I’m a researcher working on computer vision and AI. My main research objective is to introduce dynamics into machine learning which is currently dominated by statistics. Having dynamics is necessary for ML models to produce true understanding of physical reality. My recent attempt to build a machine learning system that naturally incorporates time, online processing and unsupervised learning of the physics is described here (github code here). Until dynamics is deeply incorporated in machine learning models, they will remain “statistical machines” that “statistically work” which is not enough for a real world application such as robotics. My long term ambition is to address Moravec’s Paradox. I gained majority of my experience during the six+ years (2010-2016) I've been working at Brain Corporation in San Diego.
I'd like to use this blog to share some of my thoughts on Machine Learning, from a slightly different perspective than a lot of the mainstream, namely from the perspective of actually applying these techniques to a physical device existing in physical reality. As we've learned, this is a whole different ballpark than running your algorithm on a dataset in a sterile, digital world. I hope you will find this read interesting or at least entertaining.