Apple is investing time and money in technologies artificial intelligence, it’s no longer a secret anymore. The Cupertino company will present some of its innovations at the WWDC24 of June, a perfect opportunity given that the protagonists of the event will be software, starting from iOS 18. In the meantime, however, we can be satisfied with the rumors that are shared by reliable sources, such as Mark Gurman.
Apple and LLM on-device: pros and cons of this (rumored) solution
The journalist of Bloombergthrough its newsletter Power Onat this time says that, in his opinion, the much talked about new AI features of the iPhones will be entirely powered by a large language model (LLM) offline, therefore performed entirely on the device. Looking at marketing, Apple could enrich its advertising campaign on privacy also focusing on this important factor.
The disadvantage of on-device LLMs is that, inevitably, they cannot be as powerful as models running on serverswhich instead can count on an almost unlimited number of continuously updated parameters and data.
However, as our colleagues point out 9to5macApple could derive a significant advantage from the usual optimization of your devices, strong – among other things – also with proprietary processors. Also, models that run directly on the device usually they are quicker to respond and they can work too offline in those places with limited or no network connection.
An on-device LLM could generate more sophisticated responses to messages, for example, or improve the interpretation of many requests to Siri, a virtual assistant which – everyone believes – needs a major update. Then there is the aforementioned question privacywhich in this case would be completely safeguarded.
As reliable as Gurman has proven to be over the years, to discover Apple’s strategy regarding artificial intelligence with complete certainty, all we can do is wait for the WWDC24 in June. The keynote at Apple Park will be held on June 10.
Cover image created with Microsoft Designer