Apple has revealed some of its plans for the on-device AI that will run on future iPhones, iPads, and other devices and it looks like the future of AI.
Charlie Sorrel for Lifewire:
We’re used to AI that runs on massive server farms, sucking in ten times the electricity of regular cloud servers, and of course, enjoying all the usual privacy risks associated with cloud computing…
“AI models on devices are typically smaller than those in the cloud. Their reduced size requires fewer calculations, faster responses, and less energy consumed in response to user queries. Moreover, queries and their responses do not leave the device, enhancing user privacy. And AI models on devices avoid data movement and communication with cloud and data center servers, which should improve performance and reduce energy use. Taken together, AI on devices could transform user experiences with consumer electronics,” Professor Benjamin Lee, computer scientist at the University of Pennsylvania, told Lifewire via email.
The crucial part here is that training the model takes an insane amount of computing power, but once that model is made, it can be used on regular computers. And if you shrink the model, it might not be as capable, but it can run on smaller computers, and require less power.
This brings us back to Apple’s models, called OpenELM, which are designed to run on iPhones. The entirety of a model can contain trillions of parameters, but Apple provides several smaller options, with 270 million, 45 million, 1.1 billion, and 3 billion parameters. The more parameters, the better the model is at doing the AI thing.
So why is this important? Because it lets Apple run AI on devices that have relatively limited computing power and very limited batteries. By optimizing for specific purposes, models can be much smaller and more efficient. And Apple can design its models to run on the AI hardware it has had in Macs and iPhones for years—the Neural Engine.
Support MacDailyNews at no extra cost to you by using this link to shop at Amazon.
MacDailyNews Take: The New York Times last October reported that A.I. could soon need as much electricity as an entire country, so, barring fusion power plants, Apple’s forthcoming on-device model is clearly the future of AI.
Please help support MacDailyNews. Click or tap here to support our independent tech blog. Thank you!
Support MacDailyNews at no extra cost to you by using this link to shop at Amazon.
And maybe that is why Apple seems to be „behind“ in the AI development. Apple wasn’t the first one with an MP3 player or with a smartphone but see what happened.
If they’re not ‘behind’ now, they definitely will be when Microsoft’s PHI-3 ends up on iOS devices. The same is true for Android. MS’s CoPilot is giving Google Assistant some heavy pressure to keep ahead at the moment.
Nobody has figured this out yet.
Not the user experience, not the applications, not the revenue model, not the stack. Nothing.
Apple has the best platforms for most people to use AI for most uses.
Can’t wait to see what they come up with.
Meanwhile, when I ask Siri what time a local business closes, she says, “Here’s what I found for you…” and then posts it on the screen, which I can’t read because I’m driving. She should simply say: “Lowe’s closes at 7 pm, in 47 minutes, and the nearest one is about 29 minutes away in current traffic.”
Exactly, and 7+ years ago.
“By optimizing for specific purposes, models can be much smaller and more efficient. And Apple can design its models to run on the AI hardware it has had in Macs and iPhones for years—the Neural Engine.”
So what specific purposes will Apple be optimizing for that isn’t already being done on both iOS and Android via apps from Google and Microsoft? That Neutral Engine really needs to be used more after being out for 7 years.