Microsoft Ignite 2024 Digest

Key AI/ML/Data/Analytics announcement digest

Xin Cheng
6 min readDec 18, 2024

zero day quest hacking

Microsoft Zero Day Quest invites security researchers to discover and report high-impact vulnerabilities in Microsoft AI and Cloud Bounty Programs: AI, Microsoft Azure, Microsoft Identity, M365, and Microsoft Dynamics 365 and Power Platform.

copilot is UI for AI

copilot actions: AI-powered macro that goes off and does its thing based on some fill-in-the-blank prompts.

Copilot Pages: dynamic, persistent canvas in Copilot chat designed for multiplayer AI collaboration. UI is similar to openai canvas

AI Agents in M365

Employee Self-Service Agent, Facilitator agent (real-time notes, summary of important information as the conversation happens), Interpreter agent (real-time interpretation in Teams meetings), Project Manager agent

Sharepoint Agents

Every SharePoint site comes with a “ready-made-agent”, automatically scoped to the content on that site. You can create and edit an agent to:

  • Customize its branding and purpose
  • Specify sites, pages, and files your agent should get information from.
  • Define customized prompts tailored to the purpose and scope of the agent.​​​​​​​

copilot studio agent production: publish agents to engage with your customers on multiple platforms or channels, such as live websites, mobile apps, or messaging platforms like Microsoft Teams and Facebook.

copilot analytics: component of the new Copilot Control System, with Business Impact Report to show how Microsoft 365 Copilot usage relates to KPIs (you still need to upload business impact data)

Windows 365 link: a mini PC device (endpoint) that allows users to connect to their Windows 365 Cloud PC (let employees securely stream their Windows desktop experience from the Microsoft Cloud to any device)

Azure local: Azure Local is the new name for Azure Stack HCI. Azure Local doesn’t have an acronym. (existing Azure Stack HCI customers will automatically upgrade to Azure Local) It is enabled by Azure Arc, and is cloud-connected infrastructure that can be deployed at your physical locations and under your operational control.

Azure integrated HSM: hardware security module (HSM) built directly into Azure cloud servers, designed to securely manage and protect cryptographic keys by performing cryptographic operations locally within the server environment, aligned with the vision of use of hardware-based Trusted Execution Environments (TEEs) for data privacy and the protection of workloads.

Azure boost DPU: Data Processing Units (DPUs). DPUs are optimized for tasks that require handling vast amounts of data at the network’s line rate. It integrates high speed Ethernet and PCIe interfaces along with network and storage engines, data accelerators, security features, and runs a custom, lightweight data-flow operating system.

NVIDIA Blackwell on Azure

Azure HBv5: 7 TB/s of memory bandwidth,

Microsoft Fabric Database: Azure SQL Database in Microsoft Fabric, with new AI features: native vector and RAG support, the ability to call any model via APIs from within the database, integration with popular AI platforms and services e.g., Azure AI services and Azure AI Foundry, the ability to manage a variety of data ranging from structured to unstructured, and compatibility with popular frameworks such as LangChain and Semantic Kernel

Azure AI foundry: rebranding of Azure AI Studio (not to be confused with Microsoft Copilot Studio), designed for AI software developers who want to create generative AI applications and custom copilot experiences. One of component is Azure AI Agent service

AI Model customization and experimentation: an opportunity for selected customers to collaborate closely with our engineering and research teams to develop custom models tailored to their unique domain-specific needs. Model fine-tuning with Gretel, Statsig, Scale, W&B

AI reports and evaluation: AI reports for developers to provide the information needed for cross-functional teams to assess production-readiness throughout the GenAIOps lifecycle (key project details, such as the intended business use case, potential risks and harms, model card, model endpoint configuration, content safety filter settings, and evaluation results). With online evaluation, you can continuously assess your application’s outputs, regardless of its deployment or orchestration framework, using built-in or custom evaluation metrics, with integration of tracing in Azure AI Foundry and Azure Monitor Application Insights.

Atom quantum machine: entangling 24 logical qubits made from neutral atoms and performing computations on 28 logical qubits, demonstrating reduced error rates through advanced error-correction techniques.

Ignite Book of News

AI at work

Agents

Microsoft 365 Agents SDK (Teams Copilot SDK/Teams AI library/Bot framework) provides interoperability with Copilot Studio in two ways. Developers can add functionality and extend an existing agent built using Copilot Studio using skills, allowing a maker to delegate work to other agent functionality. Secondly, a developer can connect to a Copilot Studio agent from code, providing the developers with all the functionality within the Copilot Studio ecosystem, including over 1000 connectors. Publish agents across a variety of platforms and channels

  • Microsoft Teams
  • Microsoft 365 Copilot*
  • Web
  • Email (inc. Outlook)
  • SMS (Twilio)
  • More than 10 additional message channels

copilot studio enhancement: Azure AI Foundry integrations: copilot studio can access Azure catalog of 1,800+ AI models, including industry-specific fine-tuned models, bring-your-own- model via the Azure AI model catalog, adding custom search indices as a knowledge source via Azure AI Search; image/voice support

M365 Copilot

Teams: analyze visual content presented in Teams, Copilot file summaries

PowerPoint: Narrative Builder based on a prompt and a referenced file, Presentation translation, Copilot will use images stored in asset libraries, including SharePoint Organization Asset Library and now Templafy

Excel: tell Copilot what to create, and Copilot will suggest and refine a template with headers, formulas and visuals getting them off to a great start.

Outlook: copilot for scheduling meetings, draft meeting agenda

OneNote: organize ideas pulled from a combination of typed, handwritten and voice notes on pages within a single section

Copilot in MAC: part of the Copilot Control System, which will offer data protection, management controls and reporting to enable IT to confidently adopt and measure the business value of Copilot and agents.

Copilot Prompt Gallery: discover, save and share prompts for agents provided by their organization that are tailored to specific roles, functions and tasks; latest and most popular prompts within user’s organization

Azure

Azure AI Foundry SDK, Azure AI Foundry portal (formerly Azure AI Studio), Azure AI Agent Service

Azure integration with Weights & Biases (W&B) (tracking, evaluating and optimizing models), Gretel, Scale AI (data preparation and support for fine-tuning and training models) and Statsig (configure AI apps and run powerful experiments)

AI reports: create impact assessments for their AI apps by helping developers assemble key project details, such as model cards, model versions, content safety filter configurations and evaluation metrics, into a unified AI report. Risk and safety evaluations for image content with Azure AI Content Safety

Azure AI model catalog adds industry-specific (healthcare, agriculture, manufacturing and finance) models from Bayer, Sight Machine, Rockwell, Saifr, Cerence and Paige.ai

Azure AI Search: Query rewriting, semantic ranker, will beintegrated with GitHub Models

Azure AI Content Understanding: extract information from unstructured data, such as documents, images, videos and audio, into customizable structured outputs (vs. Azure AI Document Intelligence, extraction from complex documents)

Developer Tools

Azure Container Apps: Serverless GPUs, dynamic session provides instant access to compute sandboxes for running untrusted code at scale, with each session protected by industry-standard Hyper-V isolation (modes: Python code interpreter and custom container sessions.)

Oracle Database@Azure: integrations with Microsoft Purview and Microsoft Fabric

Azure HBv5 virtual machines (HBv5 VMs), Azure Integrated Hardware Security Module (HSM), Azure Boost DPU, see above

Data and Analytics

Fabric Databases: SQL database in Microsoft Fabric is the first database engine coming to Fabric (native vector search), Copilot in Fabric which translates natural language queries into SQL and provides inline code completion alongside code fixes and explanations

AI functions in Fabric, AI skills/2 in Azure AI Agent Service integration

Open Mirroring: allowing any app or data provider to bring their data estate seamlessly into OneLake within Fabric. This capability will enable app or data providers to write change data directly into a Mirrored Database in Fabric

OneLake catalog: explore, manage and govern the entire Microsoft Fabric data estate, including structured and unstructured data, lakehouses, warehouses, machine learning models, notebooks, reports and more

SQL Server 2025: support AI apps from on-premises to cloud (vector database, model management enabled by Azure Arc)

Azure Managed Redis

Quantum

in collaboration with Atom Computing, largest neutral atom commercial system with entangled logical qubits

Azure IoT

Azure IoT Operations: a unified data plane for the edge, that runs in an on-premises Arc-enabled Kubernetes cluster.

Security Copilot

In addition to embedded summaries and insights, data security admins will be able to get comprehensive visibility in Microsoft Purview Data Security Posture Management, with GenAI-powered insights to dig into their data environment

Identity admins will be able to get AI-driven assistance in the Microsoft Entra admin center, with insights across sign-in logs, user and group details, audit and diagnostic logs and app risk to help troubleshoot, enforce least-privilege access and remediate risk for users, apps and workloads

IT admins will be able to use AI-driven Kusto Query Language (KQL) query assistance to assess data across multiple devices and platforms for rapid endpoint issue troubleshooting.

--

--

Xin Cheng
Xin Cheng

Written by Xin Cheng

Multi/Hybrid-cloud, Kubernetes, cloud-native, big data, machine learning, IoT developer/architect, 3x Azure-certified, 3x AWS-certified, 2x GCP-certified

No responses yet