Microsoft announced the acquisition on June 20 of Bonsai, a Berkeley, Calif. startup specializing in artificial intelligence (AI) technologies for autonomous systems and industrial environments, Financial terms were not disclosed.
As it turns out, Bonsai has been on Microsoft’s radar for quite a while.
The company is one of many that have received venture capital financing from M12, formerly Microsoft Ventures. In May 2017, Bonsai announced it had raised $7.6 million in a Series A round led by Microsoft Ventures, with additional backing from Siemens, ABB Technology Ventures and Samsung NEXT, the South Korean electronics company’s venture capital fund.
Microsoft was mainly interested in Bonsai efforts to make industrial-scale AI and machine learning accessible to developers.
“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,” stated Gurdeep Pall, Corporate Vice President of Business AI at Microsoft, in the June 20 announcement. “The actual training takes place inside a simulated environment.”
Pall went on to note that Bonsai is currently working on “a general-purpose, deep reinforcement learning platform” that will enable enterprises to build AI-powered applications that support a variety of use cases, namely industrial control systems, robotics, manufacturing systems and HVAC (heating, ventilation, and air conditioning). Ultimately, Microsoft envisions combining Bonsai’s technology with AI innovations from its own research arm, along with the company’s cloud-based Azure Machine Learning and IoT services, to serve as the brains of next-generation autonomous systems.
Not passing up an opportunity to rib rival Google, Pall said Bonsai’s technology bested Google DeepMind in a reinforcement-learning benchmark. Using a simulated robotic arm, whose task was to stack blocks atop one another, Bonsai’s technology got the job done 45 times faster than the Google alternative, claimed the Microsoft executive.
Bonsai’s AI-enhanced industrial-control technologies are already having a productivity-enhancing effect on how some businesses operate.
“The 30x performance improvement Siemens recently realized auto-calibrating CNC [computer numerical control] machines powered by a Bonsai BRAIN is just scratching the surface of the significant business impact deep reinforcement learning can bring to these real world systems,” stated Mark Hammond, founder and CEO of Bonsai in a blog post.
BRAIN, short for Basic Recurrent Artificial Intelligence Network, are deep learning AI agents tuned to finding solutions to challenges that they encounter.
The technology helped Siemens drastically reduce the time it takes to recalibrate manufacturing equipment compared to a human in a proof-of-concept, a first for using deep reinforcement learning to automatically calibrate CNC machines operating in the real world.
The milestone is a significant one, since CNC machines used in manufacturing typically require frequent and sustained recalibration, a task that is performed by humans, who can tire as the day drags on and are sometimes known to introduce errors that can cause an assembly line to come to a crashing halt.