Apple AI research has finally made it to publication as the Cupertino giant just published a paper on AI, contributing valuable research into improving the quality of synthetic images used for machine learning. Google, Facebook and Microsoft have long been sharing their AI research, now Apple finally relented and shared its own research, which contributes significantly to AI academic learning.

Apple AI research was submitted last Nov. 15 and got published on Dec. 22 where the paper revealed the techniques Apple used to enhance training of an algorithm and its ability to recognize images. Synthetic or computer-generated images are often used in machine learning instead of real-world images because of higher efficiency since these images are already labeled and annotated, making them easily recognized by the machine.

Apple AI research made use of synthetic images, a standard in machine learning, because these are less expensive, readily on-hand and customizable as opposed to real-world images. However, the use of computer-generated images also has its risk because in some occasions, what the algorithm has learned is not realistic enough or can be carried over to real-world scenes according to Forbes.

Hence, the Apple AI research has provided valuable insight because of its use of the Generative Adversarial Networks (GAN), which is not a novel idea, but an old one that Apple tweaked and modified for its own gains. Apple's technique employed a SimGAN variation that works in two ways - first, to minimizethe difference between synthetic and real images and second, to minimize the difference between synthetic and refined images in order to keep the image annotation, Tech Crunch reported.

In so doing, a photorealistic image is produced, bridging the gap between synthetic and real that was once problematic to Apple AI training. Moreover, Apple Ai research is able to enhance its modifications that prompts models to utilize the full history of refined images rather than confined itself to a mini-batch. In this way, the GAN can recognize computer-generated images as fake at any point in time.

Apple now joins the A-listers in AI learning research like Google, Facebook and Microsoft, which have long been contributing significantly to academic learning on AI. Relinquishing its stubborn refusal to share in the past, the Apple AI research is a good investment of the Cupertino giant into the pursuit and development of the AI technology.

Topics Apple, AI, Artificial intelligence