FSD rewrite will go out on Oct 20 to limited beta

[willow_hiller]

Because nothing like that was stated by me (or even tweets I retweeted) and because I find this notion ridiculous, I am trying to understand how come what I said could be misinterpreted in such a horrible way.

Do you agree with all of the statements made by the account you retweeted? They went on to say "Tesla's cheating with pre-loaded LiDAR-made HD maps." https://twitter.com/GretaMusk/status/1320129690322030592?s=19

I don't think there's any evidence that supports the notion that the resolution of details in the maps used by Tesla would require LIDAR to create.

 

[mikes_fsd]

I am trying to understand how come what I said could be misinterpreted in such a horrible way.

Dude, you literally quoted, via retweet posts from TSLAQ stating that Tesla pre-loaded HD Maps.
Then you reply to your own thread stating that you believe that HD maps are useful and you cannot understand "why people actively try to refuse the need for HD maps"

Did you forget what you wrote?

 

[alsetym]

Do you agree with all of the statements made by the account you retweeted? They went on to say "Tesla's cheating with pre-loaded LiDAR-made HD maps." https://twitter.com/GretaMusk/status/1320129690322030592?s=19

I don't think there's any evidence that supports the notion that the resolution of details in the maps used by Tesla would require LIDAR to create.

Regardless of the level of mapping being used, the NN is clearly doing a good job predicting road features, which is its function. Kinda funny it is 'sufficiently advanced as to seem like magic ' to make the crazies think tesla is faking the whole thing. Or do they actually even believe that ? I don't put green in that camp BTW, his argument is much more pragmatic and defensible (but probably wrong lol)

If I was going risk my multibillion $ company and fake something I'd at least make sure it worked perfectly lol.

 

[srk67]

The way I see it is teslaq is saying that tesla needs hd maps comparable to the lidar ones google uses to work, whereas green is saying that tesla is using hd maps that the fleet automatically generated, but they are only used to aid fsd and improve performance

 

[mikes_fsd]

The way I see it is teslaq is saying that tesla needs hd maps comparable to the lidar ones google uses to work, whereas green is saying that tesla is using hd maps that the fleet automatically generated, but they are only used to aid fsd and improve performance

No, green is trying to have his cake and eat it to.

He claims to only report what he "sees" but instead what we get is him trying to redefine the terms to fit his desire of "how it should be"!
Because HD Maps are "widely used" and they "help a great deal" and according to him "the need for HD maps" - he jumps to: Tesla must be using HD maps.

Talk about not keeping an open mind.

Note, his current argument, is "where did I say that"?
When he quotes TSLAQ cultists of using pre-loaded lidar-made HD maps and says, "say whatever you want"

Then when pressed, he can't accept that the NN's combined with geo-coded map data - like Karpathy said - could generate what he see in the beta videos, so to him, it must be "very precise maps" -- trying to yet again redefine "HD maps".

Yeah, all those road features, the cross walks, stop lines, traffic signs and more are recorded with great accuracy. If you reread what I said in the tweet, you can argue all you want this is not centimeter level accuracy, but it's at least multi-centimeter level or otherwise very precise.

Lastly, here is how Waymo describes their HD Maps:

Before our cars drive in any location, our team builds our own detailed three-dimensional maps that highlight information such as road profiles, curbs and sidewalks, lane markers, crosswalks, traffic lights, stop signs, and other road features.

This is what he is referring to when he says HD maps are "widely used"!

 

[outdoors]

Wonder if we could do a Tesla is using sorta HD maps or not thread. Just an idea.

 

[diplomat33]

Wonder if we could do a Tesla is using sorta HD maps or not thread. Just an idea.

We really don't. LOL.

We keep rehashing the same debates over and over again.

 

[Tiger]

You posted a video from 2014 and say “just announced”?

Well, it is the latest and most accurate information, to date, if we take the latter remarks with a grain of salt (as proven in the past).

 

[mikes_fsd]

Wonder if we could do a Tesla is using sorta HD maps or not thread. Just an idea.

I mean, this all came to the top again because FSD beta i.e. "the rewrite" was released. Seems like a fitting thread to discuss in.

 

[daktari]

Another one from Brandon with a nice perspective from a drone. Still scary as hell. Tesla should be pulling footage from his car like crazy to adjust their systems, if they’re not already.

That is a great video, and the NN in the drone is pretty good!
So in the 2 minutes and 16 seconds, it almost hits a parked car, it flyes past a pedestrian zone (a cautios person would keep speed low there), it takes a early yellow (guess OK) and it clips a parked truck very close at a decent speed, instead of keeping distance to a maybe-opening-door. So an 18 year old would probaly fail your driving exam in 2 minutes with this driving. With these sorts of fails, I wonder if they soon should pause the beta before a new version comes out.

Question is: are these fails a sensor problem? are they a driving rule problem? What would be the strategy for improving? Are Tesa employees now reviewing all this uploaded video, tagging as crazy, for then to retrain the NN with better data (or are there many NNs?)

Exciting times, that is for sure!

 

[ZeApelido]

The cars in the 25 metro areas are using the same autonomous driving software as the driverless cars in Chandler. It is the same autonomous driving software with or without a safety driver. So yes, the test cars in those 25 metro areas are autonomous cars. They don't stop being autonomous just because a person is sitting in the driver seat.

So when Waymo had a few "autonomous" cars in Mountain View 5 years ago (with much worse disengagement rate), all the others with drivers were also "autonomous"?

So, by your logic, if Tesla unleashes a few driverless cars in a suburb of Phoenix, then they would suddenly have a fleet of autonomous vehicles nationwide??

 

[ZeApelido]

This is quite impressive for the neural network to predict the intersection, and I would think the Autopilot team would take it as a compliment that people could mistake this quality for something else.

Yes, this is just like when someone asks me if I'm taking steroids.

 

[Bill Price]

new 1 hour and 24 minute video...also did not know this: "Tesla called me directly and explained that the AEB (Automatic Emergency Braking) has been disabled while using the FSD BETA toggle... I haven’t seen anyone bring this up, it’s very important IMO."

Watch Tesla Full Self-Driving Beta Drive for 1 Hour and 24 Minutes [4K VIDEO] - TeslaNorth.com

Can anyone provide Cliff's Notes on this video? No way I'm going to watch even 24 minutes of someone else driving. Oh well, unless something dramatic happened, I pretty much know what it's about.

 

[Bill Price]

Hmmm, while there is some interesting information in this thread, it is starting to sound worse than a presidential debate sometimes.

If by that you are saying that one person is only disclosing the idiocy of the other, you may have a point?

 

[MP3Mike]

it clips a parked truck very close at a decent speed, instead of keeping distance to a maybe-opening-door.

But it didn't. This is a screen grab from where he reacted like it was going to side-swipe the truck:

It looks like there is plenty of distance there, and it is going wider... I think, but am not sure, that it slowed down because it thought the truck was going to pull out on it. And that it was the sudden braking that they reacted to.

He is providing a lot of good video, but in some cases I think he is overreacting. In other cases he takes over just because it wouldn't drive exactly like he wants to. (In one case even complaining that it didn't break the road rules like he would have.)

 

[turnem]

But it didn't. This is a screen grab from where he reacted like it was going to side-swipe the truck:

View attachment 602539

It looks like there is plenty of distance there, and it is going wider... I think, but am not sure, that it slowed down because it thought the truck was going to pull out on it. And that it was the sudden braking that they reacted to.

He is providing a lot of good video, but in some cases I think he is overreacting. In other cases he takes over just because it wouldn't drive exactly like he wants to. (In one case even complaining that it didn't break the road rules like he would have.)

He's certainly very cautious - almost overly cautious but it's hard to say without being in the car.

This video in particular showed me more promise than concerns honestly. I've seen a lot of good videos this weekend!

I can't wait to experience this for myself!

 

[diplomat33]

So when Waymo had a few "autonomous" cars in Mountain View 5 years ago (with much worse disengagement rate), all the others with drivers were also "autonomous"?

So, by your logic, if Tesla unleashes a few driverless cars in a suburb of Phoenix, then they would suddenly have a fleet of autonomous vehicles nationwide??

It is not my logic. I am simply following what the SAE says on page 30 of the J3016 JUN2018 document:

"The level of a driving automation system feature corresponds to the feature’s production design intent. This applies regardless of whether the vehicle on which it is equipped is a production vehicle already deployed in commerce, or a test vehicle that has yet to be deployed. As such, it is incorrect to classify a level 4 design-intended ADS feature equipped on a test vehicle as level 2 simply because on-road testing requires a test driver to supervise the feature while engaged, and to intervene if necessary to maintain safe operation."

So it does not matter if it is a test vehicle or a production vehicle or whether it is driverless or has a safety driver. If the production design intent of a ADS is a certain SAE level, then it is that SAE level.

 

[diplomat33]

One thing that is exciting with the FSD Beta is that more people are getting a first row seat to FSD development and are also discovering the challenges with developing FSD. Up to now, FSD has been very theoretical for a lot of people. But now, with videos and hopefully wide release soon, Tesla owners will experience all the many cases that FSD has to solve and they can see first hand how difficult it is to achieve reliable and safe FSD.

 

[daktari]

Can anyone provide Cliff's Notes on this video? No way I'm going to watch even 24 minutes of someone else driving. Oh well, unless something dramatic happened, I pretty much know what it's about.

Well, the driver is complaining that no other videos on youtube shows the fails he experience. It crosses a yellow line once. It zig-zag past cones. Struggles with trafic. And some decent driving also.

 

[Huskyf]

That is a great video, and the NN in the drone is pretty good!
So in the 2 minutes and 16 seconds, it almost hits a parked car, it flyes past a pedestrian zone (a cautios person would keep speed low there), it takes a early yellow (guess OK) and it clips a parked truck very close at a decent speed, instead of keeping distance to a maybe-opening-door. So an 18 year old would probaly fail your driving exam in 2 minutes with this driving. With these sorts of fails, I wonder if they soon should pause the beta before a new version comes out.

Question is: are these fails a sensor problem? are they a driving rule problem? What would be the strategy for improving? Are Tesa employees now reviewing all this uploaded video, tagging as crazy, for then to retrain the NN with better data (or are there many NNs?)

Exciting times, that is for sure!

They arbort the system because is affraid but on drone view at 2.00 they turn normally and I think at the end it break.