Will Apple pull out artificial intelligence experts from NVIDIA and full-fledged driving car development is full swing?

ByCHRISTOPHER DOMBRES

Let's debut self-developed cars around 2020It is revealed that Apple rumored an expert in artificial intelligence and deep learning field from NVIDIA, a computer chip maker.

Apple Hires Nvidia Artificial Intelligence Director Jonathan Cohen | Re / code
http://recode.net/2015/10/23/apple-hires-an-artificial-intelligence-expert-from-nvidia-is-he-going-to-work-on-self-driving-cars/

Jonathan Cohen, who was found to have moved to Apple, was the director of deep learning research at NVIDIA. Nvidia, famous for the development of graphic boards, is also known for its focus on technological development in the field of autonomous driving in recent years. Mr. Cohen also publishes his transfers on the LinkedIn page.

Jonathan Cohen | LinkedIn


At NVIDIA, we are applying Graphic Chip, which the company has a lineup, as a chip for image recognition and radar information analysis, and in fact, we are developing an automatic pilot on-board computer based on the Tegra X1 processor "DRIVE PX"Is supplied to automobile manufacturers.

NVIDIA Tegra Automotive DrivePX | NVIDIA


One Apple, Tim Cook CEO at the event"The automobile industry is preparing for major changes"Entry into the automobile industry is strongly anticipated, such as Mr. Cohen will be supposed to support this rumor. Chris Nicholson, founder of Deep Learning Startup "Skymind.io", said: "This is an important event for Apple, as NVIDIA technology is being used for autonomous driving technology, so this drawing means It's obvious, "he says about Apple's aim.

The following movie contains a state that Mr. Cohen, who was enrolled in NVIDIA, will explain NVIDIA 's deep learning technology at the Consumer Electronics Show (CES) in January 2015. When I look at this picture, I think that it seems to understand why Apple came out of Mr. Cohen.

# CES 2015: Nvidia Drive PX Deep Learning - YouTube


Mr. Cohen, who will fit in the camera with Drive PX used for image analysis using deep learning. Deep learning reproduces human brain functionneural networkIn order to realize high throughput by stacking multiple layers of computers, which allows the computer to gradually have judgment capability by having the computer repeatedly learn.


Based on the image of the camera (30 fps) capturing the front of the vehicle, Drive PX will judge the situation. Although a white square indicating the vehicle is displayed on the display, you can see from the fact that it is perfectly tracked for the vehicle and it also recognizes the vehicle. In addition, words such as "Passenger Car (passenger car)" and "Van (van)" "SUV" are displayed on the square, and it is only the fact that not only the existence of the vehicle but also the vehicle type is recognized I understand.


It is said that the cognitive ability will become learned by showing every picture as thousands of times, tens of thousands times and learning it. By gradually learning the shape and size of the vehicle, it gradually becomes possible to grasp the characteristics of the vehicle. In the following scenes, you can see that they are perceiving patrol cars that fit perfectly behind themselves. This is encouraging.


He also boasts a high recognition rate even in night and rain running scenes. The image on the right shows the road of the night when rain falls, but you can see that it recognizes the vehicle in front. In addition, it senses that the brake lights are lit, and also has the ability to recognize that the former applied the brake.


The object of recognition is not only for vehicles but also for signals and road signs. In addition to the plate-like speed indicator standing on the side of the road, it also corresponds to signs using LEDs as follows. As the sign using LEDs repeats blinking at a high speed of 60 times per second, there are cases where the camera can not read the information well, but by proceeding with deep learning it is possible to correctly recognize even the problem by overcoming that problem It is going to be.


It is quite interesting to see what kind of movement will be caused by Mr. Cohen, who has developed such technology, transferred to Apple. Apple is expected to bring EV (Electric Vehicle) from 2019 to around 2020.

In addition, Tesla leading the EV world one step, in October 2015 we updated the vehicle software of "Model S" to support automatic operation. In addition to tracking and lane keeping of forward vehicles, we are realizing automatic lane change and the function of parking and searching for parking positions automatically, but that function is still at a stage where maturity will be advanced from now . In the following movie that contained the actual running scenery, scenes like erroneous recognition that occurred infrequently are stored, and finally it is told that it is necessary for people who are aboard to confirm safety.

Tesla drivers are having a little too much fun with auto-steer - YouTube