Ten years of IBM Watson: AI medical, a dream

Ten years of IBM Watson: AI medical, a dream

When the 2013 Waston IBM (Watson) the time for bold commitment to enter the field of health care, medical AI’s imagination is opened. From Microsoft, Google, GE Healthcare, Tencent, Baidu to countless start-ups, AI is spreading across fields such as image recognition, auxiliary diagnosis, drug research and development, health management, data intelligence, and epidemic prevention and control.

Especially in 2020, which has been shrouded by the epidemic for a long time, medical care has become the most concerned area of ​​the global capital and technology market, and domestic medical AI continues to be favored by capital. According to statistics from Caixin reporters, since June this year, at least nine artificial intelligence medical companies in China have obtained financing, involving multiple medical scenarios.

However, the fierce capital cannot change the commercialization difficulties faced by AI in the medical scene. As the first company to stand up and vigorously promote the application of artificial intelligence to the clinic, IBM Watson has proven to everyone that neither is successful in either changing the industry or creating wealth for the company.


Ten years ago, the public’s confidence in IBM was beyond doubt.

At that time, the IBM supercomputer Watson defeated the two human champions in Jeopardy (Jeopardy!) , demonstrating the power of artificial intelligence. IBM believed that this was just the beginning of a technological revolution that was about to sweep the entire society.

“We are already exploring the application of Watson skills to medical, financial, legal, and academic circles with rich and diverse languages.” IBM announced the day after Watson’s victory and entered the healthcare field with a bold commitment. The company promises that Watson’s first commercial healthcare products will be launched within 18 to 24 months.

Soon, they formed a strong circle of friends with big Vs in the medical field, including Memorial Sloan Kettering Cancer Center (Memorial Sloan Kettering Cancer Center) , Mayo Clinic, CVS Health and Johnson & Johnson.

At that time, Dr. David Ferrucci, who led the creation of the Watson team, warned: Beware of your promises.

When it comes to machine learning, massively parallel computing, semantic processing, and other fields, Watson’s great thing is to integrate these technologies into an architecture to understand human natural language . This is different from the computer “Deep Blue” in 1997, which is only doing very large-scale calculations and can only be seen as a manifestation of human mathematical ability.

Watson’s design function is to recognize word patterns and predict the correct answer to a quiz game. Ferrucci said that this is not a universal answerer that can be used in the business field. It is likely to fail the second-year reading comprehension test.

In the second year after Watson’s victory, Ferrucci left IBM to join the hedge fund giant Qiaoshui. A few years later, he founded Elemental Cognition, focusing on building a natural learning system that understands the world in a human way.

In the next eight years, IBM invested huge sums of money. For example, in 2014, US$1 billion was invested in the Watson division. The company promotes Watson as a charitable digital assistant that can help hospitals, farms, offices, and factories. IBM believes that its potential uses are limitless, from discovering new market opportunities to tackling cancer and climate change.

Not just to change the industry, Watson is also given high hopes by IBM-to provide years or even decades of growth and profits.

The seeds of IBM’s lagging behind other giants have been sown by its predecessor, Sam Palmisano. During his tenure, IBM focused on investor earnings per share figures, while ignoring the existential threat of cloud computing and other technological advancements to IBM’s business. So much so that when Ginni Rometty took over, IBM began to try to compete again, and putting the Watson mark on anything related to big data, machine learning and analytics is one of these efforts.

But these did not enhance IBM’s wealth. The stock prices of Amazon, Microsoft and Google have multiplied many times, but since Watson participated in ” Jeopardy!” , IBM’s stock price has fallen by more than 10%.

Recently, “New York Times” reporter Steve Lohr interviewed more than a dozen current and former IBM managers and scientists and found that the company’s mistakes on the Watson project began with the early focus on large and difficult projects with the goal of helping the company Bring praise and considerable income.

Manoj Saxena, the former general manager in charge of Watson’s business, said that Watson’s original goal was to do pioneering work that would benefit society, which is commendable. But this is unrealistic. “It turns out that these challenges are much more difficult than expected, and they take much longer,” he said.

Martin Kohn, the former chief medical scientist of IBM Research, recalled that he had suggested using Watson for a narrow “proof of credibility.” For example, it is more accurate to predict whether an individual will have an adverse reaction to a particular drug, rather than recommend cancer treatment.

“They said I don’t understand,” Kohn said. Current and former IBM insiders pointed out that, until recently, the company’s top executives were executives with a service and sales background, not technical product experts. They say that product people may better understand that Watson is customized for quiz shows, which is a powerful but limited technology.


In its 110-year history, IBM has introduced new technologies and sold them to companies time and time again.

Company executives are beginning to see artificial intelligence as the next wave. In 2006, Ferrucci first presented Watson’s ideas to the boss of the IBM Research Lab. He believes that creating a computer to play quiz games can promote the scientific development of artificial intelligence, namely natural language processing. In this field, scientists can program computers to recognize and analyze words. Another research goal is to advance automatic question answering technology.

After overcoming the initial suspicion, Ferrucci assembled a team of scientists to build a room-sized Watson with thousands of processors and running millions of lines of code. Its storage disk is full of digital reference books, Wikipedia entries, and e-books. Computational intelligence is a matter of brute force, and heavy machines require 85,000 watts of power. In contrast, the operating power of the human brain is equivalent to 20 watts.

When Watson won in the mind game, the market response was overwhelming. IBM customers scrambled to become customers, and executives saw huge business opportunities. However, Watson has a market, but IBM has nothing to sell.

In order to allow superstars to bring new business, IBM began to work hard to bring Watson into the healthcare industry. It is not only the largest industry in the United States but also rich in data, especially when hospitals and doctors are turning to electronic health records, and there are undeniable social benefits.

They are marketing first and products second, which makes everyone excited. Said Robert Wachter, head of the Department of Medicine at the University of California, San Francisco.

In order to find the business case for medical AI, IBM has carried out a dazzling number of projects, targeting all the different participants in the healthcare system: doctors, administrators, insurance companies, and patients. From the beginning, they anchored the project to cancer, one of the biggest challenges in medicine today, hoping to use Watson’s “cognitive” capabilities to transform big data into personalized cancer treatments for patients.

However, a series of failures quickly proved that this was an overly radical choice. Except at the company headquarters, powerful technology cannot match the complex reality of today’s medical systems.

In cooperation with one of the partners, the University of North Carolina School of Medicine, IBM technicians were frustrated by the complexity, confusion, and gaps in the cancer center’s genetic data.

“We thought it would be easy, but it turned out to be really difficult,” said Dr. Norman Sharpless, the former head of the school’s cancer center and current director of the National Cancer Institute. “We talked to each other for about a year.”

In the end, oncologists and technologists found a way that suits Watson’s strengths—quickly absorb and read thousands of medical research papers. By linking the genetic mutations mentioned in the paper to the patient’s genetic profile, Watson can sometimes point out other treatments that doctors might have missed. This is a potentially useful new diagnostic tool.

However, it turned out that it was still not useful or flexible enough to be a successful product. It’s impossible to teach Watson to read articles like a doctor—Watson’s idea is based on statistics. All it can do is collect statistics about the main results, but doctors don’t work like that. At the end of last year, IBM discontinued Watson for Genomics, a joint research project with the University of North Carolina.

IBM also shelved another cancer product, Watson for Oncology, which was developed with another early collaborator, Memorial Sloan Kettering Cancer Center, to examine a large number of cancer medical literature and real cancers with Watson’s powerful computing power. Hundreds of variables in patient health records—including demographics, tumor characteristics, treatments, and outcomes—and discover hidden patterns that are difficult for humans to discover.

Another cancer project called Oncology Expert Advisor was abandoned in 2016 due to a costly failure. This is a collaboration with the MD Anderson Cancer Center in Houston. The goal is to create a bedside diagnostic tool that can read patients’ electronic health records, a large amount of scientific literature related to cancer, and then make treatment recommendations.

There are many problems with the project. For example, during the collaboration, MD Anderson switched to a new electronic health record system, and Watson was unable to obtain patient data. Watson also had difficulty deciphering the doctor’s notes and medical history. Doctors are getting frustrated. They fight with technology instead of caring for patients. According to a public audit, after spending $62 million four years later, MD Anderson closed the project.

According to the media STAT, IBM encountered difficulties in finding buyers for Watson cancer products in the United States . Some oncologists said they believed in their own judgment and did not need Watson to tell them what to do. Others say it only recommends standard treatments that they know very well.

“They chose the highest possible standard, real-time cancer diagnosis, and immature technology,” Shane Greenstein is a professor at Harvard Business School and co-author of a recent case study of the MD Anderson Watson project. “This is a risk. High road.”

However, IBM continues to invest in the health industry, including announcing the establishment of the Watson Health division in 2015. By mid-2016, Watson Health had acquired four health data companies at a total cost of approximately US$4 billion. Most of this money seems to be difficult to recover forever.

According to media statistics, IBM has issued nearly 50 cooperation announcements aimed at developing new artificial intelligence tools. Some collaborate to develop tools for doctors and institutions; some work on consumer applications. But many collaborations have not yet produced commercial products.

Due to sluggish financial performance, IBM halted Watson’s sales in the field of drug discovery in 2019. Now, IBM is cutting Watson Health and examining the future of the business, including exploring the possibility of selling the business.


So far, the potential of AI in the medical field has been preliminarily proven in carefully controlled experiments. Only a few AI-based tools have been approved by regulatory agencies and can be used in real hospitals and doctor’s offices. These pioneering products are also mainly focused on the field of vision, using computer vision to analyze images, such as X-rays and retinal scans. (IBM does not have a product that analyzes medical images, although it has an active research project in this field.)

There are also the largest number of AI medical imaging startup companies in China. In addition to Inference, United Imaging Medical, and Yingtong, the “AI Four Dragons” including Shangtang Technology, Megvii Technology, Yuncong Technology and Yitu Technology have a large number of layouts in medical scenarios. The vast majority are concentrated in pulmonary nodules and ophthalmology.

However, with the exception of IBM Watson, even Google’s health department was exposed to crisis by the media and had to lay off and reorganize on a large scale. Google’s 2021 Q1 financial report shows that innovative businesses including DeepMind and Verily are still at large losses, and Google’s innovative business related to medical AI has not developed successfully.

Domestic medical AI companies also face similar problems. Even if it passes the approval process, it is still difficult to escape the fate of losses. Products that are only at the preliminary stage of improving the efficiency of doctors can hardly impress hospitals. The road to listing is also quite bumpy. Yitu Science and Technology Innovation Board IPO was suspended for review, and it was assumed that Technology and United Imaging Medical had announced plans to go public, but no further action was taken.

In addition to technical issues, the finer details of how technology works in practice also cannot be ignored.

DeepMind obtained the medical records of 1.6 million NHS patients through a transaction in 2015 and helped it develop an APP Streams. In December 2016, when working with the Imperial College Hospital in the UK, DeepMind obtained another 1 million patient medical data. However, the cooperation between the two has been questioned due to data privacy issues. In 2017, the UK’s privacy regulator ruled that NHS and DeepMind’s data sharing was illegal because patients did not really understand how their medical records would be used.

In 2019, Google was exposed by the media to cooperate with one of the largest medical systems in the United States on a secret project to collect and process detailed personal health information of millions of Americans in 21 states. Neither the patient nor the doctor was informed. The plan triggered an investigation by the Office of Civil Rights of the U.S. Department of Health and Human Services.

Infrastructure must also change, and healthcare organizations must agree to share their proprietary and privacy control data so that AI can learn from the millions of patients tracked over the years. However, how to protect the privacy of patients and prevent data from being leaked or used at will?


IBM described Watson as a learning journey for the company. “Innovation is always a process,” said Rob Thomas, an executive in charge of Watson’s business for the past few years. Earlier this month, he was appointed as the senior vice president responsible for global sales.

He believes that IBM’s artificial intelligence development is divided into three stages: ” Jeopardy! ” (Jeopardy!) , a large service contract that has been “experimented” for many years, now it has turned to the product business.

IBM insists that its revised artificial intelligence strategy-an ambition to streamline and change the world less-is working. The task of restoring growth was given to computer scientist Arvind Krishna. After leading IBM’s recent cloud computing and artificial intelligence business reforms, he became CEO last year. Some people even compared him with Microsoft CEO Nadella.

For a long time, many external researchers believed that Watson was mainly a branding campaign. But recently, some of them said that this technology has made significant progress.

In an analysis for The New York Times, the Allen Institute for Artificial Intelligence compared Watson’s performance on standard natural language tasks (such as recognizing the emotions of people, places, and sentences) with large technology cloud providers (Amazon, Microsoft). , Google) provided artificial intelligence services for comparison.

Watson’s performance is as good as the Big Three, and sometimes even better. “I was very surprised,” said Oren Etzioni, CEO of the Allen Institute. “IBM has taken action, of course in these capabilities.”

Watson’s business also showed vitality. Watson is now a set of software tools that companies use to build artificial intelligence-based applications-these applications simplify and automate basic tasks in areas such as accounting, payments, technical operations, marketing, and customer service. It is the workhorse artificial intelligence, and most artificial intelligence in today’s business is like this.

One of Watson’s core capabilities is natural language processing, which provides support for IBM Watson Assistant, which companies use to automate customer service queries. The company did not report Watson’s financial performance. But Rob Thomas, who now leads IBM’s global sales, pointed out signs of success.

He said that artificial intelligence is at an early stage in the enterprise market and the market opportunity will be huge. The key at this stage is to accelerate the adoption of Watson software products.

IBM says it has 40,000 Watson customers in 20 industries worldwide, more than double the number four years ago. Watson products and services are used 140 million times a month, up from about 10 million a month two years ago. Some large customers are in a healthy state, such as the large insurance company Anthem, which uses Watson Assistant to automate customer queries.

Five years ago, Watson chatted and joked with tennis superstar Serena Williams in an advertisement. Today, TV advertising has turned into this technology to help save time in offices and factory floors. As the advertisement says, Watson helps companies “automate the little things so they can focus on the next big thing.”

Compared to its initial ambitions, Watson is no longer the next big thing, but it may eventually become a solid business for IBM.

Reference link

Posted by:CoinYuppie,Reprinted with attribution to:https://coinyuppie.com/ten-years-of-ibm-watson-ai-medical-a-dream/
Coinyuppie is an open information publishing platform, all information provided is not related to the views and positions of coinyuppie, and does not constitute any investment and financial advice. Users are expected to carefully screen and prevent risks.