Thank You Dr Watson

BACK IN 2011, a computer system did something remarkably human. It beat two of the all-time best contestants at Jeopardy, one of the world’s trickiest quiz shows, where the quizmaster doesn’t even ask questions. In Jeopardy, he provides the answers, or clues, and the contestant’s task is to find the appropriate question.

In order to defeat the two flesh-and-blood champions, the machine—an IBM computer system called Watson— had to be able to understand natural language, considered one of the most difficult things to crack for a computer. It had to be able to deal with the vagaries of human speech: cultural references, subtle meanings, idioms, riddles and turns of phrase. All of this without being connected to the internet. And then it would have to work out the various clues, narrow down possible answers, and weighing each one of them, just like humans, arrive at the most likely answer. Not necessarily the absolutely certain answer, but the most probable one. All of this within seconds.

In the past, another IBM computer system, Deep Blue, had defeated the world chess champion Garry Kasparov. But with Jeopardy, Watson was playing an infinitely tougher game. Chess has complicated rules, but it requires no understanding of language, nor deep knowledge of a wide range of topics. An artificial intelligence system winning such a game was thus considered almost impossible.

The two-day game was billed as a man-versus- machine encounter. There were a range of questions, from Beatles’ songs and Michelangelo’s paintings, to Olympics history and Harry Potter. In those two days, answering in a cold computer voice, somewhat reminiscent to the sentient computer systems in films— Hal in 2001 Space Odyssey or Jarvis in the Avenger series—Watson simply drubbed its two competitors. Ken Jennings had the record of the most number of consecutive wins (74). Brad Rutter had won the most amount of money on the show ($3million). Watson got some queries wrong, but won convincingly.

When telecast, the show created a flutter online. It seemed to suggest a time when computers might be able to understand complex information and converse with people about it. During a talk later, Jennings said, “I felt obsolete… I felt like a Detroit factory worker in the 80s seeing a robot that could now do his job on the assembly line. I felt like ‘Quiz Show Contestant’ was now the first job that had become obsolete under this new regime of thinking computers.” Nobody really knew what would happen to this new computer system in the future—how it would be designed, or to what practical utility it would be put. But it seemed to suggest a promising future.

Since its exploits at Jeopardy, IBM, it turns out, has been quietly working to commercialise Watson for various businesses. And now, more than five years after it first appeared on TV, six Indian hospitals under the Manipal Hospitals’ chain have become the first few in the world to adopt Watson to solve an infinitely more challenging problem than a quiz show. These hospitals have begun using it to treat cancer.

Dr Somashekar SP, chairman of the surgical and robotic oncology department of Manipal Hospitals, remembers following the news of the development of Watson with fascination. “And then one day I began to hear that it was being developed for cancer treatment. And soon that we might be using it,” he says. “And I was like, ‘What, really?’ I was quite sceptical.”

About six months ago, after several discussions between Manipal and IBM, the chain’s six hospitals began to use the system to examine old cases. The Memorial Sloan Kettering Cancer Center in the US, considered to be one of the best in the world, has been training Watson, somewhat like a medical student, to recommend cancer treatment options. So far, Watson for Oncology has ingested about 15 million pages of medical content, including more than 200 medical textbooks and 300 medical journals on cancer. All a doctor has to do is to refer his patient’s case details, like doctor notes and lab test results, and the artificial intelligence system will go through millions of medical files and published case studies and offer an opinion on what treatment to pursue within seconds. Every recommendation comes with evidence backing it. Doctors can simply click on the recommended treatment and see which medical studies support it. Currently Watson offers recommendations on five types of cancer (breast, lung, colorectal, gastric and cervical). In the future, IBM claims, it will begin to offer advice on other forms of cancer too.

When the doctors at Manipal Hospitals began to try out Watson by checking its old cancer cases, the results, Dr Somashekar says, blew his mind. “I was unsure at first [about Watson’s efficacy]. But we found that Watson was remarkable. It agreed with almost all the steps we had taken. And in a few cases, it even offered alternatives we hadn’t quite considered. It has all this data in its possession. And within 60 seconds, it goes through all these millions of pages and gives me options,” Dr Somashekar says. “It is really quite something.”

In the last few years, IBM has increasingly come to bet on Watson to become the engine of its resurgence. It believes that the world is awash with unstructured data and the next wave of growth for the tech industry will rely heavily on analysing this ‘big data’. According to it, there have been various eras of computing. The tabulation era, in the early part of the 20th century, brought forth tools like the calculator which could perform basic arithmetic. We live now in the programming era, where computer scientists programme rules for machines to follow. The next era, which we are at the threshold of, according to IBM, will be what it calls the cognitive era, where we will begin to use systems that learn and keep getting smarter as they process more knowledge. Systems like Watson, IBM believes, will help propel consumers into that era. The tech company created its Watson business unit in 2014, and has been pumping billions of dollars into it. “The cognitive era is the next important computing paradigm,” Dr Prashant Pradhan, Watson Leader of IBM India/South Asia, says. “There is a lot of data, most of it unstructured, in various industries. If you look at retail, for instance, data on people’s preferences, goods that are trending, a lot of it exists in social media or market trend reports. But modern computers cannot understand this mass of unstructured data. We saw the need of a new class of computing… [which] can understand information, reason with it, learn from it, and then interact with us in a natural language.”

AFTER THE WIN at Jeopardy, IBM began to look at various industries where it could commercialise Watson. One of the first to be selected was healthcare, and more specifically, cancer. According to IBM executives, treatment protocols for most diseases is somewhat standardised. But when it comes to cancer—where there is an array of treatment options and a staggering amount of ongoing research every day—Watson, with its unique skills, can prove revolutionary. “In cardiology for example, if there is a blockage, doctors unblock the artery. They do an angioplasty. They prescribe statins. It is somewhat standardised. But it’s not like that with cancer. There is really no standardised treatment protocol,” Navinder Sachar, the sales leader of IBM India/South Asia’s Watson Health division says. “Added to that, every year some 700,000 papers on cancer are published globally in journals. Some 44,000 papers on cancer are published here [in India] annually. Reading all of it is impossible. It has long gone beyond human cognition.” IBM points out that the volume of all healthcare data has reached 150 exabytes. At projected growth rates, the volume of healthcare data will soon be zettabyte and yottabyte scale. Enough data, the company says, ‘to fill a stack of DVDs that would stretch from Earth to Mars’.

When IBM began to look at deploying Watson for oncology, it was particularly interested in working in India. It has been in use in the US and Thailand as well. It has also announced tie-ups with hospitals in China and South Korea, and will soon announce another tie up in Europe. “The onslaught of cancer in India is much bigger compared to other countries. According to official figures, cancer kills about 700,000 people here every year. It is expected to rise five-fold by 2020. The ratio of oncologists to patients in the US is about 1:100. In India it is something like 1: 1,600,” Sachar says. “The situation here, for doctors and healthcare, is really overwhelming.”

About three months ago, a 62-year-old man who had undergone a surgery for Stage 2 colon cancer began consulting Dr Amit Rauthan, an oncologist at Manipal Hospitals in Bengaluru. After having examined Watson’s efficacy by relooking old cases, the hospital had begun to look up the new system while treating new patients. Until then, Watson, they noticed, recommended treatment options the doctors were already considering. “We have always found each other on the same wavelength,” Dr Somashekar says.

The 62-year-old patient had undergone surgery at another centre. “The centre had advised him adjuvant therapy (to suppress secondary tumour formation) and chemotherapy,” Dr Rauthan remembers. The doctors at Manipal, Dr Rauthan recalls, were also leaning towards a few sessions of chemotherapy. He then thought of checking the case with the new system. “I thought I will just put this into Watson, and let’s see [what it says],” he says. Watson recommended something entirely different. “It was saying there was no need for chemotherapy. That there was not much benefit… It was one of those cases, I still remember it clearly, which stood out uniquely where we changed our approach,” he says.

In hindsight, Dr Rauthan reasons that the doctors at Manipal were playing it safe by initially recommending chemotherapy. “Sometimes in cancer care, there are differences of opinion. Sometimes we play safe. And Watson was saying we were probably playing safe when we don’t need to. MSK (Memorial Sloan Kettering) wouldn’t have recommended chemotherapy to a patient with the same condition.” Following Watson’s advice, the doctors have now kept the patient under observation. He will have to undergo a scan every six months to see if there is a recurrence of the tumour.

In a CBS News interview, Dr Ned Sharpless, head of the University of North Carolina’s Lineberger Comprehensive Cancer Center, revealed how in an analysis of more than 1,000 cancer patients by researchers from the university, Watson recommended the same treatments that doctors had opted for 99 per cent of the time. But Watson did better than the doctors in other ways, he said. “The… more exciting part about [the analysis] is in 30 per cent of patients, Watson found something new—so that’s 300-plus people where Watson identified a treatment that a well- meaning, hard-working group of physicians hadn’t found,” he was quoted as saying. The treatments identified by Watson which human doctors had missed were apparently those in clinical trial or based on research papers that the doctors had not read. “These were real…things…we would have considered actionable had we known about it at the time of the diagnosis.”

IBM claims it is currently in discussions with several Indian hospitals to introduce Watson into their systems. According to Sachar, they are also pursuing the Government to see if Watson can be used in smaller towns where there is a shortage of doctors. Manipal Hospitals has already taken the technology a step further. It currently allows patients, for a fee of Rs 10,000, to seek a Watson consultation online. Around 120 people so far have uploaded their medical record, and received an online report of Watson’s analysis of their case. This report comes verified by Manipal’s doctors.

Most doctors and IBM stress that the technology will never replace medical oncologists. It will act more like an assisting tool, they say. Dr Somashekar, however, is more generous in his assessment. The doctors from Manipal normally sit together in their ‘tumour board’ to discuss certain cases and take decisions on them. “To me, Watson is like another doctor on the tumour board,” Dr Somashekar says. “A doctor who provides all the evidence, and who doesn’t allow any errors or human biases from creeping in.”

artificial intelligencecancer treatmenthealthmedicinestechnology