Monetizing Cognitive Computing

Stephen DeAngelis

January 9, 2014

IBM has great hopes that its Watson computer system will someday become a huge profit center. According to Spencer E. Ante, last fall IBM’s Chief Executive Virginia “Ginni” Rometty predicted that Watson would generate “$10 billion in annual revenue within 10 years.” [“IBM Struggles to Turn Watson Computer Into Big Business,” Wall Street Journal, 7 January 2014] As Ante’s headline states, monetizing Watson’s cognitive abilities is off to a slower start than IBM had hoped. As of October of last year, Watson had generated about $100 million of revenue. That’s still a lot of money, even if it falls short of IBM’s goal. One of the problems, according to Ante, is that Watson is still struggling to prove its value in a business setting. “One of its first big projects, with the University of Texas M.D. Anderson Cancer Center,” he reports, “was ‘in a ditch’ in early 2013, [according to] Manoj Saxena, the executive overseeing Watson.” Nevertheless, Ante reports that IBM executives might be disappointed with current results but they are not depressed. He writes:

“IBM executives still believe Watson could become one of the biggest innovations in the company’s 103-year history, alongside the mainframe and personal computer. … ‘Watson has rapidly moved from an industry-first research initiative to a commercial reality’ that tackles business and social problems, an IBM spokesman said. ‘IBM is making excellent progress with clients and with partners in advancing Watson, and we are excited about Watson’s future as a cloud service and as technology that will change lives.’ Watson also helps IBM sell other technology as customers prepare to use the supercomputer. … Watson’s key distinction from other analytical software is its ability to ‘learn.’ Feed it medical cases, and Watson will rank possible treatments by ‘confidence score.’ During training, doctors tell Watson when it makes a bad recommendation, and the supercomputer learns from its mistakes.”

Watson was one of the first systems to be defined as a cognitive computing system. Characteristics of cognitive computer systems include that they are data-centric (i.e., they use Big Data) and they are designed for statistical analytics. Forrester analyst John Brand writes, “The term ‘cognitive computing’ emerged in response to the failings of what was once termed ‘artificial intelligence’.” [“Make No Mistake – IBM’s Watson (and Others) Provide the *Illusion* of Cognitive Computing,” John Brand’s Blog, 21 May 2013] As you can tell from Brand’s headline, he is skeptical about calling Watson’s brute force approach to learning a cognitive process. He concludes, “Let’s get real. Despite the fact that ‘Watson’ was trained to successfully win a game show (Jeopardy), IBM’s technology (and others to be fair) are not cognitive computing systems at all. That’s not to say they aren’t valuable – just that we shouldn’t overstate their value or capabilities.” Roger Kay agrees, “Watson, the reigning jeopardy champ, is smart, but it’s still recognizably a computer.” He believes that cognitive computing represents “something completely different.” [“Cognitive Computing: When Computers Become Brains,” Forbes, 9 December 2011] Brand’s and Kay’s remarks beg the question: What is cognitive computing?

 

Mark Smith, CEO & Executive Vice President of Research at Ventana Research, writes, “At the simplest operational level [cognitive computing] is technology for asking natural language-based questions, getting answers and support appropriate action to be taken or provide information to make more informed decisions.” [“IBM Watson Advances a New Category of Cognitive Computing,” Perspectives by Mark Smith, 11 December 2012] That definition is in line with the definition of intelligence used by Marcus Hutter, a Professor of Computer Science at Australian National University, and his team. They define intelligence as “an agent’s ability to achieve goals or succeed in a wide range of environments.” [“To create a super-intelligent machine, start with an equation,” The Conversation, 28 November 2013] Systems like Watson and Enterra Solutions® Cognitive Reasoning Platform™ (CRP) can certainly be used in a wide range of environments to achieve goals and help people make more informed decisions. That’s why I have no difficulty labeling them as cognitive computer systems. One difference between Watson and the CRP is that the latter analyzes both structured and unstructured data using ontologies as well as mathematical algorithms. The CRP is capable of addressing various commercial markets and disciplines using a generalized framework, yet it is designed so that it can be tailored to handle the disparate data sources and specific challenges found in individual industries and in different functional areas.

 

Because cognitive computing systems are capable of addressing challenges in almost every area of human activity, I share Ms. Rometty’s optimism about the future of cognitive computing as source of revenue. To demonstrate how versatile cognitive systems can be, Professor Hutter reports that his team has “successfully developed software that can learn to play Pac-Man from scratch.” Hutter believes that one drawback with systems like Watson is that they must be tailored for each new activity. Hutter explains:

“Even though IBM Deep Blue plays better chess than human Grand Masters, it was specifically designed to do so and cannot play Jeopardy. Conversely, IBM Watson beats humans in Jeopardy but cannot play chess – not even TicTacToe or Pac-Man.”

Ante agrees with that assessment, he notes, “Watson is having more trouble solving real-life problems than ‘Jeopardy’ questions, according to a review of internal IBM documents and interviews with Watson’s first customers. For example, Watson’s basic learning process requires IBM engineers to master the technicalities of a customer’s business — and translate those requirements into usable software. The process has been arduous.” Because each engagement is a kind of one off activity, Ante reports that IBM “hasn’t figured out how to generate a reliable revenue stream from Watson.” He continues:

“So far, just a handful of customers are using Watson in their daily business. With the supercomputer’s help, health insurer WellPoint Inc. determines if doctors’ requested treatments meet company guidelines and a patient’s insurance policy. Elizabeth Bigham, a WellPoint vice president, said Watson initially took too long to ‘learn’ WellPoint’s policies. Ms. Rometty, IBM’s chief executive, met with WellPoint CEO Joseph Swedish to help resolve the problems. IBM reworked Watson’s training regimen at WellPoint’s request, and the system improved. … Two years ago, doctors at M.D. Anderson began working with IBM to build a version of Watson that would recommend leukemia treatments by mining medical literature. After the initial stumbles, Anderson and IBM officials said the project is back on track. Lynda Chin, M.D. Anderson’s chairwoman of genomic medicine, says the leukemia-treatment adviser could be used later this year. It might be two more years before Watson could handle other types of cancer.”

IBM shouldn’t be too worried. The field of cognitive computing, especially in a business setting, remains relatively new. Over the next few years, I believe that both the technologies used and the people who are developing them will make great strides in this field. I don’t know exactly how much revenue cognitive computing will generate; but, I’m betting that Ms. Rometty is closer to the mark than people are currently giving her credit.