For one billion US dollars a year: Siri is likely getting a Google Gemini heart
A forthcoming LLM voice assistant from Apple will get a Google AI model, according to a report. This is not free.
Siri on different devices: Is Apple pulling off an AI coup thanks to Google?
(Image: Apple)
That Apple will not rely on its AI models for a more intelligent voice assistant, Siri, has been rumored for a long time. Now, the iPhone manufacturer seems to have actually agreed with Google to take this step. According to a report by the financial news agency Bloomberg on Wednesday evening, a corresponding deal is said to be “almost” in the bag. What is particularly astonishing is the cost: Apple is reportedly willing to pay “around one billion US dollars a year” for access to Google's Gemini models.
Well-hidden Gemini
However, the amount is significantly lower than the revenue Apple currently generates from Google for an advertising deal in the Safari browser; according to court documents, this amounts to a good 20 billion dollars per year. Whether there will be a set-off or Apple will actually hand over cash remains unclear. Apple does not intend to publicly announce that Gemini will be integrated into a language model-based Siri (“LLM-Siri”). Instead, Google is to be a “technology provider behind the scenes.”
Videos by heise
The publicly visible integration of Gemini (and presumably other models like Claude from Anthropic) in iOS, iPadOS, and macOS is also still planned via MCP, or operating system interfaces. Currently, you can ask ChatGPT via Siri (and Apple Intelligence functions), but you have to explicitly tell the voice assistant / operating system. The now planned deal will “only” use Gemini in the background to finally bring Siri into the AI era.
Apple still hopes for its technology
Apple is said to have worked for a long time to make its own AI models fit, but apparently failed initially. Meanwhile, Anthropic's Claude was also considered a candidate for Siri, but Gemini is said to have prevailed. Nevertheless, the Google technology is only intended as an “interim solution,” Apple hopes—“until its models are powerful enough,” writes Bloomberg. An implementation is reportedly already planned for spring—then Apple should already deliver a new, context-sensitive Siri. Whether LLM-Siri and the latter will appear simultaneously is apparently not yet clear.
Apple itself is working in its struggling AI department on a model with one trillion parameters (a measure of complexity), which is to be stored in the cloud. Google's Gemini model intended for Siri is said to be “ultra-powerful,” with up to 1.2 trillion parameters, and also run in Apple's Private Cloud Compute data centers (PCC). It is to be a “custom model.” Currently, Apple's current models only reach 150 billion parameters when running in the cloud. Apple executives believed they could achieve a similar quality level as the customized Google offering, it was further stated. But Google is also continuously improving Gemini. So it is unclear whether Apple can catch up with such a backlog.
Empfohlener redaktioneller Inhalt
Mit Ihrer Zustimmung wird hier ein externer Preisvergleich (heise Preisvergleich) geladen.
Ich bin damit einverstanden, dass mir externe Inhalte angezeigt werden. Damit können personenbezogene Daten an Drittplattformen (heise Preisvergleich) übermittelt werden. Mehr dazu in unserer Datenschutzerklärung.
(bsc)
