Daniel in SD
(supervised)
Everyone else in this forum is short.I can’t believe I’m the only one here who believes Elon here so far!
-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
Where do you think Tesla's FSD will be 12 months from now (April 2020)?
- Thread starter diplomat33
- Start date
Not short. But Elon's timetables are goals. Goals of doing something new have a tendency to slip. "The impossible takes a little longer".Everyone else in this forum is short.
I can’t believe I’m the only one here who believes Elon here so far!
I believe Elon on the end goal of L5 robotaxis, I am just skeptical on the timeline. I think Tesla will get there, just miss Elon's timeline.
Feature complete means that all the self-driving features are done, not necessarily good, but they are implemented. Feature complete is just the first step of course, because once you finish all the self-driving features, you still need to refine, improve, tweak etc to make it reliable enough.
Daniel in SD
(supervised)
Are there any autonomous vehicle projects that are feature complete? Waymo, Cruise, etc.
AlanSubie4Life
Efficiency Obsessed Member
Maybe you could enlighten us with the textbook definition?
It’s pretty vague, but to me it means advanced Level 2, able to do most typical surface street drives without many interventions. And able to drive from house/garage and park in the destination parking spot.
Interventions will be common for feature complete:
Simple cases (suburban commutes) - 1 intervention every 5 miles
Medium complexity (say, LA area (not downtown) surface streets): 1 intervention every mile.
San Francisco streets: 1 intervention every block.
Like I said, pretty vague, presumably intentionally so. Above is just my impression of what it means roughly.
Textbook definition:
A feature complete version of a piece of software has all of its planned or primary features implemented but is not yet final due to bugs, performance or stability issues. This occurs at the end of alpha testing of development.
------
So for Tesla, "feature complete" means all the "primary features" for FSD are done. The software is not complete, will have bugs and disengagements but the "primary features" are done and the software can move from alpha to beta.
AlanSubie4Life
Efficiency Obsessed Member
Seems fine as a definition, but makes it hard to know what that will really look like to a user. I think it will be a cool toy to experiment with at that point, according to this definition. But not useful.
To the user, "feature complete" looks like a car that can self-drive but needs to be supervised by the user. But remember that "feature complete" is an intermediary stage. It's just alpha software. So yes, it will be less useful in the beginning but the usefulness will improve over time. As Tesla continues to work on FSD beyond the "feature complete" stage, FSD will become better and will move on to beta and then public release. The public release version of FSD will be far better than the "feature complete" version.
Or is it you that doesn't understand what feature complete means when it comes to Tesla/Elon?
I can argue that NoA is not feature complete, but Elon/Tesla would disagree. I think all Tesla is going to aim for to claim feature complete is simply L2 assistance in most situations. Even if it might greatly struggle with certain basic tasks.
For the record I'm one of the 19 who claim Tesla will be still struggling to match feature complete. I say that because I challenge Tesla to my expectations, and not their expectations.
Eno Deb
Active Member
The trouble starts when you actually try to list the features. Autonomous driving is an enormously complex task and not easily packaged into a small number of "features" (and we probably don't know all required "features" yet). That's why things like the SAE levels, disengagement rates, or Mobileye's RSS model are used to judge progress. Normally in software development, features need to be reasonably clearly defined so you can make a statement whether or not they are implemented. But I haven't seen anything like this from Tesla. The promised FSD "features" beyond what's currently available (functionality formerly in EAP and now in FSD) according to the order page are:
- Recognize and respond to traffic lights and stop signs.
- Automatic driving on city streets.
There is a more comprehensive description on Tesla's Autopilot page, but it seems highly unlikely that this is what they are hoping to have "feature complete" by the end of the year:
"All you will need to do is get in and tell your car where to go. If you don’t say anything, the car will look at your calendar and take you there as the assumed destination or just home if nothing is on the calendar. Your Tesla will figure out the optimal route, navigate urban streets (even without lane markings), manage complex intersections with traffic lights, stop signs and roundabouts, and handle densely packed freeways with cars moving at high speed. When you arrive at your destination, simply step out at the entrance and your car will enter park seek mode, automatically search for a spot and park itself. A tap on your phone summons it back to you."
Last edited:
"Feature complete L5" isn't just responding to traffic signals, turning left and right at intersections. Which is what Elon is trying to sell as "feature complete L5".
But even L4 feature complete requires you to go the opposite lane to pass someone (DPV), avoiding debris, non functioning traffic signals, handling all types of constructions, detours, road closures, dead ends.
But then calling something "L5 feature complete" requires you to do that everywhere in all weather/road conditions. As elon said himself "blizzards, hurricanes, floods, fires, etc"
Its not arguable though its the facts. Its clearly NOT feature complete on the highway.
Last edited:
Everyone talks about “feature complete” and what that actually means in this context, and I’ve had thoughts about that rattling around in my head tonight. The problem that I have here is that Elon has always been consistent on level 5 autonomous driving being the goal. With level 5 the answer to “can it handle scenario X?” is going to always be “if a human can do it, yes” regardless of what X is.
With that in mind, I would argue that level 5 autonomous driving is not a scale, but a boolean. You’re either 100% level 5, or you’re not at all level 5. If there’s a bug in the system that causes it to not be able to handle a situation that a human could, you are not at level 5. Being able to handle anything is the singular feature of level 5. So it would make no sense to say “we’re feature complete but still working out the edge cases”.
As usual with Tesla, a good chunk of their problems boil down to communication. They set the bar impossibly high at level 5, so I think there’s not a chance in hell that they’ll make it. If they had said they’ll be feature complete level 3 with headway into level 4, then I’d say maybe they have a chance.
With that in mind, I would argue that level 5 autonomous driving is not a scale, but a boolean. You’re either 100% level 5, or you’re not at all level 5. If there’s a bug in the system that causes it to not be able to handle a situation that a human could, you are not at level 5. Being able to handle anything is the singular feature of level 5. So it would make no sense to say “we’re feature complete but still working out the edge cases”.
As usual with Tesla, a good chunk of their problems boil down to communication. They set the bar impossibly high at level 5, so I think there’s not a chance in hell that they’ll make it. If they had said they’ll be feature complete level 3 with headway into level 4, then I’d say maybe they have a chance.
It's scaler because humans are scaler. Some are a lot better than others.
It's scaler because over time it will be expected to achieve greater levels of safety. The biggest thing Tesla will do is lower the bar from unrealistic to something realistic. It's a lot easier to have L3, and above when you have hundreds of thousands of owners demanding it.
It's scaler because some regions (like Arizona) won't require as much safety as regions like Germany.
It's scaler because as autonomous cars get more popular it will require more and more optimization to traffic control. What good will autonomous cars be if they end up jamming up our roads with cars going from one family member to another?
It's scalar because people will accept FSD that can't drive in the snow or really bad weather conditions. As long as it can safely pull to the side of the road while the event is happening,. But, they'll expect the system to become more capable over time.
As a result I don't think of FSD in binary terms at all.
Tesla is firmly entrenching itself in being the value proposition in terms of cost/mile. Things like a million mile drive train, and a million mile battery.
Elon might promise L5 everywhere, but I doubt that's going to happen. Instead it's going to be increasing advancements of L2 in a way that no one else is doing it. Anything L3 and above will be a bit patchy at first where different states will have different rules.
Tesla doesn't have a communication problem, but a problem going against entrenched competitors. So they fight a little dirty. Where they make unrealistic claims about future capabilities as a way to gain an advantage.
The only meaningful thing to attribute "feature complete" to is various accounting rules, and how things get reported. It's also important to customers in Germany because if it's not feature complete by a certain time then they have to give money back. But, to most of us the whole idea of feature complete is pointless.
FSD or L5 driving is so linked to infrastructure, and regulatory requirements that the goal posts will always be shifting.
The best thing Tesla did was to start the fight. There is a lot to be said for at least trying to jump start something despite not having the best player in the game.
No the description on the autopilot page is not talking about "feature complete". It is talking about FSD when it is finished. Remember that "feature complete" is not a finished product, it is an alpha version of your software that just has all the basic features in.
I think we can look to NOA to see what Elon means by "feature complete". For NOA, "feature complete" was the following:
- taking exits
- auto lane changes to stay in the right lane for navigation
- auto lane changes to pass slower traffic.
So "feature complete" was the basic features needed for a car to get from A to B on the highway. Again, the features did not need to work perfectly, they just had to be in, in order to qualify as a "feature complete". So yes, NOA is now feature complete according to Elon. Keep in mind that "feature complete" does not mean that the work is done. There is still work to be done with NOA and there were other "features" that also had to go in, like handling merging traffic. But "feature complete" is just the basics that have to get done first.
Taking the same approach with "automatic driving on city streets", "feature complete" is just going to be the basic features that allows the car to get from A to B on city streets:
- auto lane changes on city streets to follow navigation directions
- Handling intersections
- responding to traffic lights and stop signs
- taking an on ramp to get on the highway
Again, they don't have to work perfectly. They are the basic building blocks for driving on city streets.
All the other critical things that Blader mentions like handling detours, debris etc are super important but they come AFTER "feature complete" for Elon because the above features have to come first. Elon is taking the approach of getting the basic building blocks done first and then building on that to get to a finished FSD system.
Similar threads
- Article
