Tesla posts FSD demo video of Model 3

[outie]

I am wondering if @outie is impressed in the slightest?

I am just seeing what I saw back in 2016, you know?

 

[scottf200]

I wish they had a none accelerated video. I’d want to watch this for 30 minutes in real time just to get the perspective of what an end customer would see.

I am wondering if @outie is impressed in the slightest?

I really wish they'd put out a realtime version of that video. I'd like to see how smooth the car takes the sharp turns and how it deals with merging lanes at various points.

You can watch it in youtube at x0.25 speed which gives you and decent idea and if you watch the steering wheel.

 

[S4WRXTTCS]

I am just seeing what I saw back in 2016, you know?

But, this time it's for real.

It's Tesla software running on Tesla hardware.

The 2016 video was done using Nvidia Software on Nvidia hardware, but they knew when they posted it that they wouldn't use Nvidia's SW because they didn't want to be vendor locked.

Of course if a company lies then there is no trust.

I trust that they demonstrated what it is actually capable of this time around. But, obviously self-driving is a complicated problem with lots of edges cases. So I expect the first 9-12 months of FSD to be quite a struggle with lots of FSD tried to kill me posts.

 

[scottf200]

I’ll be curious to start hearing reports from these Investor analyst reporters. I’ll be really interested to see what Sandy Munro (Munro & Assoc) says about it on the next Autoline program.

Tesla shorts will be like ... look the car got lost and had to turn around...

 

[outie]

But, this time it's for real.

It's Tesla software running on Tesla hardware.

The 2016 video was done using Nvidia Software on Nvidia hardware, but they knew when they posted it that they wouldn't use Nvidia's SW because they didn't want to be vendor locked.

Of course if a company lies then there is no trust.

I trust that they demonstrated what it is actually capable of this time around. But, obviously self-driving is a complicated problem with lots of edges cases. So I expect the first 9-12 months of FSD to be quite a struggle with lots of FSD tried to kill me posts.

Yea well I was already impressed back in 2016. I lost a night of sleep with my AP1 car that was delivered EOQ3 2016. I am not going to be impressed again for what appears to be the same thing on surface.

 

[chickensworth]

my car can’t even drive into the sun and stay in it’s lane... must be that new hw3.0 chip.

 

[MXWing]

my car can’t even drive into the sun and stay in it’s lane... must be that new hw3.0 chip.

Tesla is testing a remote drone pilot. My wife was driving your car.

 

[chickensworth]

Tesla is testing a remote drone pilot. My wife was driving your car.

In all seriousness, a more powerful processor is great and all, but I’ve had so many close calls and erratic behavior from EAP, NOA and all them other abbreviations while driving into the sun. Weak point being the visual cameras which to my knowledge won’t be upgraded with FSD... just the chip.

 

[Lunares]

In all seriousness, a more powerful processor is great and all, but I’ve had so many close calls and erratic behavior from EAP, NOA and all them other abbreviations while driving into the sun. Weak point being the visual cameras which to my knowledge won’t be upgraded with FSD... just the chip.

The neural network should be better with more processing and be able to deal better with the sun. More training will also help as Tesla continues to update the NN

 

[woodisgood]

Tesla is testing a remote drone pilot. My wife was driving your car.

Man, the wife can’t catch a break tonight. What’d she do?

 

[banned-66611]

So it's basically just a repeat of the fake 2016 demo. Carefully selected route, little traffic, and who knows how many takes to get it perfect.

Have they really made near zero progress since 2016?

 

[fajitamondays]

I wonder how smooth it was in real-time with the freeway interchanges. My experience with NoA on cloverleafs is that it is quite jittery (I.e. jerky steering wheel movements), and crawls at a slow pace until nearly at the gore point, where most others would have accelerated long before (E.g. "spirited drivers", motorcycles, or those with legitimately slow cars that need the extra "runway" space to get up to speed).

Anyway, these are things that NN improvements can supposedly handle, and the computing power of HW3 should certainly help smooth out. I'll probably start intervening by manually accelerating where it makes sense, to flag for review by the AP team. Wishful thinking, but they did mention they review interventions by drivers from time to time.

 

[alloverx]

Not a real driver (hence minimal movement), its a robot (wearing fancy gloves) that was trying to escape the factory.
BTW, I prefer the GM Cruise video but we cant see the dash or driver sadly.

 

[SunCharged]

Less freeway driving and more neighborhood and city driving including school zones and pedestrians, children and dogs running out in front of cars between parked cars. That's more than 50% of my driving.

Exactly. I want to see how it handles school busses letting kids on and off, school zones with reduced speed limits at certain times.

And for me, how well can it stop for deer!

 

[preilly44]

For everybody saying the video is the same at 2016.....well, it kind of is but this time there were investors and probably some Shorts actually in the cars during the demo session. They can't fake it when they have investors actually in the seats.

 

[banned-66611]

Why are there no independent videos from the journalists invited to take a ride?

We can't take Tesla's word for it after the 2016 fake video, but if a journalist were able to test it themselves with a route they selected there would finally be some credibility to these claims.

 

[Az_Rael]

For everybody saying the video is the same at 2016.....well, it kind of is but this time there were investors and probably some Shorts actually in the cars during the demo session. They can't fake it when they have investors actually in the seats.

True, but weren’t those rides on a pre-determined route? That allows for a lot of tweaks to make the software look better than it is.

 

[M109Rider]

Why are there no independent videos from the journalists invited to take a ride?

We can't take Tesla's word for it after the 2016 fake video, but if a journalist were able to test it themselves with a route they selected there would finally be some credibility to these claims.

I think the video we saw wasn’t leaps and bounds off of what we know they have now and are working on. I don’t see any basis for assuming Tesla is faking a video.

True, but weren’t those rides on a pre-determined route? That allows for a lot of tweaks to make the software look better than it is.

Of course it’s a pre determined route. That’s called planning. Any presentation, pitch, or demonstration has to be pre planned.
I would t assume that means it’s tweaked or fake.

I have more faith and excitement around an actual video with investors as witness, than speculation and assumptions on a forum.

 

[GeoX750]

Did they show any left turns into uncontrolled traffic ?

 

[BestRadar]

For everybody saying the video is the same at 2016.....well, it kind of is but this time there were investors and probably some Shorts actually in the cars during the demo session. They can't fake it when they have investors actually in the seats.

But again it was a per-determined route repeated over and over. Let the investors put in their own address and get there and then I would be impressed.