Twitter VS. Facebook What do Harvard teach from the stock price comparison?

Apple's acquisition of an AI company makes it possible to see the "future iPhone"

Apple recently acquired artificial intelligence (AI) startup Xnor.ai for $200 million. The acquisition of the company, which handles low-power machine learning software and hardware, reveals the introduction of AI into areas called "edges" that are close to users. In other words, Apple is aiming for a future where AI is installed in iPhones, Apple Watch, etc., and the terminal itself learns and becomes smarter.

Apple recently acquired a "light" artificial intelligence ( AI ) developer for $200 million. The essence of this acquisition is to maintain its superiority by advancing the introduction of AI to the side closer to the users, the area called "edge."

Xnor.ai, acquired by Apple, is a Seattle-based startup that specializes in low-power machine learning software and hardware. This field is also the main battlefield in the AI   field for Apple and other tech giants.

There, AI is gaining momentum with smartphones, smart watches, and other smart devices. These devices run computing "at the edge" rather than in the cloud. The key is to minimize power consumption.

Attempts to reduce power consumption

When an AI algorithm that requires a large amount of calculation is executed on a general-purpose chip, it tends to be large and consume a large amount of power. For this reason, startups have been appearing one after another with creative ideas, such as making AI models lightweight and running them on specialized hardware with extremely high power usage efficiency. Xnor.ai is one of them.

In March 2019, Xnor.ai demonstrated a computer chip that can perform image recognition using only power from solar cells. A research paper written by the company's founders and posted online in 2016 describes a more efficient "convolutional neural network." It is one of the machine learning methods especially suitable for image processing. The researchers succeeded in reducing the size of the network, mainly by simplifying the coupling between layers and building an approximate model.

Apple has already manufactured chips that perform certain AI operations, such as recognizing the phrase "Hey, Siri" that activates a voice assistant. However, the company's hardware will have to improve in the future without accelerating battery drain. I asked Apple for comment on this, but didn't get an answer.

Edge devices learn with AI

AI at the edge uses a trained model that can perform certain tasks, such as recognizing faces in a video or recognizing speech in a call. However, according to Mitra, edge devices may be available for learning in the near future.

This allows smartphones and other devices to improve performance without sending anything to the cloud. “It's a very exciting story,

As demonstrated by Xnor.ai, streamlining the way AI is applied to video is key for Apple, Google, and all the other mobile computing companies. Cameras and related software are important selling points for iPhones and other smartphones, and video-centric apps such as TikTok are gaining popularity among young smartphone users. Another advantage of edge computing is that you can keep your personal data on your device without sending it to the cloud.

Machine learning can also be used on Apple devices that are not currently equipped with AI, said Dave Schvemer, an analyst at research firm IDC. " I think it's feasible to run AI on an Apple Watch or AirPods, for example, to reduce noise, " says Shvmer. "There is also tremendous potential for existing products."

For further evolution of visual function

Advanced AI for videos, such as algorithms that can determine what is happening on the spot and add complex special effects, are generally executed in the cloud. This is because very large computational power is required.

“In some cases, for example, applying depth-of-field synthesis to photos requires estimating the depth of each pixel using a deep neural network,” says a professor specializing in computer vision at the Georgia Institute of Technology. Says James Hayes.

Apple can use Xnor.ai's technology in areas other than smartening iPhone cameras. Improving the ability of machines to recognize and understand the complex real world will be an important key to robotics, autonomous driving and the understanding of natural language.