Those criticisms aside, if this system went up against the current status quo in a prospective study, there is a reasonable chance it would be on par at least w

Those criticisms aside, if this system went up against the current status quo in a prospective study, there is a reasonable chance it would be on par at least with humans. Whether that makes it a worthwhile endeavour comes down to questions of cost, technical complexity and health outcomes. This last part is actually a significant barrier to 'AI' in healthcare. For that reason, I suspect most companies will prefer to sell their products integrated into assistant style software, where the value proposition is tied to reimbursement eg reporting more scans. The sensitivity is also still not as high as you would like ideally... this is a limitation of mammography.