Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register

4 months left in 2020 to make a Tesla FSD on city streets

This site may earn commission on affiliate links.
Based on what?
4 years of observation since I went all in on Tesla based on their faked “full self driving” video released in 2016.

The opinion of people developing ACTUAL full self driving at Waymo, Uber, and Cruise.

The fact that my car’s sensor suite can’t even pull off keeping itself in a lane under super difficult and uncommon situations such as “morning” and “dusk” when the sun is low in the sky.

Remember, Elon always delivers. Not always on time, but eventually he will.
This is a joke, right?

Tell that to people that bought “691 HP” P85Ds.

Or people waiting for real time battery swaps.

Or people waiting for their “top of the line P100D loaner” for every service visit.

Or people that bought the “full self driving package” as described in 2016, not the current watered down basically fancy cruise control definition.

Or the people that bought AP1 cars who were told they would be able to summon their car from across the country and it will come find them.

I could go on.

His timing has become better and he still believes FSD feature complete this year. I say January. Than the roll out for Software Stack 2.0 will begin and the fun with machine learning will really begin.

Don’t stop believin’ brother. I admire your ability to not let things like observable reality get in the way of your optimism.
 
Last edited:
LOL people claiming sensor suite is insufficient by just shouting. Based on what? There are hundreds of people working on FSD at Tesla. People much brighter than you and me. Most of R&D budget is going to FSD. Remember, Elon always delivers. Not always on time, but eventually he will. His timing has become better and he still believes FSD feature complete this year. I say January. Than the roll out for Software Stack 2.0 will begin and the fun with machine learning will really begin.

I hope you are right but I don't feel it being ready until 2025. This has to be regulated no? how will we even be allowed to go totally hands off by law?
 
  • Love
  • Like
Reactions: camalaio and cdswm3
Oh is that what it is? I thought I was drinking the Kool-aid back in 2016. :) Maybe that is why I wasn't poisoned. I was drinking snake oil.
So yea, I was sold on it too. :oops:

We weren’t the only ones brother.

Always fun to venture into the Model 3 forum and get pelted by disagrees from the young and faithful.

/shakes cane
 
Allow me to define what "Feature Complete" means.

That means the car can kind of do, in a very vague, general, and unreliable way, each of the aspects of driving that it must do, to one day achieve real FSD. But it is nowhere near ready to go out and drive around a major city, take on and off ramps, merge, handle a 4 way stop sign intersection, construction zones, etc. on its own yet.
 
I love Tesla and think Elon is an amazing visionary. I bought FSD in anticipation of greatness soon to come. Unfortunately I've lost a great deal of respect for Tesla's timeline. IMO, I don't think it's anytime soon. I've had to adjust what I think they are going to deliver to me...FSD is going to require baby sitting just like Auto pilot for the foreseeable future. I think Tesla will eventually solve this, I just have no idea when and I don't feel like my investment in FSD was particularly wise. I sure wish I would have bought TSLA with that money instead.
 
I hope you are right but I don't feel it being ready until 2025. This has to be regulated no? how will we even be allowed to go totally hands off by law?


It's already 100% legal in many US states..... which have laws on the books essentially saying "If your self driving car can obey all the same traffic rules/laws as a human driven car, it's legal to drive right now"

I mean, nobody sells a car that DOES that today, but if they did it'd already be "approved" to drive itself in those states.

You being in the EU I expect you'll be waiting...somewhat longer... for approvals there however.



Or the people that bought AP1 cars who were told they would be able to summon their car from across the country and it will come find them.

You list some perfectly valid missed promises... but he did not promise that.


Elon Musk: In Two Years, Your Tesla Could 'Summon' Itself Cross-Country

For this to happen, Musk says Autopilot-equipped Teslas would require more hardware than what's currently installed.

He was very clearly not promising AP1 cars would be able to do that.

He said possibly one in 2 years could.... (which as it turns out is a car with the current sensor suite that they STILL believe can do that eventually).

They may still turn out 100% wrong, but they never claimed AP1 cars would be able to do it.
 
Well my last trip 5 days ago one of a few events... @ 8 seconds the sudden application of the breaks causes the car to slow hard and shift to the side and I had to take corrective action. It was a hard grab of the breaks. Been doing this since 2019. Because of a shadow? Other lane keeping methods are far safer no? Running most up to date patch. It really has you looking around everywhere thinking what the actual... just triggered auto breaking.


If It can't do basic AP driving 100% by now it will be many years if ever able to drive itself on public roads.
 
Anyone that thinks FSD is coming soon is living in a fantasy land. I am firm with my belief that no car on the road today is capable of FSD based on what we have seen. The car cannot even park itself consistently.


The counter argument is that your thinking is like looking at a crawling baby and insisting anyone who think's it's going to walk is living in a fantasy land- I mean it can't even stand up so clearly it never will be able to with the same hardware right?

(also the code running today is being largely thrown in the trash sometime in the next few months to be replaced by an entirely rebuilt code- so saying anything at all about the capabilities of the system running todays code is pretty pointless)

I still don't think they're getting to L5 anytime soon either way, but something like at least L3 or possibly L4 highways should be pretty easy post-rewrite, and at least L2 city streets (which makes it feature complete per the post-3/19 definition of FSD)
 
  • Love
Reactions: pilotSteve
The counter argument is that your thinking is like looking at a crawling baby and insisting anyone who think's it's going to walk is living in a fantasy land- I mean it can't even stand up so clearly it never will be able to with the same hardware right?

(also the code running today is being largely thrown in the trash sometime in the next few months to be replaced by an entirely rebuilt code- so saying anything at all about the capabilities of the system running todays code is pretty pointless)

I still don't think they're getting to L5 anytime soon either way, but something like at least L3 or possibly L4 highways should be pretty easy post-rewrite, and at least L2 city streets (which makes it feature complete per the post-3/19 definition of FSD)

Yes. I heard the rumor that there is a whole new code base and this one will be thrown out.

1-You think a brand new code base is just going to be perfect for version 1.0, good enough to get perfect FSD? LOL.
2-From the way autopilot drives right now on the highway, it at best gets a C+/B- from me. It can't even stay centered on curves on the highway and can't stay centered on two lane roads.
3-Can't do simple avoidance. This past weekend I had to take control twice because I saw objects on the road that can damage a vehicle yet I can avoid by just shifting out of my lane a little after making sure one one was in the lane next to me. I can process and make a decision in half a sec. The car can't even process other vehicle speeds correctly.
4-The car needs to do predictions. For example, I was using autopilot on a 2 lane road. Someone pulls on to the on coming lane at an ok distance away. The autopilot breaks "hard" because it thought the car was going to hit it but I was already ready to hit the accelerator because my brain already subconsciously knew that with my vehicle speed and their vehicle speed that by the time I got to that point the other car would have pulled in to their lane and I would not have hit it. Currently the M3 can't do that, it sees car, hit breaks.
 
LOL people claiming sensor suite is insufficient by just shouting. Based on what? There are hundreds of people working on FSD at Tesla. People much brighter than you and me. Most of R&D budget is going to FSD. Remember, Elon always delivers. Not always on time, but eventually he will. His timing has become better and he still believes FSD feature complete this year. I say January. Than the roll out for Software Stack 2.0 will begin and the fun with machine learning will really begin.

The camera only system is inadequate. I expect a FSD system to be BETTER than humans, not just be on par. The two little cameras pointing forward now don't even have the same resolution as an eyeball. It doesn't see as far and can't make out objects as far as me. A FSD system needs full situational awareness. It has to "see" and process stuff 360 degrees and predict the future.
If it can't make out shadows verses real objects now, how is it going to know the difference between a blown out tire parts on the highway vs. a black plastic bag. One, I have to avoid by switching lanes or shifting a little (something that I can determine) and the other I can just drive over.
 
Yes. I heard the rumor that there is a whole new code base and this one will be thrown out.

Not a rumor, it's been discussed by Elon himself since at least January (and was already well in progress by then)


1-You think a brand new code base is just going to be perfect for version 1.0, good enough to get perfect FSD? LOL.

Not only do I not think that, I didn't SAY that.

Or anything remotely like that.

I did say based on what we know about the re-write I expect it to be significantly MORE capable than the current version because of the fundamental differences in how it works though.

Why do you think that's untrue? (and if it's untrue- why'd they bother?)



2-From the way autopilot drives right now on the highway, it at best gets a C+/B- from me. It can't even stay centered on curves on the highway and can't stay centered on two lane roads.

The system is explicitly not intended for 2-lane roads- see the owners manual.

I've never had any issue with it not holding a lane in 20,000 miles of use on highways though.


3-Can't do simple avoidance. This past weekend I had to take control twice because I saw objects on the road that can damage a vehicle yet I can avoid by just shifting out of my lane a little after making sure one one was in the lane next to me. I can process and make a decision in half a sec. The car can't even process other vehicle speeds correctly.
4-The car needs to do predictions


Both are explicitly things the re-write was intended to add the capability for.

Continuing to point out what the current system can't (and isn't even intended, per the manual) to do is not informative about what the new system that IS intended to handle those things can do.

The camera only system is inadequate. I expect a FSD system to be BETTER than humans, not just be on par. The two little cameras pointing forward now don't even have the same resolution as an eyeball.

There's 3 forward cameras, not 2. And unlike your 2 eyes they can each be focused on entirely different things at different distances all at the same time.

More significantly it can see in every direction at the same time- a human can't. Plus forward radar a human lacks.

A FSD system needs full situational awareness. It has to "see" and process stuff 360 degrees and predict the future.

Which, again, is explicitly the point of the re-write.

The current system does not, and can't, do that.

But 8 cameras plus radar should certainly be far more capable than the 2 in your head of doing it once the SW is right.





Again I'm not convinced even all that is adequate for L5 driving. But something like L3 or L4 highway should be very very doable with the sensor suite that exists today.
 
Everyone forgets the 12 ultrasonic sensors as well.

Range is quite short on those... he seemed mostly focused on seeing "further" which is kinda the opposite of what the ultrasonics do


Does the car have an accelerometer or anything to measure g forces?

Of course... you can even see it at work on cars that have track mode...not entirely sure how you're thinking it could help with holding a lane though
 
  • Like
Reactions: cdswm3