Of course, Apple is working on its own generative AI, Large Language Model (LLM) and possible ChatGPT rival called, naturally, AppleGPT. Sure, the news is based on a Bloomberg report and Apple is predictably mum on the matter but, seriously, how could the Cupertino tech giant not be working on its own AI?
According to the Bloomberg report, Apple is basing its ultra-secret project on a learning framework known as Ajax, from rival and sometimes friend Google.
The effort to build some sort of chatbot and maybe other generative AI systems has been going on since late last year but, as someone who attended Apple’s WWDC 2023 can tell you, Apple made no mention of chatbots of any kind at the June developer’s conference.
Apple’s hyper-focus on user privacy has, as I see it, somewhat hamstrung its efforts to bring any kind of LLM-based chatbot to consumers. ChatGPT, Google Bard, and Microsoft Bing are all cloud-connected and send queries out to distant servers for rapid interpretation and response (based on the LLM’s vast knowledge of how actual humans might respond under similar circumstances).
That, of course, is not the Apple way. Its Apple Silicon A16 Bionic’s Neural Network is local. It does Machine Learning on your best iPhone. Sending queries with all those possibly personal details is anathema to Apple’s privacy principles.
And yet, Apple clearly cannot afford to stay away from the siren call of generative AI. It is a revolution that is consuming the tech industry and the interests of average consumers and businesses. Even with the intense scrutiny AI development is under and the lawsuits some of it is facing, no one believes AI development is suddenly going to stop or go away.
Apple has even gone as far as, according to Bloomberg, creating its own chatbot, or AppleGPT. But that’s basically a highly limited and internal test and apparently not one that’s ever headed to consumer desktops.
What about Siri?
Where does Siri sit in all this?
Bloomberg claims that the Ajax work has already been used to improve Siri. That may be so, but the only Siri improvement we’re getting with iOS 17 (currently in public beta) is the ability to stop starting each voice assistant prompt with “Hey.”
I have no doubt that Apple is hard at work figuring out its place in the LLM AI sphere, but it’s also clear from the report that these are early days. There is no overarching strategy, and I doubt the existential question of whether or not Siri could ever host AppleGPT (or whatever it’s called) has been answered.
Ultimately, this is confirmation that Apple is just as aware of what’s going on around it and with competitors as ever. It will sample and test, develop and test, scrap and develop, and then test some more. I don’t expect Apple to tell us anything about this during the expected September launch of the iPhone 15. However, by the time WWDC 2024 rolls around, Apple might be ready to unveil a new platform. Maybe it’ll be AppleGPT-kit, AppleLLM-Kit, or even AppleGPT.
This assumes that Apple can solve its big privacy question. If not, AppleGPT could remain in Skunkworks indefinitely.