What's new

Tesla FSD 11.4: It's time for those who doubted the Tesla AI team to apologize and admit they were wrong

Tesla been talking about autonomous cars since 2015 tho

Earlier than that actually.

But remember they are choosing the hard route of relying on dynamic human-like computer vision (which BTW they are leveraging into their robots) to identify the environment.

Others are using matching real-time scanned LIDAR images to a huge pre-scanned and millimeter resolution pre-mapped database of city street (and their adjacent buildings) LIDAR images.

The Tesla one if done correctly should work in all countries without any pre-mapping (just Garmin quality road maps) while the LIDAR one will require all the streets/buildings in a country to be pre-scanned to millimeter accuracy (something which probably will not be allowed in many countries due to security reasons). To make matters worse those pre-scanned LIDAR images of a country will have to be updated frequently to handle any changes.

Tesla is gambling their strategy will be far easily accepted by almost all countries and thus giving them a self-driving monopoly quickly. This should almost be insurmountable as with over a million cars currently on the road learning new edge cases every day it will be too far ahead for a newcomer to surpass.

 
Last edited:
.
Tesla been talking about autonomous cars since 2015 tho

Another pro about the vision system is it will work in areas with no internet. This will be impossible for a LIDAR system which needs to stream in the pre-calculated map data.
 
. . . .
It is pretty cool, but It requires absolutely picture perfect roads with clear lanes, landscaping, black tar, stop signs, traffic lights etc - orderly traffic without any bikes or cyclists, only in California can you find that or maybe Tokyo or Seoul or couple of cities in Germany and China. Even in many parts of the US it would fail to work and lead to accident, let alone other parts of Asia, Latin America, Europe etc.

People will have to be very careful using this, assume you are going from rich town in California to some backwoods hood town, you might assume everything is going to go well, and then on the way as you enter the poorer part roads start deteoriorating, signs disappear and all, hope its going to throw you a warning that it cant keep up.

I dont see too many countries investing money anytime soon so that AI can drive, several European nations - UK, France, Spain, Italy, Portugal etc are fully developed nations but dont have roads like this. This car will remain limited to california for a long long time... it requires typical american suburban setting - suburbs with orderly roads connected by freeways..etc. Most of the world wasnt build this way.
 
Last edited:
.
orderly traffic without any bikes or cyclists, only in California can you find that or maybe Tokyo or Seoul or couple of cities in Germany and China. Even in many parts of the US it would fail to work and lead to accident, let alone other parts of Asia, Latin America, Europe etc.


There's 400,000 people in the US and Canada that have access to the FSD BETA. They are not all in California. But yes since Tesla is in California we should expect it to be trained more heavily on California's streets.

Tesla's SELF DRIVING Beta Avoids Deer At Night on Dirt Roads | Also Drives Without Headlights!
 
Last edited:
. .
And how many of them trust it and actively use it?

Well they paid like an extra ~$10K for it. If they aren't using it then why pay for it.

Well it's now $15,000
TeslaFSD.png
 
. .
I still prefer nVidia solution over Tesla's.

No doubt Tesla AI will manage the road well only if it's able to read the image correctly.

An additional 3D scanner is crucial, it adds more safety procedure if the AI fails to recognize the image.

It's aware there's something on the road despite it doesn't know what it is.
 
.
I still prefer nVidia solution over Tesla's.

No doubt Tesla AI will manage the road well only if it's able to read the image correctly.

An additional 3D scanner is crucial, it adds more safety procedure if the AI fails to recognize the image.

It's aware there's something on the road despite it doesn't know what it is.

Tesla was using the Nvidia Drive PX 2 but switched to a in-house custom chip as they were dissatisfied relying on apis they didn’t have full control over.

LIDAR is fine if it is used as a backup to cameras instead of vice versa. LIDAR is not going to know when a light is green, what a speed limit sign says, or if the guy ahead of you has their turn signal on. Roads are vision orientated. Elon Musk said automakers should be concentrating on getting their vision systems top notch instead of taking shortcuts with relying on LIDAR.
 
Last edited:
.
It wasn't China because there are no consumer cars in China that have the ability to drive on roads that have not been pre-mapped with LIDAR.

You have no consumer cars that can drive on other countries roads.

The only consumer car you do have that can currently drive on city streets in China is XPENG and the only reason it can do that is because they use an off-the-shelf NVIDIA Drive system solution.

The same NVIDIA Drive system in a Mercedes.

Even the Dongfeng, which sells for as little as $12,000, has autopilot.

It's not high end technology.

 
. .

Tesla to license FSD to other OEMs, allow transfer of FSD to new cars​

Tesla CEO Elon Musk confirmed Wednesday that the automaker is “in discussion with major OEMs about using Tesla FSD.”

Tesla Full Self-Driving is the automaker’s beta advanced driving assistance system (ADAS) that can automate driving tasks on highways and urban streets. If other automakers were to adopt FSD technology, they would need to onboard both Tesla’s software and hardware suite. Tesla’s approach to ADAS, and ultimately to autonomy, is to rely only on computer vision processing, or cameras, rather than a range of sensors like lidar and radar.

During Wednesday’s second quarter earnings call, Musk also said Tesla will allow the transfer of FSD software to new vehicles, but only in the third quarter.
 
.
Back
Top Bottom