Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register

Autonomy Investor Day - April 22 at 2pm ET

This site may earn commission on affiliate links.
Where are the other investor reports on the test rides, can someone provide links ?

Here is the Morgan Stanley report.

D47JjL6UYAALbSA.jpg:large
 
That was sarcasm of course but then again there could be some truth to it. I guarantee you we'd see them all over the news if one of the test rides crashed. Here are three who did take test drives.


I don't doubt that if something went really bad that it would repeated everywhere, but do you think if Tesla didn't put a misleading FSD demo video out in 2016 that this in-person demo would get significantly more traction/press coverage?
 
Based on the patents and demonstrations from Tesla, I think it was slightly disingenuous of Tesla to say they aren't using HD maps. You saw the 3D radar rendering in the demo, very neat. They've patented a way to improve GPS location using cameras, again very clever. They clearly are doing what you suggest- the data is there. Elon said they used to use HD maps. What they found, taking Elon's word at face value, is that the HD maps sometimes impaired driving. I think the confusion is Tesla's emphasis that the mapping data cannot be a primary source of information; that it is, as you suggest, there for a fail safe only, but the vision should always be most trusted.

I also suspect there's a terminology discrepancy for the sake of marketing. HD mapping created using vision+radar+magic instead of LiDAR is something Elon puts under the banner of vision and refuses to call it "HD maps" to differentiate it from competitors. Again, it's not a lack of data. They've shown that.

Was it @verygreen that said Tesla’s current traffic light detection is also tied to mapping info? How generalized is it if your traffic light detection relies on maps?
 
Here is the Morgan Stanley report.

D47JjL6UYAALbSA.jpg:large

So even though Tesla was able to pick their own route, the driver still had to step in?

I mean we have already seen them do a fully autonomous drive back when AP2 came out. And nobody had to step in, in the video.

I would have imagined that at least this 20 minute drive with Analysts would work perfectly. What is this? Amateur hour? 20 minutes of geo fenced driving should be doable...
 
So even though Tesla was able to pick their own route, the driver still had to step in?

I mean we have already seen them do a fully autonomous drive back when AP2 came out. And nobody had to step in, in the video.

I would have imagined that at least this 20 minute drive with Analysts would work perfectly. What is this? Amateur hour? 20 minutes of geo fenced driving should be doable...
The difference was the 2016 demo was a one off success for video purposes, whilst this route was done multiple times with different passengers in the car.
 
  • Like
Reactions: wesley888
I agree the 2019 video and test drives certainly are more meaningful than the 2016 video which turned out to be fixed/fake demo/third-party code probably. I do believe Tesla has and is testing — and was demoing — NoA-like urban street system this time around and that is a meaningful difference.

How capable that turns out to be in actual fact we’ll learn when feature complete Level 5 no geofence hits by the end of 2019.
 
  • Helpful
Reactions: rnortman
The whole presentation was... interesting. The things that worry me are that they're looking at trying to do a city robotaxi first -- possibly the hardest possible use case fora vision-primary system. And, that they've ditched any kind of HD maps, or realtime HD mapping efforts.

It also reaffirmed to me that here in the UK, I doubt we'll ever see an FSD solution from Tesla. In America? Sure. Driving is relatively easy (even in the fabled "LA traffic") in the US; it's a country that matured with the motor car. It's all a lot more sensible. I'm thinking that none of the Tesla executive team have driven anywhere other than LA, San Fran or possible New York. I get the impression that if I were to get into a car (any car!) with Elon, and have him chauffeur me from Wimbledon to Tottenham Hale at Friday 5pm, we'd probably die.

I'd rather they focus all their efforts on making long-distance travel safe. City driving is stressful, boring and complex... but I'd rather do it. It's low speed and sort-of low risk (at least, to life). Highway driving is mindnumbingly boring, simple, yet highly dangerous. I'd rather a computer do that... we're not at the stage where it can do that without it causing additional stress - mostly thanks to the new "cut-in" NN!
 
The difference was the 2016 demo was a one off success for video purposes, whilst this route was done multiple times with different passengers in the car.

Sure, but the base premise was probably the same. Fixed route, heavily mapped, probably hard coded portions for what to look for/how to act.

Back then they had 550 miles and about 186 disengagements. Would be interesting to see how many miles and disengagements they had this time. Google for example only had to 0.19 disengagements per 1000 miles driving, back in 2016. So with the Google tech back then, there would have been a high chance that non of the test drives would have had a disengagement. And in 2018 Waymo was already down to 0.09.

If Tesla really wants to show off their lead, they should start to test on CA roads and report their disengagements. Not just drive a geo fenced route with some analysts on board. But at least we will get a disengagement report next year for those drives and the testing before.
 
  • Like
Reactions: Kant.Ing
@R.S

It is obvious Tesla does not have any currently demonstrable lead (as in demonstrable by a test drive), otherwise they would have demonstrated it.

What they have is an OTA updateable fleet with certain features and a vision-based machine learning project to go along with that... and a plan to exceed others by the end of 2019 after which they should be able to demonstrate their Level 5 prowess (with a safety driver).
 
  • Like
  • Helpful
Reactions: croman and rnortman
The whole presentation was... interesting. The things that worry me are that they're looking at trying to do a city robotaxi first -- possibly the hardest possible use case fora vision-primary system. And, that they've ditched any kind of HD maps, or realtime HD mapping efforts.

I was surprised about that, too. Mapping should be one of their biggest advantages, since they have tons of vehicles on the road.

A NN can be trained in a simulation, that’s not really something you need lots of cars for, especially since those cars can’t make their own decisions until they are good enough, which limits the training value. But if you have millions of cars, you can easily make real time HD maps.
 
I would have imagined that at least this 20 minute drive with Analysts would work perfectly. What is this? Amateur hour? 20 minutes of geo fenced driving should be doable...

I am not sure it was geofenced in the strictest sense. Tesla did not modify the software just for that specific route. They did not hard code anything. They just picked the route, that's all. It was a live test drive with no tricks. And they did the route over and over again for different investors. So there was bound to be a test drive with one disengagement.
 
I am not sure it was geofenced in the strictest sense. Tesla did not modify the software just for that specific route. They did not hard code anything. They just picked the route, that's all. It was a live test drive with no tricks. And they did the route over and over again for different investors. So there was bound to be a test drive with one disengagement.

Why not allow the investors in the car to pick a route within a fixed radius?
 
  • Like
Reactions: R.S and croman
That was sarcasm of course but then again there could be some truth to it. I guarantee you we'd see them all over the news if one of the test rides crashed. Here are three who did take test drives.

How about as another explanation -- the route they chose to drive was so bland and easy that even if it performed reasonably well it would not have been newsworthy. This has all been done before. When they handle dense urban environments (or even a reasonable surban environment with bicyclists and pedestrians), heavy rush hour traffic on poorly marked highways with short accel/decel lanes, and bad weather -- without constant driver interventions because the car screwed up -- then they have a demo of something that nobody has done before (without lidar).
 
  • Like
Reactions: croman
@R.S

It is obvious Tesla does not have any currently demonstrable lead (as in demonstrable by a test drive), otherwise they would have demonstrated it.

You don't think that the custom toolset they used to build this, the new chip, the fleet, the training data, the breakthroughs in NN vision processing ... none of this demonstrates a lead?

Tough crowd.