People and machines are vessels for different species of intelligence. Comparing them invites disparaging labels: natural versus unnatural, real versus artificial.But in reality, machine is whatever people make of it. It is not truly intelligent
People and machines are vessels for different species of intelligence. Comparing them invites disparaging labels: natural versus unnatural, real versus artificial.
But in reality, machine is whatever people make of it. It is not truly intelligent because it has no ability to function entirely on its own. Ultimately, it depends on people for its application. Take the use of computer-aided detection in medical imaging.
Social and economic concerns have guided the evolution of this technology. Algorithms are applied after people have had a first whack at the data, creating a "second read" capability-a backup that does not threaten the ego or position of the radiologist while compartmentalizing the technology for reimbursement.
But subordinating machine intelligence by forcing CAD programs to operate in the background is extraordinarily inefficient. Why make a radiologist read an image twice to look at what might have been missed the first time? Some people say that having the computer analyze the image first might risk missing something important and, by highlighting other areas, distract the radiologist from what would otherwise be seen. The question is, why must there be a first or second reading at all? Why not a simultaneous read with the computer serving as a digital assistant, presenting processed and unprocessed images side by side?
This is the idea behind "computer-assisted reading," a technology that has passed FDA review and is being readied for the U.S. market. Its first application will detect signs of lung lesions or nodules buried in the digital signals of CT scans. Its developer, Medicsight, has more such programs in development for detecting polyps in the colon and calcifications in the heart.
Their acceptance will depend on radiologists. And that will take a major change, not only in the practice of medicine but in human nature.
Intelligent computing platforms will reach their potential only when their human operators accept them as partners. They cannot be treated as rivals or tools on the same evolutionary scale as a shovel. They must be afforded a unique status, something in between, that supports an active partnership of people and machines.
Without innate intelligence and common sense, computing platforms do nothing more than process zeroes and ones. That realization should be enough to quell any deep-seated fears of being supplanted by technology-that and knowing "off" is a flip of the switch away.
The Reading Room: Artificial Intelligence: What RSNA 2020 Offered, and What 2021 Could Bring
December 5th 2020Nina Kottler, M.D., chief medical officer of AI at Radiology Partners, discusses, during RSNA 2020, what new developments the annual meeting provided about these technologies, sessions to access, and what to expect in the coming year.