(Bloomberg / Getty Images)
Hi, Backchannel nation. Steven here. Happy end-of-summer (though who’s happy about it?).
This week I wrote about Apple and AI. If you’ve been following Backchannel at all, you know we have been closely covering the rise of machine learning, as well as the fascinating issues raised by increasingly successful artificial intelligence efforts. It was great to hear, and share with Backchannel readers, Apple’s accomplishments and thoughts on ML and AI.
Though Apple took me through its AI implementations in some detail, I had plenty of time to talk with the three key leaders who briefed me — senior vice presidents Phil Schiller, Eddy Cue, and Craig Federighi — about the company’s philosophy and practices regarding AI. One of the most fun discussions, though I had no room to squeeze it into what was already a 5000-word story, involved recommendations, particularly when it came to music.
Apple has used algorithms in the past to recommend things to users (remember the “genius” function?), but with machine learning, it is stepping up its game even more. For instance, Cue says that without machine learning, the Apple News product, which finds articles that Apple think you’re eager to read, is unimaginable. And of course algorithms are involved in what Apple figures you will like to hear musically.
This opens up, to me at least, a philosophical paradox. Specifically in its music product, Apple has emphasized the human element. It has legions of music editors — former rock critics, deejays, and musicians — who go to work each day compiling playlists and identifying hot tracks. (I wrote about these folks in an article earlier this year.) Part of this was an outgrowth of its purchase of Beats, a company whose ethos springs from the creative impulses of its leaders. The company even hired Nine Inch Nails frontman Trent Reznor as Chief Creative Officer.
Apple has mocked competitors who rely on solely algorithms for recommendations. It makes sense that a fellow human being who has spent years embedded in the music scene, someone whose fingertips are callused from pulling out just the right vinyl at the right time, would prevail over a digitized neural net in delighting you with brilliant segues.
Yet AI specialists are generally agreed that deep learning techniques will inevitably exceed human capabilities in this as well as just about a million other things. Surely beating Trent Reznor can’t be that much harder than crushing a Go champion.
Cue has been involved in Apple’s music efforts for many years, so I asked him about this. Haven’t you planted a flag in the ground with your ‘human touch’ claim? I asked him. What happens now?
Cue wouldn’t concede that his company actually did plant that flag. Right now, he says, Apple recommendations depend on a combo of humanity and machine — flesh-and-blood music editors whose selections are mixed with algorithmic results, increasingly honed by machine learning. “I don’t think machine learning will get some of the human curation things that we want for a long, long time, if ever for that matter,” he says. “But there are a bunch of things we can’t humanly curate because we don’t have enough people. Like we can’t humanly curate all of your music, for example.”
And what if the day arrives when deep-learning-powered algorithms are demonstrably better at making playlists than even the hippest homo sapiens? Adios, deejays. “If the machine can do it better, we’ll do it better,” says Cue.” I don’t have a problem with it. We’re just saying, let’s make sure it’s great before you jump over.”
The week in Backchannel:
The iBrain is Here. And It’s Already on Your Phone. This is the story I mentioned above. Apple’s secrecy had previously engendered speculation around whether the company really “got” AI. In this story its executives reveal the extent to which the firm is already using advanced machine learning in its products.
Div Turakhia Just Became A Billionaire. Our Jessi Hempel offers a rich — very, very rich — portrait of a 34-year-old guy from a middle-class Mumbai family who just sold his ad-tech company to Chinese investors. In many ways, Turakhia embodies the American dream, a dream that hardly works for Americans anymore, if it ever did.
Amazon’s Bet on Twitch Was Ahead of Its Time. In this week’s Follow-up Friday, Jeremy Hsu looks back at Amazon’s acquisition of Twitch in 2014. At the time, few people could understand what was so attractive about a platform that let people watch live streams of other people playing video games, but two years later, Twitch is facing stiff competition from YouTube and Facebook Live. Read it now, while prose is still a thing.
