Microsoft’s Project Bonsai to train industrial control systems using machine learning

by Sebastian Moss
Article Image

Machine learning for machines to learn

Microsoft has released a platform for training autonomous industrial control systems.

Project Bonsai, available in public preview on the Azure cloud service, aims to use "machine teaching" to help industrial robots interact with the world, based on technology Microsoft acquired when it picked up California-based startup Bonsai back in 2018.

Train your replacement

Here's how Project Bonsai works: first, subject matter experts without an AI background are asked to break up their workflow into simple steps and tasks, that are then taught to the AI agents.

Those AI agents then act out the lessons in a simulated environment, and the experts are able to provide feedback and guidance as the system learns. Once it reaches a sufficient level, it can be integrated into real-world manufacturing scenarios.

“This approach bridges AI science and software to the traditional engineering world, enabling fields such as chemical and mechanical engineering to build smarter, more capable, and more efficient systems by augmenting their own expertise with AI capabilities,” Mark Hammond, Microsoft general manager for autonomous systems, said.

The company has already signed up a few customers, including the chemicals division of Siam Cement Group. “Our problem was relatively complex, due to the nature of chemical processes,” Surachate Chalothorn, CTO for Olefins and Operations of SCG's Chemicals Business, said at the now-virtual Microsoft Build.

“To achieve our goal for operational excellence, having a good simulation model was not enough. We believed that chemical process simulation needed to integrate with AI technology to solve our complex problems.”

SCG said that the model it developed with its own team, on top of Project Bonsai, can run its operating procedure in simulation '100,000 times a day.'

Specifics on what this actually means for performance improvements were not shared, but Chalothorn claimed that it enabled the company "to make a breakthrough to the next level in process optimization by achieving better, faster, and safer operations, simultaneously.”

Practitioner Portal - for AI practitioners


Hesai and Scale AI open-source LiDAR data set for autonomous car training


Scale claims this is the first time such data has been released with zero restrictions


IBM adds free AI training data sets to Data Asset eXchange


Big Blue has something for you

Practitioner Portal


More EBooks

Upcoming Webinars

More Webinars

Experts in AI

Partner Perspectives

content from our sponsors

Research Reports

More Research Reports


Understanding the advantages of AI chatbots over rule-based chatbots

Infographics archive

Newsletter Sign Up

Sign Up