Ann: ResApp announces positive results of COVID-19 screening test, page-479

My basic understanding of machine learning is that there are different approaches that you can take depend on the situation / use case.

Supervised ml:
feed in a dataset
run experiments
review results
human input to guide / refine algorithms

Unsupervised ml:
framework setup, whereby a computer teaches itself to solve a puzzle or navigate a maze etc.

In resapps case I would imagine that any refinement to the algorithm(s) would be done using human supervision with robust processes in place.

To your point though KrisM I agree that accuracy should improve over time, as more clinically validated data gets fed in. Unlikely that the algorithms would self learn though.

Happy to stand corrected though and hope that this helps.