Elon: FSD Beta tweets

[diplomat33]

I think it's more about money than being lazy about sensor fusion. As an engineer, if I proposed something that was just as good but saved money, I got a few solid attaboys. In contrast, if I proposed something that was slightly better but cost more, I got a polite thank you. In my experience, both incremental improvements and cost savings associated with applying new ideas to an existing product were always modest.

Using more sensors sounds great, but it adds both cost and complexity. The latter generally reduces reliability. It's the same logic that Tesla used for its automatic wiper tech. Sure, a wetness sensor costs next to nothing but it also needs a connection to the wiring harness. Doing everything based on vision has obvious drawbacks, but it definitely saves money.

Good sensor fusion adds better performance and reliability. The question is does the increased reliability and performance outweigh the increase in cost. The answer is pretty clear, yes IMO. AVs with sensor fusion like Waymo and Cruise have far superior perception and FSD than Tesla. Elon is just hoping that with enough data and training that he can achieve the same performance as Waymo or Cruise without the extra cost because if he can, then it would be a huge win.

 

[EVNow]

I never said that they have all of self-driving figured out. I simply said that they have good sensor fusion. Not the same thing. "solving FSD" is more than just doing sensor fusion.

LOL. You have said before that Waymo has solved FSD. Infact I remember having a “discussion” with you on that.

 

[jebinc]

LOL. You have said before that Waymo has solved FSD. Infact I remember having a “discussion” with you on that.

Maybe his thinking has evolved?

 

[Knightshade]

Good sensor fusion adds better performance and reliability.

Can you cite where any of these folks you think have "good sensor fusion" have tested their systems against vision only?

Because if you can't I wonder how you can possibly state with such certainty they've 'added better performance' via "good sensor fusion" without having any data to back that claim up?

Maybe they just have REALLY BAD vision and AI and need the extra sensors as a crutch for it?

 

[Daniellane]

Could 2021.44.25.5 be the 10.8 Beta FSD/Holiday…? Appeared on a few cars within the last few hours on Teslascope

 

[aronth5]

Good sensor fusion adds better performance and reliability. The question is does the increased reliability and performance outweigh the increase in cost. The answer is pretty clear, yes IMO. AVs with sensor fusion like Waymo and Cruise have far superior perception and FSD than Tesla. Elon is just hoping that with enough data and training that he can achieve the same performance as Waymo or Cruise without the extra cost because if he can, then it would be a huge win.

I suspect you'll agree there will be room for different solutions targeted at different needs with different cost points. For example- The solution for someone who wants to use their car for true Robotaxi services is different than someone who is elderly and wants to keep their car. Unfortunately too many people get hung up on the various levels (L1-L5), get bogged down in the detail and forget the larger picture.

 

[EVNow]

Unfortunately too many people get hung up on the various levels (L1-L5), get bogged down in the detail and forget the larger picture.

Absolutely. Those levels are a red herring and as Hotz says there is no difference between levels - just a matter of error rates.

 

[Daniellane]

Could 2021.44.25.5 be the 10.8 Beta FSD/Holiday…? Appeared on a few cars within the last few hours on Teslascope
View attachment 747442

Got the answer from another thread…

Yes, this version is FSD Beta 10.8 w/ holiday update.

https://twitter.com/x/status/1474007106252468228

 

[TIppy]

Do you really think Elon just did eenie meanie miney mo?

No. But they did just pick one. He said they don't know which one is correct if they disagree, so what do they do now when vision is wrong? They have no way to detect this. At least if they had radar they would know they disagreed and that something was wrong rather than just blindly proceeding along. Better still they could have three sensors and pick the two that agree.

 

[EVNow]

They've been using valhalla and mapbox for routing for years now.

No - they have switched over to TomTom. See my sig.

ps:

Couple of issues I saw near my home (that I've previously mentioned here and corrected on TomTom) - look very different on MapBox.
- Extra road segment that TomTom/Bing had are not in MapBox (which just looks like OSM). But the car definitely wanted to take that non-existent road.
- One of the roundabouts that was in TomTom is not in MapBox. But the car definitely shows that roundabout - behaves like it is navigating the roundabout (putting on turn signal, stopping and waiting - and most importantly shows the roundabout in the navigation)

To me this is quite definitive that Tesla is not using MapBox "maps" - but TomTom. Now, Tesla could be using MapBox APIs with their own version of some kind of combined map that uses TomTom + OSM the basis. If Tesla uses a custom integrated map - we don't know how they handle conflicts and differences. Seems to me if they are doing that - they give priority to TomTom and supplement with OSM in some areas ?

 

[qdeathstar]

Good sensor fusion adds better performance and reliability. The question is does the increased reliability and performance outweigh the increase in cost. The answer is pretty clear, yes IMO. AVs with sensor fusion like Waymo and Cruise have far superior perception and FSD than Tesla. Elon is just hoping that with enough data and training that he can achieve the same performance as Waymo or Cruise without the extra cost because if he can, then it would be a huge win.

I disagree. Waymo can only perceive in a small geographic region. Waymo sees a painting, FSD beta sees the earth. FSD sees much more that waymo does

 

[diplomat33]

I disagree. Waymo can only perceive in a small geographic region. Waymo sees a painting, FSD beta sees the earth. FSD sees much more that waymo does

No. Waymo is geofenced but the camera vision, lidar and radar can perceive objects everywhere.

EDIT: the geofenced areas are just where Waymo has deployed a service for the public. It is not the limit of what the FSD system can do.

 

[Daniel in SD]

I disagree. Waymo can only perceive in a small geographic region. Waymo sees a painting, FSD beta sees the earth. FSD sees much more that waymo does

Arizona is the 8th most popular state for tourism. I bet they can also perceive people and vehicles from out of state.

 

[EVNow]

Arizona is the 8th most popular state for tourism.

Didn't know geofenced parts of Chandler, AZ were so popular

 

[EVNow]

It is not the limit of what the FSD system can do.

We just don't have independent public confirmation of what they can do.

 

[Daniel in SD]

Didn't know geofenced parts of Chandler, AZ were so popular

Chandler | Visit Arizona

Chandler is home to top resorts, restaurants and recreational opportunities that attract visitors year-round. Discover all the things to do in Chandler, AZ.

www.visitarizona.com

Chandler has long been an attractive outpost for relaxation and a hotspot for family and cultural activities.

 

[GlmnAlyAirCar]

Can you please take this diversion back to the Waymo thread so I can safely ignore it again?

 

[qdeathstar]

No. Waymo is geofenced but the camera vision, lidar and radar can perceive objects everywhere.

EDIT: the geofenced areas are just where Waymo has deployed a service for the public. It is not the limit of what the FSD system can do.

Based on dot dot dot

If they has good auto labeling and sensor fusion they wouldn’t have to be geofences.

 

[GlmnAlyAirCar]

Based on dot dot dot

If they has good auto labeling and sensor fusion they wouldn’t have to be geofences.

Please knock this *sugar* off.

 

[qdeathstar]

Please knock this *sugar* off.

Well would you look at me, PLEASE!?