Data and pricing to determine IBM Watson’s success in India
Services of the IBM Watson cognitive system don’t come cheap; the model fails if historical data is not well-organized
Latest News »
- BJP, BJD compete to mine Odisha’s history for electoral gain
- Google reaffirms commitment to digitally empower small and medium businesses
- Market round-up: Allocation to India by Asia ex-Japan funds down
- CBDT clarification on Ind-AS makes tax calculation harder for companies
- Greece arrests Russian suspected of running $4 billion bitcoin laundering ring
International Business Machines Corp.’s (IBM) Watson, which the company describes as a “cognitive” system that uses artificial intelligence (AI) technologies, is best known for analysing mountains of data generated by institutions and companies, mostly in healthcare and education. It does so with its multiple application programming interfaces (APIs), which enable software programs to talk to each other.
Globally, for instance, IBM partnered with New York-based Memorial Sloan Kettering Cancer Center (MSKCC), billed as the world’s oldest and largest private cancer centre, to have Watson trained by its clinicians and analysts using the rich cancer patient data it has accumulated over the years (bit.ly/1P9wXmI). Closer home, one of the first institutions to implement Watson was Manipal Health Enterprises, which announced plans to adopt Watson for Oncology—a flavour of the platform optimized for treating cancer (ibm.co/1pIFZ22)—two years ago.
More From Livemint »
Earlier this year, though, Pune-based Maharashtra Institute of Technology used Watson—the system that even beat Jeopardy players in 2011—for a very different task. It used Watson and Bluemix APIs to train its humanoid, named Chintu, to serve as a senior citizen companion. Bluemix is an implementation of IBM’s open cloud architecture based on Cloud Foundry—an open source platform as a service.
Chintu is being trained as a part of IBM’s global Shared University Research grant programme. “Our aim is do something that has a social impact,” said Vrushali Kulkarni, professor and head of the computer engineering department who is leading this project at the institute.
The hardware platform for Chintu, or the physical humanoid, is based on NAO—the 2 feet, 5kg humanoid robot procured from French robotics firm Alderban Robotics, which is a unit of SoftBank Group Corp. “It (Chintu) can read newspapers for senior citizens and do things like set alarms and reminders. For instance, if an elderly person needs to take some medicine at a certain time, a reminder can be set and Chintu will remind them accordingly,” Kulkarni said.
The entire interaction with Chintu, she added, can be conversation-based and the answers to a question are not necessarily pre-fed but “based on the context of the question”. For this, the institute uses the speech-to-text API under Watson. “It (Chintu) is still in a prototype stage,” she said, and the organization will continue to test it for another couple of years. Asked about its commercial potential, Kulkarni said, “Commercial product development is not on our mind. It is academic research.”
The students in her department who work on Chintu typically use the programming language Python. Citing the lack of technical skills in different programming languages proposed for use by IBM, she said, “We need to understand what Watson is, the architecture, and then work around it.”
Subscribe to Our Newsletter »
Spreading watson’s wings
Chintu is a case in point that Watson is keen on penetrating sectors other than healthcare and education. In India, for instance, IBM is working with a bunch of start-ups in India and helping them implement their AI plans.
Consider the case of Bengaluru-based Talview, a five-year-old human resources technology start-up, which uses IBM’s mentorship and technologies to automate the hiring process using video assessments and algorithms.
Similarly, Shivaam Sharma runs a Bengaluru-based start-up, Trans Neuron Technologies Pvt. Ltd, which he describes as a “skill-tech and ed-tech company that has AI and ML at the heart of its cloud platforms”.
The company has developed an analytics engine and an analytics platform to harness the power of Big Data and in-house data to help firms and institutions (bit.ly/2puPAcs).
Chennai-based Textient Analytics, uses Watson to develop its own “cognitive analytics platform that provides strategic brand and consumer insights in the form of unstructured and text data,” according to Nagarajan Sankar, co-founder of the company. Textient specializes in understanding “consumer psyche” by collating and analysing unstructured data from the web, social media and online video and deliver insight reports tailored to the needs of its clients.
Using these reports, brands can get a better handle on attributes such as “brand appeal, attractiveness quotient or brand stimuli”.
“We have to understand data and human behaviour not only in quantitative but qualitative terms as well,” said Sankar. Textient uses concepts underlying the Customer-Based Brand Equity (CBBE) model proposed by Kevin Lane Keller, a marketing professor at the Tuck School of Business at Dartmouth College.
Explaining how this model is combined with Watson, he said, “Keller suggested developing branding and marketing campaigns around how customers think and feel about your product. We take the industry conversation around a brand—for example, everyone wants to talk about Flipkart. We formulate our brand communication based on what they are thinking, searching, and buying, why they are interested in certain products and so forth. Then these insights are used for developing brand strategies. This model takes into account about 20 attributes. We take one at a time and then map it to the brand-based model developed through machine learning on top of Watson.”
Textient, then, generates a report based on these findings and other industry developments.
But watson is not cheap
It was only three years back that IBM established a separate unit for Watson and committed $1 billion to it. But companies that implement Watson have also to justify the return on investment.
J.P. Dwivedi, chief information officer (CIO) at the Delhi-based Rajiv Gandhi Cancer Institute and Research Centre, points out that Watson does not come cheap. “We were in talks with IBM for deploying Watson but found out that the cost of Watson consultation was too expensive for most of our patients,” he said.
Nevertheless, the organization decided to use Watson APIs available on IBM’s Bluemix.
Others like Ajay Bakshi, managing director and chief executive officer (CEO) of Manipal Health Enterprises—which runs a network of hospitals across India—do not appear to be seeking immediate returns. According to Bakshi, the “underlying philosophy” for the organization has been to “take a patient in Manipal meet Watson for free”. He added, “We are not charging a rupee extra for these Watson services, although it is a multi-million dollar contract signed with IBM.”
Bakshi points out that Watson is “not a diagnostic tool”. Rather, it helps both doctors and patients consider various treatment options and come to the right decision.
“Whether you are in Goa or Vijayawada, we have created a system with the help of IBM Health wherein details of the patient will be entered into Watson and Watson will then generate a 70-80 page report which outlines the best treatment options for the patient based on their cancer and history,” he said. The outcome could be, for instance, three options for treating a particular cancer in a patient (Watson covers around 60 different types of cancer).
Manipal, however, does charge a nominal fee for generating the Watson report on its website for patients in remote regions who want to benefit from Watson’s capabilities.
“There it is not free because we have spent a lot on technology to put this system in place. What you get is the entire Watson report and video consultation with one of our oncologists. So, this (facility) is available to all the patients in India regardless of whether or not they are Manipal patients,” he reasoned.
Besides pricing issues and lack of adequate technical skills, platforms such as Watson also face challenges in terms of market confusion and availability of “rich historical data”, according to industry analysts.
“There is a lot of confusion in the eyes of the clients, because all these words—AI, cognitive, machine learning and others—are often used very loosely or interchangeably without the context. Cognitive and AI are the umbrella terms in our view. In real life, you apply a specific technology for a specific purpose,” said Milan Sheth, partner and national leader of robotics and cognitive practice at consulting firm EY India.
Talking about the adoption and usage challenges for Watson, he said, “The most important part in Watson—and we have worked on Watson with a few of our clients—is that if you don’t have historical data or the historical data is not well-organized, the entire model will not work. It just fails.”
The problem, he added, is that historical data is not “just the way you are capturing it, but the way you wanted it to perform”. According to him, data has to be organized in a manner that the desired result will show. “So data supply chain, which is the term I typically use, is a big challenge,” he concluded.
Watson also has competition, primarily from Google’s DeepMind AI unit (whose AlphaGo system beat the world’s top player of Go, a complex board game, in May), Microsoft Corp.’s AI and Research Group, Amazon AI Services, Facebook AI Research (FAIR) and OpenAI, a non-profit lab partly funded by Elon Musk of Tesla.
There are many other competitors, many of which operate in certain areas within AI, including IPSoft’s Amelia, Wipro’s Holmes, TCS’s Ignio and Infosys’s Mana.
Despite the presence of many companies and continuing investments in the space, AI offers a huge opportunity since it is still a nascent field with 2016 global revenues estimated at just $644 million. According to data from Tractica and Statista, worldwide AI revenue will reach $37.8 billion by 2025 (bit.ly/2tIvCP8).
IBM, on its part, is aware of the issues around Watson and has been trying to acquire data in multiple ways—such as partnering with the MSKCC or acquiring companies, a case in point being its purchase of certain parts of The Weather Co. (see interview). But, as a recent article in MIT Technology Review pointed out, not all such attempts are successful: “A heavily promoted collaboration with the M.D. Anderson Cancer Center in Houston fell apart this year,” it noted (bit.ly/2shCAs3).
The wider success of Watson, and other AI platforms, may well depend on how the different contenders in this space are able to meet these challenges.