We are excited to bring Transform 2022 back in-person July 19 and virtually July 20 - 28. Join AI and data leaders for insightful talks and exciting networking opportunities. Register today!
Ahead of its annual Build developer conference in Seattle, Washington next week, Microsoft today announced enhancements to Azure Machine Learning, its service that enables users to architect predictive models, classifiers, and recommender systems for cloud-hosted and on-premises apps.
“Furthering our commitment to building the most productive AI platform, we’re delivering key new innovations in Azure Machine Learning that simplify the process of building, training, and deployment of machine learning models at scale,” wrote Microsoft cloud and AI group executive vice president Scott Guthrie in a blog post. “Today we’re delivering innovative Azure services for developers to build the next generation of apps. With 95% of Fortune 500 customers running on Azure, these innovations can have far-reaching impact.”
Specifically, Azure Machine Learning — which already boasted support for AI frameworks such as Facebook’s PyTorch, Google’s TensorFlow, and scikit-learn, in addition to automated hyperparameter tuning — now features a more intuitive automated ML interface designed to make model creation simpler, along with a drag-and-drop visual machine learning dashboard. Also new are a suite of MLOps features that tie into Azure DevOps, Microsoft’s end-to-end software development toolkit, intended to promote reproducibility, auditability, and automation in AI model design.
“[There is a category of AI practitioners] who are learning machine learning concepts, they want to make their own models, but they are not coders. This could be IT professionals, or folks with background in statistics or mathematics,” said Microsoft’s director of artificial intelligence Bharat Sandhu. “For those customers, we’re offering experience[s] to make models visually.”
That’s not all that’s in tow with this week’s bundle of upgrades. The ONNX (Open Neural Network Exchange) Runtime, an open source AI ecosystem developed by Microsoft, Facebook, IBM, Huawei, Intel, AMD, ARM, and Qualcomm that lets developers seamlessly switch between frameworks, now supports Nvidia’s TensorRT and Intel’s nGraph for high-speed inferencing (i.e., prediction) on Nvidia and Intel hardware. Additionally, Microsoft is making generally available hardware-accelerated models that run on field-programmable gate arrays (FPGAs), or integrated circuits designed to be configured after manufacturing.
Microsoft also announced the broad launch of cognitive search for Azure Search, its fully managed hosted cloud search service, which allows users to apply algorithms to extract insights from structured and unstructured content. And it previewed a capability that enables developers to store those insights to create BI visualizations or machine learning models.
Lastly, the company revealed that it’s now an active contributor to the MLflow project, an open source format for packaging data science code in a reusable and reproducible way. “AI and machine learning can turn developers into heroes, for their ability to deliver really personalized, super-immersive experiences to customers,” said Microsoft’s director of operational databases and Blockchain product marketing Wisam Hirzalla. “We want to make it easy for any company to use the technology.”
Microsoft launched Azure Machine Learning back in June 2014, and made it generally available in December 2018. Along the way, it introduced new tools like Machine Learning Workbench, which makes it easier for developers to manage machine learning models, and an AI model generation feature that automatically selects and optimizes algorithms for target scenarios. It also rolled out model explainability, a bias-mitigating solution that helps customers to identify which input features weigh heaviest on a system’s predictions.
Last fall, Azure Machine Learning gained a Python software development kit and interoperability with Power BI, a business analytics service that facilitates the creation of reports, dashboards, and more. As of November, AI models built in Azure can be shared within Power BI, which autonomously discovers models that each user has access to and automatically creates a point-and-click user interface to invoke them.
Azure Machine Learning complements services like Microsoft’s Azure Bot Service, a scalable chatbot platform with which 400,000 digital agents have been created to date (with 3,000 coming online each week). And it dovetails with Azure Cognitive Service images, a set of APIs, software development kits, and container images that enables developers to inject apps with AI.