This article is more than 1 year old
Apple's 'KGB level of secrecy' harms its AI projects – but don't worry, it's started a blog
Boffins still turned off by iMaker's closed culture
Analysis Continuing its campaign to court AI boffins, traditionally tight-lipped Apple has tiptoed further toward engagement with the outside world through the publication of its first research blog.
"Welcome to the Apple Machine Learning Journal," the initial post says. "Here, you can read posts written by Apple engineers about their work using machine learning technologies to help build innovative products for millions of people around the world. If you're a machine learning researcher or student, an engineer or developer, we'd love to hear your questions and feedback."
Apple calls its website a journal because it's focused on machine learning. However, it lacks some common characteristics of an academic journal. There are no details about submitting research papers or peer review procedures, and the initial post has no named authors. Nor is there any code to explore, something seen on other research blogs, not to mention Apple's Swift blog, introduced in 2014.
"I think it's strange to call it a 'journal' when it appears to be a research blog," said David Duvenaud, assistant professor of computer science and statistics at the University of Toronto, in a phone interview with The Register.
The blog's single post, not counting its welcome message, is based on an academic paper submitted for publication last November by Apple computer scientists Ashish Shrivastava, Tomas Pfister, Oncel Tuzel, Josh Susskind, Wenda Wang, and Russ Webb.
The paper's appearance in December attracted significant media attention because it suggested a cultural change of course for Apple, which has tended not to celebrate or share its technical advances.
Earlier that month, Russ Salakhutdinov, a professor of machine learning at Carnegie Mellon University hired by Apple in October, told attendees at the Neural Information Processing Systems conference in Barcelona that Apple would begin publishing its AI research.
Salakhutdinov's hiring was widely seen a sign Apple intended to make its culture more inviting to AI researchers.
In 2012, Apple CEO Tim Cook promised to double down on product secrecy. And the company has done so to some extent. But in the years that followed, Silicon Valley's awakening to the value of AI-oriented technology has made secrecy problematic.
"At this point they've taken it to a KGB level of secrecy," said Chris Nicholson, CEO of Skymind, an enterprise AI company, in a phone interview with The Register. "They've hired people from the three-letter agencies to keep a lid on things. That breaks one of the incentives to join a company [among academics]. You don't just go for money. You go for recognition."
Duvenaud said it's difficult to tell how well Apple's effort to recruit AI researchers is going, but among those he knows, the prospect of working there isn't particularly appealing. However, he acknowledged that Apple has hired some talented people.
"I think their closedness really hurts them in the recruiting game," he said. "Most people I know don't even interview at Apple. I personally also chose not to interview there for this reason."
AI and associated disciplines – machine learning, deep learning, reinforcement learning, and so on – have become something of an obsession among technology companies. Amazon, Baidu, Facebook, Google, and Microsoft talk about it incessantly, to the point that actual value of AI research tends to get lost amid the marketing hype.
Apple aims to compete in AI too, with Siri, its forthcoming HomePod, and its self-driving car research, among publicly known initiatives. Its forthcoming iOS and macOS releases will include support for a new machine learning framework called Core ML.
"Apple really needs AI," said Nicholson. "They really need a good AI strategy and they're executing on that much more slowly than other companies. They're going to pay a price for that."
Duvenaud suggested it will take another year to see if Apple's courting of AI experts amounts to anything. "It's very, very hard to hire any sort of machine learning researchers right now," he said. "Everyone has already figured out you need to pay high salaries."
Top researchers, he said, want freedom to publish, which is available at companies like Google or Facebook, with minimal red tape. Apple researchers now have... a blog, without bylines.
"If Apple continues to have the same hiring troubles, which I think they will, they'll be forced to become more open," said Duvenaud. ®