Microsoft’s annual developer conference, Microsoft Build, went ahead despite the COVID-19 pandemic, albeit virtually.

The 48-hour digital forum made several important announcements about the trajectory of the company to over 100,000 registrants, the largest crowd ever to ‘attend’ the conference. “While this year feels different, coming together as a community is more important than ever. There are tens and thousands of you tuned in from dozens of countries around the world, it’s fantastic to see,” Microsoft CEO Satya Nadella said in his opening keynote.

The company did make several important announcements, unveiling tools for Machine Learning (ML) and emphasising on the importance of building more responsible Artificial Intelligence (AI) now that the technologies play important roles across industries, applications and governance. “Microsoft is committed to the advancement of AI and machine learning (ML), driven by principles that put people first, and tools to enable this in practice,” wrote Eric Boyd, the company’s Corporate Vice President, in a blog.

Boyd mentioned that the Aether Committee, an internal cross-company board, is helping Microsoft build more responsible ML abilities. The research has helped Azure Machine Learning develop more responsible capabilities. Boyd also invited attendees to review their open-source toolkits, InterpretML and FairLearn, that will equip data scientists and developers have better control on creating more responsible and fair ML and AI.

“Model interpretability capabilities in Azure Machine Learning, powered by the InterpretML toolkit, enable developers and data scientists to understand model behaviour and provide model explanations to business stakeholders and customers,” stated Boyd. Model Interpretability will help the community to build ML that are more accurate, understand a variety of model behaviours, and produce effects of different feature values on model predictions by running what-if analysis.

FairLearn will help Azure Machine Learning prioritise building fairer AI systems.

“Using Fairlearn with Azure Machine Learning, developers and data scientists can leverage specialised algorithms to ensure fairer outcomes for everyone,” wrote Boyd. Fairlearn will help ascertain model fairness during the training phase as well as deployment, optimise model performance while reducing unfairness and compare models dealing with unfairness through interactive visualisations. Among other important announcements, Boyd also introduced WhiteNoise, Microsoft’s differential privacy toolkit. “Using the new WhiteNoise differential privacy toolkit with Azure Machine Learning, data science teams can build ML solutions that preserve privacy and help prevent re-identification of an individual’s data,” announced Boyd. The toolkit was developed in collaboration between developers and researchers of Microsoft and Harvard’s Institute for Quantitative Social Science (IQSS) and School of Engineering.

Kevin Scott, the Chief Technology Officer, introduced Microsoft’s new supercomputer. The supercomputer, built in partnership with and exclusively for OpenAI, is meant to train AI models at OpenAI. “The supercomputer developed for OpenAI is a single system with more than 285,000 CPU cores, 10,000 GPUs and 400 gigabits per second of network connectivity for each GPU server. Compared with other machines listed on the TOP500 supercomputers in the world, it ranks in the top five,” states a Microsoft Blog.