IBM is giving Watson a major promotion. The prized computing system, which once beat two "Jeopardy" champions on the game show, is fast becoming the strategic focal point for the company’s growth in emerging, revenue-rich businesses.
For the past year, Watson has been quietly sorting through the genetic profiles of cancer patients. The mighty processor has learned a lot along the way, according to International Business Machines Corp. and its partners at the New York Genome Center. Now, IBM is planning to open access to the specialized interface and software built for Watson’s genomic studies to researchers at about a dozen institutions, in hopes that they may see it as a valuable scientific collaborator that will be worth a hefty subscription fee.
The opportunity is one of the ways IBM is attempting to leverage Watson for software products, consulting services and specialized analytics. The company has staked much of its future on the potential for Watson and related technologies to be spun off into customized products and services licensed by businesses, universities and research centers.
IBM desperately needs Watson to deliver. The once-dominant computer company has experienced 11 consecutive quarters of declining revenue and continues to struggle even after recently shedding some of its least profitable ventures, including its server business and semiconductor manufacturing unit. The hardware components that were once the bread and butter of sales now constitute less than 10 percent of its business.
At its annual investors’ conference in February, IBM said it expects cloud, data analytics, security and mobile technologies to make up around 44 percent of its sales by 2018, up from 27 percent last year. Executives also stated that the company intends to invest an additional $4 billion into these areas over the next three years.
Still, the company's transformation will take some time. "We believe that data analytics is critical to IBM's long-term growth story, and a big piece of that is Watson,” Bill Kreher, a technology analyst at Edward Jones Investments, says. "IBM is committing more money and resources toward solutions like the genome project than anyone else. I think patient investors will be rewarded as IBM continues to progress down the cognitive computing front."
IBM's revenue projections are partly based on the potential of cognitive computing, one of Watson’s specialties. Cognitive computing uses artificial intelligence to predict patterns and make inferences on queries rather than simply spit out all the possible results for a search. Systems built on it can incorporate feedback in a manner which has been said to mimic human learning. Cognitive computing offers a particularly elegant approach to sorting through unstructured data by recognizing hierarchies and identifying relationships within vast volumes of information. The International Data Corporation estimates that 90 percent of all digital data is unstructured, and leads to lost revenue, higher costs and slower customer response times for companies.
The collaboration between IBM and the genomic research center in Manhattan, which the partners announced last year, is one of Watson’s most public relationships. So far, researchers have run the genomes of a few hundred patients with various cancers through Watson to see what it could make of the 3 billion DNA base pairs and 20,000 genes contained within each.
To researchers, Watson's value comes in synthesizing an immense variety of information that is far beyond the scope of their own abilities to fully grasp. Sequencing a human genome has become faster and cheaper in recent years, but scientists remain overwhelmed by the amount of data that each genome generates and the challenge of matching the intricacies of any single genome to advances in current medical knowledge, which is doubling every five years.
"A genomics expert obviously can make sense of all the literature, but do they have the time to read the 6,000 or more papers related to cancer that get published every year?” asks Dr. Ajay Royyuru, director of computational biology at IBM Research. “That's the comprehensive cognitive aid that we think Watson genomics can be.”
After a year of testing, IBM has told International Business Times that Watson is now ready to serve a beta group of institutions that will each run 100 or so more queries through the system over three months. The goal is for Watson, which is the size of three stacked pizza boxes and named for IBM’s founder Thomas Watson, to apply the information it has gleaned from reams of scientific journals and clinical trials to pick out problem points along the genomes of cancer patients. If it can do that, the collaborators think it might point them toward treatments or even a cure.
Their challenge, though, is to leverage Watson’s intellect and processing power in a way that truly leads to scientific advances or clinical applications. “I see them having a little bit of difficulty scaling this to other genome centers,” says Michael Shanler, a life sciences analyst at Gartner Industry Research. “Routinely using Watson to solve complex genomics problems will be a hell of a lot more difficult than winning at 'Jeopardy.'”
Shanler points out that the scientific literature on which Watson relies is often riddled with mistakes, and says the system will likely run up against some of the same limits that researchers do. Furthermore, Watson’s ability to track the details of one type of cancer may not smoothly translate to others, and building a customized platform for each vein of cancer research could be costly.
"It's not a question of whether or not it can work -- I think most people believe that this can work with enough data and enough attempts,” Shanler says. “It's just going to require more effort than most end users and IBM realizes in order to scale the technology." He adds that IBM is probably better positioned to meet this challenge than the many smaller companies that aspire to build similar systems.
Eventually, IBM envisions Watson as a diagnostic aid that could be licensed by hospitals around the country. Clinicians could send a patient’s genome, a biopsy and other records to a technician who would run it through Watson to recommend personalized treatments from a suite of drug and gene therapies. A clinician would ultimately decide which treatments to administer, but Watson would bring the scope of options best suited to a particular patient to the doctor’s attention. “We’re building a reasoning engine that has the ability to drill down and say why treatment A might be better than treatment B for that one patient,” Royyuru says.
Earlier this week, IBM acquired AlchemyAPI, which specializes in sorting unstructured data, in a bid to surround Watson with even more computing power and the expertise of 40,000 new developers, according to a statement. The company has also extended access to Watson as an open development platform and says developers at large have already built 7,000 apps that play off of Watson’s skills. Seven leading universities recently announced that they will begin cognitive computing courses this fall in which students will attempt to figure out how best to use Watson to answer pressing questions in the retail, health care and travel industries.