After years of near-silence, Apple is slowly starting to make a bit of noise about its work on artificial intelligence. Last December the iPhone maker shared its first public research paperon the topic; this June it announced new tools to speed up machine learning on the iPhone; and today it started blogging. Sort of.
The company’s new website, titled “Apple Machine Learning Journal,” is a bit grander than a blog. But it looks like it will have the same basic function: keeping readers up to date in a relatively accessible manner. “Here, you can read posts written by Apple engineers about their work using machine learning technologies,” says the opening post, before inviting feedback from researchers, students, and developers.
As the perennial question for bloggers goes, however: what’s the point? What are you trying to achieve? The answer is familiar: Apple wants more attention.
It’s clear that the recent focus on AI in the world of tech hasn’t been kind to the iPhone maker. The company is perceived as lagging behind competitors like Google and Facebook, both in terms of attracting talent and shipping products. Other tech companies regularly publish new and exciting research, which makes headlines and gets researchers excited to work for them. Starting a blog doesn’t do much to counter the tide of new work coming out of somewhere like DeepMind, but it is another small step into public life. Notably, at the bottom of Apple’s new blog, prominently displayed, is a link to the company’s jobs site, encouraging readers to apply now.
What’s most interesting, though, is the blog’s actual content. The first post (actually a re-post of the paper the company published last December, but with simpler language) deals with one of the core weaknesses of Apple’s AI approach: its lack of data.
Much of contemporary AI’s prowess stems from its ability to sieve patterns out of huge stacks of digital information. Companies like Google, Amazon, and Facebook have access to a lot of user data, but Apple, with its philosophy of not snooping on customers — in favor of charging megabucks for hardware — has rather tied its hands in that regard. The first post on its machine learning blog offers a small riposte, describing a method of creating synthetic images that can be used to train facial recognition systems. It’s not ground-breaking, but it’s oddly symbolic of what needs to be Apple’s approach to AI. Probably a blog worth following then.