Bonsai software

Bonsai AI, Inc. was a Berkeley, California-based artificial intelligence startup founded in 2014 by Mark Hammond and Keen Browne, both former Microsoft employees. Hammond, the CEO, brought a deep interest in cognitive science and neural systems, holding a degree in the field from Caltech and having worked in the Yale Neuroscience Department. Browne, who served as Head of Product, contributed his experience from four years in Microsoft's developer division. The founders' vision was to simplify the complex field of AI, making it accessible to developers and subject matter experts who lacked deep machine learning expertise.

The company developed a general-purpose, deep reinforcement learning platform designed for enterprises, particularly those utilizing industrial control systems. Its primary market included sectors such as robotics, energy, HVAC, and manufacturing. Bonsai's business model revolved around abstracting away the low-level complexities of artificial intelligence, allowing developers to program AI models without needing to build neural networks from scratch. This was achieved through a novel technique called "Machine Teaching," which combined human expert knowledge with machine learning to guide the training process. The platform featured a special-purpose programming language called Inkling and an AI Engine that would automatically generate, train, and host the necessary models. This allowed users to deconstruct complex problems and curate training data, often within rich simulation tools, to teach AI "brains" how to perform control and optimization tasks.

Prior to its acquisition, Bonsai had raised $13.6 million from investors including Microsoft's venture arm M12, NEA, ABB, Siemens, and Samsung Next. The platform demonstrated significant efficiency gains, such as training a simulated robotic arm 45 times faster than competing approaches and helping Siemens auto-calibrate a CNC machine 30 times faster. In a significant milestone, Microsoft acquired Bonsai in June 2018 for an undisclosed amount, integrating its 42 employees into the AI and Research group. The acquisition was a strategic move for Microsoft to create an end-to-end toolchain for building autonomous systems. Bonsai's platform was combined with Azure Machine Learning, Azure Cloud, and Project Brainwave to empower developers to build, operate, and manage "brains" for autonomous systems deployed via Azure IoT.

Keywords: machine teaching, deep reinforcement learning, autonomous systems, industrial AI, control systems optimization, AI platform, simulation-based training, Mark Hammond, Keen Browne, Microsoft acquisition, Project Bonsai, Azure AI, robotics AI, manufacturing AI, reinforcement learning platform, Inkling programming language, low-code AI, intelligent systems, AI model generation, developer tools for AI