It seems Apple is enlisting Google’s help for its new and improved Siri. According to a report by Bloomberg’s Mark Gurman, the iPhone maker is currently finalising an agreement with the search engine giant. This deal will allow Apple access to a custom version of Gemini for a price of around US$1 billion (~RM4.2 billion) a year.
Google’s model uses 1.2 trillion parameters for its processes, far exceeding Apple’s own models. For comparison, the cloud-based model for Apple Intelligence has 150 billion parameters. Apparently, the company will only use the custom Gemini for some of Siri’s functions, namely the assistant’s planning and summarising capabilities. These features allow it to synthesise information and decide how to perform complex tasks. Meanwhile, some other functions will continue to use Apple’s in-house models.
Aside from that, this custom Gemini will run on Apple’s own Private Cloud Compute servers, so no user data will be sent to Google. The report noted that the bitten fruit company has already allocated AI server hardware for powering the model.
Gurman also noted that Apple does not view this partnership with Google as a long-term solution. Instead, this is a temporary measure while it develops its own in-house models that are up to par. The company had previously considered using other AI models, including OpenAI’s ChatGPT and Anthropic’s Claude, before settling on Gemini.
And as Apple is working to roll out the new Siri early next year, its models team is also developing a cloud-based model with one trillion parameters. According to the report, the brand hopes the model will be ready for consumer applications next year.
(Source: Bloomberg)
