Apple has published its first research paper about Artificial Intelligence. The study focused on AI learning and how to train models.
Quartz reported that the paper is authored by six Apple researchers: Ashish Shrivastava, Tomas Pfister, Oncel Tuzel, Josh Susskind, Wenda Wang and Russ Webb. It focused more on how to create enough data to train AI instead of mere interaction with the iPhone.
The paper was published through Cornell University Library. It is entitled "Learning from Simulated and Unsupervised Images through Adversarial Training."
The researchers noted that, with today's technology, training AI models on synthetic images has become more practical. However, this type of learning is not enough due to a gap between synthetic and real image distributions.
They proposed another type of learning named the Simulated+Unsupervised (S+U) learning. This is when a model is tasked to "improve the realism of a simulator's output using unlabeled real data, while preserving the annotation information from the simulator."
The method utilizes a different network that is similar to Generative Adversarial Networks (GANs). However, it will use synthetic images as inputs rather than random vectors.
Apple's research focused on having AI models identify hand gestures and detect where people are looking. To solve this, researchers took established data sets of synthetic images and utilized a neural network to refine and make them look more realistic.
Afterwards, the system compares the refined and real images. It then decides which picture is real and updates itself.
According to Tech Crunch, the research presents high potential and high risk. This is because minor glitches in the training material can cause major negative implications on the final product.
Fortune noted that Apple has long been criticized by the scientific community for being secretive about its findings especially with Artificial Intelligence. As a result, the company's reputation as a leader in technology has been questioned. Moreover, it also impaired the tech giant's ability to attract top talent.
