Microsoft is buying Bonsai, an AI startup based in in Berkeley, California as part of its vision to make it easier for developers and subject matter experts to build the “brains”— machine learning model for autonomous systems of all kinds.
Bonsai has developed a novel approach using machine teaching that abstracts the low-level mechanics of machine learning, so that subject matter experts, regardless of AI aptitude, can specify and train autonomous systems to accomplish tasks. The actual training takes place inside a simulated environment.
The company is building a general-purpose, deep reinforcement learning platform especially suited for enterprises leveraging industrial control systems such as robotics, energy, HVAC, manufacturing and autonomous systems in general. This includes unique machine-teaching innovations, automated model generation and management, a host of APIs and SDKs for simulator integration, as well as pre-built support for leading simulations all packaged in one end-to-end platform.
“What I find exciting is that Bonsai has achieved some remarkable breakthroughs with their approach that will have a profound impact on AI development. Last fall, they established a new reinforcement learning benchmark for programming industrial control systems. Using a robotics task to demonstrate the achievement, the platform successfully trained a simulated robotic arm to grasp and stack blocks on top of one another by breaking down the task into simpler sub-concepts,” Gurdeep Pall, Microsoft’s Corporate vice president, Business AI said in the company’s blog post.
“Their novel technique performed 45 times faster than a comparable approach from Google’s DeepMind. Then, earlier this year, they extended deep reinforcement learning’s capabilities beyond traditional game play, where it’s often demonstrated, to real-world applications.”
Using Bonsai’s AI Platform and machine teaching, subject matter experts from Siemens, with no AI expertise, trained an AI model to autocalibrate a Computer Numerical Control machine 30 times faster than the traditional approach. This represented a huge milestone in industrial AI, and the implications when considered across the broader sector are just staggering.