>>12541962you underestimate the power of "good enough"
concrete science takes time, $ and very delayed ROI
ML is the "good enough" option with significantly less time and $ investment.
if you think from big cooporation and small corporation perspective why not just use the "good enough" innovation and rake in the profits and then rake in more profits from newer "good enough" innovation instead of waiting for concrete science
That is the reason even though Google Alphafold 2 is not concrete science most companies and pharma companies will start gravitating towards such models for quick approximate discoveries and quick profits
Infact most likely opposite of what you said is likely going to happen and most of the concrete science efforts will be replaced by - dump huge data into model, check validity and novelity, if all yes then use it for new product.