>It sounds like he published some theoretical musings back in 1990s without any real practical implementation

He published working models back then, the problem was compute power was very limited. In the past decade deep learning took off, people took his models, renamed then and ran them on vastly more powerful computers, to great success, then failed to cite him.