Following up on my recent blog about artificial intelligence, I wanted to share some expert commentary from Science (July 7, 2017) in a paper by Jennifer Sills summarizing some of the most interesting uses of AI by both health care providers and the world of commerce:
- Robotic surgery may be ideal for helping to prevent human surgical fatigue and lead to faster recovery of patients. Nevertheless, there are often untoward issues that arise during surgeries which require expert assessment and rapid decision making. Equipping robots for all eventualities is probably impossible and in the end, impractical. Finally: who is responsible for robot error? The machine or the live surgeon?
- Using unmanned aerial vehicles to survey terrain and wildlife number is probably more precise than human observation. They become particularly valuable in tracking the nature and extent of fires, making managing fires more effective.
- A computer analysis of patented chemical reactions from 150,000 patent documents filed over a 40 year period revealed “bigger, greasier and flatter” synthesized products over that period. Unfortunately, the actual methodologies used to produce these items varied so much across investigators that program creators pointed out the need for standardized terminology and the inclusion of essential information about procedures. Until then, strict interpretation of accurate AI assisted data isn’t possible.
- Using machine learning to analyze students’ understanding of concepts based on their written answers in response to questions about the topics they were taught was touted as a way to standardize the quality of educational efforts. Almost predictably, computerized results varied about as much as human “expert” scoring. As one instructor observed, it is difficult to characterize and evaluate student comprehension of complex topics. At the very least, though, computer assessment saves faculty time.
- Using AI to predict outcome for cancer patients is an attractive idea, but as always, correctly programming the computer for accuracy in the myriad of possibilities that characterize the experience of cancer is the limiting factor: “garbage in, garbage out”. The oncology community is not ready to count AI’s usefulness in predicting precise outcomes and emphasizes the importance of patient-doctor interaction and communication in defining the elements of what determines outcome.
- The use of AI in exploring protein to protein interactions in plant life to improve survival in periods of drought is increasingly employed by biologists, apparently to an extent and with an accuracy that will make it possible to do with much fewer human laboratory biologists: a universal issue in the transitioning of human effort to AI.
- Help for close inspection of astronomers for the myriad of events that fill our stratosphere is at hand in the use of machines to point out the many intriguing objects that would benefit from more detailed scrutiny by human astrologers. The computer as a triaging machine is a useful concept.
Here’s what emerges: programming the computer completely and accurately for any given task is the first issue at hand, no matter what its intended use. Second, when AI does prove as or more accurate than human effort, we have to be ready for the humans who finds themselves redundant and begin to rethink our employment opportunities in this transitional era. The present expansion of AI in our lives is very analogous to the industrial revolution when we needed many fewer agricultural workers to grow and harvest our food and had to retrain those “left out and left over” for absolutely new,technically based jobs.
Specialist in internal medicine and primary care.