Source: DeepTech
On May 21, 2024, Pacific Time, Microsoft held the Build Developers Conference in Seattle, the United States, mainly focusing on AI PC, Copilot, small language models, data and analysis and other fields to launch a number of technological innovations.
Next, DeepTech will introduce the most concerned technologies one by one.
Copilot+PC: Transform PC Industry with AI
In the face of the development wave of generative AI, Microsoft has been committed to redesigning Windows PCs as early as 2023. At this conference, the company launched a Windows PC category called "Copilot+PC".
(Source: Microsoft) It is understood that AI exists on every layer of this PC product, so it is known as the "fastest and most intelligent Windows PC ever" [1].
Specifically, the PC neural processing unit it carries has a processor computing capacity of more than 40 trillion operations per second; In terms of running AI workloads, the performance and efficiency are 20 and 100 times higher than those of traditional PCs.
This new Windows PC is planned to be launched in June 2024, first equipped with Qualcomm Snapdragon X series processors, and then Microsoft will introduce more such devices based on Intel and AMD.
In addition, it is worth mentioning that Microsoft has also released many powerful functions related to Windows, hoping to provide more software tools for developers in the AI transformation era to meet their various development needs.
Team Copilot: extend the capabilities of Copilot to the organization
Today, Copilot has made important contributions to helping individuals improve productivity. In order to further release commercial value, Microsoft provided a set of new functions on the basis of this product at this conference.
First of all, release Team Copilot, expand the ability of Copilot beyond personal assistants, so that they can participate in organizational work, thus helping the team improve work efficiency and collaboration.
For example, it can replace the project manager and other work roles and be used in Microsoft Teams, Microsoft Loop and other places where collaboration can be achieved.
Secondly, a custom Copilot can be built to independently complete many tasks, including executing automated long-term business processes, learning based on user feedback, recording abnormal requests and seeking help.
Finally, the introduction of Copilot extensions and Copilot connectors can not only help users get applications in their own business systems, but also promote developers to create Copilot extensions more quickly.
Multi modal small model Phi-3-vision: with language and visual functions
Phi-3 series models, including Phi-3-mini, Phi-3-small and Phi-3-medium, are small language models launched by Microsoft a few months ago.
Now, the model family has another 4.2B parameter multimodal model, Phi-3-vision. Functionally, the model combines language and visual functions, and can infer images, extract and infer text from images, generate insights and answer questions according to charts.
For the Phi-3 series models, after benchmark tests of language, reasoning, mathematics, coding, etc., its performance can surpass models of the same scale or even larger scale. The newly released Phi-3-vision is no exception.
According to the evaluation results provided by Microsoft, it is superior to Claude-3 Haiku, Gemini 1.0 Pro V and other larger models in general visual reasoning tasks, optical character recognition, table and chart understanding tasks.
Obviously, the continuous enrichment of the Phi-3 model family provides more options for Microsoft Azure customers to develop generative AI applications.
Mini language model Phi Silica: specially built for neural processing unit in Copilot+PC
Not only that, Microsoft also launched a small language model Phi Silica with 3.3B parameters, which is specially built for the neural processing unit on the "Copilot+PC".
According to the company, this model will not occupy a lot of resources and can fully free the CPU and GPU of the PC to perform other computing tasks.
A Microsoft spokesman told the media: "The uniqueness of Phi-3-Silica is that it is the first locally deployed language model for Windows. It is optimized for the neural processing unit of 'Copilot+PC', which can bring lightning speed of local reasoning to the device [2]."
Introducing real-time intelligence into Microsoft Fabric to help enterprises make more informed decisions and actions
In the AI era, data driven world has become a reality. What follows is that enterprises from all walks of life must improve the speed and accuracy of operations to better upgrade business models and meet customer needs, so as to reduce costs and increase revenue.
Based on this, at this conference, Microsoft announced the introduction of real-time intelligence in Microsoft Fabric.
Real time intelligence combines real-time analysis of Synapse, the best functions of Data Activator and a series of preview functions to help enterprises make more intelligent decisions based on the latest insights.
Specifically, by providing a complete set of end-to-end solutions for collection, processing, analysis, visualization, monitoring, early warning and taking action against events, users can quickly and easily build solutions, provide high-value insights and early warning, and achieve real-time operation scheduling.
So far, the Build 2024 developer conference has In progress, interested parties can go to the official website of the conference to watch it online.
reference material:
one https://blogs.windows.com/windowsdeveloper/2024/05/21/unlock-a-new-era-of-innovation-with-windows-copilot-runtime-and-copilot-pcs/
two https://venturebeat.com/ai/microsoft-introduces-phi-silica-a-3-3b-parameter-model-made-for-copilot-pc-npus/
https://azure.microsoft.com/en-us/blog/new-models-added-to-the-phi-3-family-available-on-microsoft-azure/
https://www.microsoft.com/en-us/microsoft-365/blog/2024/05/21/new-agent-capabilities-in-microsoft-copilot-unlock-business-value/?ocid=FY24_soc_omc_br_x_TeamCopilot
https://blog.fabric.microsoft.com/en-US/blog/introducing-real-time-intelligence-in-microsoft-fabric/
https://build.microsoft.com/en-US/home#live -stream
Operation/typesetting: He Chenlong
