OpenAI has introduced Operator, an AI agent that performs tasks directly in a browser, like filling out forms, booking services and placing orders. Currently available as a research preview for Pro users in the US, Operator uses a new Computer-Using Agent (CUA) model that combines GPT-4’s vision capabilities with advanced reasoning.
Operator interacts with websites by clicking, typing and scrolling, and pauses for user input on sensitive tasks like passwords or payments. Users can customize workflows for specific sites and save prompts for recurring tasks.
Safeguards include user data protection, privacy controls and restricted handling of financial transactions. OpenAI aims to refine Operator based on feedback before expanding access and integrating it into ChatGPT. Pro users can try it now at operator.chatgpt.com.
Source. Android Authority
