When will we see a major new update? I can't complain of what I have 13.2.2 but excited about our future!
FSD 13.3 or 13.4
- Thread starter 2024YSATX
- Start date
Seems like there was a lot of pressure to get v13 features out with the holiday update, and now that 12.6.1 is out for HW3, Tesla isn't telling us what's coming next or when. Perhaps the FSD team got a week off.
It could be that the next 'big' release will be getting the Cybertruck up to true v13.2.2 speed, with reverse capability, etc.
Aside from the 'coming features' in the release notes, I have no idea what Tesla is working on for the next release, although I assume it has to do with parking.
It could be that the next 'big' release will be getting the Cybertruck up to true v13.2.2 speed, with reverse capability, etc.
Aside from the 'coming features' in the release notes, I have no idea what Tesla is working on for the next release, although I assume it has to do with parking.
Looks like 13.2.6 went out to quite a few cars and now has slowed down to a stop. Maybe that means a decent jump is in the works ‘soon’ ? 13.2.2 was very good today going around a car being towed on a narrow curvy mountain road with a double yellow. It seemed to understand that we were being waved along into the oncoming traffic lane, did so very naturally
. That almost makes up for it trying to make a left out of a parking lot exit that was right turn only every time.
. That almost makes up for it trying to make a left out of a parking lot exit that was right turn only every time.I'll speculate. It's been since September since newer features of V13 were last updated in the release notes. The fact that we not only haven't seen these updates (3x this and that), we haven't even heard a whisper about them. I think we have a combination of issues.
First, the team has been focused on other FSD promises since September, including foreign country releases, and then robotaxis. There must be a ton of enforced learning there that will have to benefit the rest of us when we get it.
Second, that last part of V13 sounds like it was tougher than expected - save the hardest parts for last. I think parking lots alone and parking in general is probably a lot harder that driving on the highway. This part doesn't really stress the vision system or need speedy reflexes. It needs a competent planner, which it obviously lacks in 13.2.8. I still don't know how they do neural net training on a particular task (like mastering parking lots), but perhaps their initial plan didn't get the results needed.
That said, what's next for us? I think v13.3 - or jump to v14.x. If the robotaxi work pays off, perhaps jump right to it.
First, the team has been focused on other FSD promises since September, including foreign country releases, and then robotaxis. There must be a ton of enforced learning there that will have to benefit the rest of us when we get it.
Second, that last part of V13 sounds like it was tougher than expected - save the hardest parts for last. I think parking lots alone and parking in general is probably a lot harder that driving on the highway. This part doesn't really stress the vision system or need speedy reflexes. It needs a competent planner, which it obviously lacks in 13.2.8. I still don't know how they do neural net training on a particular task (like mastering parking lots), but perhaps their initial plan didn't get the results needed.
That said, what's next for us? I think v13.3 - or jump to v14.x. If the robotaxi work pays off, perhaps jump right to it.
I think it is about time we get another FSD version, but being that we are only weeks away from June. The more likely scenario is that we get a v14 or a newer version of v13 with the initial RoboTaxi release. That is, if they even release it right away to the public, and if they don't also simultaneously release both a supervised and unsupervised version with different pricing than currently. The future isn't looking bright for FSD for consumers.
I'm thinking V13.x comes out with much improved and extensive end-of-drive performance. Basically finish the roadmap, and layer on the user interface for a robotaxi, if making a car into a robotaxi. Add more features to the performance we already have.
I think if Tesla makes significant changes to the planner, as in it starts to understand traffic patterns - THAT would be v14. Something just plain smarter.
With Model Y robotaxis, I do wonder a lot how Tesla will handle people who don't know how to open the doors. Plus you cannot see the door handles at night. Same wonder about opening the trunk for luggage. I don't think driving will be much of an issue, if Tesla can master parking and dropping off.
I think if Tesla makes significant changes to the planner, as in it starts to understand traffic patterns - THAT would be v14. Something just plain smarter.
With Model Y robotaxis, I do wonder a lot how Tesla will handle people who don't know how to open the doors. Plus you cannot see the door handles at night. Same wonder about opening the trunk for luggage. I don't think driving will be much of an issue, if Tesla can master parking and dropping off.
I'm hopeful but skeptical, especially about timeline or timeframe rather. I think we will still be waiting a long while for a new FSD version, especially because the tests they have conducted so far are in fairly controlled environments with a fairly small number of cars (at least to my knowledge). And seeing the recent posts about them possibly wanting to raise the price of FSD, it also has me concerned on whether it will be viable for me to continue subscribing.
I do agree, that whatever version they release next will likely be primarily focused on end-of-drive performance, but I also don't think that it will have many more of the promised features. Hard to tell with Tesla, but they tend to take their time, which I appreciate from a safety standpoint, but hate from an enthusiast standpoint.
My main question, however, is how will they handle poor weather conditions? Will they just stop operating the network if it becomes too hard for the car to see? Will any operating vehicle just pull over on the side of the road with the passengers still inside? How do they plan to address different driving surfaces like snow, ice, gravel, dirt, etc.?
My experience in poor weather is that it works just as well if not better than I did but that remains to be seen.
Not to change the subject but- what's happening with the Tesla-Fi stats listed on this site? Looks like 80%+ is only 800 or so cars now when it used to be thousands.
Not to change the subject but- what's happening with the Tesla-Fi stats listed on this site? Looks like 80%+ is only 800 or so cars now when it used to be thousands.
Remember so very long ago that Elon showed a clip on X of our first look at end-end neural net V12 FSD? That was about 2 years ago. Now we've had v12 without NN highways, then V12 with highways, park assist, and ASS. Then we (HW4 owners) got V13. I suspect by July of this year we'll be on V14.
As for testing in a controlled environment, there's nothing really controlled when you are out among regular car drivers. I suspect the extensive mapping and testing is to a) solve that destination-end-point (parking, dropoff) problem, b) work out the user interface, c) work out the back end of the user interface and all the fleet management stuff, d) determine if there are roads the cars should NOT go down, for whatever reason, e) find the repeating edge cases and fix them (no turn on red, oddball lane usage/markings, and be sure some of the known edge cases (no turn on red, school zones, school busses, etc) are fixed.
Why do people think this will take a year just because they went 2 months without an FSD version change? Consider how much parking lot data has to be collected to produce training videos. But once Tesla has come up with the NN model for parking, they simply have to feed that into the training system and refine the result.
I do think FSD updates in the US have been delayed a bit while 13.2.x is rolled out in other countries. (Can't wait to see right hand drive FSD.)
Poor weather? It's either drivable (slower) or not. Tesla starts, like Waymo, in places in the southwest where there is no snow, not much rain and lots of sunny days. If the weather there gets really bad, like if a haboob (dust storm) is blowing through that temporary holds/delays are put on the business in the area. Just like airlines delay flights due to bad weather.
Does that mean eventually Teslas can reliably drive though the winters in Fairbank? Probably, but I don't know. I do think eventually FSD will drive better than a human in any weather condition.
As for testing in a controlled environment, there's nothing really controlled when you are out among regular car drivers. I suspect the extensive mapping and testing is to a) solve that destination-end-point (parking, dropoff) problem, b) work out the user interface, c) work out the back end of the user interface and all the fleet management stuff, d) determine if there are roads the cars should NOT go down, for whatever reason, e) find the repeating edge cases and fix them (no turn on red, oddball lane usage/markings, and be sure some of the known edge cases (no turn on red, school zones, school busses, etc) are fixed.
Why do people think this will take a year just because they went 2 months without an FSD version change? Consider how much parking lot data has to be collected to produce training videos. But once Tesla has come up with the NN model for parking, they simply have to feed that into the training system and refine the result.
I do think FSD updates in the US have been delayed a bit while 13.2.x is rolled out in other countries. (Can't wait to see right hand drive FSD.)
Poor weather? It's either drivable (slower) or not. Tesla starts, like Waymo, in places in the southwest where there is no snow, not much rain and lots of sunny days. If the weather there gets really bad, like if a haboob (dust storm) is blowing through that temporary holds/delays are put on the business in the area. Just like airlines delay flights due to bad weather.
Does that mean eventually Teslas can reliably drive though the winters in Fairbank? Probably, but I don't know. I do think eventually FSD will drive better than a human in any weather condition.
I don't think it's people thinking it is going to take a year. It is more to do with Tesla (more on Elon's side) of having a repeated track record of extremely optimistic timelines, and a few times that they have overpromised and underdelivered. Call it pessimism if you will.
It is obvious that a new version will come at some point. But, because it has been so long between updates, the expectations for the next update and what it should deliver in terms of performance and improvement are already very high. And those expectations will only continue to climb as the time between updates widens. You can't really blame people for those high expectations, either.
That is the most reasonable approach. However, the confidence in just how bad the weather can get before that happens is what is being questioned, more so than what happens when it is already bad. Again, if the car is currently on a route, what happens then? Does it just pull over and let the client out? Does it finish the route? Does it delay it? Some of these aren't really optimal, and could lead to a lot of unhappy customers.
Either way, we know FSD is still being worked on and that a new update is likely underway, but the expectations are high because of the delay in even incremental improvements and lack of communication on this front for regular FSD users.