On Thursday, OpenAI introduced ChatGPT Agent, an innovative feature enabling its AI assistant to perform multi-step tasks using its own web browser. This update combines functionalities from the previous Operator tool and the Deep Research feature, allowing tasks such as creating PowerPoint presentations, meal planning, and updating spreadsheets. Positioned within the realm of “agentic AI,” ChatGPT Agent can autonomously navigate websites, run code, and manage documents while keeping users in control.
Users can closely monitor the AI’s actions inside a private sandbox environment, ensuring security as it utilizes a virtual operating system for internet access. Importantly, the Agent requires user permissions for tasks like purchases and offers options to interrupt or oversee actions at any time. Despite its advanced capabilities, the AI’s effectiveness in completing complex tasks may depend on specific scenarios due to its limitations in problem-solving and reliance on prior training data. The previous Operator feature will remain available for a short time.
Source link