The only thing Tesla has over Mach-E

[HuntingPudel]

<SNIP>
Also not sure I’m ambitious enough to remove the tub just to take a picture.
<SNIP>

I just had my 2024’s frunk tub out for the 5th time. I can’t count the number of times my 2021’s tub was out. It’s super easy on a 2024+ car. Pop the three beauty panels, undo 6 screws, undo two electrical connections, pull tub. It takes me more time placing the beauty panels and tub in places where they won’t get scratched up than it does to remove everything. ??

Sponsored

 

[tobiasjef]

No, it was purely because they didn’t have a way to safely monitor the driver to make sure they’re paying attention.

Incorrect. Tesla has had an interior camera for driver monitoring for years. They never used it because autopilot is not hands free and FSD was pushing the limits of what was being done in a car.

Tesla was abusing a loophole to claim their system is level 2 when it is not. Tesla has been testing a level 4 system but in order to make it appear that you are more than a safety driver, they have required hands on the wheel.

That's why tesla was also very slow to release "old" features like actually smart summon. Tesla has also had every element of self driving (starting from park, auto parking, summon, banish, etc.) since 2023 but they have never connected those features into a hands free, parking lot to parking lot driving experience.

Now in 2025 you can finally use FSD in a way in which you can go without touching the steering wheel and pedals for a year.

In fact tesla has the strictest driver monitoring of any system. You can't crawl in the backseat because new cars have in-cabin radar which can detect humans in the front seat.

 

[tobiasjef]

Multi-sensor data fusion is a perfectly valid way to develop an understanding of an environment

Sensor fusion is a process of combining sensor data or data derived from disparate sources so that the resulting information has less uncertainty than would be possible if these sources were used individually.

Ford is NOT using sensor fusion in that they are sensing lane lines with cameras and sensing cars with radar. That's the least safe way to do it.


Cruise control radar has very low resolution and it is only useful for detecting objects of a similar speed to what you are moving. That means no stationary objects.

With such a fundamentally dangerous limitation of radar, if 99% of the time, cameras and radar see the same depth, but 1% of the time there is a disagreement and radar is the one that is wrong, what is radar providing? Nothing.

The engineers at tesla only wanted radar to remain in the cars because radar helps with the data labeling they do as they get the ground truth easier for labeling their videos.

Secondly, tesla IS using sensor fusion if they are only from cameras. They are using a technique called vidar which is commonly used in self driving where pixels are triangulated based on overlapping camera views. Tesla has overlapping in the form of front cameras as well as a bumper camera. In the side, the repeater camera, 180 degree front camera, and the b pillar camera all overlap.

IF you look at the disengagement stats for FSD, 80% of disengagements would be solved by better map data. That is an easy fix as maps can be done using machine learning. You also do not need to remap everything as only disengagements would be fixed.

The other 20% are just sign reading issues or issues reacting to school buses and emergency vehicles.

Fix those two big ticket items and tesla's miles per disengagement will shoot way up. Signs can also be done using mapping as not every non-read road sign is a problem but only the ones which cause disengagements.

 

[Jimrpa]

Sensor fusion is a process of combining sensor data or data derived from disparate sources so that the resulting information has less uncertainty than would be possible if these sources were used individually.

Ford is NOT using sensor fusion in that they are sensing lane lines with cameras and sensing cars with radar. That's the least safe way to do it.


Cruise control radar has very low resolution and it is only useful for detecting objects of a similar speed to what you are moving. That means no stationary objects.

With such a fundamentally dangerous limitation of radar, if 99% of the time, cameras and radar see the same depth, but 1% of the time there is a disagreement and radar is the one that is wrong, what is radar providing? Nothing.

The engineers at tesla only wanted radar to remain in the cars because radar helps with the data labeling they do as they get the ground truth easier for labeling their videos.

Secondly, tesla IS using sensor fusion if they are only from cameras. They are using a technique called vidar which is commonly used in self driving where pixels are triangulated based on overlapping camera views. Tesla has overlapping in the form of front cameras as well as a bumper camera. In the side, the repeater camera, 180 degree front camera, and the b pillar camera all overlap.

IF you look at the disengagement stats for FSD, 80% of disengagements would be solved by better map data. That is an easy fix as maps can be done using machine learning. You also do not need to remap everything as only disengagements would be fixed.

The other 20% are just sign reading issues or issues reacting to school buses and emergency vehicles.

Fix those two big ticket items and tesla's miles per disengagement will shoot way up. Signs can also be done using mapping as not every non-read road sign is a problem but only the ones which cause disengagements.

First, congratulations on successfully using Google to parrot a definition of multi-sensor data fusion.
Second, I never said that the current iteration of Ford BlueCruise implements robust multi-sensor data fusion. In fact, about the only comment I made was to the effect that Ford doesn’t have a very diverse sensor suite in the current iteration of BlueCruise.
You’ve made your perspective abundantly clear and all I can say is, as I said before, enjoy your Tesla.

 

[tobiasjef]

Second, I never said that the current iteration of Ford BlueCruise implements robust multi-sensor data fusion

I'm saying ford is NOT using sensor fusion at all. And ford does have a perfectly adequate sensor suite. It's called cameras.

The issue is doing what tesla is doing is not easy and the only one attempting it right now is rivian. Rivian intends on doing what tesla is doing as they are collecting 100s of gigabytes a month of video from cars.

It's also worth mentioning that a large percentage of tesla's "training" comes from simulated videos. Something no automakers are doing

If tesla can solve sign recognition via mapping then they can do unsupervised already. They also need to limit the car from driving across train tracks just like waymo does. The behavior is not good enough.

 

[Jimrpa]

I'm saying ford is NOT using sensor fusion at all. And ford does have a perfectly adequate sensor suite. It's called cameras.

The issue is doing what tesla is doing is not easy and the only one attempting it right now is rivian. Rivian intends on doing what tesla is doing as they are collecting 100s of gigabytes a month of video from cars.

It's also worth mentioning that a large percentage of tesla's "training" comes from simulated videos. Something no automakers are doing

If tesla can solve sign recognition via mapping then they can do unsupervised already. They also need to limit the car from driving across train tracks just like waymo does. The behavior is not good enough.

I’m tapping out. Whatever you say. If you want to hear “you win”, then there you go. Elon Musk is the most brilliant person who’s ever lived and any thought he has is perfect and infallible.

 

[tobiasjef]

I’m tapping out. Whatever you say. If you want to hear “you win”, then there you go. Elon Musk is the most brilliant person who’s ever lived and any thought he has is perfect and infallible.

What does this have to do with elon? Elon is not doing anything related to self driving.

The only decision tesla made was to not use lidar which is an aesthetic decision among others.

Tesla has limitations with camera placement, but it's not a lidar issue.

The other issue of tesla is actually a compute limitation where tesla is doing self driving on only 150w of compute.

Waymo uses 4000w of compute which is why they have a "tech demo" that will never be profitable

 

[Ahlarict]

Cobbling things together from the existing parts library is what Ford engineers do best.

I’m sure Munro or someone like that will give us a good look as the ‘25s start Landing in customer hands. While I’m curious, I’m sure they’ll be pictures and videos of the new heat pump before I get my car. Also not sure I’m ambitious enough to remove the tub just to take a picture.

I’m always perplexed how people seem to trust BlueCruise or GM’s SuperCruise when they don’t trust Tesla FSD. Also can’t understand how the NHTSA and IIHS give Ford and GM a pass to be hands-free, while Tesla can not be. None of them should be hands-free, nor trusted to fully drive without driver supervision and awareness.

God love 'em for trying, but I'm expecting miles and miles of inexpensive-but-not-particularly-elegant-nor-efficient rubber tubing snaking their way through every possible corner of the car if the Lightning's thermal system is anything to judge by...

 

[tobiasjef]

One random one I forgot. Tesla seems to be one of the few EV brands with no shark fin antenna.

hyundai hides the shark fin antenna in their advertising pics as if they are ashamed of it

 

[Kamuelaflyer]

Meanwhile, we Europeans are waiting for anything remotly ressembling an Auto Pilot...

 

[Ahlarict]

Shirley that'll happen soon!

 

[tobiasjef]

People joke about the name "autopilot" but have you used any autopilot systems, i.e planes and boats? They don't do very much

i don't see how tesla's name is very misleading

 

[Mustangor]

People joke about the name "autopilot" but have you used any autopilot systems, i.e planes and boats? They don't do very much

i don't see how tesla's name is very misleading

Really ? I thought planes were able to land without any human intervention... Looks pretty cool to me.

 

[GreaseMonkey]

One random one I forgot. Tesla seems to be one of the few EV brands with no shark fin antenna.

hyundai hides the shark fin antenna in their advertising pics as if they are ashamed of it

You work on commission, bruh?

 

[MacherAWD]

Sensor fusion is a process of combining sensor data or data derived from disparate sources so that the resulting information has less uncertainty than would be possible if these sources were used individually.

Ford is NOT using sensor fusion in that they are sensing lane lines with cameras and sensing cars with radar. That's the least safe way to do it.


Cruise control radar has very low resolution and it is only useful for detecting objects of a similar speed to what you are moving. That means no stationary objects.

With such a fundamentally dangerous limitation of radar, if 99% of the time, cameras and radar see the same depth, but 1% of the time there is a disagreement and radar is the one that is wrong, what is radar providing? Nothing.

The engineers at tesla only wanted radar to remain in the cars because radar helps with the data labeling they do as they get the ground truth easier for labeling their videos.

Secondly, tesla IS using sensor fusion if they are only from cameras. They are using a technique called vidar which is commonly used in self driving where pixels are triangulated based on overlapping camera views. Tesla has overlapping in the form of front cameras as well as a bumper camera. In the side, the repeater camera, 180 degree front camera, and the b pillar camera all overlap.

IF you look at the disengagement stats for FSD, 80% of disengagements would be solved by better map data. That is an easy fix as maps can be done using machine learning. You also do not need to remap everything as only disengagements would be fixed.

The other 20% are just sign reading issues or issues reacting to school buses and emergency vehicles.

Fix those two big ticket items and tesla's miles per disengagement will shoot way up. Signs can also be done using mapping as not every non-read road sign is a problem but only the ones which cause disengagements.

Don't they have an extremely high accident rate and high insurance? Clearly the systems are not that safe when in actual use. The fact that in a lab they are some of the safest tested but when in actual use are some of the most dangerous shows clear design flaws. And they don't release data on if FSD was in use just before an accident, so not sure I would trust their numbers.

Sponsored