IBM Watson promised to change the medical sector with its artificial intelligence (AI) Watson. However, the company has bet too high and cannot deliver on its promises.
This criticism was voiced by several doctors and professors to IEEE Spectrum. IBM Watson was introduced in 2014. During a presentation, Watson managed to analyse a series of symptoms of a patient and provide a number of possible diagnoses based on these. A certainty percentage was given for each of these diagnoses and referred to supporting medical sources for substantiation.
Successful and less successful results
In the ensuing years, Watson continued to be developed, achieving successes. In July 2017, for instance, researchers at IBM Research announced that they had succeeded in analysing genome data 960 times faster using IBM Watson. Manually analysing an average genome test takes around 160 man hours, while Watson managed to perform this process in just 10 minutes. This without compromising on quality.
Other projects, however, were less successful. In 2012, for instance, the Memorial Sloan Kettering Cancer Centre partnered with IBM to train Watson to diagnose and treat cancer patients. In this project, the AI did not deliver the desired results. Thus, the website STAT managed to get its hands on IBM documents showing that Watson regularly gave incorrect advice and did not live up to expectations. As an example, it describes a trial in which a cancer patient with heavy bleeding was prescribed medication by Watson that would make the bleeding worse. IBM stresses that this was a test based on fictitious information rather than real patient data.
A second project that was not a success was a collaboration with the MD Anderson Cancer Center. This hospital invested $62 million in a project around Watson, which was later cancelled. Watson would not meet the requirements of a functional product.
Critical
Several doctors and professors are critical of Watson, arguing that the AI has failed to deliver on its promises. In practice, Watson is said to function mainly as an AI assistant and is capable of performing certain routine tasks. "In terms of reputation, I think they are in trouble," Robert Wachter, chief executive officer of the medical division of the University of California tells IEEE Spectrum. Wachter points out that IBM is the first company to invest heavily in implementing AI in healthcare. This presents challenges, which IBM, as the frontrunner, is the first to face.
Herbert Chase, professor of Medical Sciences and Biomedical Informatics at Columbia University, was involved in the initial development of Watson. "Doctors - especially those who are on the front line, primary care - go to work every day knowing that they cannot possibly know everything they need to know to deliver the best, most efficient and most effective care," Chase said. The professor saw potential in Watson as a tool to support clinical decision-making and help doctors deliver better care.
However, Chase, who has not been involved in Watson's development since 2014, says he is disappointed with Watson's slow development in the healthcare field. "I am not aware of any spectacular home runs," the professor said.
'IBM not leading the way'
Eliot Siegel, a professor of radiology and board chair for information systems at the University of Maryland, has also been involved in Watson's development. Siegel, too, is critical. The professor argues that AI-based tools will be invaluable to doctors within a decade. However, he is not convinced that these tools will be developed by IBM. "I don't think they are at the forefront of AI," Siegel said. "The most exciting things are happening at Google, Apple and Amazon."
The full article can be found at IEEE Spectrum.
Author: Wouter Hoeffnagel
Source: IEEE Spectrum
Source: IBM Research
