I think his contribution at that time was mainly clever ways to drastically scale up existing neural net techniques, but just scaling things up 100x or 1000x seems to be a large part of the magic of deep learning.
And TBH, it was him, who knows general computing from circuits to planet scale distributed systems, really enabled the AI as it today, at least inside Google.
I think his contribution at that time was mainly clever ways to drastically scale up existing neural net techniques, but just scaling things up 100x or 1000x seems to be a large part of the magic of deep learning.