Introducing Manus: The AI General Agent – YouTube
See more
Chinese researchers have demonstrated an artificial intelligence (AI) “agent” that can make decisions on its own without the need for specific instructions from a human operator.
The Manus AI agent was created by Chinese startup Butterfly Effect. The company claims it is the world’s first general-purpose AI agent, with a level of autonomy that is lacking in current AI models. The scientists behind Manus say it represents a potential breakthrough in what artificial general intelligence (AGI) might one day be capable of.
This new type of AI responds to text commands, similar to chatbots like ChatGPT or DeepSeek. However, unlike chatbots, it can perform a variety of tasks without the need for frequent step-by-step instructions.
Manus is not yet available to the general public, but a limited distribution of invitation codes has allowed some users to gain access, generating a lot of interest online. Some users have apparently created playable video games from simple commands, while others have used Manus to develop and launch websites.
However, Manus is still in its early stages, with some users reporting crashes and other issues, such as a tendency to get stuck in an endless feedback loop. The company is also aware of Manus’ comparatively high failure rates compared to ChatGPT. Chief Scientist Peak Gee acknowledged that this is part of the early challenges of launching a new tool on the X social media platform.
Reporter Caiwei Chen, who gained access to Manus for MIT Technology Review, said the experience was “like collaborating with a very smart and effective intern.” However, Chen also added that Manus sometimes didn’t know what to do, made the wrong judgments, and took shortcuts. “In the end, it’s promising, but not flawless,” Chen said.
Is the era of artificially intelligent agents coming?
Manus AI is the second Chinese AI tool to make waves in the tech industry this year. DeepSeek changed the AI landscape in January by achieving comparable or better results than its American competitors, reportedly at a lower cost. Manus, meanwhile, may represent the first example of an entirely new generation of AI.
Chatbots typically operate on a single large language model (LLM), while Manus uses multiple LLMs and other independently running software to perform various tasks autonomously, MIT Technology Review reports. This is also called a multi-agent architecture, where multiple components interact and collaborate to perform tasks.
General-purpose artificial intelligence agents promise to be more human-like than previous versions of AI.
In a competition between Manus and ChatGPT, the agent often offers more detailed answers than the chatbot, which the company confirmed with GAIA testing data. However, Manus also takes much longer to provide those answers because it conducts more in-depth research, Tom's Guide reports.
In other words, Manus decides how to respond to a request on its own, rather than relying on step-by-step instructions like a chatbot. It’s designed to initiate tasks on its own and dynamically adjust its approach as needed, Forbes reports.
Give Manus one query and it will explore the network, write code, and analyze data to respond without requiring any further intervention, unlike a typical LLM-based chatbot.
Examples demonstrated by Manus AI on its website include planning a detailed vacation itinerary, analyzing the stock market, and reviewing resumes for job openings. The difference between this and a tool like ChatGPT is that Manus is able to break down and execute complex tasks without the need for constant input. The system also runs from the cloud, so users can close their computers at any time and Manus will continue to function in the background.
The emergence of the general-purpose AI agent raises a new wave of ethical questions and considerations related to the use of AI. Agents behave more humanely, but unlike humans, they can work quickly and continuously – with
Sourse: www.livescience.com