In the summer of 2014, Apple completely changed how Siri works.
The secretive Cupertino company adapted Siri’s voice recognition to use a cutting edge artificial intelligence technique called neural networks, and switched it over on July 30, 2014, according to an in-depth feature by Steven Levy.
Neural networks is a type of AI inspired by the human brain, which has become especially useful thanks to today’s powerful computers. Before that, Siri recognised human voices using more rudimentary AI techniques that have been around for decades.
It was the biggest change to Siri since it launched in 2011. And almost nobody noticed — there wasn’t a any press coverage beyond a few people noticing that Apple had begun hiring neural network experts. And to users, Siri continued to work the same. It just was better at understanding what you said.
“This was one of those things where the jump was so significant that you do the test again to make sure that somebody didn’t drop a decimal place,” Eddy Cue, Apple’s internet services boss told Backchannel.
In fact, Siri made half the errors using the new neural network than it had in the past, according to Alex Acero, who leads the Siri speech team. “The error rate has been cut by a factor of two in all the languages, more than a factor of two in many cases,” Acero said.
Apple has been criticised recently for seemingly falling behind rivals like Alphabet, Facebook, Microsoft, and Amazon at emerging artificial intelligence techniques like neural networks, which is presumably why it opened the kimono and gave Levy access to many of its top AI experts.
The feature reveals several interesting facts about Apple’s AI operations, including:
- The entire size of the “AI brain” on an iPhone is about 200 MB.
- Apple’s buying a ton of small AI companies as acqui-hires — 20 to 30 companies a year, according to Cue.
- Apple has decided to use GPUs for its AI. Other companies like Intel and Microsoft are pushing different approaches.
- Apple tends not to hire established researchers for its AI efforts, instead hiring smart people and having them learn on the techniques at Apple.