On the early morning of May 22nd, the Microsoft Build 2024 Developer Conference was held in Seattle, USA. This technology giant has released over 50 updates in one go, including the construction of AI infrastructure and the implementation of model products. Among them, the most noteworthy content includes:
Team Copilot, personal intelligence assistant becomes team assistant
Microsoft Fabric, providing real-time intelligent SaaS solutions
Azure AI brings new cutting-edge models and multimodal features
Cobalt chip, self-developed AI chip update
Ultraman's Final Spoiler OpenAI New Model
Microsoft CEO Satya Nadella talked at the beginning about Microsoft's two dreams for computers: "One is to make computers understand us, not us to understand them; and the other is to let computers help us effectively reason, plan, and act based on information in a world of constantly increasing information."
The day before the conference, Microsoft had just released Windows 11 AI PC. With the powerful performance of a new NPU chip that can achieve over 40 trillion operations per second (40+TOPS), as well as providing long-lasting battery life and supporting advanced AI models, Windows 11 AI PCs can achieve functions that ordinary PCs cannot achieve, such as real-time audio translation and subtitles, making Windows a more open AI platform and developer development platform.
With the support of OpenAI, Microsoft has become a leader in the AI industry, gradually transforming from being involved in PC hardware, AI infrastructure construction, and model product implementation to becoming a comprehensive AI solution provider.
The Team Copilot that can assist the team in completing their work has arrived
At the press conference, Microsoft officially launched Team Copilot, expanding Copilot for Microsoft 365 from a personal AI assistant behind the scenes to a team assistant.
Users can call Copilot in collaboration tools such as Teams, Loop, Planner, etc. Team Copilot can serve as a meeting host, manage meeting agendas, track meeting time, and record meeting points. It can also act as a collaborator in chats, providing important information, tracking action items, and resolving unresolved issues. It can also serve as a project manager, helping to ensure the smooth progress of each project and promptly notifying the team for input. These features will be released in a preview version later this year.
Microsoft Copilot Studio has launched a new Agent Agent feature, allowing developers to build proactive response data and events based on specific tasks and features. Based on the new features, developers can understand context, infer operations, and input through memory and knowledge, and learn based on user feedback, thereby independently managing complex and long-term running business processes.
Microsoft Fabric provides real-time intelligent SaaS solutions
The new Real Time Intelligence in Microsoft Fabric provides end-to-end software as a service (SaaS) solutions.
According to Interface News, real-time intelligence has now enabled previews, helping data analysts achieve a simple low code or no code experience while also benefiting professional developers through a code rich user interface. For example, the Dener Motorsport team has been using Microsoft Fabric to support real-time analysis, storage, and reporting, ensuring optimal performance and maintaining the car in good repair condition, making drivers safer. The team plans to incorporate real-time intelligence as part of their winning strategy to gain insights during the race.
For developers, building applications requires a certain degree of flexibility, customization, and efficiency to ensure that the development work is practical and feasible. The new Microsoft Fabric Workload Development Kit has achieved this.
Azure AI brings new cutting-edge models and multimodal features
As the latest flagship model released by OpenAI, GPT-4o can now be used in Azure AI Studio or as an API. This multimodal model integrates text, image, and audio processing, setting new standards for both generative and conversational AI experiences.
The AI Small Language Models (SLMs) Phi-3 series developed by Microsoft has also released a new multimodal model, Phi-3-vision, which is now available for use in Azure. The Phi-3 model is relatively cost-effective and has been optimized for personal devices. Phi-3-vision can input images and text, and receive text replies. For example, users can ask questions about charts or open-ended questions about specific images.
Developers can experience these cutting-edge models in Azure AI Playground and start building and customizing models in Azure AI Studio.
Among them, OPPO will pilot the application of Azure AI's voice to text, fast transcription, and Azure AI text to speech technologies on its new smartphones.
Launch Cobalt chips optimized for cloud scale application performance
Microsoft has adopted its own AI infrastructure system approach, which includes hardware and software provided by Microsoft and its partners, all of which have been optimized to run AI workloads on a large scale and fine tuned according to customer needs.
At the press conference, Microsoft announced the launch of Cobalt chips specifically designed for performance optimization in large-scale cloud applications, which have now provided support for billions of conversations with services such as Microsoft Teams. Meanwhile, Microsoft announced the launch of a new Cobalt 100 Arm based Virtual Machine (VMs) preview version based on customized silicon chips released in November last year.
The Cobalt 100 Arm based virtual machine is the first generation of virtual machines to use Microsoft's new Cobalt processor. It is customized based on the Arm architecture and optimized for efficiency and performance when running general workloads and cloud native workloads. Compared to similar Azure virtual machines, the performance improvement of Cobalt 100 Arm based virtual machines can reach 40%.
Ultraman's Final Spoiler OpenAI New Model
OpenAI CEO Ultraman made his final appearance at the press conference, giving developers two suggestions and previewing the new model.
Ultraman stated that the industry has not experienced platform changes for some time, but is now going through such a period. He suggested developers to make good use of it, "Now is not the time to postpone plans or wait for the next opportunity. This is a special moment, and in the coming years, many things will happen, and many great new things will begin to emerge."
"When the mobile phone revolution began or truly emerged in 2008 and 2009, people would say, 'We are a mobile company, we have mobile applications.' But just a few years later, no one said they are a mobile company because it has become the foundational standard. 'Ultraman pointed out that generative AI is a remarkable new technology, but it does not free you from the hard work required to build great products, great companies, or great services.' You must put in this effort."
At the end of the speech, Ultraman exposed the new model of OpenAI, stating that the new modality and overall intelligence will be the key to the next model of OpenAI. He said the model will become smarter, more powerful, and safer, and the GPT-4o will be faster and cheaper.
Regarding the impact of big models on the future, Nadella also believes that, just like Moore's Law driving the information revolution, Scaling Law and model architecture together have changed the way humans use and generate data.
The breakthrough progress in artificial intelligence has brought profound changes to Microsoft and extended to the developer community. In the past two years, GitHub Copilot has become the most widely used AI developer tool with 1.8 million paying users. It can be expected that as Microsoft gradually grows into a comprehensive AI solution provider, more developers will use Microsoft's products, technologies, and services to develop applications.
The transformation brought about by AI has just begun.