Apple could use Google Gemini to power some iPhone AI features

In addition to using Gemini to power features in its apps and services, Google offers its LLM to third-party developers. Apple is reportedly in talks with Google to license Gemini for the iPhone.

According to Bloomberg, there are “active negotiations to allow Apple to license Gemini, Google’s suite of generative AI models, to power some of the new features coming to iPhone software this year.” Apple has also been in talks with OpenAI, which powers Microsoft’s AI capabilities.

Apple is specifically looking to partner on cloud-based generative artificial intelligence, with today’s report citing text and image generation as examples of what Gemini could be used for. At the same time, Apple is working to bring its own AI models and capabilities to the device with the upcoming iOS 18 release.

Discussions are still ongoing, and it is unclear how the AI ​​agreement will be branded. That would be a significant extension of the existing relationship — search engine default — between the two companies.

Looking at the rest of the industry, Google announced a partnership with Samsung in February to have Gemini power compression features in the note-taking and voice recording apps on the Galaxy S24, as well as the keyboard. Samsung also uses Imagen 2 text-to-image diffusion for the generative editing feature in the photo gallery app. All of these features require server-side processing, but Samsung also uses an on-device version of Gemini.

Google offers Gemini in three sizes, and the Pro is used by most first- and third-party apps. Gemini 1.0 Pro runs the free version of gemini.google.com, while 1.0 Ultra is used in the paid Gemini Advanced tier.

Gemini 1.0 is available in a stable version, but in mid-February Google previewed Gemini 1.5 with a significantly expanded context window that allows it to absorb more information. This can make the “output more consistent, relevant and useful”.

  • Gemini Ultra: The largest and most capable model for very complex tasks
  • Gemini Pro: The best model for scaling in a wide range of tasks
  • Gemini Nano: The most efficient model for on-device tasks

Bloomberg he doesn’t expect the deal to be announced until WWDC in June, with Apple settling on OpenAI or even the multiple providers offered as possibilities in today’s report.

More about Gemini:

FTC: We use automatic affiliate links to earn money. More.

Source link

Leave a Reply

Your email address will not be published. Required fields are marked *