Microsoft’s Copilot is gearing up to unleash a game-changing feature: Copilot for OneDrive. This innovative tool allows users to effortlessly query and retrieve information about files stored in the Microsoft cloud. While the functionality is undeniably powerful, its execution demands considerable computational resources, making it a challenging feat to replicate on local machines.
Scheduled for launch in May 2024, Copilot for OneDrive will be accessible exclusively to enterprises and schools equipped with a Microsoft Copilot for Microsoft 365 license, priced at $30 per user per month. Notably absent from the announcement is any mention of availability for consumers subscribed to Copilot Pro, the $20/user/month service, leaving uncertainties regarding broader accessibility.
Why is this announcement significant? Because it tasks Microsoft’s cloud servers with the arduous task of comprehending the contents of your files—a formidable endeavor in itself. At the core of artificial intelligence lie two fundamental functions: machine learning and inferencing. While inferencing enables user interactions with AI, such as generating essays or images, machine learning involves establishing intricate relationships within data sets—a computationally intensive process.
During a demonstration, I tasked an AI language model with “learning” a 113-page PDF detailing the U.S. Code governing the office of the president. The learning process alone consumed close to two hours, underscoring the resource-intensive nature of the endeavor, even on robust hardware configurations like a 14th-gen Core HX laptop with a GeForce RTX 4090 GPU.
However, once the learning process concluded, the AI exhibited remarkable proficiency, enabling detailed inquiries akin to consulting a domain expert. This capability extends beyond legal documents to encompass various file types, offering unprecedented convenience and insight.
While running an AI model locally ensures the privacy of sensitive documents, offloading this task to Microsoft’s cloud servers raises concerns about data confidentiality. Yet, the time-consuming nature of document learning may compel users to opt for cloud-based solutions, despite potential privacy implications.
As Microsoft prepares to roll out Copilot for OneDrive, the absence of restrictions on the number of files that can be learned raises intriguing possibilities. With capabilities spanning file types from documents to presentations, this feature promises to revolutionize productivity and information retrieval within OneDrive.
