Chinese scientists have introduced an autonomous artificial intelligence (AI) "agent" called Manus, developed by the startup Butterfly Effect. This is touted as the world's first general AI agent, showcasing a level of autonomy absent in existing AI models. Manus offers a glimpse into the potential future of artificial general intelligence (AGI). Unlike traditional chatbots like ChatGPT, which require step-by-step instructions, Manus can independently execute various tasks based on simple text prompts. Although it is not yet publicly available, limited access has generated significant interest, with some users creating playable video games and launching websites using Manus. However, users have reported operational glitches, including crashes and a tendency to fall into infinite feedback loops.

The head of the project, Peak Ji, noted these issues are common during the early stages of deployment. Caiwei Chen, a reporter for MIT Technology Review who tested Manus, described it as working with an intelligent and efficient intern, though she noted it occasionally misinterprets tasks and makes errors. Manus is the second AI tool from China to make waves this year, following DeepSeek, which delivered competitive chatbot results at a lower cost. Manus operates on a multi-agent architecture, utilizing several large language models (LLMs) and other software to manage tasks independently. It tends to provide more detailed answers than ChatGPT but takes longer due to its in-depth research approach. This agent can autonomously navigate the web, write code, and analyze data based on a single prompt, distinguishing itself from conventional LLM-driven tools. Users can close their computers while Manus continues its tasks in the cloud. The introduction of a general AI agent raises new ethical considerations, as these agents can operate swiftly and tirelessly, resembling human behavior but functioning without fatigue or the need for continuous guidance.