Apple hasn’t been very vocal about AI so far (surprise, surprise!), however, a recent paper submitted by the company tells us the real story of how very advanced the Apple AI research actually is.
Quartz reports —
While it’s clear Apple has been working on machine learning for years, it’s notable that the company’s first paper on the subject took a top conference prize.
Titled”Learning from Simulated and Unsupervised Images through Adversarial Training“, the paper deals with creating better fake images (of people) and training AI systems to better recognise facial patterns.
The paper came second at the international conference it was submitted at, and was beaten only by Facebook, whose facial recognition is perhaps the strongest non-military software out there. The fact that Apple’s paper doesn’t deal with iPhone specific issue, but rather, with a much more varied and complicated issue shows the range of Apple AI research.
Of course, the research does have an iPhone application. Quartz explains —
“Apple’s published research focuses on those two examples: identifying hand gestures and detecting where people are looking, examples of basic image recognition problems that could be applied to anything from tracking user behaviour to a wave-to-unlock iPhone feature”
The report goes on to lay out the basic questions/problems that the Apple AI research team is focusing on:
a) recognizing and processing images,
b) predicting user behaviour and events in the physical world,
c) modelling language for use in personal assistants, and,
d) trying to understand how to deal with uncertainty when an algorithm can’t make a high-confidence decision.
Bigger fish to fry
All this shows that Apple, while maintaining their sagely silence on the issue, are secretly working on something much, much bigger. In retrospect, even the Apple Car seems a little childish and largely inconsequential. For now, everything seems quite hush-hush. What secret treasure lay at the heart of Apple AI, only time call tell. Or Tim Cook.
