Google DeepMind is reinventing one of the oldest tools in computing: the computer mouse cursor.
Alphabet’s AI research division said it is developing an AI-enabled pointer powered by its Gemini model that can understand not only what users are pointing at on a screen, but also what they are trying to do. The goal is to make AI interactions more natural and less dependent on typing detailed prompts into separate chatbot windows.
For now, it only works in Google Chrome and its new AI laptop, Googlebook.
In experimental demos, users could point to an image of a building and ask for directions, highlight a recipe and request doubled ingredients, or hover over a data table and generate a pie chart. Google said the system combines pointing, speech and surrounding context so the AI can better infer intent.
The effort reflects a broader push across the tech industry to embed generative AI directly into workflows rather than treating chatbots as standalone tools. Google said the technology is being integrated into Chrome and its upcoming Android-based Googlebook laptops through a feature called “Magic Pointer.”
The company outlined four design principles behind the project: keeping users in their workflow, reducing the need for complex prompts, enabling shorthand commands like “fix this,” and turning on-screen pixels into actionable objects such as maps, dates or shopping links.