffic light detection NN that has been deployed now on AP2 is not Tesla's best. Tesla has better in their development. When Tesla does deploy "feature complete" on AP3, you will need to update your chart because Tesla will have deployed a lot more than just the AP2+ features on that list. Come back to me when we have "feature complete" deployed on AP3 and then we can compare AP3 to EyeQ4.
All the features have actually been deployed to all cars, mass market too. They are not all been used for control in these mass market cars, less than 5% of the capability of EyeQ4 is currently being used in mass market cars. But if you want to see their use, look at Mobileye's robottaxi fleet which uses eyeq4. Their lack of feature use in mass market cars is due to tier 1s and traditional automakers. But the features are deployed and Mobileye is using it to crowd-source HD maps from 2+ million cars from around the world.
I will update it later on this year when AP3 is "FSD feature complete". But it doesn't change the fact they are years behind. They would be at where EyeQ4 was 2 years ago, later on this year and this is with EyeQ5 being on the horizon (just under a year away).
diplomat33
Average guy who loves autonomous vehicles
I concede that Tesla is behind in development. But I think they will catch up with AP3. And if AP3 "FSD feature complete" is what Tesla is saying it will be, then Tesla will jump ahead in terms of features actually deployed to the most mass market cars. As you yourself pointed out, a lot of auto makers are not taking advantage of what Mobileye has deployed to their EyeQ4 firmware. If Tesla does enable all their "feature complete" features on AP3 to the entire Tesla fleet, then Tesla will jump ahead to other auto makers in terms of deployed features.
If there were no limits on the reporting, there would be video of it.
willow_hiller
Well-Known Member
Okay, now you're being absurd. There would be massive backlash if Tesla tried to censor opinions. The only limit was video recording. Tesla did not swear all of the investors to secrecy on some sort of convoluted contract where they're only allowed to post positive news about the event.
There is a massive backlash. Outside of the reality distortion field people are calling out yet another fake FSD demo.
willow_hiller
Well-Known Member
What's your agenda here, @banned-66611 ? 90% of your account history is liking comments about paint defects, 5% dragging up old Elon Musk lawsuits, and 5% blatant lies like the above.
There is no backlash outside of the TSLAQ fanatics.
There is no backlash outside of the TSLAQ fanatics.
What's your agenda here, @willow_hunter ? 90% of your account is Tesla fanboy nonsense and kissing Musk's arse.
I'm interested in the truth here. Tesla's claims about their autonomy are dubious. They seem to have had some major quality issues, which was something of a concern before I cancelled by Model 3 pre-order.
I distribute likes and informative ratings to people who post useful information and insightful comments. If that offends your fanboy sensibilities, too bad.
I'm interested in the truth here. Tesla's claims about their autonomy are dubious. They seem to have had some major quality issues, which was something of a concern before I cancelled by Model 3 pre-order.
I distribute likes and informative ratings to people who post useful information and insightful comments. If that offends your fanboy sensibilities, too bad.
boonedocks
MS LR Blk/Blk 19”
@banned-66611 What year/model Tesla do you own...…...I'll answer for you.....You don't.....I figured as much...
willow_hiller
Well-Known Member
He's just a troll, @boonedocks . And if his goal was to derail this conversation, he succeeded for a while.
Long story short, a lot of us are excited for city NOA by the end of this year. Once NOA works on city streets, I don't see why anyone couldn't complete a coast-to-coast drive.
Long story short, a lot of us are excited for city NOA by the end of this year. Once NOA works on city streets, I don't see why anyone couldn't complete a coast-to-coast drive.
boonedocks
MS LR Blk/Blk 19”
I knew that...just funny how people love to get in a conversation of which they know nothing about. Like I would not be able to go over to his Honda forum board and speak about anything with knowledge since I don't and have never owned one likewise I would never go over to the Bugatti Veyron forum board and troll there just because I wish I had one but don't.....
Last edited:
Nah you'd only be bold enough to do that if you were shorting Honda (or VW for the Bugatti).
NONE of the investor passengers who were riding in the demo thought it was any close to fake.
If I were a mod, i would give you one chance to explain why you think that FSD demo was fake?
What is a "fake" FSD demo anyway? There are a million FSD demos out there, what constitutes a fake demo?
I think that the places the FSD demo drives were done had been tested over and over again, insuring that they were going to go well. The weather was good as well. This takes nothing away from what Tesla's engineering team has accomplished, but it in no way represents any kind of worse case. A real "test" would be to randomly pick a destination that maybe has never been seen before. And do it in bad weather. At night or right at sunset. Or with a blocked off road with stunt drivers not following the rules of the road.
I don't think anyone is close to running that kind of test.
The reality is that we don't need to actually get that good at autonomous driving before it is better than the current 30K+ deaths per year in the US.
Check out these videos from Cruise.
Two years ago:
Sure, those are good. Even very good. But the weather is perfect; and all in the day time.
Before lots of people will give up their cars and get in the back seat of robotaxi it has to work all the time.
Many people would give up working in every location -- as long as it works around home for commuting and shopping.
Do you think these are good enough, today or in 2019, to drive in the snow at night over Tioga pass?
Similar threads
