Gemini 現在可在 Android 上自動化多步驟任務

🔑 增強重點摘要

• •The feature is currently in beta, limited to Pixel 10, Pixel 10 Pro, and Samsung Galaxy S26 series devices, and initially available only in the U.S. and Korea.[2][3][5]
• •Gemini performs tasks in a secure virtual window processed in the cloud, with real-time notifications for monitoring progress, allowing users to view, intervene, or stop anytime.[2][5][6]
• •Supported apps include Uber for rideshares, DoorDash and Grubhub for food delivery, with Gemini handling navigation, selection, and input but requiring manual confirmation for final order placement.[4][5]
• •Activation requires long-pressing the power button and explicit user command, with safeguards ensuring no access to the rest of the device and opt-out for screenshot reviews by human trainers.[7][8]

🛠️ 技術深入

• •Codenamed 'Bonobo' in beta builds, relies on screen automation where Gemini scrolls, taps, and types within a secure virtual window of the target app.[8]
• •Virtual window processing occurs in the cloud, limiting access to only the specific app and preventing interaction with other device data.[5][6]
• •Includes live progress notifications and interface options to 'Stop task' or 'Take control', with descriptions of actions like 'Setting the pickup time'.[5]

🔮 前景展望AI analysis grounded in cited sources

⏳ 時間線