
Google's latest voice assistant is built on the company's
Gemini advanced AI model with advanced reasoning that can manage daily routines, many of which begin at home.
The model provides inference and search capabilities in an effort to make the technology more powerful and easier to use than Google
Assistant.
To access Gemini for Home, users still say, “Hey Google,” to get started, similar to the company’s assistant, but interactions will feel fundamentally new. Gemini
will understand the context of the request and complete it.
Google Senior Project Manager Niranjan Hira, who recently joined the company from Amazon, and Google Distinguished Engineer Fenghui
Zhang explained the concept of inference. The closest concept, "pattern matching." In other words, AI models match patterns to predict what you want the person making the query wants.
advertisement
advertisement
For example, in the U.S., the word "jelly" usually matches with the words "peanut butter." The ability to match or infer also could be considered a prediction based on human behavior. This is
what's known as inference in speech patterns. It's something that deep learning AI models can infer, especially Google's latest AI voice model. It also can make recommendations.
As of
August 21, 2025, ads do not serve up in Gemini AI assistant conversations, according to AI Mode, but Google CEO Sundar Pichai has suggested in the past that the company is considering ways to
incorporate ads into the voice system.
Gemini Live has memory and
delete functions, and will allow for ongoing conversations. The features are intended to give users more personal options because it can reference past chats to learn preferences and provide
personalized responses.
Google anticipates a gradual rollout that will begin early in October, which will replace Google Assistant on existing speakers and displays, with free and paid
versions.
Apple may also have an upgraded voice assistant, but it likely won’t ship until Spring 2026, according to Bloomberg.
An upgraded version of App Intents could make Siri
the actual hands-free controller of a user’s iPhone. This means voice will control the phone for even things such as taking a photo and editing it before sending it to a friend.