Google, Meta, and Microsoft, as well as startups like OpenAI and Anthropic, all had well-developed AI manufacturing strategies by the time Apple finally announced its push in June. Conventional wisdom suggests that this door is the wrong way to drink.
Apple denies it. Its leaders say the company is coming at the right time — and that it has been secretly preparing for this moment for years.
That’s part of the message I got from talking to Apple executives this fall about how they created what’s now called Apple Intelligence. Senior vice president of software engineering Craig Federighi is a regular character in an ongoing web series in the tech world known as key product launches. Not publicly known is senior vice president of machine learning and AI strategy John Giannandrea, who previously ran machine learning at Google. In a separate interview, I spoke with Greg “Joz” Joswiak, Apple’s senior vice president of worldwide marketing. (These conversations helped me prepare for my sit-down with Tim Cook, which I did the next day.) All the executives, including Cook, emphasized that despite AI’s massive disruptive potential, Apple would manage this game-changing technology. the same transparency and thoroughness the company is known for. To paraphrase the song of some of the musicians who founded a company called Apple, the employees of Cupertino were always waiting for this moment to appear.
“We were doing intelligence in 2015, like predicting which apps you’ll use next and helping predict routes on maps,” Joswiak said. “We didn’t always talk about it publicly, but we were also ahead of the curve.”
In 2018, Apple poached Giannandrea from Google, a move Cook told me showed that Apple anticipated the coming AI revolution. The company has made him a new senior VP position, an unusual move for Apple that breaks with its traditional hiring practices. When he arrived, Giannandrea was struck by how much Apple was already leveraging advanced AI in some of its most popular products. “Face ID is a feature that you use every day, many, many times a day to unlock your phone, and you don’t know how it really works,” he said. “There’s a lot of deep learning that happens in private on your phone to make that feature work. But for the user, it just disappears. “
Federighi says that testing OpenAI’s GPT-3 model, released in 2020, blew his mind. He says: “Things that seemed to be on their way to becoming possible suddenly became possible.” “The next real question was whether it was possible to use the technology in Apple’s way.”
Apple soon had multiple teams working on transformer-based AI models. So when ChatGPT went global in November 2022, there was no need for Apple to assemble an internal team to develop AI products—work was already underway to build features that would “just disappear.” “We have ways to integrate technology that works across the organization to make big changes to products,” Federighi said. “When it comes to taking a big step in the community, we pulled a lot of those wires together, in a way we’re very familiar with at Apple.”
Source link