Its now been ~2 months since the last major Autopilot autonomy update

[mitchellh3]

No real purpose for this post other than to highlight this.

• Its been almost 2 months since 17.26.76 which primarily added perpendicular parking.
• Its been 3 months since 17.17.4 which brought max speeds up to AP1.
• I'm not counting "silky smooth" since it wasn't in update notes and feedback seems hit or miss.
• Yes, we got 17.28.wtf a month ago but no major updates, and nothing of value in the release notes.

Other Tesla, Inc. factors:

• The Model 3 is now in production.
• Q2 is closed, reported, and Q3 is under way.
• Elon's twitter has been quiet on AP.

When is the next major autonomy update? (rhetorical)

Given the length of time since the last one, do we think they're now working on 8.2 rather than incremental 8.1 updates? There is much to be desired still on AP2 stability and vision capabilities (its still sketchy for me in a handful of clearly marked CA highways).

I'm not trying to complain. I love my car. Just making a note.

 

[calisnow]

Who the heck knows at this point...

 

[MXWing]

Probable that any resource that can assist on Model 3 has been diverted to those efforts.

What we see released to our cars is probably months behind what they are working on. They have to find a build that they want to lock on and validate that across X number of miles. I would suspect a new build has been decided upon next for release but still undergoing validation.

 

[AnxietyRanger]

Good question, who knows. And where is that December 2016 "8.1" (now 8.2?)... the one with the new browser and AP1 ramp to ramp or whatever it was...

On a related note, I think Elon's "silky smooth" episode was kind of telling how Tesla operates. Elon of course is said to run the latest development release in his own car (how latest? and anyway isn't he chauffered these days? anyway...) and was gushing the silky smooth and its imminent release on Twitter.

Yet what was released right after that obviously did not yet have the smoother steering control algorithm, which came later. (And IMO constitutes as a fairly major feature change, even though the actual road vision still has its reported hit and misses. They obviously replaced the AP2 ping-pong steering code with something new.)

So we kind of get a sense that Tesla has various revisions and possibly potential release candidates at various stages of readiness, various places in the development and validation pipelines. What gets released and when can change of course, but in any case the releases are probably never showing the latest developments - just one of the development points that 1) matured for release and 2) was chosen for release.

Elon - to use him as an example - is probably always driving something completely different that is shipping to cars.

 

[mrkisskiss]

We actually haven't had any major autonomy updates since, I guess, January 2016.

AFAIK, that was the 7.1 release that added summon, improved auto steer, auto parallel parking etc etc...

Since then, I don't think they've actually added any major autonomy features... They've basically just been fine tuning existing AP1 features, and trying to rebuild the exact same feature set on HW2 (which is still missing rain sensing for everything, and even auto high beams outside of the US).

As far as I can tell, it's been more than 18 months since we've seen any new autonomy capabilities from any Tesla car.

 

[AceSkywalker]

Typical Elon Standard Time. Musk should just be like the studio behind the upcoming video game "The Artful Escape" with their launch date.

1:09 if you're impatient

In my opinion the "near" future of not owning a vehicle and being driven around by autonomous taxis for cheaper than the TCO of a personal vehicle is MUCH further than predictions I've seen from fans of autonomous car technology.

 

[lux_cars]

In my opinion the "near" future of not owning a vehicle and being driven around by autonomous taxis for cheaper than the TCO of a personal vehicle is MUCH further than predictions I've seen from fans of autonomous car technology.

Given the slow pace of progress with AP2, my fear is that Tesla is actually further behind on autonomy than other companies and may not even get there by the time the others, who don't set Elon standard time deadlines, do in 2019/2020. Esp. with current sensor suite and without the help of third party such as Mobileye.

 

[Bebop]

I predict some major updates in September/October. Based on the previous software release history, it seems like we are due for a 9.0 in that time frame. Next 1-2 months. Like it's been kind of dead for a while and all of a sudden a bag of bricks to the face will happen in updates. It will be unleashed in a huge horde is my guess.

 

[AnxietyRanger]

Given the slow pace of progress with AP2, my fear is that Tesla is actually further behind on autonomy than other companies and may not even get there by the time the others, who don't set Elon standard time deadlines, do in 2019/2020. Esp. with current sensor suite and without the help of third party such as Mobileye.

There has been a growing, nagging feeling similar to this in my head, given how AP2 has unfolded thus far. That said, we have not seen where the rumored FSD branch currently is.

What I do think set's Tesla's goal apart, though, is doing it all in vision only basically (while others are gunning for camera/lidar/radar three-way fusion). That is basically a somewhat easier task, I think. (Not commenting on reliability, just the easiness of getting to basically work.)

So, I still think they have a chance at getting that done at a comparable pace to the rest of the industry (or even faster) - whether or not that is sufficient for regulators may be market dependent and different, of course.

What I do think is the biggest hurdle are the rules of the road type of stuff. Has Tesla got all that figured out already sufficiently...?

The second hurdle, of course, is the question is the suite sufficient to do FSD in the first place.

 

[Bebop]

The second hurdle, of course, is the question is the suite sufficient to do FSD in the first place.

Let's assume it is because otherwise, Tesla will be in a heap of trouble for those who bought an EAP/FSD car being advertised on the custom order page (even today) as being able to drive itself and then to realize the hardware is not enough. Major problems will happen and I expect many many class action lawsuits. I highly doubt Tesla wants that to happen and so I trust everything is okay.

 

[AnxietyRanger]

Let's assume it is because otherwise, Tesla will be in a heap of trouble for those who bought an EAP/FSD car being advertised on the custom order page (even today) as being able to drive itself and then to realize the hardware is not enough. Major problems will happen and I expect many many class action lawsuits. I highly doubt Tesla wants that to happen and so I trust everything is okay.

I do believe Tesla (or at the very least Elon believes it to be sufficient. What I'm not so sure about is what will unfold once all that (still on-going) development work is completed and they can really see for themselves... not to mention when that meets the regulators.

Tesla of course has disclaimers about the latter. What they don't have disclaimers for is if the suite is not sufficient at all (i.e. they can't even deliver Level 2 full self-driving with the driver still responsible).

For me the biggest question mark is bad weather. Mere fixed cameras without anything more than heat to protect them seems like a bad recipe for rain and snow, but we shall see...

 

[cwerdna]

Given the slow pace of progress with AP2, my fear is that Tesla is actually further behind on autonomy than other companies

Fear? It's almost a certainty they're way behind at least Waymo (current name for Google's effort that has been going on for many years) if not several others in this space.

Have you looked at the California DMV disengagement reports?
Autonomous Vehicle Disengagement Reports 2015
Autonomous Vehicle Disengagement Reports 2016

Will be interested to see what each company submits once 2017 is over. It's been discussed at 2016 California Autonomous vehicle disengagement reports and numerous other threads.

Background info at Autonomous Vehicles in California and Testing of Autonomous Vehicles.

Google released this video in 2012:

. Google's Waymo just started this earlier this year: Google's Waymo gives free self-driving car rides in Phoenix.

Nissan released Autonomous Nissan Leaf Video recently. It also had a pointer to GM's efforts further down (wish we had some details on those videos).

Baidu (Chinese search giant) released

in Dec 2015.

admittedly isn't that impressive but at least they took a reporter along for the ride.

 

[Bladerskb]

Will be interested to see what each company submits once 2017 is over. It's been discussed at 2016 California Autonomous vehicle disengagement reports and numerous other threads.

Easy. Tesla will accumulate a bunch of miles in highway situation (which has already been solved by others) so they can claim they are at 1 in 10,000 miles disengagements give or take.
My bet is on them not sharing the road type of where the miles occurred on spurring news article and coverage from naive journalists proclaiming Tesla the leader of FSD tech.

 

[cwerdna]

spurring news article and coverage from naive journalists proclaiming Tesla the leader of FSD tech.

And, unfortunately, many in the general public and those that follow Tesla only a surface level also have this belief.

I've heard the same the same perceptions and beliefs from a few coworkers and friends of mine. For one friend, I pointed him to a lot of the same things I just posted (+ in 1 of the threads). That changed his mind. Need to send those coworkers the same thing. I've only mentioned it to them, so far.

Gotta love how Tesla's all about hype and often confusion/steering perceptions.

 

[alcibiades]

I recently was at a party with a VP from Here. I overheard him laughing about how f'ed Tesla was, telling someone that there was no way Tesla'd be able to hit L5 with current set-up.

 

[Bebop]

I recently was at a party with a VP from Here. I overheard him laughing about how f'ed Tesla was, telling someone that there was no way Tesla'd be able to hit L5 with current set-up.

VP of what? Is he close within Elon's executive circle?

 

[182RG]

VP of what? Is he close within Elon's executive circle?

HERE Data-Driven Location Solutions: Real-Time Location Services

 

[lux_cars]

HERE Data-Driven Location Solutions: Real-Time Location Services

I think he might be asking for what function/group is this individual the VP at Here.

 

[ItsNotAboutTheMoney]

Fear? It's almost a certainty they're way behind at least Waymo (current name for Google's effort that has been going on for many years) if not several others in this space.

Well, yes and no.

Tesla's system is far less capable than several other companies' systems. However, those other companies' systems have approximately $0 in sales.

 

[alcibiades]

Well, yes and no.

Tesla's system is far less capable than several other companies' systems. However, those other companies' systems have approximately $0 in sales.

If you want the argument to be that Tesla has the best Ponzi scheme, then I guess you have a point. Google has yet to try to monetize Waymo. GM has yet to try to monetize Cruise.

That said, both of those systems are driving people around.

My Tesla breaks for overpasses and ghosts (bumper sticker idea!), and it has a regressive algorithm for lane identification. So yes, I'm a sucker. Congrats to Tesla. I gave them $.