Blog Musk Touts ‘Quantum Leap” in Full Self-Driving Performance

A “quantum leap” improvement is coming to Tesla’s Autopilot software in six to 10 weeks, Chief Executive Elon Musk said a tweet.

Musk called the new software a “fundamental architectural rewrite, not an incremental tweak.”

https://twitter.com/x/status/1294374864657162240

Musk said his personal car is running a “bleeding edge alpha build” of the software, which he also mentioned during Tesla’s Q2 earnings. 

“So it’s almost getting to the point where I can go from my house to work with no interventions, despite going through construction and widely varying situations,” Musk said on the earnings call. “So this is why I am very confident about full self-driving functionality being complete by the end of this year, is because I’m literally driving it.”

Tesla’s Full Self-Driving software has been slow to roll out against the company’s promises. Musk previously said a Tesla would drive from Los Angeles to New York using the Full Self Driving feature by the end of 2019. The company didn’t meet that goal. So, it will be interesting to see the state of Autopilot at the end of 2020.

 

[FringeGlock]

I am suppressing my "inner fanboy" because if we set expectations too high, we are likely to be disappointed. Let's not forget that Elon called Smart Summon "magical" and said it would "blow our minds". And we all know Smart Summon fell way short of that when it was first released. Elon is great at hype. And he admits that he cannot quite do his work-home commute free of disengagements so we know the rewrite won't be able to handle everything. My guess is that the rewrite will be a big improvement over what we have now and it will do turns at intersections. So some simple commutes will have zero disengagements. But I suspect there will be plenty of situations that the rewrite cannot handle.

Oh, yes! I paid $5000 for updates when I bought the car. It is a fantastic automobile! Every car should drive like this one. However, this auto pilot needs some very serious engineering and thoughts. I have engaged the auto pilot several times, with sometimes disturbing or hair raising reactions. The first one was on a two lane road with a 45 MPH speed limit posted. The machine, the car, thought it was much higher and headed for that speed. This was scary. It may help to know that the back roads in our area are winding and undulating with NO shoulders. On another occasion, as the vehicle was approaching a curve, it started to cross the yellow double center line. Good thing no opposing traffic was coming. With autopilot and navigation engaged, approaching a right turn, off divided four lane US 19E, going south into Elizabethton, Tennessee, the vehicle made no effort to slow down for the right turn. That was hair raising. I hit the brake pedal.
Using only the speed control on the back roads is another weird issue. As the road goes up and down and curves left and right, the vehicle acts as if it has a serious lack of brains, braking hard when it "sees" the road going up, again braking hard at curves, actually too late. I had just finished reading The Robots of Dawn by Isaac Asimov. We have a very long way to go.

 

[Silicon Desert]

However, this auto pilot needs some very serious engineering and thoughts. I have engaged the auto pilot several times, with sometimes disturbing or hair raising reactions. The first one was on a two lane road with a 45 MPH speed limit posted. The machine, the car, thought it was much higher and headed for that speed. This was scary.

In this particular situation, you might want to pay closer attention to the display BEFORE putting into AP. It will show you what it thinks the speed limit is in that area, thus, the speed it will attempt when going into ap. And if you changed the default settings to allow going faster than the speed limit, then it will do that too.

 

[Silicon Desert]

Is there anyone who can honestly say they'd rather go back to ICE?

Well yes, me. BUT I only drive it in extremely remote areas where there is practically no charging. And I have a bigger fuel tank in the ice for that reason. So I rather use the ice for now until I can get decent fast charging in all the places I want to go. Hopefully soon

 

[Silicon Desert]

A "fundamental architectural rewrite" sounds like new code with new bugs we've never seen before.
Watch your car CAREFULLY when it is on AutoPilot for the first few months!

yea, typically what happens. That is why I dare not be one of the first people to download and use it. I'll let the folks whom can't resist it to test it for me for a few months before I consider trying it and reading people's notes.

 

[Redbrick]

I love my Model S and all t can do but lets be realistic. Elon has made many, many wild claims about Full Self Driving being just around the corner. In the mean time development has moved forward at a much more realistic pace and Elon's claims have, time and time agian, proven to be wildly optimistic to the point of irresponsibility. If we ever do get to a point where Full Self Driving is a reality it is extremely unlikely to be with the hardware we have right now either because it is not capable or because it is so old it has reached the end of its useful life.

It is often touted that the life span of a Tesla is very much more than an Internal Combustion Car but the reality is that most often the reason why a traditional car reaches the end of its life is not because of the engine but because of all the rest of the systems which are either exactly the same or at least very similar to a Tesla. Wheel bearings disintegrate, suspension collapses and the interiors fall apart long before the engine and transmission dies on a gasoline car.

Full self driving is an extremely difficult problem to crack and even though Tesla may have some fo the best engineers on the planet working on it they are not even close to having something ready for the world at large. Smart summon simply doesn't work in the vast majority of situations and autopilot is like a nervous learner, not even close to the most mediocre human driver. It is an expensive gimmick that is nice to use to take some of the stress out of a journey but is years from being able to drive without constant supervision. Auto park is not even worth playing with. If you cannot park substantially better than a Tesla in auto park mode then you should not be driving at all.

Elon has to keep making these wild claims and predictions, which probably, hopefully, make him a little embarrassed. It is a requirement of running the Tesla organisation that he keeps the hype alive to retain enthusiasm from owners and prospective customers but his completely unrealistic time lines have proven to be ridiculously over optimistic for as long as Tesla has been a company.

I had a car ..the engine and it various auxillary components failed first....got rid of it.

 

[Redbrick]

The only thing that matters to whom? Progress doesn't mean much really if you're nowhere near. Think about this, I shoot model rockets into the sky with my kids sometimes. Last month we shot one higher than ever, close to 300 ft probably. I am making progress towards the moon since I did get closer than ever before, right? And I have no competitors either trying to shoot a $25 rocket to the moon either. All that matters is progress, right? Want to invest?

Oh, you'll say you need more expensive rockets to get to the moon and I should be considering Space X a competitor because they are closer? Ok, then Tesla should consider Waymo a competitor. Waymo self driving (actually self driving without a driver, not Tesla FSD) is not close to Tesla, they left them in the dust long ago. Before you come back saying Waymo is not selling their cars to consumer, well SpaceX is not selling their rockets to consumers either. The rockets I shoot are available in hobby stores all over the world. How much can I put you down for a consumer $25 rocket to moon company? I'm pretty sure I can get even closer by shooting them higher than 300 ft off the ground, so there will be progress which is all that matters to you, right?

I stand by what I said 4 years ago, when people like yourself were telling me how FSD2.0 will drive kids to school, be able to be summoned across the country, etc - it will never happen. None of the cars Tesla sold the original FSD (before redefining it to consist primarily of features sold as EAP before in order to deliver FSD Texas sharpshooter style) will never happen via OTA or any free upgrades (since the FSD price supposedly included all the hardware that the car will EVER need to be fully self driving). Not that Tesla does thins, but even if they did, being able to handle even 60% of situations on the road means you're about 1% there to actual full self driving.

I wouldn't doubt spaceX starts selling rockets soon...wonder what letter designation that would come with?

 

[KLee]

A “quantum leap” improvement is coming to Tesla’s Autopilot software in six to 10 weeks, Chief Executive Elon Musk said a tweet. Musk called the new software a “fundamental architectural rewrite, not an incremental tweak.” The FSD improvement will come as a quantum leap, because it’s a fundamental architectural rewrite, not an incremental tweak. I drive...
[WPURI="https://teslamotorsclub.com/blog/2020/08/15/musk-touts-quantum-leap-in-full-self-driving-performance/"]READ FULL ARTICLE[/WPURI]

Check out the video on the Topic. I think its really good information on autopilot history and 2.5d to 4d. What do you guys think?

Tesla Autopilot ReWrite 4D! & 3D labeling & Competition
https://youtu.be/kLukc3AtO-8

 

[Stratman]

I am suppressing my "inner fanboy" because if we set expectations too high, we are likely to be disappointed. Let's not forget that Elon called Smart Summon "magical" and said it would "blow our minds". And we all know Smart Summon fell way short of that when it was first released. Elon is great at hype. And he admits that he cannot quite do his work-home commute free of disengagements so we know the rewrite won't be able to handle everything. My guess is that the rewrite will be a big improvement over what we have now and it will do turns at intersections. So some simple commutes will have zero disengagements. But I suspect there will be plenty of situations that the rewrite cannot handle.

Suppressing inner fanboy too and and a bit of historical reality.., Agree, "smart" summon is dodgy at best. Once the rewritten "FSD" is released on mass, it will have more bugs than a locust convention. Over time , no doubt it will be improved, but essentially the whole optimisation process starts all over again and it will need HW4 or 5...6

 

[goRt]

I was driving my wife's S recently and watching the AP2.5 display of what cars it sees. It cannot even reliably track cars going in the opposite directions - most of the time I would see them just fine, passing my car, but a significant number of times I saw the cars just disappear and appear back once they passed my car. Visibility was perfect, no blinding sun or rain or darkness. It doesn't see cross traffic almost ever either. So no, Tesla AP is nowhere near what you describe.

PS> My AP1 car still drives better on AP on the highway (lane keeping basically) than my wife's AP2.5 - so many phantom braking events on the AP2.5 that I turned it off not to annoy drivers around or cause an accident. My wife keeps AP disabled completely after trying it for a week - she does not like it at all. An no, the phantom brakes were not slamming on brakes kind of event causing pile-ups, but completely unnecessary slow downs when the road curves or cars are near. So, maybe the new rewrite will at least match the AP1 quality for simple lane keeping? Not holding my breath though.

You're confusing the delay in displaying in the display with what the car's actually processing.

 

[Thumpernator]

Hell, I just want them to fix the "phantom braking" issue and the amount the car slows down for curves. It's beyond ridiculous.

 

[nukequazar]

I had a car ..the engine and it various auxillary components failed first....got rid of it.

What kind of car was that?!?

 

[nukequazar]

I love my Model S and all t can do but lets be realistic. Elon has made many, many wild claims about Full Self Driving being just around the corner. In the mean time development has moved forward at a much more realistic pace and Elon's claims have, time and time agian, proven to be wildly optimistic to the point of irresponsibility. If we ever do get to a point where Full Self Driving is a reality it is extremely unlikely to be with the hardware we have right now either because it is not capable or because it is so old it has reached the end of its useful life.

It is often touted that the life span of a Tesla is very much more than an Internal Combustion Car but the reality is that most often the reason why a traditional car reaches the end of its life is not because of the engine but because of all the rest of the systems which are either exactly the same or at least very similar to a Tesla. Wheel bearings disintegrate, suspension collapses and the interiors fall apart long before the engine and transmission dies on a gasoline car.

Full self driving is an extremely difficult problem to crack and even though Tesla may have some fo the best engineers on the planet working on it they are not even close to having something ready for the world at large. Smart summon simply doesn't work in the vast majority of situations and autopilot is like a nervous learner, not even close to the most mediocre human driver. It is an expensive gimmick that is nice to use to take some of the stress out of a journey but is years from being able to drive without constant supervision. Auto park is not even worth playing with. If you cannot park substantially better than a Tesla in auto park mode then you should not be driving at all.

Elon has to keep making these wild claims and predictions, which probably, hopefully, make him a little embarrassed. It is a requirement of running the Tesla organisation that he keeps the hype alive to retain enthusiasm from owners and prospective customers but his completely unrealistic time lines have proven to be ridiculously over optimistic for as long as Tesla has been a company.

These are not wildly optimistic claims, they are lies. These cars with current sensor hardware and available computing power will never get past Level 2 Autonomy. They would need LIDAR and constantly updating 3D HD maps to get to "full self driving."

 

[S0XY]

These cars with current sensor hardware and available computing power will never get past Level 2 Autonomy. They would need LIDAR and constantly updating 3D HD maps to get to "full self driving."

Prove it

 

[DanCar]

These are not wildly optimistic claims, they are lies. These cars with current sensor hardware and available computing power will never get past Level 2 Autonomy. They would need LIDAR and constantly updating 3D HD maps to get to "full self driving."

Agree with the lies part. Do agree that improved sensors will eventually be needed. This could be more and different cameras. Don't agree with the rest. I'm hopeful Tesla will be able to do level 3 in stop and go traffic on limited access freeway with current sensor suite.

 

[diplomat33]

Prove it

Companies using lidar and HD maps who have FSD:
Waymo
Cruise
Aurora
Mobileye
Baidu
Yandex
Aptiv

Just to name a few....

Companies not using lidar or HD maps who have no FSD:
Tesla

 

[diplomat33]

If you are hoping Dojo will be a magic bullet that will allow Tesla to "solve FSD", I am afraid you are going to have to wait a bit longer. Elon says Dojo V1.0 is a year away still:

 

[DanCar]

Companies using lidar and HD maps who have FSD:
Waymo
Cruise
Aurora
Mobileye
Baidu
Yandex
Aptiv

Just to name a few....

Companies not using lidar or HD maps who have no FSD:
Tesla

Very early in the driverless game. No winner yet. Some of us predict the no lidar camp will win. To give an example of the folly of the masses: No one was doing reusable rockets, until SpaceX came around. Now everyone is talking about reusable rockets. It took 15 years for SpaceX to make its point. So perhaps by 2030 we will see that Lidar is helpful, but not absolutely necessary.

 

[diplomat33]

Very early in the driverless game. No winner yet. Some of us predict the no lidar camp will win. To give an example of the folly of the masses: No one was doing reusable rockets, until SpaceX came around, now everyone is talking about reusable rockets. It took 15 years for SpaceX to make its point. So perhaps by 2030 we will see that Lidar is helpful, but not absolutely necessary.

I agree that it is early still and that there is no definitive winner yet. I am just pointing out that right now, the "lidar camp" appears to be in the lead since they have demonstrated the most capabilities in terms of FSD.

I am not sure your reusable rockets is a valid analogy. Reusable rockets required perfecting existing technology. But in the case of FSD, we have companies that are already doing FSD. And with companies like Waymo and Cruise, we already know that FSD is possible with cameras, radar, lidar and HD maps. The question is whether you can do FSD with less hardware. So a better analogy might be can we do the same reusable rockets but with a lighter rocket that has less hardware than others are using.

 

[MXLRplus]

Eh? How do you think the Space Shuttle worked?
The entire concept of the Space Shuttle is based on reuse.
Even Russia tried to copy it.

 

[mikes_fsd]

Very early in the driverless game. No winner yet. Some of us predict the no lidar camp will win. To give an example of the folly of the masses: No one was doing reusable rockets, until SpaceX came around. Now everyone is talking about reusable rockets. It took 15 years for SpaceX to make its point. So perhaps by 2030 we will see that Lidar is helpful, but not absolutely necessary.

I think we will know in about 24 to 36 months whether or not Lidar is helpful needed or not. (could be less than that, but given "Elon time" I will stick with this estimate)