At I/O 2024 in May, Google showcased two examples of agent-based experiences that users would access through Gemini. In December, the company may provide further details about agents functioning within Chrome, powered by Gemini 2. 0, as part of Project Jarvis. “I envision [agents] as intelligent systems capable of reasoning, planning, and retaining information.

They can think several steps ahead, navigate various software and systems, all to accomplish tasks on your behalf, while most importantly operating under your guidance. ” —Sundar Pichai discussing AI agents According to The Information, Google is “creating an artificial intelligence that can take control of a user’s web browser to perform tasks such as conducting research, making purchases, or booking flights. ” “Project Jarvis” — inspired by J. A. R. V. I. S. from Iron Man — will function within Google Chrome and aims to offer a consumer-oriented feature (as opposed to an enterprise one) designed to “automate routine web-based tasks. ” The article does not clarify whether this will be available on mobile or desktop devices. During I/O, Pichai demonstrated how “Gemini and Chrome collaborate to assist you with various preparations: organizing, reasoning, and synthesizing on your behalf. ” This showcased event was generically conducted through gemini. google. com without displaying any other user interface compared to a prior demonstration on Gemini for Android. Upon receiving a command or action, Jarvis operates by taking “frequent screenshots of the user’s computer screen and interpreting those images before executing tasks such as clicking buttons or entering text. ” The current report reveals that Jarvis “functions at a relatively slow pace because the model requires a few seconds to process before performing each action. ” Therefore, it is likely that this system is not yet operational on-device and still depends on cloud processing.