According to "The Information," Google might introduce its own version of Rabbit's big model action concept as early as December. Codenamed "Project Jarvis," it is a tool that promises to do tasks on the behalf of the users themselves, including "gathering research, purchasing a product, or booking a flight" as three sources with knowledge of the project have pointed out.

Apparently, the tech is only accessible in a web browser and only for Chrome. It is also reportedly powered by a future version of Google's Gemini.
The tool is meant to "automate everyday, web-based tasks" by analyzing then interacting with on-screen elements, including clicking buttons or entering text following screenshots. As of this point, the tool "takes a few seconds" to do each task, according to reports from The Information.

This is just one of a series of choices by leading AI companies which are looking to improve on the automation of digital tasks. For example, Microsoft Copilot Vision will allow conversation with webpages one is reading, and Apple Intelligence promised to handle all screen-based actions across all apps this coming year. Another AI called Anthropic released its beta version model known as Claude and described it as cumbersome and error-prone; OpenAI is developing a model similar to that one.

However, "The Information" cautions that Google could change the preview for Jarvis still in December. It might distribute it as a limited release to a set of testers for its efforts to find and correct the issues that arise with such a radical change.