For a long time, Apple segregated itself from the rest of the artificial intelligence community, refusing to attend conferences, or to let its researchers publish their work in existing journals.
Today, Apple got around this second problem in the most Apple way possible: by launching a machine learning journal of its own.
The blog, bearing the simple title Apple Machine Learning Journal, is a new repository for Apple’s machine learning research. Apple describes it as a place that will host posts written by Apple engineers, “about their work using machine learning technologies to help build innovative products for millions of people around the world.”
Interestingly, the usually-secretive Apple also encourages machine learning researchers, students, engineers, and developers to contact the company at the email address email@example.com, and notes that, “we’d love to hear your questions and feedback.”
At present, Apple has one paper available on the page. Titled “Improving the Realism of Synthetic Images,” it describes an alternative to supervised training of neural networks that involves using synthetic images from a simulator. As Apple writes:
“This is cheap as there is no labeling cost, but the synthetic images may not be realistic enough, resulting in poor generalization on real test images. To help close this performance gap, we’ve developed a method for refining synthetic images to make them look more realistic. We show that training models on these refined images leads to significant improvements in accuracy on various machine learning tasks.”
This builds on work Apple previously discussed in what was, to date, its only other published article about artificial intelligence.
A new commitment to AI
Until fairly recently, Apple was widely viewed as lagging behind its contemporaries when it came to machine learning, one of the most exciting current fields in tech. Despite Apple’s status as the top company in tech, its approach to ML was considered detrimental in terms of bringing on board new researchers.
For the past couple of years, Apple has been trying to reverse this perception. In addition to opening up about its research in academic circles, it’s also been more public about discussing the plethora of Apple products which now involve machine learning. Prior to this, Apple used machine learning (for example, for getting the keyboard to work on the original iPhone), but not with the enthusiasm of a company like Google.)
This year, it seems to have turned a particularly sharp corner. WWDC was full of references to machine learning, while Apple used the event to debut new machine learning tools and APIs — called Core ML — designed to let developers incorporate face tracking, face detection, landmark recognition, and more into their apps.
Apple is also rumored to be developing a chip designed specifically for handling machine learning task, which would make iPhones and iPads even more adept at carrying out these types of functions locally.
Hopefully this blog becomes the central hub of Apple’s new dedication to AI. Let’s hope it receives regular updates…