Microsoft makes Azure smarter with a flurry of new machine-learning services

Microsoft makes Azure smarter with a flurry of new machine-learning services

The company has announced a swathe of new services for Azure, ahead of Microsoft’s Build 2019 developer conference next week.

Microsoft is doubling down on building machine learning into the services it offers across its Azure cloud platform.

The company announced a swathe of new services for Azure, ahead of Microsoft’s Build 2019 developer conference in Seattle next week.

“AI use across Azure is growing rapidly. More than 1.3 million developers are using Azure Cognitive Services, and nearly 3,000 new bots are created each week”, said Frank Shaw, Microsoft’s corporate VP for communications, adding that Microsoft’s goal is to build Azure into “the best cloud for AI”.

 

Microsoft has added a new Decision category to the Cognitive Services it offers via Azure. According to Shaw, this will allow firms to build a “sophisticated decision process” into services that observe the world, such as image recognition.

This category includes Content Moderator, the recently announced Anomaly Detector, and a new service called Personalizer. Available in preview, Personalizer aims to provide companies with the analytical and decision-making tools to tailor services to individual users.

Microsoft also unveiled two new offerings in Cognitive Services’ vision category: Ink Recognizer for recognizing handwritten text using a digital pen; and Form Recognizer, which automates data entry by extracting text, key-value pairs, and tables from documents.

Under Azure Cognitive Services’ speech category, Microsoft is releasing a new transcription service, which Shaw says offers “advanced text-to-speech capability”, designed to transcribe meeting conversations in real time, including the ability to indicate which person said what.

In the language category, the new Language Understanding service will have a language analytics dashboard to evaluate the quality of language models.

More of Microsoft’s Cognitive Services are also being made available on-premises and on devices at the edge of corporate networks, thanks to the addition of container support for Speech-to-Text, Text-to-Speech, Form Recognizer and Anomaly Detector.

There are also new updates to Azure services that simplify the building, training and deployment of machine-learning models.

These include a new ‘intuitive UI’ in Microsoft’s AutoML service, which Shaw says “makes designing high-quality models easier”, via features such as a drag-and-drop interface for creating machine learning models without using code and new machine-learning notebooks for use when writing code to create models.

Microsoft also highlighted Azure MLOps, which is aimed at helping increase the rate at which firms can spin up the infrastructure needed to support large-scale machine-learning deployments. Azure MLOps helps companies build continuous integration and continuous delivery pipelines for use with machine-learning projects.

This provides developers with “reproducibility, auditability, and automation of the end-to-end machine-learning lifecycle”, says Shaw, with features including profiling and monitoring.

New hardware-accelerated Azure machine-learning models are also being released that run on specialized chips (FPGAS) that are optimized for machine-learning inference. Microsoft’s ONNX Runtime for ONNX-compliant machine learning models also gets support for Nvidia TensorRT and Intel nGraph models.

Microsoft also announced the general availability of the Azure Cognitive Search, which uses machine-learning models and other smart functions such as OCR to extract insights and structured information from content stored on the cloud platform. Cognitive Search also gains the ability to interrogate data stored in a wider range of Microsoft services, such as Power BI visualizations.

Also new is the HoloLens 2 Development Edition, a $3,500 release of Microsoft’s latest augmented-reality headset, which also includes Azure credits and free trials of Unity Pro and Unity PiXYZ plug-in for CAD.

Edge computing services

New services are targeted at edge-computing devices, the low-power computing devices that sit at the edge of a corporate network, such as on a factory floor.

Azure SQL Database Edge is being released in private preview and Microsoft’s Shaw says it will “complement our existing support for AI and IoT workloads at the edge, by running on Arm devices and bringing new capabilities like data streaming and time-series data within database, machine learning, and graph” said Shaw. Code written to work with SQL Server can be “easily” used with Azure SQL Database Edge, he added.

Meanwhile IoT Plug and Play offers developers an open modeling language to connect IoT devices to the Azure cloud, certified to work with more than a dozen devices at launch, with more devices due to be added over time.

Microsoft is also focusing on Blockchain, with the release of Azure Blockchain Services, which “simplifies the formation, management and governance of consortium blockchain networks, allowing business to focus on workflow, logic, and app development”. Microsoft is selling the service as offering a managed consortium network that can be deployed using a few clicks, with built-in governance for common tasks, such as adding new members, setting permissions, and authenticating user applications. JPMorgan’s Quorum platform is the first ledger available on the Azure Blockchain Service.

Also new is a Blockchain extension for the VS Code editor, designed to simplify the process of building and extending applications using Blockchain.



Source link