Special Reports

Apple’s First Artificial Intelligence Paper Published: Simulated & Unsupervised Learning Key In Boosting Realism Of Simulated Images


It is a relief for the artificial intelligence research community that finally Apple, branded as a secretive company, has now published its very first AI paper on Dec. 22. It can be noted that artificial intelligence development and researches lately have increasingly becoming topics of interest including other tech giants like Microsoft, Uber, Google, Facebook, IBM, Skype, Salesforce, and Shell as it is seen to be the next big thing in the tech world.

Apples's contribution to the AI research community through the paper describes a technique on improving the training of an algorithm's ability to recognize images using computer-generated images rather than real-world images, Forbes reported. In machine learning, using of synthetic images like those from a video game to train neutral networks, which are designed to think in the same way as human brain are much efficient compared to using real-world images.

Reports said that it would be easier for the network to learn synthetic image data since it is already labeled and annotated compared to using real-world images where everything still needs to be exhaustively labeled, Business Insider reported. According to Apple, since synthetic image data is often not realistic enough, it leads network to learn details only present in synthetic images and fails to generalize well on real images.

To improve with synthetic data, Apple researchers lead by Ashish Shrivastava suggests a simulated and unsupervised learning to boost realism of a simulated image. This technique combines unlabeled real image data with annotated synthetic images. Apple also pits two neutral networks against each other to generate photorealistic images in its modified version of a new machine learning technique called Generative Adversarial Networks.

The move of Apple to finally release research results specifically on its AI researches was widely commended, as this could help in the increasingly indispensible role of AI to cutting-edge innovations sprouting in the world today.

© 2024 University Herald, All rights reserved. Do not reproduce without permission.
Real Time Analytics