Official benchmark results released by Google or DeepMind

Google's Gemini-Powered Project Mariner Scores 90.5%, Redefines Web UX with AI Automation

Dec 11, 2024, 05:03 PM

Google has unveiled Project Mariner, its first-ever AI agent capable of performing tasks on the web autonomously. Developed by DeepMind and powered by Gemini 2.0, the AI agent can navigate websites, control the Chrome browser, move the cursor, click buttons, and fill out forms, mimicking human interactions. Project Mariner represents a significant shift in user experience (UX) design, described as a 'fundamentally new UX paradigm,' allowing AI to automate labor-intensive tasks and interact with web content. The AI agent achieved a 90.5% score on the WebVoyager benchmark, showcasing its ability to handle complex tasks such as booking flights. Currently available as a research prototype, the technology is integrated into an experimental Chrome extension, potentially transforming how users engage with the internet and disrupting industries reliant on web traffic.