[ACCI-CAVIE] Rather than wiping the slate clean, Apple is quietly fine-tuning its artificial intelligence tools, and the end result is not far off. Apple is working on a multimodal language model that will be integrated into the Mac and iPhone environment, which should be the subject of the next conference in June 2023.

Who said Apple wasn’t interested in artificial intelligence? The American firm is working on a multimodal language model that could soon power most of its tools, from its intelligent assistant Siri to Photos. Called Ferret, it is the counterpart to Gemini, recently unveiled by Google.

2023 was undoubtedly the year of artificial intelligence. Yet one company, and not the least, has been particularly discreet on the subject: Apple. At the start of 2024, the Apple company is still not communicating on the subject, but specialised American media are reporting that last autumn an open source version, accessible to all (via GitHub), of a multimodal language model developed in collaboration with Cornell University, in the state of New York, was put online.

Ferret, Apple’s AI, still in the development phase

Ferret (“ferret” in French) is a model designed to simultaneously understand text and images, the idea being to be able to identify different elements to be analysed as part of queries. For the moment, no concrete use has been made of it in Cupertino, but this research could soon be integrated into the Mac (through macOS) and iPhone (iOS) environments for practical uses.

Once it has reached maturity, Ferret could, for example, help Siri to respond to more specific requests. It could also help recognise and sort photos, as well as creating images or text. For the visually impaired, it could even help to describe images or landscapes.

Ferret is still in the development phase. Apple could therefore wait a while before making any official announcements, and take advantage of its annual conference for developers (WWDC), scheduled for next June, to unveil its future uses.

By FS