4月5日，为庆祝微软成立50周年，该公司一次性发布了Copilot的诸多新功能，使其在功能上与ChatGPT或Claude等竞品更为相近。
Copilot现已支持记忆、个性化、基于Web的操作、播客创建、摄像头和屏幕分析以及深度研究等功能，被官方誉为“面向每个人的AI伴侣”。它能追踪用户的偏好、兴趣乃至生日等详细信息，并据此定制回复和建议，甚至主动提供帮助。微软表示，用户可选择希望Copilot记住的信息，或完全退出此功能。这是微软努力使Copilot更加个性化的一部分，尽管目前还处于早期阶段，但用户很快就能为Copilot定制外观。
微软AI首席执行官Mustafa Suleyman表示：“Copilot不仅仅是一个AI，它是你的。每个Copilot都将拥有其独特的风格和属性组合，实现个性化定制。”
除了个性化之外，Copilot的功能也变得更加强大。一项新的动作功能使AI能够使用网络浏览器执行任务，如预订演出门票、餐厅和购物等。特别是当与Copilot的研究产品并寻找折扣和促销活动的新购物功能结合使用时，这一功能尤为实用。
2024年12月开始在网页工具中推出的Copilot Vision现已扩展到Windows和移动应用。在Windows上，Copilot能够“看到”用户屏幕上其他应用和文件中的内容，以回答问题或与文件和内容交互。而在iOS和Android上，它则能够讨论从用户的手机摄像头或相册中的照片中可见的任何内容。
深度研究功能使Copilot能够分析大量文档或在线资源以应对复杂项目，其研究能力正与Bing集成，以便在搜索引擎内提供AI驱动的响应。与其他AI工具一样，Copilot现在还可以利用其研究生成播客风格的音频来解释一个主题，而新的Pages功能则允许它将文档中的笔记和研究组织到一个单独的画布上。
微软官方表示，许多新功能将从今天开始以“初始版本”的形式推出，并在接下来的几周和几个月内不断改进。功能的可用性因功能、平台和市场而异。尽管这些功能并非首创——ChatGPT去年增加了记忆功能，谷歌Gemini有视觉模式，而且每个AI公司都想建立个性化的AI智能体——但微软一次性添加了所有这些功能。
On April 5th, in celebration of Microsoft’s 50th anniversary, the company has unveiled a multitude of new features for Copilot, aligning it more closely with competitors like ChatGPT and Claude in terms of functionality.
Copilot now supports memory, personalization, web-based operations, podcast creation, camera and screen analysis, as well as in-depth research capabilities, earning it the official title of “AI companion for everyone.” It can track user preferences, interests, and even details like birthdays to tailor responses and suggestions, even offering proactive assistance. Microsoft states that users can choose which information they want Copilot to remember or opt out of this feature entirely. This is part of Microsoft’s ongoing effort to make Copilot more personalized; although it’s still in the early stages, users will soon be able to customize the appearance of their Copilot.
“Copilot is not just an AI; it’s yours. Each Copilot will have its unique style and attribute combination, achieving personalized customization,” said Mustafa Suleyman, CEO of Microsoft AI.
Beyond personalization, Copilot’s functionality has also become more robust. A new action feature enables the AI to perform tasks using a web browser, such as booking event tickets, making restaurant reservations, and shopping. This feature is particularly useful when combined with Copilot’s new shopping capability, which searches for products and looks for discounts and promotions.
Copilot Vision, which was first introduced in web tools in December 2024, has now expanded to Windows and mobile applications. On Windows, Copilot can “see” content from other applications and files on the user’s screen to answer questions or interact with documents and content. On iOS and Android, it can discuss anything visible in photos from the user’s camera or album.
The in-depth research feature enables Copilot to analyze vast amounts of documentation or online resources for complex projects. Its research capabilities are being integrated with Bing to provide AI-driven responses within the search engine. Like other AI tools, Copilot can now also use its research to generate podcast-style audio explanations of topics, while the new Pages feature allows it to organize notes and research from documents onto a separate canvas.
Microsoft has announced that many of these new features will start rolling out in their “initial version” from today, with continuous improvements expected over the next few weeks and months. The availability of these features will vary depending on the function, platform, and market. Although these features are not entirely new — ChatGPT added memory capabilities last year, Google Gemini features a visual mode, and every AI company aims to build a personalized AI agent — Microsoft has introduced all these features simultaneously.