Starting with Machine Learning on devices like MCU - Part 15 (Why TensorFlow on MCU, What is the benefit?)

While we talk of AI/ML and using huge processors and data centers, why should we choose a MCU to do AI activities? What benefit do we get?

• Local computing which we call Edge AI, no need of connecting to expensive serves, everything processed at the point of data generation, quick decision and less resources, more secure
• Not all requirements are resource intensive, for a small sensor based read and decision, we need not use a huge processor
• If the same achieved on bigger processors can be achieved with a low cost solution and tinkered as per our requirement, what more do we need?
• Better power management, lesser power consumption, compact solution
• Once a daemon for embedded engineers, AI/ML becomes more accessible for embedded engineers and help them provide real time solutions
• Quantization available to make models less complex and used on devices like MCU, enabling usage of AI/ML models