Your PC Can Be Managed by Microsoft’s New On-Device AI Model

Microsoft’s most recent AI model is capable of running natively on your PC and is intelligent enough to complete tasks for you, such as online product purchases, in a possible preview of the future. The experimental Fara-7B AI model was made available by Microsoft on Monday. It was described as the company’s first “agentic” small language model “designed specifically for computer use,” including controlling the mouse and keyboard. Fara-7B has 7 billion parameters, which is a lot less than OpenAI’s GPT-3 model from 2020, which had 175 billion. (The precise parameters of OpenAI’s most recent models are unknown to us.) Microsoft’s experimental model, on the other hand, “achieves state-of-the-art performance within its size class and is competitive with larger, more resource-intensive agentic systems that depend on prompting multiple large models,” in spite of its relatively small size. This includes performing better than OpenAI’s GPT-4o when set up for online browsing. The new AI model from Microsoft works “by visually perceiving a web page,” allowing it to comprehend and act over a PC’s desktop. Microsoft went on to say that “it uses the same modalities as humans to interact with the computer” because “it does not rely on separate models to parse the screen, nor on any additional information like accessibility trees.” When a user is required to consent to certain actions, such as entering account login information, Fara-7B appears to respond and act more slowly than a human. Nevertheless, the demonstrations provide a glimpse of how, as AI models become smarter, they could automate and carry out numerous everyday tasks for users.