The Macalope asks, do you want AI fast, or do you want it right?
The Washington Post, it seems, wants it fast.
“Apple is behind in AI and killed its self-driving car project. What’s next?”
Based on the implications here, we can safely assume it’s one or more of the following:
The Post’s piece is a virtual confetti gun of Apple doom, ranging from the company’s quarterly results to some unit sales figures Tim Cook takes exception to. Meanwhile, The Verge’s David Pierce is slightly more positive on Apple’s chances of recovering from its devastating behindness in AI.
“Better Siri is coming: what Apple’s research says about its AI plans”
It would be easy to think that Apple is late to the game on AI.
And, apparently, get it published in The Washington Post.
But over the last few months, rumors and reports have suggested that Apple has, in fact, just been biding its time, waiting to make its move.
AI, as with all new technologies, is largely treated as nothing more than a foot race. One must deliver something, anything, to be considered in the race. Who cares how many answers it gets wrong, whether or not you took the source material without licensing it, what kind of junk you’ve wrapped it up in, whether or not you’re just a repurposed crypto company, or whether or not your answers go on to kill people with its flawed results. Shipping is all that matters.
Capitalism: not caring about anything since Adam Smith.
IDG
IDG
IDG
What is lost in the discussion is whether or not there’s a right way and a wrong way to ship AI. Others have complained that on-device AI, like math homework, is simply too hard. “Plus, if it’s on the device, how will we scrape all the data?! It’s like you’re not even thinking!” That part’s just a coincidence, of course. It’s really because it’s too hard. Doesn’t work. Not even worth trying.
Apple, as Pierce notes, thinks it can do it.
By taking advantage of the most inexpensive and available storage on your device, they found, the models can run faster and more efficiently.
The killer app here, of course, is making Siri better. But beyond just a Siri that’s more responsive and intelligent is the ability to privately synthesize your information for you. Apple greatly improved autocompletion by putting it on the device, so it would try to anticipate the words you might want next, not what some clown in Hoboken wants. (The Macalope’s not trying to deride Hoboken, it’s just that there’s an actual clown who lives there who was greatly skewing the autocompletion data. Thanks, Carl.) Imagine your device knowing you better based on your data… and then not sharing that with advertisers. Seems worth a little wait to the Macalope.
None of this is unusual. People said Apple was late to other janky technologies, too, like netPCs, netbooks, and, well, smart speakers. Smart speakers persist, of course, but the market potential wasn’t exactly one Apple needed to rush to get into.
It obviously remains to be seen if Apple can make on-device AI work. Maybe it can’t deliver. But it should at least get some credit for trying to implement AI in a way that’s better from the perspectives of privacy, licensing and targeted results.
