The entirety of AI innovation is driven by hardware improvements. Almost all AI research is done at the edge of our compute capabilities, and most of the models which produce results today existed in some form 20 years ago.
We didn't leap-frog our predecessors presumptions, we built exactly what many envisioned but were unable to put to fruition with the computers of their time.
The point is that we retried something that was previously deemed infeasible. Had you been too much aware of the history, you might not have done that.
We didn't leap-frog our predecessors presumptions, we built exactly what many envisioned but were unable to put to fruition with the computers of their time.