Exclusive look at how AI and Machine Learning work at Apple

On July 30, 2014, Siri has had a brain transplant.

Three years ago, Apple was the first major tech company to integrate a smart assistant into its operating system. Siri is a corporate adaptation of a standalone application they bought, with the team that created it, in 2010. The initial reviews were great, but for months and years later, people The user becomes impatient with its shortcomings. Usually, it misinterprets commands. Edits will not be edited.

So Apple moved Siri voice recognition to a neural network-based system for US users later that July (it appeared worldwide on August 15, 2014.) Previous specs still work – if you’re keeping the score at home, this includes the “hidden Markov model” – but now the system uses machine learning techniques, including deep neural networks (DNN), the cumulative neural network, the long-term memory unit, the periodic unit, and the n-gram. (Glad you asked.) When users do the upgrade, Siri still looks the same, but now it’s enhanced with deep learning.

As an example with tacit advances that could reveal its mind to competitors, Apple has not made this development public. If the user notices, it is only due to fewer errors. In fact, Apple now says the results in the accuracy improvement are excellent.

Senior Vice President of Internet Software and Services Eddy Cue.

Eddy Cue, Apple’s senior vice president of internet software and services, said: “This is one of the things where the leap is so significant that you have to take the test again to make sure someone isn’t. drop a decimal.

This story of Siri’s transformation, revealed for the first time here, could shock many in the world of artificial intelligence. Not that the neural mesh has improved the system – of course they’ll do it – but Apple has been doing it very quietly. Until recently, as Apple’s hiring in the AI ​​sector accelerated and the company made a number of well-known acquisitions, observers viewed Apple as the laggard in the fiercest competition. industry: the race to the best. make use of those powerful AI tools. Since Apple has always been quiet about what goes on behind badge doors, cognoscenti AI has no idea what Apple is doing in the field of machine learning. “It’s not part of the community,” taught Jerry Kaplan courses at Stanford about the history of artificial intelligence. “Apple is the NSA of the AI.” But the AI ​​Brahmins realized that if Apple’s effort was as significant as Google’s or Facebook’s, they would hear it.

“At Google, Facebook, Microsoft, you have the best machine learning leaders,” said Oren Etzioni of the Allen Institute for AI. “Yes, Apple has hired some people. But who are the five leaders of machine learning working for Apple? Apple has voice recognition, but it’s not clear where else [machine learning] help them. Show me your product where machine learning is being used!

“I’m from Missouri,” said Etzioni, who is actually from Israel. “Let me see.”

Well, earlier this month, Apple has shown where machine learning is being used in its products – not for Etzioni, but for me. (Oren, please read.) I spent a better time during the day in the One Infinite Loop meeting room at the Cupertino headquarters, receiving a core summary of the company’s activities in the field. AI and machine learning from Apple’s top executives (Cue, worldwide senior marketing vice president Phil Schiller and senior vice president of software engineering Craig Federighi), as well as two scientists key Siri learning. As we sit down, they give me a dense, two-page program that lists Apple products and services with machine learning in mind – already shipped and coming. – which they will discuss.

Message: We were here. Player. Nobody is second to none, but we do it our way.

Leave a Comment