Apple Releases AI Research Paper, Focus On Computer Vision

Apple has been working on its artificial intelligence technology for years and it finally allowed the world to get a glimpse of its progress by publishing a paper on the subject. The paper was published with the help of Cornell University Library.  Based on the data, it would seem that much of the company’s focus is directed at computer vision. This directly ties in with Apple’s interest in autonomous driving.

At its core, the research paper unveils a rather novel concept that could change how AI are trained. Apple is essentially creating a program that can produce fake images that are practically indistinguishable from the real thing in order to train a machine how to recognize things like faces and objects, Fortune reports.

This addresses one of the underlying problems of training machines, which is the vast amount of resource needed in order to provide programs with enough data to learn from. For example, if researchers wanted to train an AI to recognize faces, they would have to flood it with millions upon millions of samples, which can be costly and takes a lot of time.

However, by having a machine to create all of the different facial features, colors, structures, and even texture, it would make it a lot easier to train an AI. This is exactly what Apple is proposing through its paper and calling it "Simulated+Unsupervised (S+U)."

The paper was published by six authors from Apple’s research division, Quartz reports. The findings also indicate that by using this type of training method, machines can also start recognizing things like different hand gestures and the direction that a person is looking at.

It would also seem that Apple is following what appears to be an emerging trend within the AI community. Instead of simply focusing on teaching machines how to learn, researchers are now focusing on creating new teaching methods.