Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register
This site may earn commission on affiliate links.
More shameless self-promotion by buttershrimp.... AP2 continues to improve a lot for me. I'm going to try to put the car in tricky situations in my videos. If you give me ideas, music requests, and I'll try to do it as long as it is pretty safe and also if I'm curious and bored enough to see what happens. My FSD videos will not be fake, however, they likely will contain rap music. Perhaps we should start a new thread? Hmmmmmmmmm....

 
Remember, they promised it would come out of your garage and to your front door, find a parking space by itself etc.

Sure. Elon also used the example of Summon being able to pick you up from the other side of the country.

Certainly hard to imagine from the EAP we have seen so far. A possible/probable separate FSD codebase is an unknown of course.
 
  • Like
Reactions: 1 person
Two (mostly) different codebases. The code will be layered, with the upper control using lower layers, such as the visual object recognition NN (VOR-NN). The VOR-NN and associated routines is substituting for the Mobileye chip in AP1. So, AP1 = Mobileye chip + driving-logic. AP2/2.5 is VOR-NN + driving-logic. FSD = VOR-NN + mapping/routing-logic + driving-logic, and I would not be surprised if the driving logic uses a second deep NN.

So the lower level share the current NN stuff, so that won't be lost in a transition to FSD.

Also, why have two product names, EAP and FSD, if they are the same but for capability? Why not just have EAP, EAP+, EAP2, EAP-MAX, EAP-Ultra?
But for FSD you agree VOR-NN must be of excellent quality or will "mapping" compensate?
 
Ait, I chose to start this thread with a reddit post by "PM_YOUR_NIPS_PAPER".

I feel weird to post his/her/their reddits here, but I feel he/she/they is legit. And *hsey" isare posting some intriguing stuff. Now this:

Q/Comment:
4-5 years and no FSD would be a massive failure on Tesla part
especially considering Waymo is "finished"
The director of sales at Tesla said to expect great things about AP in the coming months... but he's also in sales. Tesla will probably be done before 12 months.

A (PM_YOUR_NIPS_PAPER):
It's going to take much longer than 12 months. I work (or worked? I'm going to be vague) on Autopilot. There are so many workaround hacks in place, even members of the team are not getting EAP. The head of Autopilot didn't even order FSD.
 
Right now we only have two benchmarks to indicate how far off FSD is.

1. Promised coast-to-coast FSD demo by the end of the year. With the weather getting worse and only a month to go, it looks like this promise will be broken which indicates that FSD is well behind where it was expected to be.

2. Current state of EAP. Can't even drive along a highway reliably yet. Hasn't even reached AP1 parity.

Conclusion: FSD is a long way off, and AP2 may have been a mistake. At the very least, AP2 has set them back over a year now.
 
Sadly, I think the workaround claim is true.

Why else would we get such strange bug introductions with each FW release? In 17.36 my DRL didn't work, in 17.42 my left front three sonic sensors don't work. Then there's the weird Air Suspension false positive bug in 17.42(?) as well.

However, let's all remember back 10 years when Steve Jobs demoed the original iPhone at WWDC'07. It is widely known that the demo was scripted in a very specific order or else the phone's firmware would crash. It was very similar in my eyes to the FSD demo as it was scripted and not (yet) indicative of real usage.

That being said, Apple delivered a functional device 6 months later and 10 years later is dominant.

Maybe Tesla can convert from the "workaround hack" EAP and "scripted" FSD functionality starting soon and deliver on their vision (pun). I am forced to happily wait.

Then again, everything at Tesla lives in the future and nothing lives in the present:

- FSD/EAP/Summon+
- Model 3
- Model Y
- Semi
- Roadster
- Service/Parts Availability
- Tesla Ride Share Network
- 3D Tunnels

Lots of promises I want to have realized, but I'm getting over burdened with new promises faster than anything already promised is being realized.
 
Ait, I chose to start this thread with a reddit post by "PM_YOUR_NIPS_PAPER".

I feel weird to post his/her/their reddits here, but I feel he/she/they is legit. And *hsey" isare posting some intriguing stuff. Now this:

Q/Comment:


A (PM_YOUR_NIPS_PAPER):

Who knows if this person is legitimate or not. I poked through that person's post / comment history and it does seem that he has some experience with machine learning in academia and industry, but that's it. I also note that he seems to have a massive ego with a thing for shitting on others / others' projects. My tendency is not to put too much weight into anything he says regarding his purported experience at Tesla. (But that doesn't necessarily mean that the claims are false. Some of them are simply well-known, independent of this guy.)
 
Last edited:
Congrats, Your work is done!

Apparently the work is only beginning, as I don't exactly share this excessively pessimistic mood.

I will lighten things up with this conclusive message.

giphy.gif
 
Conclusion: FSD is a long way off, and AP2 may have been a mistake. At the very least, AP2 has set them back over a year now.
My understanding is that Mobileye was itself transitioning from hard-coded object-recognition to VOR-NN, and they wanted to keep the data stream fro themselves. Tesla did not want to lose control to Mobileye, and developed their own VOR-NN. Mapping will help with the VOR-NN.

"especially considering Waymo is "finished"
I very much doubt Waymo is "finished". They operate in very restricted pre-mapped environs, and depend on relatively slow LIDAR mapping, and therefore are restricted to max speed. Tesla is working on a general solution.
 
I very much doubt Waymo is "finished". They operate in very restricted pre-mapped environs, and depend on relatively slow LIDAR mapping, and therefore are restricted to max speed. Tesla is working on a general solution.

This assumes Waymo would somehow stop at where they are - or not be working on the next step. That is IMO a pretty bold assumption.

Being pessimistic about AP2/FSD is probably premature. But being pessimistic about Waymo's autonomous credentials is IMO preposterous.
 
2. Current state of EAP. Can't even drive along a highway reliably yet. Hasn't even reached AP1 parity.

If a few here keep repeating this 10 times each day, doesn't mean that is true. From the videos I am seeing from Buttershrimp and others, I can see for all intents and purposes AP2 has reached parity and even forged ahead of AP1 in some scenarios.
 
Last edited by a moderator: