>but they wouldn't have been possible without the academics that kept working on neural networks while that field was actively scorned during the 90s-2000s AI winter
The bottleneck was on compute power. Industry would have also worked on neural networks once the compute power for it existed.
I think the point is that universities and public research broadly speaking tolerate a lot of risk that private institution are not built to shoulder. I'm no AI historian but ANNs were an intellectual backwater after the 1st/2nd AI Winter and the only entities "foolish" enough to keep investing in them were Canadian public research universities. If it weren't for Hinton's dogged quasi-autistic special obsession with them ... no for profit entity would have bothered to invest in "computer power".
The bottleneck was on compute power. Industry would have also worked on neural networks once the compute power for it existed.