Deal: Google Gemini becomes foundation for Apple's AI models
Briefly

Deal: Google Gemini becomes foundation for Apple's AI models
"Apple and Google have signed a multi-year deal under which the Cupertino-based manufacturer will use Gemini models as the basis for its own AI models. It has also been announced that Siri will receive the long-delayed AI upgrade later this year. In a brief joint statement, both companies announced that they have entered into a multi-year partnership. Google's Gemini models will be used as the basis for Apple's Foundation models. These are the iPhone maker's 'core models' for artificial intelligence, also known as Apple Intelligence."
"The end user will probably not notice anything immediately because Gemini will not be visible or noticeable. Everything will be packaged in an Apple flavor, and Apple's own servers may be used to run the AI model instead of Google's. Gemini is a closed-source model that typically runs only on Google Cloud, but there may be room for an exception here."
"The upgrade will likely allow Siri to work with a larger language model than before. This will make the assistant smarter, tasks will be performed faster, and all output will be much more personalized. Exactly how this works is not yet known. Apple could use Gemini to train its own smaller models. This process is called "distillation," in which a compact model learns from the complex answers of a large model. Apple could have reached an agreement with Google through an official license."
Apple and Google have entered a multi-year partnership under which Google's Gemini models will be used as the basis for Apple's Foundation Models, also called Apple Intelligence. The new personalized Siri will be built on those Foundation Models and is scheduled to arrive later this year on iPhones, Macs, and iPads, though no exact date has been announced. Users may see the upgrade as faster, more intelligent responses and more personalized output. Gemini is closed-source and typically runs on Google Cloud, but Apple may run models on its own servers or distill Gemini into smaller models under a licensing agreement.
Read at Techzine Global
Unable to calculate read time
[
|
]