New news, IBM intends to sell Watson health division.
This also means that IBM layout ten years of AI medical road is over!
The sale, also with the previous series of IBM divestiture, spin-off operation as the same, are to streamline the company’s business, more focused on the Cloud Computing field.
Watson Health department, is IBM layout AI medical window, mainly using AI to help hospitals, insurance companies and pharmaceutical manufacturers to manage data and assist in diagnosis.
But established six years ago, annual revenue is only $1 billion, accounting for less than 2% of the company’s total revenue, has not yet been profitable. And this department before the acquisition alone spent more than $4 billion.
It is no wonder that IBM is thinking of selling.
Plus because of cloud computing, its competitors – Microsoft and Amazon valuation has soared more than ten times.
So this once known as AI medical firm “long runner”, how will go next, still do not know.
IBM’s ten years of AI healthcare
Watson, originally a computer system capable of answering natural language questions, was named after Thomas Watson, the founder and first CEO of IBM.
It was originally developed to answer questions on the American variety show “Dangerous Edge”.
Watson gained fame in 2011 when it won the show’s champion human contestant, taking Home a $1 million prize.
The blue giant IBM then bet heavily on AI healthcare.
The day after Watson won the championship, it announced its ambitious plan to make Watson an AI doctor in the future and also promised to launch its first healthcare products in the next 18-24 months.
In February 2013, IBM announced Watson’s first commercial application, to be used in collaboration with WellPoint (now Anthem) for utilization management decisions in lung cancer treatment at a New York City cancer center.
From there, IBM focused its research on this piece of oncology treatment.
In 2015, IBM established a dedicated division, Watson Health.
At that Time, the former CEO of IBM, Virginia Rometty, also called Watson Health the company’s “moon landing plan”, which shows IBM’s determination to lay out medical care.
In 2016, Watson Health, which was established only two years ago, has already reached more than 10,000 employees.
But from a series of cooperation, it is still a big thunder but little rain.
Among them, there is a pilot program launched with MD Anderson Cancer Center, the purpose of the “mission to eradicate cancer”, after spending $ 62 million, because it did not meet expectations and terminated.
According to 2019, IEEE Spectrum, of the 25 representative projects IBM Watson has worked on with other organizations since 2011, only five collaborations have resulted in AI healthcare products.
In addition to collaborations, Watson Health has another way of thinking, and that is to acquire, for a not insignificant amount.
Some of the better known brands are Merge Healthcare, which analyzes mammograms and MRIs, with an acquisition value of nearly $1 billion.
ePhytel, which assists with patient communication, for which IBM paid about $230 million.
Then there is Truven Health Analytics, which is used to analyze complex medical data, with an acquisition value of $2.6 billion.
But insiders say the acquired companies didn’t make much of a difference.
The product coming out of IBM’s Watson Health division is completely different from the AI doctors once envisioned.
It is more like an AI assistant that can perform certain daily tasks, but is not actually involved in the medical diagnosis process.
The controversy
From its inception, Watson Health has suffered from multiple controversies.
In an effort to find a business case for AI healthcare, IBM undertook a dizzying array of projects targeting different players in the healthcare system: doctors, administrators, insurers and patients.
Among the multiple attempts, IBM chose to let Watson’s NLP understand medical texts rather than engage directly with medical image diagnosis.
Even Turing Award winner Yoshua Bengio is not optimistic about this AI medical model, arguing that
In medical text documents, AI systems cannot understand their ambiguity or the subtle clues that human doctors notice.
In medical text documents, AI systems are unable to understand their ambiguity or the subtle clues that human physicians notice.
The results were naturally less than ideal, with Watson unable to extract its own insights from new medical history events in the medical literature, and the researchers also found that it was unable to mine information from patients’ electronic health records as expected.
The researchers realized that Watson was unable to compare new patients with cancer patients for whom hidden patterns had previously been discovered. IBM had hoped that the AI would mimic the ability of their expert oncologists to use previous results and experience to develop treatment strategies for new patients.
Herbert Chase, a professor at Columbia University’s Department of Medicine, had worked with IBM on a prototype of a diagnostic tool they hoped would become a “clinical diagnostic support” tool. But the two sides parted ways in 2014 because of Professor Chase’s frustration with Watson’s slow progress.
On top of that, Watson itself has made its fair share of mistakes.
In 2018, Watson was revealed to have prescribed the wrong drug to patients, which could have killed people in serious cases.
And the drug has been used by more than 230 hospitals worldwide, including 67 Chinese hospitals such as Peking University Cancer Hospital and Xuanwu Hospital, in 22 provinces.
At the World AI Conference 2019, Alok Gupta, vice president of IBM Watson Health Division, also put up a large PowerPoint presentation that they have been involved in all six areas.
At that time, this division was rated as a firm “long runner” in AI healthcare.
Now it’s facing the fate of sale, which is also saddening.
Recent Comments