Google is developing an AI agent that can independently use a web browser to help users automate everyday tasks. In doing so, the tech giant will compete with Anthropic, which recently presented a similar feature in beta.
According to The Information, Google is secretly working on ‘Project Jarvis’. This project involves a “computing” AI agent that can use a web browser, in this case Chrome of course, to perform everyday tasks of end users.
Jarvis functionality
The AI agent responds to end users’ prompts by regularly taking screenshots of what is visible on their computer screen. This is somewhat similar to Windows’ Recall feature on Copilot+ PCs. Then the AI agent interprets these screenshots before taking action by (automatically) clicking a button or entering text in a field.
Jarvis’ functionality initially focuses on research, online shopping, and flight booking. The feature may come into preview as early as December this year and represents another extension of Google’s Gemini LLM. The latest iteration of Gemini will also be presented in December.
Competition with Anthropic
With the arrival of Jarvis, Google is directly competing with Anthropic. This AI vendor recently introduced an update to its Claude 3.5 Sonnet LLM, which introduced the option to partly take over an end user’s computer.
This particular agent can also search the Internet, view local files, and perform button-clicking actions. The AI agent can then open files or programs for end users.
Also read: Claude 3.5 Sonnet can use your entire computer after update