Microsoft’s AI Moves at Ignite 2019 Deserve a Closer Look

Zooming in on big data, responsible AI and robotics

Intrepid attendees at Ignite 2019 navigated more than 1,500 sessions and a mind-numbing flurry of announcements as Microsoft flexed the full breadth of its technological capabilities in Orlando, Florida. The show revealed progress in multiple areas: cloud and edge computing, security and management, software development and modern workplace to name a few. And although announcements in these areas grabbed much of the limelight, a crucial topic that received less notice was artificial intelligence (AI).

Let’s explore the moves and discuss what they mean for Microsoft and the nascent enterprise AI market.

The Big Headlines from Microsoft Ignite

Firstly, a quick recap of some major announcements and areas that took most of the attention this year.

Azure Arc

Azure Arc was Ignite’s most eye-catching announcement. It revealed that Microsoft is now homing in on hybrid and multicloud management, representing the biggest shift yet in Azure’s strategic evolution. Azure Arc follows similar investments made by Google Cloud with Anthos and IBM Red Hat over the past 18 months. It’s a sign that the competition in cloud computing is shifting to the control plane. It will be fascinating to see if Amazon Web Services (AWS) follows suit with an update to Outposts next month at re:Invent.

Power Platform

Power Platform, a tool for “citizen developers”, also received a lot of attention from Microsoft. The low-code, no-code platform has grown to become one of its most important strategic assets. Ignite showed how Power Platform is helping Microsoft reach new enterprise developers and helping to enhance the customization and “stickiness” of its software-as-a-service products such as Teams and Dynamics 365. It will be a vital weapon against Salesforce, SAP, Google G Suite and Slack in the future.

Microsoft 365

And as expected, Microsoft 365 was a dominant topic at the event as well, with Teams alone generating over 50 breakout sessions. The arrival of Project Cortex and Microsoft Endpoint Manager into Microsoft 365 are examples of how the service is progressing from the commercial bundle launched in 2017 to a more cohesive platform.

AI Announcements You May Have Missed

Against this backdrop of critical areas, it’s unsurprising that it felt like AI took a back seat this year. But that’s not to say that Microsoft’s AI updates were any less important.

Big Data and Analytics: Azure Synapse Analytics

One of the main blockers of AI in enterprises at present is data quality. Microsoft CEO Satya Nadella’s keynote presentation featured the launch of Azure Synapse Analytics, a new big data analytics service that combines cloud-based serverless computing resources with on-premises infrastructure for a unified experience in data ingestion, preparation and management. In a breakout session, a Microsoft product team leader said the solution is a staggering 75 times faster than Google’s BigQuery service and three times faster than Amazon’s Redshift. It also integrates with Azure Machine Learning.

Performance and trust in AI depends entirely on the quality of the data tools and practices that companies feed into it. The new service is an important move that should become a crucial focal point for Microsoft’s AI initiatives over the next 12 months.

AI and Industrial Robotics with Autonomous Systems

Another important area was the expansion of Microsoft’s preview programme for Autonomous Systems, its incubation project for its machine teaching and reinforcement learning solutions aimed at mechanical and chemical engineers who build industrial systems and robotics.

Based on Microsoft’s acquisition of Bonsai in 2018, the Autonomous Systems platform helps engineers build intelligent machines that go beyond basic automation to be able to sense, learn and respond to changes in their environments through AI. For example, researchers at Carnegie Mellon University used the platform to train perception and object detection models deployed on robots involved in search and rescue operations in mines. Microsoft demonstrated the technology at the Autonomous Systems stand with a robot playing air hockey, which you can see in action below.

Microsoft also announced several partnerships in this area, including with MathWorks, AnyLogic and CGTech, whose simulators are used by engineers worldwide. Other partnerships with solution providers Fresh Consulting and Neal Analytics, and enterprise drone software maker 3DR were also announced.

Autonomous Systems is a fascinating part of Microsoft’s business and one that’s worth watching. Along with the traditional developers, data scientists and business users, who have been the main focus of cloud suppliers, engineers will be important users of AI tools in the future. When combined with its hybrid cloud, edge and Internet of things offerings, the platform is part of a bigger play that could make Microsoft more relevant in industrial environments, plant operations and in operations technology. This is a large, untapped opportunity for Microsoft.

But Microsoft isn’t alone in eyeing the potential of this area. AWS is using the warehouse operations of its retail arm to incubate its AI for industrial environments. Its 2018 launch of AWS RoboMaker illustrated Amazon’s intent here. Additionally, Google’s DeepMind is concentrating research into this field.

As the worlds of industrial robotics and AI collide over the coming years, it will be interesting to see which cloud service provider will best marry the IT and operations technology environments (see also What Ignite 2018 Signalled for Microsoft and the Cloud Wars).

Azure Machine Learning Goes Deeper into Responsible AI

The biggest area of development, however, was in Azure Machine Learning, which saw several updates, of which the most notable were in responsible AI.

One of the main reasons why AI hasn’t yet shifted out of the labs and into everyday workplace life is that companies need more support with its governance (see Taking AI from the Lab to Real Life). Many businesses I speak to need better answers to important questions before they fully operationalize the technology, such as: how do I design ethical applications for my business? How do I avoid bias in my data? How can I ensure the models I build are explainable and interpretable to others, including business users, and not just my developer or data science teams? How can I guarantee that algorithms I deploy respect privacy and are secure and compliant for the industry I operate in?

Aimed at solving these challenges, Microsoft announced a set of new governance features in its life-cycle management platform MLOps in Azure Machine Learning. These included role-based access controls, virtual network security, dataset and model traceability and auditing, data drift monitoring and above all, new solutions for interpretability and bias detection.

Microsoft said it will integrate its open-source tool for model explainability, Interpret ML, into Azure Machine Learning and showcased several new dashboards for model interpretability that are now in the platform. Finally, it announced Fairlearn, a new toolkit for bias detection and mitigation in GitHub, which, through a set of visualizations, helps uncover insights into fairness in model predictions.

Responsible AI and governance have become the most important topics in the AI market in 2019. According to CCS Insight’s IT Decision-Maker Workplace Technology Survey 2019, the level of transparency of how systems work and are trained and the ability of AI systems to ensure data security and privacy are now the two most important requirements when investing in AI and machine learning technology, cited by almost 50% of respondents. The survey also found that for 43% of respondents, tools that support AI operations and life cycle management are the biggest current gap in the market for suppliers of AI platforms.

Microsoft’s push into these key arenas is currently ahead of several of its main rivals, but to lead the market in trusted AI, the company must work harder to raise awareness of these capabilities.

Microsoft’s Flywheel Is Gathering Pace

Ignite showcased the enviable “flywheel” effect occurring throughout Microsoft, as its tightly aligned cloud infrastructure, security and applications businesses continue to accelerate.

At the centre of this flywheel is AI. And although it may not have garnered as many headlines this year, I believe that, altogether, Microsoft’s progress in data analytics, autonomous systems and responsible AI was the dark horse of the event’s announcements.

Microsoft is making ground against the competition, having reported in its latest quarterly results that 20,000 customers are now using Azure AI, including more than 85% of Fortune 100 companies. This marked the first time the company shared its AI customer numbers.

Microsoft needs to shout louder about its achievements, before the competition closes the gaps.