At some point in our lives, advances in AI have made our lives easier and more advanced. Whether it's facial recognition, AI job interviews, or self-driving cars, technology has brought us convenience and efficiency in many areas. Here's a story about how IBM is using Watson to develop an AI doctor to maximize the convenience and efficiency of the healthcare system.
IBM Watson The rise of AI as a physician assistant
Diagnostic assistive machines, which help doctors by comparing and analyzing visual data such asX-rays, MRIs, and CTs taken at medical sites, are widely used in practice. The Cheonho Health Center in Gangdong-gu, Seoul, is the first in Korea to introduce an AI-assisted diagnostic machine for retinal and cardiovascular diseases.
The retinal and cardiovascular disease diagnostic assistant has already been certified as an innovative medical device by the Ministry of Food and Drug Safety. It allows patients to assess the risk of eye diseases and cardiovascular diseases through a simple eye test. It helps prevent and detect complications of metabolic syndrome and chronic diseases at an early stage.
In addition, the Ministry of Food and Drug Safety has selected the 'Class 2 Medical ImageDetection and Diagnosis Assistance Software' product, which analyzes patients'CT images and biometric information with AI technology to assist in the diagnosis of sleep apnea, as an innovative medical device and is supporting its commercialization.
The world is already accelerating the application of advanced technologies such as AI, biotechnology, and nanotechnology to existing medical devices and treatments.Industry and government departments continue to support the development and commercialization of medical devices with significantly improved safety and efficacy.
However, due to the nature of medical devices, there are cases where AI is subject to errors due to incorrect data. IBM's watson is a prime example of this, as we'll see below.
A chronology of IBM's artificial intelligence Watson
IBM's watson was unveiled in 2005. It came to prominence in 2011 when it defeated two previous champions on the popular American quiz show Jeopardy by an overwhelming margin.Like the 2016 Go match between AlphaGo and Lee Sedol, many people were intrigued by this exciting story of a machine transcending humanity.
In response to this achievement, IBM announced that Watson would eventually be applied to finance, law, and healthcare - at the time, it was powerful enough to process tens of thousands of books' worth of data in a single second.
High hopes for AI doctors
There were also high hopes that Watson, if put to good use, could help overcome cancer, the number one killer of humans. In fact, a cancer research center in the U.S. installedWatson to analyze 70,000 papers in one month to find protein elements that could be targeted for new anti-cancer genes.
Up until this point,Watson was an AI promising to revolutionize healthcare. IBM invested $62million in Watson to develop an AI that could help fight cancer. But in May2018, IBM officially labeled Watson's healthcare business a failure. That's because it created a dilemma: it got in the way of doctors' work and made them less efficient.
Is Watson Public Enemy?
"The cancer data was too complex, and data tainted by misdiagnoses, doctors' personal expressions, and more prevented Watson from performing as well as it could," says the New York Times in its analysis of Watson's failures.
A doctor at JupiterHospital in Florida who worked with Watson in real life said, "Watson was a complete failure."
Watson's inability to learn and adapt to clinical data, which varies from country to country, was another reason for its failure, as was the time it took to analyze the fast-moving academic literature and train its skills.
For example, Watson recommended a cancer patient with severe bleeding to use a drug that could actually make the bleeding worse - the most catastrophic error an AI can make in healthcare. There were other instances of Watson suggesting treatments that could have endangered the patient's life. Eventually, it was concluded that while it could be extremely efficient when fully commercialized, it was a risky decision that could kill people based on faulty data.
To summarize, their failure to fully familiarize themselves with cancer data that varies betweenEast and West, age and ethnicity, and their inability to keep up with cancer journals and science as it changes over time, led to their failure.
AI doctors make mistakes
In Korea, Watson's inability to reflect the realities of health insurance led to over treatment with costly antibiotics, and doctors were unable to incorporate the results into their practice.
"Even if the drugs recommended by Watson are available in Korea, they cannot be covered by insurance or are only licensed to treat certain cancers," said ChoiJong-kwon, a professor of hematology-oncology at Konyang University Hospital.Watson, which was supposed to revolutionize healthcare, ended up being a failed project due to unrefined data.
IBM Watson - Becoming an AI medical assistant, not an AI doctor
In the end, the industry's expectations of an AI doctor are dwindling, and Watson is increasingly recognized as an "AI medical assistant.
A professor at a university hospital who is using Watson pointed out that "doctors who treat patients have to worry every time they make a decision against Watson, and hospitals are not sure about the future of Watson because they have to pay high royalties to IBM every year while not receiving any revenue fromWatson".
*Reimbursement: A system that pays for medical care based on usage and price by setting a fee for service (the total amount of money a medical center receives from patients and health insurers for providing medical services covered by health insurance).
For AI doctors to really be possible...
In the end, the voice and demand for complete data is becoming a necessity, not an option. The technology that creates AI is important, but so is the quality of the data that feeds it.
Some experts wonder what would have happened if the quality of the data Watson was trained on had been higher. It's important to note that it's the refined data that underpinsAI that makes it viable. As a result, the need for complete data is becoming more of a necessity than an option.
We need to give voice to the fact that it is diverse data and smart algorithms that create accurateAI. At Datahunt, we will continue to strive for more complete data by using human-in-the-loop technology and more.