Google's new experimental AI agent can browse the web for you

Google's Gemini 2.0 with Mariner automates web tasks like data retrieval and shopping, offering transparent logic and user intervention.

: Google unveils Gemini 2.0, featuring the Mariner AI agent that automates web browsing tasks like data retrieval and shopping. Mariner, tested in an experimental Chrome extension, can execute multi-step commands such as finding company emails or placing items in online carts. It shows its logic chain in a chat window, allowing users to intervene before completing actions. Although not extremely fast or accurate yet, Mariner is part of Google's broader AI strategy and might influence the future of web use.

Google has introduced Gemini 2.0, an advanced iteration of its GenAI toolchain, which includes agents like Mariner that automate web browsing. This AI agent, an extension for Chrome, can independently retrieve information, make purchases, and manage various online activities. Although designed to assist with routine digital tasks, Mariner's effectiveness remains under evaluation.

In one demonstration, Mariner could find contact emails for multiple companies and even identify artworks by famous impressionists like Van Gogh. The agent demonstrated its capacity by searching the web and placing art supplies in an Etsy cart, albeit without making final purchases. Its operations are transparent, displaying logic steps in a chat interface for user supervision.

Google acknowledges Mariner's slow speed and accuracy issues, as showcased in tasks taking several minutes. However, Mariner's technology highlights the potential for AI in transforming web interactions. Discussions include its resemblance to Microsoft’s AI concepts, suggesting web browsing could become a background task managed by AI agents. The AI's future adoption remains contingent on overcoming its current limitations.