FSD beta waiting game

[diplomat33]

Yes, and this strikes me as incredibly intellectually lazy. VW is jumping to the 'fun' part of designing the interior of a driverless car, without dealing with the actual HARD problem of building full autonomy. It's like a kid who can't focus on the difficult math test so they just start doodling in the margins.

That is a bit harsh IMO. VW is partnered with Argo.AI which has very good autonomous driving. And VW with Argo have been testing L4 on the ID Buzz vehicle. So they have been putting work into building full autonomy. And since they know full autonomy will happen someday, they figured they would do a little exercise conceptualizing what a future L5 vehicle might look like. There is no harm in that. They are still working on fully autonomy with Argo.

 

[Daniel in SD]

I will say their press release is really misleading: "The all-electric powered Innovation Experience Vehicle (IEV) is a real prototype that drives autonomously (Level 5)". Uh, it does?

The term "drive autonomously" is commonly used for vehicles that still have a safety driver. They are prototypes of "Autonomous Vehicles"
For example as far as I know Zoox has never operated on a public road without a safety driver but they also claim to drive autonomously.
Interestingly the SAE explicitly says that they do not use the word "autonomous" in their definitions.

 

[dckiwi]

That is a bit harsh IMO. VW is partnered with Argo.AI which has very good autonomous driving. And VW with Argo have been testing L4 on the ID Buzz vehicle. So they have been putting work into building full autonomy. And since they know full autonomy will happen someday, they figured they would do a little exercise conceptualizing what a future L5 vehicle might look like. There is no harm in that. They are still working on fully autonomy with Argo.

This is good context, thanks - I wasn't aware of the Argo.AI partnership.

 

[dckiwi]

The term "drive autonomously" is commonly used for vehicles that have safety driver. They are prototypes of "Autonomous Vehicles"
For example as far as I know Zoox has never operated on a public road without a safety driver but they also claim to drive autonomously.
Interestingly the SAE explicitly says that they do not use the word "autonomous" in their definitions.

Except this press release states it "drives autonomously (Level 5)". And not only is it not level 5, but this specific prototype (GEN.TRAVEL) doesn't actually exist.

 

[diplomat33]

I will say their press release is really misleading: "The all-electric powered Innovation Experience Vehicle (IEV) is a real prototype that drives autonomously (Level 5)". Uh, it does?

My guess is they have a prototype that is fully functional, ie it has sensors that work and probably the Argo FSD stack. So it can drive autonomously. You probably would not actually take it on a road trip but it does "work".

But this is not the first time a company has demo'ed a concept driverless car. Mercedes did the same thing a few years back with their F015 "luxury in motion" concept driverless car that was technically driveable, although you probably would not actually take it on public roads.

 

[Daniel in SD]

Tesla also claimed to have an autonomous vehicle in 2016 (but they couldn't legally operate without a driver because California did not allow driverless testing until 2018).

Except this press release states it "drives autonomously (Level 5)". And not only is it not level 5, but this specific prototype (GEN.TRAVEL) doesn't actually exist.

I sounds like it exists, they just haven't shown it publicly yet. Obviously it doesn't actually work reliably enough to be driverless (just like FSD beta).

 

[dckiwi]

My guess is they have a prototype that is fully functional, ie it has sensors that work and probably the Argo FSD stack. So it can drive autonomously. You probably would not actually take it on a road trip but it does "work".

But this is not the first time a company has demo'ed a concept driverless car. Mercedes demo'ed their F105 prototype driverless car that was driveable, albeit on a closed empty road.

I'll try to be less harsh, but this video strikes me the same way. Lots of time and energy spent on slick videos, with well dressed actors sitting in a reimagined interior while the prototype performs simple lane keeping. But are these manufacturers (or their partners like Argo) tackling the hard problems of autonomy like Tesla is? Are they struggling with occupancy networks, accurately predicting the distance and velocity of cross traffic while performing ULTs, understanding intersection and lane topology, etc? Because until they do the hard work, they'll never be able to actually build these fancy concept cars.

 

[dckiwi]

I sounds like it exists, they just haven't shown it publicly yet.

This sounds like Tesla circa 2016-2020. And just like Tesla during this period, they are not showing anything because they have nothing to show.

 

[Daniel in SD]

But are these manufacturers (or their partners like Argo) tackling the hard problems of autonomy like Tesla is?

Yes. Everyone is investing absurd amounts of money into this absurdly hard problem. The marketing videos are pocket change.

 

[Daniel in SD]

This sounds like Tesla circa 2016-2020. And just like Tesla during this period, they are not showing anything because they have nothing to show.

Most of these companies show stuff at academic conferences (as does Tesla).

 

[calikid]

You will get what you paid for. The full FSD has not actually been released yet. FSD beta is simply an "early access" to the beta version. But everybody who paid for FSD will get the full version when Tesla releases it publicly.

What about the people who paid for FSD years ago and their leases ended? What did they get?

 

[diplomat33]

But are these manufacturers (or their partners like Argo) tackling the hard problems of autonomy like Tesla is? Are they struggling with occupancy networks, accurately predicting the distance and velocity of cross traffic while performing ULTs, understanding intersection and lane topology, etc? Because until they do the hard work, they'll never be able to actually build these fancy concept cars.

These automakers are outsourcing FSD to their partners. They will simply build the vehicles. But the partners are solving the hard problems of autonomy, yes. Earlier this year, Argo launched driverless testing in Miami and Austin. Ford-backed Argo AI launches driverless testing in Miami, Austin

 

[diplomat33]

What about the people who paid for FSD years ago and their leases ended? What did they get?

Yes, they got totally screwed.

 

[DelPhonic1]

I don't know how you can possibly say that. Their L4 is way more sophisticated than Tesla's L2. They have more sensors than Tesla. They have advanced sensor fusion. They also have way more advanced neural nets for perception, behavior prediction and planning. Their L4 can literally drive safely and reliably on their own with no human driver because they have incredibly detailed and accurate perception from dozens of sensors that can track 100's of objects, and can accurately predict behavior of objects, understand body language of pedestrians, hand signals from law enforcement. etc... and determine safe path in real-time, all from deep neural nets. It is way more advanced than Tesla's FSD.



It is not a meaningless distinction. Is the differences between an airplane and a boat meaningless to you? They might not be selling you the car but it matters if the system can drive on its own without supervision or if it requires driver supervision. That makes a big difference. If you don't understand the difference between autonomous and not autonomous, how can you even say that nobody knows how autonomous driving works? And you can't say that nobody knows how "the tech" works and then when people try to explain what the tech is, just go "I don't care, the tech is meaningless to me".

Can you post a link of a vid that shows a company like this and the car is actually doing it?

 

[diplomat33]

Can you post a link of a vid that shows a company like this and the car is actually doing it?

Yes. Here is Waymo driverless ride for over an hour:

Here is another 90 mn Waymo driverless ride:

Here is a 40 mn driverless Waymo ride:

 

[clydeiii]

To do L4, it would have to. And I see no reason it couldn't. Waymo has sophisticated vision that can read addresses on the side of buildings, road signs, street names etc... It would be a matter of training the vision to understand the special signs.

It's trivial to make it read the sign. It's hugely challenging to make it understand the sign and even more challenging to make it react appropriately in heavy traffic. Are you saying it can not just read, but understand and react, 100% of the time? Seems...unlikely. My guess is they're geofenced right now so that they can feed the brain the right answer in a limited geographical area, but I admit I haven't read much about Waymo's tech.

 

[diplomat33]

It's trivial to make it read the sign. It's hugely challenging to make it understand the sign and even more challenging to make it react appropriately in heavy traffic. Are you saying it can not just read, but understand and react, 100% of the time? Seems...unlikely. My guess is they're geofenced right now so that they can feed the brain the right answer in a limited geographical area, but I admit I haven't read much about Waymo's tech.

The geofencing is purely for ride-hailing. Geofencing is not to "feed the brain the right answers in a limited geo area".

But why would it be unlikely? Waymo has powerful HD cameras that can read signs 500 meters away and powerful NN. Machine Learning is quite powerful now. And Waymo also uses HD maps that can have information about signs, just in case the car is not able to see the signs because they are occluded or something. Yes, I believe Waymo can understand and react correctly to the signs in heavy traffic, maybe not 100% but very very close, like 99.99999%.

Waymo has advanced camera vision that can read and neural nets that can understand what signs mean. And Waymo has cameras, lidar and radar to accurately and reliably track every object in the road, read signs, but also has developed very advanced behavior prediction neural nets to predict what heavy traffic will do. Waymo has a NN called StopNet that can predict behavior of hundreds of objects.

A whole-scene sparse input representation allows StopNet to scale to predicting trajectories for hundreds of road agents with reliable latency

Research — StopNet: Scalable Trajectory and Occupancy Prediction for Urban Autonomous Driving — Waymo

We introduce a motion forecasting (behavior prediction) method that meets the latency requirements for autonomous driving in dense urban environments without sacrificing accuracy. A whole-scene sparse input representation allows StopNet to scale to predicting trajectories for hundreds of road...

waymo.com

 

[crusader2007]

Anyone might care to speculate what is the current queue for FSD beta? It has been a few weeks since last announced 10.69.2.1 and they said they had 60K slots to make a total of 160K on Beta. Is it worth waiting or chances are dismal at this time and better just play the lottery or wait for the next rollout? Just curious.

 

[LudicrousMeme]

Anyone else still waiting for FSD Beta? My safety score ranges between 97 - 99, which is really tough to achieve in Los Angeles! I average 200 miles of driving per week. I've been waiting 6 months, driving like a granny, desperately maintaining my safety score...

Does anyone else have this experience? The FOMO is killing me!

Thanks for any insight or therapy you guys can provide!! : )

 

[SilverString]

Anyone else still waiting for FSD Beta? My safety score ranges between 97 - 99, which is really tough to achieve in Los Angeles! I average 200 miles of driving per week. I've been waiting 6 months, driving like a granny, desperately maintaining my safety score...

Does anyone else have this experience? The FOMO is killing me!

Thanks for any insight or therapy you guys can provide!! : )

You might want to look at this thread. If your current software branch is higher than 2022.20 then you won't get the current beta release. Most non-beta cars are now on 2022.28 or 2022.36. If you look the TeslaFi data the only new added to the 2022.20.18 beta came from 2022.20.8 and .9, and there were only a few of those, about 1% of the update.

So you could just turn off the beta and re-apply once news of a new FSD update starts to leak. Clearly Tesla has backed off giving 10.69.3 to the 80 and above SS crowd.

PSA You will NOT get into the FSD beta if you already have 2022.24 or higher.

Iv seen many posts about people asking about how Tesla picks who gets in and who doesn't, so I thought I'd try and clear up one point that i keep seeing brought up. A big factor for people who say they have high scores and have alot of autopilot miles seems to be their software version. FSD...

teslamotorsclub.com