Top AI Announcements at Microsoft Build 2024 

Via Unwind AI

https://blogs.microsoft.com/blog/2024/05/21/whats-next-microsoft-build-continues-the-evolution-and-expansion-of-ai-tools-for-developers/

Top AI Announcements at Microsoft Build 2024 

“Microsoft Build 2024 kicked off with a host of new announcements to enhance AI and developer tools across their Copilot and Azure AI ecosystem. The conference revealed new AI capabilities and features that promise to improve productivity and collaboration. Among the standout announcements were the new Teams Copilot, Copilot Studio, new Phi-3 models, GitHub Copilot extensions, and more. Here are the top 10 AI announcements:

1. Team Copilot: The new Teams Copilot extends Copilot for Microsoft 365 from a personal assistant to a team-oriented AI assistant. It’s your new teammate that can manage meeting agendas, take notes, show important information in chats, track action items, and manage project tasks and deadlines, improving team collaboration and communication.
2. Copilot Studio: This new platform enables developers to build custom Copilots that act as agents for specific tasks.
   1. You can create your own Copilot agents which can independently carry out tasks, automate business processes, have memory and knowledge, learn based on user feedback, and ask for help when needed.
   2. To ground these custom Copilots’ responses in organizational data, you can bring together data from Microsoft’s various data sources like Microsoft Graph and Fabric, combining all with a visual connector.

3. Phi Silica: A transformer-based, 3.3B parameter local language model optimized for Copilot+ PCs. It is the first small language model locally deployed on Windows, offering fast first-token latency performance. Developers can access the Phi Silica API to deliver user experiences across the Windows ecosystem.
4. New Phi-3 Models: Microsoft has released three new models as part of Phi-3 models family – Phi-3-vision is Microsoft’s first small vision language model with 4.2B parameters, Phi-3 small model with 7B parameters, and Phi-3 medium model with 14B parameters. Phi-3 mini, small and medium models are generally available on Azure.
5. Real-Time Intelligence: Microsoft Fabric is a unified platform for data integration and analytics. The new Real-Time Intelligence feature, in preview, allows you to act on high-volume, time-sensitive data with tools like the Real-Time hub, event streams, and AI-powered insights for anomaly detection.
6. GitHub Copilot Extensions: GitHub’s new Copilot extensions integrate tools and services from third parties within GitHub’s IDE. Developers can customize their GitHub Copilot experience by integrating their preferred tools like Azure, Docker, and Sentry, where the Copilot Chat will now have context from all these tools to help you through your workflow without switching tools.

7. Azure AI Search: The new Azure AI Search, a full-featured information retrieval system, is designed for superior RAG and enterprise search. It offers enhanced information retrieval with new vector search capabilities, hybrid search improvements, and seamless integration with Microsoft Fabric’s OneLake connector.
8. Azure AI Studio: Microsoft’s pro-code platform for developing and deploying custom AI models now is generally available and has a more user-friendly interface alongside code-first capabilities.
   1. Code-First Development Experiences – Integration with Azure Developer CLI and the AI Toolkit for Visual Studio Code streamlines the development process, facilitating LLMOps and accelerating code-to-cloud workflows.
   2. Opensource Foundation Models – The studio will have more models from companies like Nixtla, Cohere, and AI21 Labs.
   3. Other features: It provides a comprehensive toolchain for developers, including features like collaborative project hubs, experimentation playgrounds, data retrieval with Azure AI Search, fine-tuning capabilities, agent-based orchestration, and tracing for debugging.

9. Azure OpenAI Service: This provides access to advanced AI models, including OpenAI’s latest GPT-4o model, now generally available in Azure AI Studio and as an API. It also provides fine-tuning capabilities for customized outputs, and an Assistants API for advanced virtual assistants.
10. Cognition + Microsoft: Microsoft will use Cognition’s AI software engineer Devin to help its developers achieve more, starting with code migrations and modernization, and bring Devin to its customer base. Devin will be powered by Azure.”