Apple researchers have come up with an improved way to train artificial intelligence algorithms, an achievement that's less significant for its scientific value than for the fact that it may be a sign of more Apple AI research to come.
While tech giants like Facebook, Google and Microsoft have amassed considerable brainpower from AI academics and engineers and regularly advertise their findings, Apple has remained largely silent. The new algorithm training technique, detailed in a paper published last week, is Apple's first major public contribution to artificial intelligence research. It comes two months after Carnegie Mellon professor Russ Salakhutdinov joined the company as its director of AI research, signalling a shakeup in Apple's AI priorities.
The paper itself doesn't exactly turn heads outside of research circles: it proposes a new method of using machine learning to more efficiently train neural networks, the building blocks of artificial intelligence that are present in everything from chatbots to self-driving cars. But it is sponsored by a company that until now has kept private virtually all of its research, AI or otherwise.
The shift is due in part to Salakhutdinov's hiring. He is no stranger to the tech industry, having previously served as a Microsoft Research Faculty Fellow and picked up awards from Google and Nvidia. His background in academia, meanwhile, which prides itself on collaboration, means that he's less likely to subscribe to traditional Apple-style secrecy.
Immediately after Salakhutdinov was hired, he remarked that he was "excited" about the new position and indicated that he would be looking to hire even more AI engineers. Beefing up its AI team would put Apple in a league with some of its traditional rivals like Microsoft, as well as Facebook, which is reserving large amounts of computing power from its data centers to power its AI research.
This story originally appeared on PCMag
