Welcome to Tesla Motors Club
Discuss Tesla's Model S, Model 3, Model X, Model Y, Cybertruck, Roadster and More.
Register

Tesla is dumping Mobileye???

This site may earn commission on affiliate links.
This seems quite surprising considering past statements about how crucial Mobileye's technology was. In one article I researched again this afternoon, Elon said that Mobileye's technology is the best in the world & that is why Tesla uses their tech. I wonder who will supply the AP camera & chip now.
Some addition details on the MobileEye camera from George Hotz, who's famously not a fan of MobileEye.

He said the MobileEye camera is an RCC digital camera, meaning RED-CLEAR-CLEAR. Implying it's a grayscale camera that only sees the luminosity in the RED channel. He stated that THIS was the reason why it couldn't distinguish a white truck against a bright blue sky. The camera in comma.ai's system will be a full color camera, and I would think Tesla may use a color camera in the future as well.

The interview is worth watching if you're a fan of autopilot and technology in general.

 
Suggest changing the thread title...

It was MobilEye who dumped Tesla. Not the other way around. MobilEye opted not to renew Tesla's contract, citing concerns with Tesla's implementation of Autopilot using sensors that were never designed to handle a lateral crash such as the one that occurred recently resulting in the death of a Tesla owner.

http://www.bizjournals.com/sanjose/news/2016/07/26/mobileye-and-tesla-part-waysamid-autopilot.html

Except that they did not - you're quoting bizjournals who is quoting marketwatch. If you go to the original source says that both companies did not see a way forward and both did not want to say much about dissolution of their partnership. Marketwatch as well other sources note however that Tesla was not happy with Mobileye.
 
They may be jumping onto the Cruise Automation system! Fully autonomous vehicles are already on the road operated by Cruise in the Phoenix area. GM owns Cruise, purchased for $1B in March of this year!

The founder is the same kid that challenged Elon 2 years ago that his system was better and Elon said to him that if he proves it then he'd be seeing TESLAs business immediately! I'd say GM spending $1B on them may just be the "proving it" part...
 
Based on Google news:
In the wake of a fatal crash, Tesla will quit using Mobileye's chips for Autopilot vision
The car company will likely work on its own image-recognition technology.
BY JOHANA BHUIYAN @JMBOOYAH JUL 26, 2016, 2:23P
Weeks after a fatal Tesla accident involving the company’s semi-autonomous tech sparked probes from federal agencies, Tesla parted ways with Mobileye, the startup that supplies the chips that enable Teslas to recognize images and to work with the car’s Autopilot system.

This also means Telsa is likely working on its own chips to improve the semi-autonomous features.

Mobileye CTO Ammon Sashua made the announcement during a call discussing the company’s second-quarter financial results.

“We continue to support and maintain the current Tesla Autopilot product plans. This includes a significant upgrade of several functions that affect both the ability to respond to crash avoidance and to optimize auto-steering in the near term, without any hardware updates,” the company said in a statement. “Nevertheless, in our view, moving toward more advanced autonomy is a paradigm shift both in terms of function complexity and the need to ensure an extremely high level of safety. There is much at stake here, to Mobileye’s reputation and to the industry at large. Mobileye believes that achieving this objective requires partnerships that go beyond the typical OEM / supplier relationship, such as our recently announced collaboration with BMW and Intel. Mobileye will continue to pursue similar such relationships.”

In the hours after the announcement, Mobileye’s stock dropped more than 7 percent.

When asked why the two companies parted ways, Sashua pointed to the companies’ respective responses to the fatal accident. Mobileye’s technology is only capable of helping to avoid accidents with cars in front of it, not trucks crossing the highway laterally, as was the case in this accident.

“This incident involved a laterally crossing vehicle, which current-generation AEB systems are not designed to actuate upon,” the company wrote.

Tesla said Autopilot, which combines proprietary and third-party technology, is supposed to be able to recognize “any interruption of the ground plane in the path of the vehicle” but “the high, white side of the box truck, combined with a radar signature that would have looked very similar to an overhead sign, caused automatic braking not to fire.”

While it’s not clear which party ended the relationship, Comma.ai CEO George Hotz — whom Tesla CEO Elon Musk once challenged to build better software than Mobileye’s — said: “Without a doubt it’s Tesla who parted ways.”

Hotz, who also challenged Musk to a race of their respective semi-autonomous cars when his was done, said Mobileye isn’t innovative enough for Tesla.

“Their stated mission is to lower the safety rating of cars that don’t have their technology,” he told Recode. “So when you have a company like that versus a company like Tesla who actually wants to build incredible cars, there’s absolutely no reason Tesla needs Mobileye.”

It’s likely Tesla — which recently tapped former Advanced Micro Devices chip engineer Jim Kellerto be its vice president of Autopilot — will begin working in-house on the technology for which it once depended on Mobileye. And according to Hotz, it won’t be hard.

“Tesla has a lot of trouble attracting machine-learning talent. They don’t have a cafeteria, for example. So people are going to Google and Apple,” he said. “But to be honest, the Mobileye system is so easy to reproduce; you don’t need the best talent. I did it in a couple of months. Our software can already do more than Mobileye’s.”

According to the company, Mobileye will continue to pursue partnerships like its recently announced deals with BMW and Intel to create autonomous cars. BMW has yet to ship any semi-autonomous tech. But Hotz says its proclivity for working with established players is exactly why he thinks Mobileye isn’t concerned with innovating.

“I met with some automakers and all they seemed to be about was their brand,” Hotz said. “They say, ‘Well, no people buy — insert crappy car Y here — because their parents had crappy car Y. That’s not what sells cars anymore. What sells cars really is the car experience. When Apple comes out with their car and even the Tesla Model 3, [automakers] will see they’re not in the same ballpark. And then a company like Mobileye wants to work with these established players.”

“It’s like you see the same thing in mapping — all the losers are trying to group up together,” he said.
 
Wow, this is almost insane. Mobileye has very sophisticated chips that run complex neural network algorithms to understand what the camera is seeing. Tesla must now source (or build, gulp) chips from someone else and develop equivalent neural networks to recognize cars, paths, people, etc..

3. Tesla has the unique advantage of high speed, constant connection with their cars, with rapid update capability, which none of the other MobileEye customers has. This again allows Tesla to collect more data and evolve status quo quicker, while MobileEye is having a hard time catching up to them.
Elon Musk is very far-sighted, and he had wisely decided to NOT share the data collected from Model S/X with MobileEye, which means the in house DNN Tesla trains currently maybe leaps and bounds ahead of MobileEye now.

yes whatever mobileye used to train their DNN, Tesla has much better data now with so many cars on the road. they could build a much better DNN with that data that no one can touch.
 
...Mobileye has very sophisticated chips...

Current Mobileye system is too dependent on road markings.

As shown in youtube by another thread, the Autopilot was centering the lane fine until there's a break of lane markings coming up.

Look at the left red mark that locates the left yellow lane marking before the road marks are erased (with unmarked fresh concrete):

tc3zOWh.png


After the Autopilot encountered the missing lane markings and passed over them, look at the left yellow lane marking and the left red mark, Autopilot then veered toward the right and hit the upcoming right traffic cones.


vqVtH2X.png



Other system like Nvidia can learn how to steer even in the absence of road markings.

It can understand the rules of engagement between vehicle and road.
 
Sometimes, the little things matter, especially to Millennials. Don't discount free gourmet food.

It helps that Google and Apple are wildly profitable and stable. Tesla, not so much.

The little things matter to everyone. Do you know how many 60 year old doctors I have seen leave the comfort of their home to go get free food at the hospital cafeteria?

Everyone likes to feel appreciated and if you commit to working on Tesla's autopilot team you are likely committing to 70-90 hour work weeks. You will want to be very comfortable at work and the lack of a comfortable environment which you will spend the next years of your life in would be a big deal to anyone. :-D

George is a smart guy but probably not the best at articulating things. The broad point he was trying to make is valid though. What he said about recreating MobilEye's software in a few months is also true. It is not absurdly complicated.
 
This will alter people's expectations of AP 2.0 from early next year to late next year IMHO.

There won't be any new AP hardware till at least the release of the 3, and that doesn't even mean it will be activated. There is a chance that the model 3 will be released with the current hardware and they will try to push it to the limits.