Self-supervised learning is an approach to machine learning that uses data itself to create labels and generate insights without relying on human supervision or external feedback. It has the potential to bring AI closer to human-like learning through observation, association, and abstraction. Our award-winning Autonomous Speech Recognition (ASR) engine uses self-supervised learning to improve its understanding of voices by training on vast quantities of data, including 1.1 million hours of audio. This approach enables our ASR to develop a more comprehensive understanding of voices and build common sense through incremental learning, ultimately making it more accessible and human-like.