Kodez | Your trusted consulting partner.
May 22, 2020

Microsoft Build 2020 : Azure AI Announcements

The global pandemic situation has changed the world's view on technological interventions and innovations. Carrying their corporate mission of “Empowering every person and every organization on the planet to achieve more”, the big tech giant Microsoft hold their annual developer conference “Build 2020” as a 48-hour virtual event.

Being virtual didn’t get down the excitement the conference creates among the developer community as well as within the enterprises. Here are few exciting announcements Microsoft did on Build 2020 related to AI, the hottest topic in the table right now.

Responsible ML in Azure Machine Learning

Responsible ML address model understanding, protection and control

Since AI is becoming a critical part across all industrial domains, safety, privacy and responsible use of AI related intelligent applications is a key area to focus on. Responsible ML allows the developers to control and protect the machine learning models develop through Azure Machine Learning while ensuring the human interpretability of the predictive models.

Here are the main open source libraries Azure AI has introduced under Responsible ML

  • InterpretML - For performing machine learning model interpretations
  • Fairlearn - For accessing the fairness and biasness of machine learning models
  • WhiteNoise - Differential privacy toolkit on Azure Machine Learning

Bot Framework and Azure Bot Service updates

Overview of the services leverage by a virtual assistant

Bot Framework facilitate for developing intelligent chatbot agents. Microsoft is making adaptive dialog capability generally available by enabling bots to switch contexts within a conversation. These are some main features came out recently related to bot development.

  • General availability of Bot Framework composer.
  • Enabling developers to reuse the skills of other bots in the same organization.
  • Enabling human hand-off feature in Azure bot service giving them the ability to keep the human in the loop conversations in a chatbot workflow.

Enhanced Azure Cognitive Services features

Enhanced LUIS interface

Azure cognitive services democratize AI, by reaching out every developer to leverage the power of complex machine learning applications just through a few clicks. Several new features on Cognitive services have been announced with Build 2020.

New Cognitive Service capabilities which are generally available

  • Advanced text extraction in computer vision – Introducing Read 3.0 by expanding the text reading language coverage to French, German, Portuguese, Italian, and Dutch other than English & Spanish.
  • Cognitive service containers – allows the developers to deploy Cognitive services from cloud to Edge with containers.
  • Language Understanding (LUIS) – A revamped labeling experience has introduced for LUIS making it easier to understand complex language structures.

Other than these updates, speech to text in Cognitive service is coming up with 30% improvement in accuracy and 15 new voices in speech.

New Cognitive Service capabilities which are in preview

Azure Personalizer Apprentice mode feature – Personalizer delivers personalized and relevant experience for your application user through an AI powered intelligent backend. The new apprentice mode allows the Personalizer API to learn in real time alongside existing solutions without being exposed to users until it delivers performance results according to desired KPI goals.

Azure Cognitive Search updates

Azure Cognitive Search overview

Azure Cognitive Search is a fully managed search-as-a-service offered by Microsoft. It provides developers APIs and tools for adding a rich search experience over private, mixed content in web, mobile, and enterprise applications. With Build 2020 Azure search comes with Azure machine learning integration and custom search ranking as preview features enabling the developers to add the same natural language stack capabilities Bing and Microsoft Office are equipped with for their application development.

New optimization techniques built for Turing model for ONNX Runtime

ONNX Framework

ONNX runtime is an open source machine learning model runtime which helps to train and inference machine learning models on any platform, hardware or underlying framework. Microsoft have announced some key optimization techniques for Turing model, which is the largest model in the world. The open source library is now available on GitHub to be optimized by ML engineers.

Project Bonsai – public preview

Project Bonsai is a machine teaching service for building and operating autonomous systems. This is going to be the next revolution of Industry 4.0 era, where the AI powered process automations going to take place with Azure Machine Learning. Project Bonsai allow engineers to apply their subject matter expertise to accelerate the development of intelligent control systems without the need for data science skills.

Kodez | Cloud. DevOps. AI. Mobility.