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.

 

[DanCar]

Don't understand why people argue about such vague arguments... unless there are details you can't say something like this is right or not right. Arguable we've had city autosteer since AP1

City autosteer implies turning on city streets and navigation on city streets.
The actual term used on Tesla's website is automatic driving on city streets.

 

[DanCar]

[/quote]

I think we will be feature complete, full self-driving, this year. Meaning the car will be able to find you in a parking lot, pick you up and take you all the way to your destination without an intervention. This year. I would say I am of certain of that, that is not a question mark. However, people sometimes will extrapolate that to mean now it works with 100% certainty, requiring no observation, perfectly, this is not the case.

 

[Snuffysasa]

City autosteer implies turning on city streets and navigation on city streets.
The actual term used on Tesla's website is automatic driving on city streets.

Why? Highway autosteer does not imply taking exits, or highway interchanges...

why would city autosteer imply taking tuns on streets?

 

[Mardak]

being able to do roundabouts well (say 90% of the time) - end of 2021 - 0.1% on progress to True self driving

A single percentage is easy to communicate but it hides a lot of details especially when considering Level 4 vs Level 5. For example, that 90% number could have various meanings:

• 90% of worldwide roundabouts are successful 100% of the time (and remaining 10% of roundabouts always fail)
  
• 95% of worldwide roundabouts are successful 95% of the time
• 100% of worldwide roundabouts are successful 90% of the time

In the context of Level 4, the first item is desired as one would geofence deployment to the regions that have 100% success rate. The types of failures could be static, e.g., is the shape of roundabout and lane entrances/exits understood, or dynamic, e.g., other vehicles moving at certain speeds/angles or lighting conditions. (And separately, there's always the correctness vs natural/smooth-ness, e.g., yielding too long.)

For a California Tesla owner, having Autopilot successfully navigate the few roundabouts (if any) that one might encounter on a daily basis is practically full self driving in that aspect for that person's commute. And similarly for all other aspects of full self driving, a given person's commute requiring no intervention is effectively Level 4 for many of the miles the vehicle would see.

As you point out, successfully handling some number of commutes without intervention is far from the many potential routes that Level 5 would need to handle for "True self driving." But Tesla's approach allows them to release things as Level 2 and actively measure their self-driving progress within a commute, city, region, country, worldwide. And ideally other people's commutes making progress towards no intervention makes it more likely to be successful when traveling not-your-usual routes.

 

[powertoold]

[/QUOTE]

Well, at least he began his comments by saying "I think." Lol

He thinks he's certain, but he was overly optimistic.

 

[Knightshade]

Elon definitely overhypes FSD, but I don't remember him saying anything factually wrong. Many of his statements are related to overly optimistic timelines or opinions about the utility of certain features.

For example, dumb summon is magical, but it's not to the level of human standards. For a computer to use cameras to generate and navigate an occupancy map is very impressive, but for the layman, smart dumb summon isn't useful.

if he says he's confident they will have X by Y date, and they don't- was he factually wrong? Or does he have literally until he dies to deliver X and he's still ok?

Because there's a LOT of good examples there.

 

[mikes_fsd]

Funny... you bolded for emphasis so I wanted make my own emphasis.
this is from 9:59 timestamp of that video

I think we will be feature complete, full self-driving, this year. Meaning the car will be able to find you in a parking lot, pick you up and take you all the way to your destination without an intervention. This year. I would say I am of certain of that, that is not a question mark. However, people sometimes will extrapolate that to mean now it works with 100% certainty, requiring no observation, perfectly, this is not the case.

at 9:20 of that same video -- which is BEFORE that quote.

I still think the last 10% of autonomy is extremely difficult or even the last 1%

 

[Daniel in SD]

It's all the robotaxi stuff that's ridiculous. Also, any claims that they're close to L3 Autopilot.
Supervised self driving is relatively easy and tons of companies have done it. I'm honestly surprised how long it's taken Tesla to get there.
I expect they'll do the cross country stunt next year which will be cool but completely useless IMHO.

 

[mikes_fsd]

I expect they'll do the cross country stunt next year which will be cool but completely useless IMHO.

I suspect many of those bashing Tesla approach to Autopilot are in the same boat as you and will try to diminish the usefulness of such a demo once it happens. (funny as for the past year this has been used as the defacto "proof" that Tesla is behind on FSD - but it will just be "useless" if they do it "late")

I look forward to the cross country demo.

 

[AlanSubie4Life]

I suspect many of those bashing Tesla approach to Autopilot are in the same boat as you and will try to diminish the usefulness of such a demo once it happens. (funny as for the past year this has been used as the defacto "proof" that Tesla is behind on FSD - but it will just be "useless" if they do it "late")

I look forward to the cross country demo.

If they can do a cross-country demo (full video without speed up provided) with absolutely no interventions at all, except on entrance and exit to the actual location of the charging stops (not the off-ramp; it needs to drive into the parking lot of the charging location!), using production release software, in the next year, I will be SUPER impressed. And I think such a demonstration with production software would be indicative of a useful product. I'm sure I would still have complaints about the details of its operation and implementation choices, but that sort of fundamental capability would be significant, even if it was still not actually Level 3.

 

[mikes_fsd]

If they can do a cross-country demo with absolutely no interventions at all, except on entrance and exit to the actual location of the charging stops (not the off-ramp!), using production release software, in the next year, I will be SUPER impressed. And I think such a demonstration with production software would be indicative of a useful product.

I know you are saying that because you're skeptical.
But this will indeed be a huge accomplishment - I will be super impressed as well!

Hence, why I do not think it is "useless" as @Daniel in SD is already trying to frame it.

 

[AlanSubie4Life]

Hence, why I do not think it is "useless" as @Daniel in SD is already trying to frame it.

I think when framing these things, it's important to be as specific as possible about what one is actually expecting to see. I suspect that @Daniel in SD and I are in agreement on being super impressed with what I have described if they deliver it, though I could be wrong. I suspect he thinks there will be some significant caveats to the demo which make it fall outside of what I have described.

But again, have to emphasize (even though I would be super impressed!) that such a demo would not mean that Level 3 (don't even need to discuss Level 4/5) is necessarily imminent at that point.

But to me the main thing is that such an accomplishment would mean MUCH better performance and reliability of Autosteer than where it's currently at. 2500+ miles on both freeways and surface streets without a single intervention is quite unlikely unless the software & system is really quite good.

And of course, as I've described it, it's fundamentally not a demo - any owner could do the trip and expect to achieve the same result at least 50% of the time. Obviously that exact % is important too - it's not much good if they cherry-picked a single trip that happens successfully only 2% of the time...but the way this works, I suspect that it's going to have to be really good to even get to 50% of the time...which is why I'd be impressed by such a release.

 

[mikes_fsd]

But again, have to emphasize (even though I would be super impressed!) that such a demo would not mean that Level 3 (don't even need to discuss Level 4/5) is necessarily imminent at that point.

I love how you edited to hedge your statement. That is fine.

Supervised self driving is relatively easy and tons of companies have done it. I'm honestly surprised how long it's taken Tesla to get there.

My problem is that if "others have done it" and it is "relatively easy" why has no one else done this "stunt"?
Specifically a cross-country, no intervention, fully autonomous ride!

The smugness is reaching 11 very quickly.

 

[AlanSubie4Life]

I love how you edited to hedge your statement. That is fine.

It's normal. There were no changes to my thinking; I was just making sure the parameters and my thinking were clear. This is simply what I think; a cross-country demo with no interventions does not mean L3 is imminent, and I've thought that for a very long time at this point.

 

[Daniel in SD]

I think when framing these things, it's important to be as specific as possible about what one is actually expecting to see. I suspect that @Daniel in SD and I are in agreement on being super impressed with what I have described if they deliver it, though I could be wrong. I suspect he thinks there will be some significant caveats to the demo which make it fall outside of what I have described.

Stunts are fun! It will be impressive as a stunt but I will not take it to mean that Tesla is close to FSD. When they can do it on surface streets 1,000 times in a row then I will believe they are close. Elon Musk himself has said the obvious, it's that last 1% that's hard. This is something that every autonomous vehicle company has discovered and Tesla hasn't even started working on it.

 

[AlanSubie4Life]

Stunts are fun! It will be impressive as a stunt but I will not take it to mean that Tesla is close to FSD. When they can do it on surface streets 1,000 times in a row then I will believe they are close. Elon Musk himself has said the obvious, it's that last 1% that's hard. This is something that every autonomous vehicle company has discovered and Tesla hasn't even started working on it.

As I've made clear, I'm just saying I think it's an impressive Level 2 driving feat. Quite far from FSD I would agree. I'm just saying the utility increase of being able to reliably accomplish a cross-country trip is significant (obviously it would have to be reliable for them to demonstrate it - they're not going to do hundreds of cross country trips until they happen to get one that works, and program the software to specifically deal with a very specific route to make the demo work - that obviously wouldn't "count" under my original statement (I'm not going to go back and edit it to make that clear - it goes without saying) - it's not something any user could reproduce, which was clearly required).

 

[Daniel in SD]

My problem is that if "others have done it" and it is "relatively easy" why has no one else done this "stunt"?
Specifically a cross-country, no intervention, fully autonomous ride!

The problem is supervised self driving is useless in my opinion. If you're trying to develop robotaxis, like most autonomous vehicles companies are, it makes sense to get them working in a single location first.

Delphi claims to have done it 2015 but only on the highway.
This Is Big: A Robo-Car Just Drove Across the Country

Levandowski (the guy who stole from Waymo and went to Uber resulting in a giant lawsuit and jail time) claims to have done it in 2018. I don't trust that guy though!
Former Google engineer, once barred from autonomous-driving industry, says he just completed first cross-country self-driving car trip

 

[mikes_fsd]

And of course, as I've described it, it's fundamentally not a demo - any owner could do the trip and expect to achieve the same result at least 50% of the time. Obviously that exact % is important too - it's not much good if they cherry-picked a single trip that happens successfully only 2% of the time...but the way this works, I suspect that it's going to have to be really good to even get to 50% of the time...which is why I'd be impressed by such a release.

If they send 3 cars from Fremont or LA to Times Square and even 1 makes it on first try, I would be super impressed.
If all 3 made it on first try I would expect to see a wide release within the ~90 days of the demo.

And that is the beauty of Tesla approach, every "demo" is fundamentally not a demo, because when they release the software to the fleet everyone with the car can go and try it out.
Can you point to ANY other provider of FSD that can say the same today?

Stunts are fun! It will be impressive as a stunt but I will not take it to mean that Tesla is close to FSD. When they can do it on surface streets 1,000 times in a row then I will believe they are close.

You absolutely have no idea what you are talking about or what is actually being accomplished.

Tesla hasn't even started working on it.

Please do tell me what Tesla has and hasn't started working on! With some actuall facts / sources to back it up.

 

[mikes_fsd]

As I've made clear, I'm just saying I think it's an impressive Level 2 driving feat. Quite far from FSD I would agree.

Ah the smell of back peddling!

 

[AlanSubie4Life]

Ah the smell of back peddling!

Huh? How is this backpedaling? This is all 100% consistent. It's the same statement as post #734. And #732. And #730.

And that is the beauty of Tesla approach, every "demo" is fundamentally not a demo

I get the impression that the famous "demo" video from a few years back was a demo, as was the demo after Autonomy Day. Tesla does demos - and these are not covered by my original statement.

The problem is supervised self driving is useless in my opinion.

I agree it's useless as a means to accomplish robotaxis - quite a different problem scope. But I'd say extremely good supervised self-driving is quite useful for long trips and has substantial market value - unless of course someone else has a bunch of robotaxis or a better system at the same price. I'd like my car to be much better at supervised self-driving!