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.

 

[Bladerskb]

From where I stand, MobileEye got a glimpse of Tesla's internal development while still working on AP1 stuff and freaked out and copied Tesla's sensor suite verbatim.
But when it came time couldn't provide the sufficient processing power to use the suite (a la custom chip at Tesla).

Is that why Mobileye had dozens of automakers testing out their 8 camera and 3 camera trifocal system in 2014?

Blog - Musk Touts 'Quantum Leap" in Full Self-Driving Performance

My my, mobileye must be the ultimate espionage. After tesla came begging them for their tech (for some weird reason), they stole Tesla's tech, went back in time with their secret time machine and patented it all and started marketing it to other automakers as their own.

Either that or you are...

 

[mikes_fsd]

Given the videos and the progression from .10 to .11 and now to .12 with the updates rolling out and with UI enhancements, here is my speculation and prediction:
Thoughts, observations and speculation:

1. Elon is sandbagging FSD Beta - a little - out of the gate - speculation: to stretch the wow factor for a longer time.
2. The updates have shown direct correlation to actual user issues (from .10 to .11 Kim in Rhode Island first, was NOT able to go through a round about after the update was able to do them consistently and pretty well.
3. Same with watching the new videos from .11 to .12 the left hand turns seem more confident than they were before.
4. The fact that updates are rolling out and @verygreen said that they have to reload firmware to update NN's (or at least that was the case) I think part of the rewrite was to allow much greater flexibility on what you can tweak in the stack without loading a whole new version of the NN's

Then you add in the recent tweets:
https://twitter.com/elonmusk/status/1322431151986925568
https://twitter.com/elonmusk/status/1322433989261611009
https://twitter.com/elonmusk/status/1322435476280758272
https://twitter.com/elonmusk/status/1322435853348671488

Prediction:
Game over for the "competition".
This will be evident in months not years.
Tesla has found the solution/approach that will solve the FSD problem with real world as the proving grounds, in real users hands. Kind of like building the machine that builds the machine, but for NN building and AI training.

As @powertoold said, this is contingent on Tesla rolling FSD Beta out to wide release!

 

[powertoold]

Game over for the "competition".
This will be evident in months not years.
Tesla has found the solution/approach that will solve the FSD problem with real world as the proving grounds, in real users hands.

Yup, unless Tesla is faking the improvements by releasing a crippled beta, it's game over. If they're actually able to improve the beta by using real world tester data this quickly, they've got an agile and effective stack that will hopefully allow them to widely deploy the FSD beta to all cars. Once the FSD beta is widely deployed, it's game over for sure. Tesla can't and won't widely release this FSD beta until it's at least as good as an average human. There's no reason to, as the liability would be greater than any FSD revenue benefit.

Tldr: if Tesla is *ever* able to widely release the FSD beta, it's game over for any competition.

 

[DanCar]

The game over discussion is funny. The pie is very big. There will be several winners. But like I've said for more than a year, Tesla is leading the pack. I hope Mobileye can get level 3-4 deployed soon and I wish the same for everyone else. Competition is good.

 

[powertoold]

The game over discussion is funny. The pie is very big. There will be several winners. But like I've said for more than a year, Tesla is leading the pack. I hope Mobileye can get level 3-4 deployed soon and I wish the same for everyone else. Competition is good.

True, but Karpathy and Elon consistently emphasize that any improvement in Autopilot, even minor, is a huge deal because it's country / continent / worldwide. That's a big deal. If Tesla widely releases FSD beta, likely for all USA owners, then consider the implications of that confidence... along with how many features are packed into the FSD beta.

 

[mikes_fsd]

The game over discussion is funny. The pie is very big. There will be several winners. But like I've said for more than a year, Tesla is leading the pack. I hope Mobileye can get level 3-4 deployed soon and I wish the same for everyone else. Competition is good.

I think anyone dependent on HD Maps is done for without a "fundamental rewrite of their system" (note: not green's definition of "HD Maps" but Karpathy's definition of it)

But I wholeheartedly agree on the need for competition.
We need a minimum of 2 more real competitors to Tesla's solution.
MobileEye is a possibility.
Maybe someone like Comma AI - that thinks way outside of the box - is a possibility as well (but that's a stretch).

Who else is there? That will not be half a decade (or more) behind at this point?

 

[powertoold]

But I wholeheartedly agree on the need for competition.
We need a minimum of 2 more real competitors to Tesla's solution.

Sadly, I think Tesla will be so far ahead, no one will catch up. Not only does Tesla have a giant fleet, but they have a fleet of future-proof cars with incredible safety record and performance. They have everything, from the mfging capability, to the software, and the engineers. It's mind-boggling what they're doing. Who can stop them?

 

[masterxel]

To put it a different way, a successful global release of FSD could force a change in strategy from other competitors. Honestly reading the sudden rush of all the other self driving car articles posted recently alongside the FSD beta announcement, I feel like it already has.

 

[mikes_fsd]

Sadly, I think Tesla will be so far ahead, no one will catch up. Not only does Tesla have a giant fleet, but they have a fleet of future-proof cars with incredible safety record and performance. They have everything, from the mfging capability, to the software, and the engineers. It's mind-boggling what they're doing. Who can stop them?

You and I are on the same page.

But I do think - just like Elon's comment about California getting complacent because of winning for too long - if you have no competition to egg each other on, things just naturally slow down in advancement!

Look at the space race!

 

[Daniel in SD]

It's 100% game over if Tesla achieves better than human performance on HW3 vehicles.

More likely scenario is that "regulatory issues" prevent or quickly cancel the release of beta city FSD to the whole fleet. Then Elon and his disciples will claim that "regulatory issues" are preventing the gathering of the data necessary for exponential improvement. He will move the testing to Mars which will not be subject to Earth's laws. FSD on Mars will be far safer than human drivers on Earth and the promise of FSD will be fulfilled.
Another likely scenario is that it turns out beta city FSD is perfectly safe because it remains bad enough to keep testers extremely vigilant and most people won't use it (like Smart Summon).

The suspense is killing me! How much does it cost to ship a Model 3 to Mars? (looks like about $10 million on a Falcon Heavy, making the $10k price for FSD really look like a bargain!)

 

[tmoz]

Upping the price to $10K for the feature - I would think this only happened when Tesla became very confident in its performance.
It would be nice if there were a price for robo taxis and a price for the non-commercial use of the technology.
My model S is a 2015 and at the time of purchase, there was talk of the model S going from LA to NY with no driver, at least from the sales person who sat in the car with me on one of my many test drives (5 or 6 total). It didn't have auto lane change at the time, but got it shortly after purchase, so I didn't know how many more features of auto pilot I was in store for.

My next purchase from Tesla will be either the Cyber truck (or model Y if my 2012 plug-in Prius bites the dust before the cyber truck is ready). My current pickup truck is 2004.

I'm not sure I'd spring for FSD if I had to stay alert while using it. If I could confidently read a book on a 4 hour trip. I'd pay the 10K in a heart beat.

Some musings.

 

[Florafauna]

It's 100% game over if Tesla achieves better than human performance on HW3 vehicles.

More likely scenario is that "regulatory issues" prevent or quickly cancel the release of beta city FSD to the whole fleet. Then Elon and his disciples will claim that "regulatory issues" are preventing the gathering of the data necessary for exponential improvement. He will move the testing to Mars which will not be subject to Earth's laws. FSD on Mars will be far safer than human drivers on Earth and the promise of FSD will be fulfilled.
Another likely scenario is that it turns out beta city FSD is perfectly safe because it remains bad enough to keep testers extremely vigilant and most people won't use it (like Smart Summon).

The suspense is killing me! How much does it cost to ship a Model 3 to Mars? (looks like about $10 million on a Falcon Heavy, making the $10k price for FSD really look like a bargain!)

Not sure about the cost of sending an M3 to Mars but I believe there is a Roadster on its way right now...

 

[banned-66611]

Given the videos and the progression from .10 to .11 and now to .12 with the updates rolling out and with UI enhancements, here is my speculation and prediction:
Thoughts, observations and speculation:

1. Elon is sandbagging FSD Beta - a little - out of the gate - speculation: to stretch the wow factor for a longer time.
2. The updates have shown direct correlation to actual user issues (from .10 to .11 Kim in Rhode Island first, was NOT able to go through a round about after the update was able to do them consistently and pretty well.
3. Same with watching the new videos from .11 to .12 the left hand turns seem more confident than they were before.
4. The fact that updates are rolling out and @verygreen said that they have to reload firmware to update NN's (or at least that was the case) I think part of the rewrite was to allow much greater flexibility on what you can tweak in the stack without loading a whole new version of the NN's

Then you add in the recent tweets:
https://twitter.com/elonmusk/status/1322431151986925568
https://twitter.com/elonmusk/status/1322433989261611009
https://twitter.com/elonmusk/status/1322435476280758272
https://twitter.com/elonmusk/status/1322435853348671488

Prediction:
Game over for the "competition".
This will be evident in months not years.
Tesla has found the solution/approach that will solve the FSD problem with real world as the proving grounds, in real users hands. Kind of like building the machine that builds the machine, but for NN building and AI training.

As @powertoold said, this is contingent on Tesla rolling FSD Beta out to wide release!

Don't mistake some quick gains for them being close to done. They are fixing the easy stuff first, it gets exponentially harder as they get closer to level 5 autonomy.

They are still at least 3-4 years away from having a working robotaxi that meets the specification they sold back in 2016. And that's assuming things go well for them, but their poor attitude towards safety means they will likely suffer legal setbacks.

 

[AlanSubie4Life]

He will move the testing to Mars which will not be subject to Earth's laws. FSD on Mars will be far safer than human drivers on Earth and the promise of FSD will be fulfilled.

I think you are definitely onto something here. You can tell Elon the Chessmaster is 10 steps ahead here, with his Starlink user agreement. Smart.

 

[AlanSubie4Life]

Tldr: if Tesla is *ever* able to widely release the FSD beta, it's game over for any competition.

Other than the open-ended timeframe, I basically agree with this, when taken in conjunction with your statement that they won’t release it until it is safer than a human driver.

So IF they release it relatively soon in safer-than-human final form, it IS game over.

But it’s not game over if that takes 5 years, obviously.

Guess we will see! Exciting times. If successful in a timely manner with their strategy, Tesla will do extremely well. Lots of “ifs.”

 

[Bladerskb]

I think anyone dependent on HD Maps is done for without a "fundamental rewrite of their system" (note: not green's definition of "HD Maps" but Karpathy's definition of it)

The delusion of grandeur of Tesla fans are uncanny. Why would any company do a rewrite to go back to where they were in the process 5 years. Tesla rewrite which wasn’t even a rewrite at all. Because It’s using the same NN with just added control algorithm. But whatever they say u buy it hook line and sinker and give no original thought to it.

Why would a crowd sourced hd map of the world which Mobileye has means they are done. Use your brain. With a crowd sourced HD map you don’t need some BEV network that isn’t even 99% accurate when u are looking for 99.99999% accuracy.

Sadly, I think Tesla will be so far ahead, no one will catch up. Not only does Tesla have a giant fleet, but they have a fleet of future-proof cars with incredible safety record and performance. They have everything, from the mfging capability, to the software, and the engineers. It's mind-boggling what they're doing. Who can stop them?

So far ahead by using the same NN from 2018 that they deployed to catch up to 2017 eyeq4? Seriously. Again use some logic.

 

[tmoz]

Regardless. 20 yrs ago, this was just a dream. It will be solved and sometime in the next 5, 10, or 20 yrs, every car will have this capability.

 

[powertoold]

So far ahead by using the same NN from 2018 that they deployed to catch up to 2017 eyeq4? Seriously. Again use some logic.

I honestly have never understood your "point" or logic. If you're a real software engineer, you'd know that one company's implementation of a programming technique or tool may be vastly superior to another's. This has played out in real life because Mobileye is BEHIND Tesla in deploying features. Tesla is FAR AHEAD in consumer-deployed automation, period. No one's going to argue with that (you tried with your deceptive posts regarding BMW's traffic light feature), and it's evidence of a superior implementation.

 

[Daniel in SD]

it's evidence of a superior implementation

Or lower quality standards. Mobileye doesn't release beta features to the public AFAIK.
You need to compare Mobileye's non-beta features with Tesla's non-beta features.

 

[tmoz]

Since it is a beta, and updates are being issued to decrease the number of interventions, that would indicate that it is not FSD at the moment. Calling it a beta, conveys the risks of the code only to a small percentage of the Tesla's owners - definitely my opinion here as I have zero data, and it is just a guess on my part based upon the techie and non techie people I know. It would be better to call it something like Pre-FSD or MSD (mostly self driving). Also, for those people paying for FSD, I would think that 50% now and 50% later would be a better payment schedule. FSD as defined by I can take a nap in the back seat while driving from San Diego to LA, or LA to SF, or Phoenix to Tucson for example.