Anthony Levandowski beats Tesla's Elon Musk to first Auto Cross country (3k Miles, 0 disengagement)

[MP3Mike]

There was on rumor on Reddit or TMC or such early this year that Tesla indeed was attempting the drive with freshly painted Supercharger exits and all that, but it was too brittle. I recall the rumor indeed said it was pretty much hardcoded but it still didn’t fully work out. I think it is at least a possibility they were trying to ”game it”, but even that did not work out.

Just a rumour, but I do recall reading about and afterwards Elon’s comments made sense...

But what part of it was failing for Tesla? The highway driving, or the local driving for the refueling stops that was completely manual in Pronto.AI demo?

 

[electronblue]

But what part of it was failing for Tesla? The highway driving, or the local driving for the refueling stops that was completely manual in Pronto.AI demo?

Fair question. The painted Supercharger stops would at least suggest Tesla tried to take such exits automatically though could still be somewhat hardcoded of course. How much did Pronto.ai drive manually of those stops, I know they drove some part (did it include taking an exit etc)...

 

[MP3Mike]

Fair question. The painted Supercharger stops would at least suggest Tesla tried to take such exits automatically though could still be somewhat hardcoded of course. How much did Pronto.ai drive manually of those stops, I know they drove some part (did it include taking an exit etc)...

I haven't tried to analyze the video, and the article just said "he did not touch the steering wheels or pedals, aside from planned stops to rest and refuel." So he could have gone manual for the whole exit, refuel, and return to the freeway.

Of course there are "weasel" words in there. OK, he didn't touch the steering wheels or pedals during the trip. What about the mouse, keyboard, turn signals, joystick, and/or any other buttons they added to the car? (And really there was more than one steering wheel? That seems weird.)

 

[TheTron]

By current firmware do you mean Autopilot or Tesla's still unreleased FSD software? Elon said a while ago Tesla's FSD software can already do coast to coast run but it does not want to game the system. Gaming the system I believe means running a pre-mapped route with practice runs probably just what Levandowski has done. A true capability is you tell the car at the driveway where, including anywhere in the other coast, you want to go and it will take you there. That's what every Tesla will need to have. Once again you can't compare someone's "demo" to Tesla's cars that everyone could have one in hand and scrutinize its capability. There will be no engineers there to help you to plan and run every trip.

I suspect if Tesla had *anything* that could complete a coast to coast in a similar fashion to Levandowski, gamed or not, they would be showcasing it as a demo of what is to come from EAP in the near future. Their silence on their current capability speaks volumes to me.

 

[electronblue]

I suspect if Tesla had *anything* that could complete a coast to coast in a similar fashion to Levandowski, gamed or not, they would be showcasing it as a demo of what is to come from EAP in the near future. Their silence on their current capability speaks volumes to me.

I actually took the rumour from year ago and Elon’s comments as pretty much as information that Tesla tried to game the coast to coast and failed. The rumour said there had been multiple attempts and all sorts of hardcoding and preparation going on to make it work. This rumours preceded Elon’s comments and Tesla postponing the drive indefinitely. Then they gave up and Elon spinned it a little to make it sound better than it was — is what it felt like to me. Just my take on it from the outside looking at these comments.

 

[CarlK]

I suspect if Tesla had *anything* that could complete a coast to coast in a similar fashion to Levandowski, gamed or not, they would be showcasing it as a demo of what is to come from EAP in the near future. Their silence on their current capability speaks volumes to me.

In what similar fashion? No one even knows details of Levandowski's run. If Tesla is going to do a demo it will at least be pre-announced and will allow third party monitoring. Remember all Tesla's "demo's" are run by third party drivers, aka customers. If Levandowski is confident with his capability and wanted to prove it to the world he could have arranged a run with journalists on board from start to end.

 

[CarlK]

I haven't tried to analyze the video, and the article just said "he did not touch the steering wheels or pedals, aside from planned stops to rest and refuel." So he could have gone manual for the whole exit, refuel, and return to the freeway.

Of course there are "weasel" words in there. OK, he didn't touch the steering wheels or pedals during the trip. What about the mouse, keyboard, turn signals, joystick, and/or any other buttons they added to the car? (And really there was more than one steering wheel? That seems weird.)

Waymo cars did have a kill button. That steering wheel or pedal description does sound weird. We just have no idea how that demo was conducted.

 

[Cosmacelf]

This article from the Guardan says that Pronto can only be used on freeways, not city streets. So coast to coast, except for pit stops to get off the freeway for refueling. This isn’t that big a deal. Also, take a look at Pronto’s hardware in the picture in the article. Pronto should be called “Proto”, as in prototype.

Self-driving car drove me from California to New York, claims ex-Uber engineer

 

[electronblue]

If Tesla is going to do a demo it will at least be pre-announced and will allow third party monitoring.

You really think Tesla is going to pre-announce their coast-to-coast attempt? From what was heard from Tesla and through the grapeview around end of 2017 and early 2018 the attempts were made and nothing was pre-announced. Tesla never seemed to suggest a pre-announcement either?

 

[CarlK]

You really think Tesla is going to pre-announce their coast-to-coast attempt? From what was heard from Tesla and through the grapeview around end of 2017 and early 2018 the attempts were made and nothing was pre-announced. Tesla never seemed to suggest a pre-announcement either?

I have no idea what that rumor is about but they did not post-announce anything either. There is also a rumor that Tesla AI team has already done in inclement weather of what Levandowski says he has done. Rumors are of course even worse than demos on edited video tapes. I don't know how Tesla will do it or even if it ever will but It does not count the same if it's not pre-announced and monitored. Whatever just don't distract from the the point how ludicrous it is for people to use such "demo" to compare to what in real cars everyone can, and will, try and test.

 

[J1mbo]

How did we all miss this from @verygreen ?

 

[TheTron]

I have no idea what that rumor is about but they did not post-announce anything either. There is also a rumor that Tesla AI team has already done in inclement weather of what Levandowski says he has done. Rumors are of course even worse than demos on edited video tapes. I don't know how Tesla will do it or even if it ever will but It does not count the same if it's not pre-announced and monitored. Whatever just don't distract from the the point how ludicrous it is for people to use such "demo" to compare to what in real cars everyone can, and will, try and test.

I agree you can’t compare a demo with production hardware/software, however Tesla haven’t even demo’d something as impressive as this cross country drive let alone released something capable of doing it. I suppose they could be keeping quiet then unleash something truly amazing out of nowhere but they tend to over hype and under deliver, if anything, so I assume they are actually behind.

Hope I’m wrong though...

 

[electronblue]

@TheTron Yep, this is a constant point of contention between myself and @CarlK... I think it is false equivalency to pool all ”demos” into one worthless group. Merits of demos need to be assessed realistically on a case by case basis because not all are the same.

 

[CarlK]

I agree you can’t compare a demo with production hardware/software, however Tesla haven’t even demo’d something as impressive as this cross country drive let alone released something capable of doing it. I suppose they could be keeping quiet then unleash something truly amazing out of nowhere but they tend to over hype and under deliver, if anything, so I assume they are actually behind.

Hope I’m wrong though...

Can't you see this "demo" of "something as impressive as this cross country drive" was not that impressive at all? I'm sure Tesla is able to do a demo like that anytime it wants. There was nothing but freeway driving and he used very vague words to describe what autonomous capabilities were included. There was also no mention of what the on board computer was doing including if an operator could use the computer to control the car in some way including, say, to slow the car down or stop which otherwise we would call disengage. @MP3Mike mentioned it's weird that he just says he did not touch the steering wheel or pedal. I find it also weird that he says the run could have been done without a driver in the car. You don't need to be physically inside a car to operate the computer. A lot of weasal words indeed. All he needed to say is simply there was no direct or indirect human intervention in the entire trip.

One thing that differetiates the two is Tesla has no need to put up a demo to prove anything, at least not until when full autonomous capacity could be demostrated. People will always look at its cars to judge what EAP or FSD could or could not do. Start ups or new comers have different priorities. They need to use demos to estabilish credibility. That is the most important thing to them. Their survival is depending on whether it could convince investors to put the next round of financing.

@TheTron Yep, this is a constant point of contention between myself and @CarlK... I think it is false equivalency to pool all ”demos” into one worthless group. Merits of demos need to be assessed realistically on a case by case basis because not all are the same.

There is little, if any, "merits" of a demo like that other than its PR value. Any so so basketball player could produce a demo tape showing great or perfect 3 point shooting. That says nothing about if he is better or worse than Steph Curry.

 

[electronblue]

@CarlK My point wasn’t so much this demo but demos in general. Not all demos are equal and should not considered equally worthless...

 

[TheTron]

Can't you see this "demo" of "something as impressive as this cross country drive" was not that impressive at all? I'm sure Tesla is able to do a demo like that anytime it wants. There was nothing but freeway driving and he used very vague words to describe what autonomous capabilities were included. There was also no mention of what the on board computer was doing including if an operator could use the computer to control the car in some way including, say, to slow the car down or stop which otherwise we would call disengage. @MP3Mike mentioned it's weird that he just says he did not touch the steering wheel or pedal. I find it also weird that he says the run could have been done without a driver in the car. You don't need to be physically inside a car to operate the computer. A lot of weasal words indeed. All he needed to say is simply there was no direct or indirect human intervention in the entire trip.

One thing that differetiates the two is Tesla has no need to put up a demo to prove anything, at least not until when full autonomous capacity could be demostrated. People will always look at its cars to judge what EAP or FSD could or could not do. Start ups or new comers have different priorities. They need to use demos to estabilish credibility. That is the most important thing to them. Their survival is depending on whether it could convince investors to put the next round of financing.



There is little, if any "merits" of a demo like that other than its PR value. Any so so basketball player could produce a demo tape showing great or perfect 3 point shooting. That says nothing about if he is better or worse than Steph Curry.

I saw it as a demo of NOA level autonomy that was actually reliable enough to travel decent distances without intervention. I see that as being impressive given the current state of Autopilot. You assume Tesla has this nailed already but hasn’t shown it to us. I don’t have that level of faith in Tesla.

Frankly I’d take some cooked up demos at this point over a bunch of tweets from Elon that subsequently turn into memes.

All that said, Tesla do have the best ADAS currently available on the market to my knowledge (some might disagree). However they have massively failed to deliver on the hype from the AP2 launch two years ago.

 

[CarlK]

@CarlK My point wasn’t so much this demo but demos in general. Not all demos are equal and should not considered equally worthless...

True. As I said pre-announced and monitored demo does have some merit even that it is still not the same a what's in a product everyone can try. The "demo' that is the subject of this thread is neither.

I saw it as a demo of NOA level autonomy that was actually reliable enough to travel decent distances without intervention. I see that as being impressive given the current state of Autopilot. You assume Tesla has this nailed already but hasn’t shown it to us. I don’t have that level of faith in Tesla.

Frankly I’d take some cooked up demos at this point over a bunch of tweets from Elon that subsequently turn into memes.

All that said, Tesla do have the best ADAS currently available on the market to my knowledge (some might disagree). However they have massively failed to deliver on the hype from the AP2 launch two years ago.

I don't think anyone here has used "bunch of tweets from Elon that subsequently turn into memes" in the discussion other than using it to imply a negativity. Yes there is no question whatever are in Tesla's cars is still ahead of those in any other car on the market. Using the same logic (that demo's have some merits) why is anyone who has such an advanced technology not putting something in a real product that could generate profits and really beat Tesla? The proof is always in the puddings.

 

[electronblue]

The "demo' that is the subject of this thread is neither.

Of course. This demo, as any, needs to be evaluated on its merits as always — and there are controversial parts to its merits for sure.

 

[Bladerskb]

In what similar fashion? No one even knows details of Levandowski's run. If Tesla is going to do a demo it will at least be pre-announced and will allow third party monitoring. Remember all Tesla's "demo's" are run by third party drivers, aka customers. If Levandowski is confident with his capability and wanted to prove it to the world he could have arranged a run with journalists on board from start to end.

Really? just like their two FSD demo in 2016 right? Or how they flew dozens of journalists to a particular city (New Jersey) to do EAP demo on one route specific so that hot pieces can be written? Why not let the journalists from all over go to their respective tesla store and do a EAP demo there and selecting any highway route they want?

LOL gotta love tesla fans creating qualifiers to disqualify others while their own gawd company can't live up to.

 

[Vitold]

Journalists noted that pronto. AI could not merge on their own and had to be assisted.