Microsoft’s Magnetic-One can perform complex tasks using AI
Microsoft on Monday introduced a new multi-agent artificial intelligence (AI) system called Magnetic-One. The tech giant called it a powerful system that can activate multiple AI agents to perform complex tasks via web browsers locally on a device. It is based on a new framework that allows an AI model to access multiple modalities and capabilities to perform tasks such as booking a ticket, purchasing a product online or editing a document stored on the device stored. In particular, Microsoft’s Magnetic-One is an open source project and accessible to researchers and developers.
Microsoft introduces Magnetic-One
Generative AI has made a quantum leap in machine intelligence and the ability to generate output in text, images, audio and video formats. However, while modern AI systems are great at information retrieval, they still remain poor at reasoning, especially when it comes to solving problems and completing tasks.
This is why AI agents, which can be thought of as miniature software that can perform an action, have become an important extension of large language models (LLMs). Microsoft’s Magnetic-One also works on the same principle, as described in a research paper. The company describes it as a “high-performance generalist agentic system” designed to perform complex multi-step tasks such as software engineering, data analysis, scientific research and web navigation.
Magnetic-One has a multi-agent architecture, meaning one LLM can activate multiple agents to complete a task. For this, the AI system activates a master agent called the Orchestrator. It supervises four other agents, with each agent specialized in one task.
Magnetic One’s workflow
Photo credits: Microsoft
For example, if the system is asked to book a ticket to a movie, the Orchestrator can activate a vision agent that can look at the screen and process the visual information. Another may have knowledge of web browsers and can navigate them. The third could break the prompt down into actionable steps, and the fourth could handle financial transactions. Dividing the task among several such specialized agents increases both accuracy and speed of completion.
The open-source Magnetic-One AI system is available on GitHub and can be accessed here. It is available to researchers and developers and can also be used for commercial purposes under a modified Microsoft license. In addition, Microsoft has also released AutoGenBench, a tool that evaluates the performance of AI agents. It comes with built-in repeat and isolation controls to thoroughly test the agents.
(Except for the headline, this story has not been edited by NDTV staff and published from a press release)
