OpenAI's ChatGPT Agent Criticized for Slow Performance and Errors in Initial Use

OpenAI's latest AI agent, ChatGPT Agent, has come under fire for significant performance issues and operational errors during its initial rollout. Despite its intended capabilities to streamline tasks such as scheduling and ordering food, reports highlight a troubling lack of efficiency.

Early users reported that the agent took nearly an hour to complete a simple task of ordering cupcakes, raising questions about its practicality in everyday applications. Isa Fulford, the research lead, noted the complexity of developing such an agent, but the slow response time has bewildered many users. Furthermore, a demonstration showcased the agent planning a trip to a fictional baseball stadium in the Gulf of Mexico—an area devoid of any sports facilities—adding to its growing list of operational inaccuracies.

The ChatGPT Agent integrates features from earlier OpenAI AI agents and operates using a virtual computer to assist users effectively. However, a safety measure requires users to approve significant actions, reflecting concerns over the current reliability of AI technologies. For now, the agent is accessible to Pro users under a limit of 400 prompts per month, with Plus and Team subscribers set to get access later, albeit with a reduced prompt allowance of 40. There remains no clear timeline for when free users might gain access.

This initial reception highlights the challenges OpenAI needs to address as it seeks to improve the agent's efficiency and reliability over time, demonstrating the ongoing conversation around the limitations of AI capability.