Microsoft’s annual developer conference kicks off on Monday, and the company will no doubt have all manner of things to announce for Azure and, if we’re lucky, Windows. To whet our appetites, the company has unveiled a crop of new Azure and Internet-of-Things services with, as we should no doubt expect these days, a focus on machine learning and blockchain.
First up are some new capabilities under the cognitive-services banner. These are the services that are most similar to human cognition: image recognition, speech-to-text, translation, and so on. Microsoft is adding a new category of service that it’s calling “Decision.” In this category are services that make recommendations to aid decision-making. Microsoft is putting some existing services into this category: Content Moderator (which tries to automatically detect offensive or undesirable text, images, and video) and Anomaly Detector (which examines time series data to find outlier or anomalous events). To these, Microsoft is adding Personalizer, which learns about a user’s preferences and makes recommendations accordingly.
Microsoft is also offering previews of its Ink Recognizer (which turns handwriting into machine-readable text) and Form Recognizer, which can extract structured data from hand-filled forms. Cognitive Search, which uses machine learning to enable searching across disparate data types (such as OCR-scanned images, PDFs, and handwritten notes) is being promoted to general availability.
To make machine-learning-driven services easier to develop, the company has built a new drag-and-drop tool for building ML systems without requiring any code. Machine learning should also be easier to integrate into existing DevOps workflows with the new MLOps, which is described as DevOps for machine learning. It offers reproducible, auditable, automatable machine-learning lifecycles. And to actually run machine-learning workloads, FPGA acceleration is now generally available, as is using Nvidia’s TensorRT and Intel’s nGraph hardware.
Away from the cloud, Microsoft is beefing up its edge-compute offerings, too. Azure SQL Database Edge, now in private preview, runs on ARM and x64 processors and puts a lightweight database engine that supports the same SQL queries as SQL Server and Azure SQL Database. This makes it easier to do data processing on local devices rather than in the cloud. The company has also announced IoT Plug and Play, which automates device registration and connection to cloud services.
And finally, Azure Blockchain Service provides a system for developers to create, manage, and govern consensus-based blockchain networks. Microsoft will also offer JPMorgan’s Quorum, a ledger built on the Ethereum platform, to Azure users.