News Summary

OpenAI has launched a new feature for its AI assistant, ChatGPT, called the ChatGPT Agent. This innovative tool allows users to complete multi-step tasks autonomously by controlling its own web browser. The agent integrates functionalities from previous tools like OpenAI’s Operator and Deep Research, enabling it to navigate websites, run code, and create documents such as spreadsheets and PowerPoint presentations. Users can instruct the agent to perform tasks ranging from assembling outfits for purchase, planning meals, to updating financial spreadsheets. The system operates within its own virtual sandbox, ensuring user privacy and control. Additionally, the agent includes a ‘Watch Mode’ for tasks requiring active oversight, like sending emails. OpenAI’s research lead demonstrated the agent’s versatility by automating tasks such as planning a date night and analyzing financial data. While the agent boasts impressive capabilities, its performance may vary based on task complexity, as it is primarily a master imitator trained on specific examples.