Apple plans to release ‘Siri’, a revamped AI assistant based on Gemini

The next-generation version of Apple's artificial intelligence (AI) assistant 'Siri' is expected to be released in September.

Apple plans to release a revamped version of 'Siri' based on Google's AI model 'Gemini' in September, The Information, an American information technology (IT) media outlet, reported on the 3rd (local time), citing multiple sources.

The next-generation ‘Siri’ is based on on-device (device-type) AI that performs calculations within Apple devices such as iPhones even without an Internet connection.

To this end, Apple is conducting 'distillation' work to create a small model that can be run in devices such as smartphones based on the Gemini model.

This kind of on-device AI has excellent security features such as personal information protection because it does not transmit data to the outside, but it is difficult to perform complex tasks due to performance limitations of hardware such as semiconductors.

Therefore, Apple is expected to choose a complex structure that leverages the power of Google Cloud, which is powered by Nvidia's AI chip, for some complex tasks.

However, Apple plans to maintain security standards through NVIDIA's 'Secret Computation' function, which encrypts data even when using Google Cloud.

Apple originally intended to use a 'private cloud compute' server running on self-designed AI semiconductors such as M series chips for AI calculations, but it is said that the plan was revised considering that it was not enough to run the huge model Gemini, which has trillions of parameters.

This decision is in contrast to Apple's existing strategy of directly retaining control of technology.

Apple is expected to unveil specific features of the new Siri at the Worldwide Developers Conference (WWDC) held at its headquarters in California on the 8th.