wesley888
Member
Actually, when AP2 first came out, they advertised that EAP will use 4 camera and FSD will use all 8. So EAP supposed to just use the 3 in the front and 1 in the back, and not use any of the side cameras. I figure they just can't do it safely, even just for EAP without the side cameras.
diplomat33
Well-Known Member
Putting aside the levels of autonomy, hopefully, we can all agree that V9 is a step in the right direction towards eventually getting to self-driving.
V9 is a step in the right direction to eventually delivering EAP's On-Ramp to Off-Ramp feature. "FSD" is really a different beast, but you could consider EAP to be a prerequisite so in that sense it brings them closer to FSD. But another way of looking at it is that EAP is on one side of the Grand Canyon and FSD is on the other side. We just moved closer to the canyon, but we're on the EAP side.
V9.0 might be the first time the (reduced) FSD code has been released as EAP. Previously, I felt the EAP branch and FSD development were separate.
diplomat33
Well-Known Member
I would say that v9 takes us to about 90% of the on ramp to off ramp feature. With v9, we are really really close now.
I used to think that Tesla was pursuing two separate paths, one for EAP and one for FSD. But I have changed my mind on that. I now think that Tesla is working one single path towards FSD and EAP will simply gain new features and gain competency over time until it eventually becomes FSD. So in that sense, I think V9 definitely moves us one step closer to FSD.
A slow rollout getting slightly faster and random people getting invited to the early access program? I think it's time to grab a couple boxes of donuts and head to my local service center...
RyanDe
Member
Can anyone chime in that has v9 about if autopilot is better overall too? I've seen and read a bunch about how it changes lanes etc but is it just better overall now? Better at staying in the lane, not stopping/accelerating too quickly, recognizing cars coming into your lane?
@dennis_d earlier in the thread described it "as if it's a new system". Someone on Reddit mentioned it taking cloverleafs and local streets it was not able to take before.
I think it's safe to say yes, it is better overall. Features aside, activating more cameras (or the separate codebase that's been in parallel development) has made AP2 perform even better than before.
Just send it already. The NN could make a difference in avoiding an accident or other negative outcome. Its clearly as safe as the best of 8.1 or better, so that should be it? They usually patch their issues in upcoming firmware updates anyways. Get it over with, Tesla.
Yeah ditto. Totally willing to watch this new neural net mature itself over time, but I want it now! We already have to be fully attentive anyway.
So, after several days with 9.0 it's quite nice and it sorta takes clover leafs and such, but it takes them incredible slow compared to normal traffic but the shadow braking is really really bad. I can't go for a highway drive for 20 minutes without 2 - 3 shadow braking events. I was hoping the final calibrations would have resolved it, but not so much, so hopefully that's part of what they are refining out of the system.
I'll also share this observation, yesterday, I was driving down a highway in the center lane, and the car in front of me dove to the right to reveal a stopped construction truck in the middle of the highway and shocking to me at least, the car began to slow down for a stop, so it clearly saw the stopped car in front that wasn't present before. But, this was a one off occurrence, so I can't really say they have solved the untracked stop car problem, but this was definitely a solid result.
I'll also share this observation, yesterday, I was driving down a highway in the center lane, and the car in front of me dove to the right to reveal a stopped construction truck in the middle of the highway and shocking to me at least, the car began to slow down for a stop, so it clearly saw the stopped car in front that wasn't present before. But, this was a one off occurrence, so I can't really say they have solved the untracked stop car problem, but this was definitely a solid result.
As with everything in software, and even moreso with AI/autonomy, the first 90% of features takes 90% of the development time, and the remaining 10% takes the other 90% of the development time.
In autonomy I'd go so far as to say the first 90% takes the first 90% of the time, the next 9% takes the next 90% of the time, the next 0.9% takes the next 90% of the time, and so on. And before you can trust this I think we need at least four 9's. We just got our first 9.
To call this even L3 they have to handle the corner cases. And there are sooooo many corner cases.
Pet peeve alert: Artificial neural nets do not "mature themselves over time". They are static binary blobs delivered to the car during software updates which do not modify themselves. They only improve via new releases from the mothership. Improvements may come from additional data collected from the fleet or they may come from human-driven modeling algorithm changes.
Edit: Typo, original said "human-driving modeling" which is not at all what I meant to say.
Last edited:
diplomat33
Well-Known Member
Well, according to Musk's latest tweet, they are working on the corner cases.
I trust they will get there. Autopilot is not L3 yet but it will get there. We are certainly a lot closer than we were last year, that's for sure.
Az_Rael
Supporting Member
Yeah, the main reason I bought FSD was to gain access to the other 4 cameras. Figured that would make for more stable EAP even if FSD never materialized. Guess that was a $3k donation to Tesla's coffers if EAP ends up using all the cameras too. Oh well.
FSD will still have more driving features than EAP. However I expect EAP (and even stock AP) will have all the safety features available in FSD.
Holy-moley! That's the biggest single download I've heard of. You sure that's just the Tesla and now your whole network? Everyone has been saying maps is a little over 5GB.
Similar threads
- Poll
