How will Tesla demo FSD?

[Bladerskb]

I am not championing anything. I am just asking questions. When I watch the video, I will let you know my thoughts. If the video is underwhelming, I will certainly say so.

No you won't.

 

[diplomat33]

No you won't.

I am telling the truth. If the video is underwhelming, I will say so.

 

[Daniel in SD]

This is what companies are trying to master today. This is where the state of the art is...

That video is amazing. Watching a third party video of Waymo's vehicles it seemed much less impressive in real life than their demo videos would suggest. I don't see any third party videos on youtube of how good Cruise vehicles really are.

 

[diplomat33]

This is what companies are trying to master today. This is where the state of the art is...

Yes, it is impressive. That is why Cruise and Waymo are the top leaders. They are testing the state of the art. Nobody denies that.

 

[Inside]

The impressiveness of Tesla's video (if they release one) would not only depend on how difficult the route is, but would also have to be gauged by how many disengagements took place before they had a clean run. Sadly, we'll probably never know. I wonder if the Monday video will be as impressive as the 2016 video.

 

[ZeApelido]

MobileEye EyeQ4 can do it. The problem is that doing FSD demos isn't all that useful.

So which car models?

 

[Dutchie]

Yes, it is impressive. That is why Cruise and Waymo are the top leaders. They are testing the state of the art. Nobody denies that.

Why is that impressive? It doesn't show that the car is driving by itself. It is just a video looking out forward from car driving. For all intention purposes it is just to show what a self driving car should be capable of doing while someone is driving the car.

 

[diplomat33]

Why is that impressive? It doesn't show that the car is driving by itself. It is just a video looking out forward from car driving. For all intention purposes it is just to show what a self driving car should be capable of doing while someone is driving the car.

No, I believe the car is driving itself in that video.

 

[diplomat33]

@Bladerskb Considering that Tesla has already given us NOA promos with other cars on the road and have teased that they have traffic lights and intersections done (not to mention the 2016 FSD demo was better), I would be very very surprised, and yes disappointed, if the entire FSD demo is just a boring highway drive with no other cars even allowed on the road with no traffic lights or intersections. If I am wrong and that is all we get, and we don't get any demo with intersections, then yes, I will admit that I was wrong and disappointed. Deal?

 

[Daniel in SD]

So which car models?

Good question! They claim to be shipping but I haven't seen info about which models they're in.
As far as I know EyeQ4 is fairly worthless at the moment since they're nowhere near getting FSD working. They did make a pretty awesome demo though! Audi's "traffic jam pilot" level 3 system uses EyeQ3 but they still haven't released it so I'm guessing that it probably doesn't work yet.

 

[Daniel in SD]

Why is that impressive? It doesn't show that the car is driving by itself. It is just a video looking out forward from car driving. For all intention purposes it is just to show what a self driving car should be capable of doing while someone is driving the car.

Yeah, if that's a human driving the car it's not all that impressive
That's why I asked if there is some third party video of Cruise's vehicles.

 

[Bladerskb]

@Bladerskb Considering that Tesla has already given us NOA promos with other cars on the road and have teased that they have traffic lights and intersections done (not to mention the 2016 FSD demo was better), I would be very very surprised, and yes disappointed, if the entire FSD demo is just a boring highway drive with no other cars even allowed on the road with no traffic lights or intersections. If I am wrong and that is all we get, and we don't get any demo with intersections, then yes, I will admit that I was wrong and disappointed. Deal?

I'm sure they will have streets scene in the video before they were spotted getting on the highway and after they get off.
But again if stopping for stop signs and lights and making turns in a demo was impressive, where were you in 2010? or heck Tesla's own 2016 video?

 

[diplomat33]

But again if stopping for stop signs and lights and making turns in a demo was impressive, where were you in 2010? or heck Tesla's own 2016 video?

You are missing the point or maybe we have different expectations. I don't think I ever said that stop signs and lights would be "impressive", I am looking to see what the AP3 software can do NOW. Remember that AP3 is brand new and Musk is making bombastic promises about "full autonomy" but we haven't seen AP3 in action and we haven't seen anything from Tesla about AP3's FSD capabilities. Right now, we have nothing. The April 22 event will be our first look at what Tesla has done so far with AP3 software. So I am just looking to see what AP3 software is able to do right now.

 

[favo]

I'm sure they will have streets scene in the video before they were spotted getting on the highway and after they get off.
But again if stopping for stop signs and lights and making turns in a demo was impressive, where were you in 2010? or heck Tesla's own 2016 video?

One possible big difference is if they announce that the demonstrated version will be available "soon." Now of course there are all kinds of caveats we don't know the answer to yet. What exactly are the capabilities, aside from what might be demoed. What is the definition of "soon". Obviously, Elon has a lot of trouble with time estimates, but one thing we do know is that HW3 is a shipping product. They may demo something that they say will only come to cars with HW3 and FSD. Or maybe some new functionality/improvements will be available to all HW2.x/HW3 cars. They may say they are targeting Q2 (or Q4 or whatever) for release. They may even name a specific date (probably not). We shall see.

Another big difference from some other demos (like Waymo) is that the software will be running on hardware that actual customers have and can buy right now. It seems the problem can be attacked on two fronts, hardware and software (putting aside modifying the environment). Waymo is using a lot of expensive hardware, while they refine their software. It works well so far. I think this made a lot of sense when they started, because machine learning was quite a bit less promising than it is now. Tesla now seems to be all in on the software side and NN/machine learning in particular, with special hardware tailored to help, but at a reasonable cost. Also, they are focusing mostly on vision (no pun intended ).

It remains to be seen whether it is harder to make the more expensive sensors (e.g., LIDAR) cheaper vs. making the mostly vision-dependent software good enough on relatively cheap hardware. Time will tell. Sometimes first-movers get left behind because new technology/methods become available, and they remain committed to the old way, because it got them pretty far. (Like, say, I dunno, ICE manufacturers.) People in the AI world have been talking about "solving" vision for so long, it's almost a joke. However, I think there is a convergence of software (NN/ML, etc.) and hardware (powerful but affordable, purpose-built chips) that will usher in an unprecedented rate of progress.

It never ceases to amaze me how far some can jump to conclusions or claim to know what actually motivates someone else. It's also kind of sad that we feel the need to attack and defend more than listen and understand. This is a pretty amazing time. A lot of smart people are working on autonomous transportation, and there are multiple approaches that are showing both promise and concrete results. Most of us alive today will not have to worry about getting old and stuck at home with no way to go somewhere with (relative) independence, as our eyesight and coordination fail us (medical advances will help too!). I, for one, am looking forward to Tesla's presentation on Monday. I'm also looking forward to future announcements and progress from MobilEye, SuperCruise, Waymo, ProPILOT, and the rest. Some of the other companies seem to be dragging their feet (q.v., recent German alliance or Toyota--WTF?!). I wish them all good luck. In the end, we will probably learn something from all approaches, failure and success alike. Sorry, I'll step off the soap box now.

 

[CarlK]

Why is that impressive? It doesn't show that the car is driving by itself. It is just a video looking out forward from car driving. For all intention purposes it is just to show what a self driving car should be capable of doing while someone is driving the car.

It's not impressive even if the car did drive itself. It's only an edited video. Like the redacted Mueller report the meat is in the part they don't want you to see. Nothing short of a third party observer in the test drive can prove anything. Tesla is the only company that (has to) play it fair. It is putting cars in everyone's hands.

 

[diplomat33]

Folks, let's remember what Tesla wrote in the "Autonomy Investor Day" announcement:
"Investors will be able to take test-drives to experience our Autopilot software first-hand, including features and functionality that are under active development."

Don't fall for the FUDsters saying that it will just be some lame highway FSD demo. We know there will be actual test drives of new features!

 

[Daniel in SD]

It's not impressive even if the car did drive itself. It's only an edited video. Like the redacted Mueller report the meat is in the part they don't want you to see. Nothing short of a third party observer in the test drive can prove anything. Tesla is the only company that (has to) play it fair. It is putting cars in everyone's hands.

Well they are testing on public streets so people can observe. They should be required to have a light on top of the vehicle saying it’s in autonomous mode. I would say that if the car can navigate those situations even some of the time it’s pretty impressive. I like the video because it shows how ridiculously difficult FSD is.

 

[S4WRXTTCS]

That's like saying "Can we stop comparing Teslas to other EVs?"
Waymo and Cruise are pretty clearly the leaders in autonomous vehicle technology. It's hard to avoid discussing them on an "Autonomous Vehicle" forum.

The problem with Waymo is they're likely always going to be fleet vehicles. You won't ever have the opportunity to buy one so I'm not sure why we would compare a solution for most everyone versus a solution only for commercial vehicles.

Obviously Waymo can use considerably more advanced technologies on their vehicles, and their budget is significantly higher.

Cruise is possibly a good comparison if GM ever intends on bringing the technology to consumer hands, but I haven't heard of them actually selling autonomous bolts to end customers.

For myself I'm keeping my own comparisons between Audi, and Tesla.

They're really the only two companies where I'll have a chance purchasing a long-range L3 or better EV.

I'm also keeping an eye somewhat on autonomous vehicles in China. There is good reason to believe that they'll achieve mass scale autonomous driving well before anyone else does.

 

[Electroman]

Autopilot software first-hand, including features and functionality that are under active development

That phrase, "that are under active development" is the key here.

So don't expect an end to end FSD demo. It will show case a few critical capabilities that highlights the promise of near-future FSD, that will take more time to develop and mature into a production product.

I say that is perfectly fair.

 

[Daniel in SD]

The problem with Waymo is they're likely always going to be fleet vehicles. You won't ever have the opportunity to buy one so I'm not sure why we would compare a solution for most everyone versus a solution only for commercial vehicles.

Obviously Waymo can use considerably more advanced technologies on their vehicles, and their budget is significantly higher.

Cruise is possibly a good comparison if GM ever intends on bringing the technology to consumer hands, but I haven't heard of them actually selling autonomous bolts to end customers.

For myself I'm keeping my own comparisons between Audi, and Tesla.

They're really the only two companies where I'll have a chance purchasing a long-range L3 or better EV.

I'm also keeping an eye somewhat on autonomous vehicles in China. There is good reason to believe that they'll achieve mass scale autonomous driving well before anyone else does.

Waymo's initial plan was to develop a level 3 system which would be for consumer vehicles. They decided, rightly or wrongly, that a level 3 system would not be safe so they decided to work on a commercial service. Their goal has always been to get a product to market as soon as possible and they decided that the fastest way to get a product to market was to work on a robo taxi service. I would be very surprised if they have no plans to release a consumer system and there is no indication that they don't.
People here seem to focus a lot on the vision system but I think the path planning part of the problem is way more difficult (Watch that Cruise video!) and will be applicable no matter what your vision system is. And many of the truly difficult parts of the vision system (say, recognizing hand directions from a construction worker or path predicting pedestrians) are done with cameras on all systems.
The distinction between a fleet vehicle and a consumer vehicle seems besides the point to me. Tesla's system will have to work just as well as Waymo's.