The health care industry suffers from an ailment known as information overload. It's a condition that makes it difficult for doctors and insurance providers to quickly determine the procedures that will be required and covered for each patient. The individual's history, years of case files and clinical evidence must be considered -- no small feat for a human. But it's as easy as saying "aah" for Watson -- IBM's artificial intelligence computer system that processes natural language questions against a deep well of data to compute evidence-based answers in a matter of seconds.
Watson can sift through the data equivalent of about 1 million books, analyze the information and provide precise responses to complicated questions in less than three seconds. Insurance provider WellPoint Inc. doesn't have to imagine what that could mean to the health care industry. Through a partnership with IBM that began two years ago, the Indianapolis-based company is working with more than 3,000 physician offices in its network to provide patient coverage and treatment options almost instantaneously.
We trained Watson to think like a nurse or physician on staff.
Elizabeth Bigham, vice president of health IT strategy at WellPoint Inc.
"We trained Watson to think like a nurse or physician on staff," said Elizabeth Bigham, WellPoint's vice president of health IT strategy. "It receives requests from providers, finds medical policies, compares what the provider said in the request, determines who the patient is, what they want to do and why, and renders a recommendation to our staff."
Watson, Bigham said, is a game changer in the medical field. Other companies are also developing Watson-powered applications for health care providers and consumers -- for example, software startup Welltok Inc.'s CafeWell Concierge, which will offer personalized, location-based guidance on diet, exercise and preventive services.
And it's not just health care. In January, IBM launched a Watson Group business unit to ignite new commercialization efforts in a range of industries, including financial services, travel, telecom and retail. To help fuel its efforts, the new unit was given $100 million to invest in third-party software developers; it made an initial investment in Welltok in February. IBM also is making Watson services available in the cloud and providing software developers with access to its Watson application programming interface to build new kinds of cognitive apps. Welltok's project is one of those efforts; another is a Watson-powered "smart adviser" self-service applications, which can understand natural language, read and interpret text and learn from other types of smart technology, such as virtual personal assistants. Also, Fluid Inc., which makes software designed to improve online shopping, is developing an app that makes product recommendations based on information provided through natural dialogue. So talking to a smart device, like an iPad, could deliver the same kind of experience a shopper would have with an in-store sales associate.
A bountiful mind
It is this cognitive capacity -- the ability to mimic the human brain, to learn and to understand in context and to be more assistant than tool -- that will revolutionize computing as we know it, IBM officials said.
That's according to IBM, of course. But Dan Miller, founder and senior analyst at Opus Research, doesn't disagree, calling the technology "transformative."
"IBM was out to demonstrate deep computing's ability to do things like understand recognize a topic quickly, discern irony and satire," he said. "It also helped uncover some of the constraints of the brute force approach. Scientists have long known that the answers get better and quicker if the systems address a specific topic or domain. And that's where IBM Watson is taking it."
In health care, for example, where the volume of data is doubling every five years, it could take a nurse 20 minutes to collect the data needed to make a treatment assessment. Cognitive technology, coupled with big data, is delivering that same evidence-based information in a matter of seconds.
"We know Watson makes us more efficient and is helping us turn around requests faster," Bigham said. "It also ensures we are consistent in our application of medical policies and guidelines."
IBM is also working with pharmaceutical companies to understand drug interactions. Using IBM's new Watson Discovery Advisor service, which makes connections across millions of articles, journals and studies, drug researchers can formulate conclusions that today can take months in just hours.
Of course, cognitive technology is not a new concept. Artificial intelligence emerged as a hot topic in the 1960s, when computer scientists set out to build systems that were as intelligent as humans. The 1980s saw a flowering of "expert systems" from companies like Symbolics and Lisp Machines as well as the highly publicized development of a massively parallel processing supercomputer aimed at AI applications by Thinking Machines -- but those efforts quickly ran out of steam. Other supercomputer makers -- even IBM -- also dangled the idea of intelligent machines in front of government agencies and research laboratories, but their focus was on solving grand scientific problems, such as modeling the global climate and mapping the human genome. Those things didn't require cognitive capabilities, just colossal computing power.
Today, processing power and storage are not the big issues they once were. When Watson was introduced in 2011, it ran on 90 servers and 20 terabytes of disk. But the current system is 90% smaller and 20 times faster, said Steve Gold, IBM's vice president of marketing and sales operations for Watson systems.
And the cognitive technology has evolved as well.
Watson is an amalgamation of artificial intelligence, machine learning and natural language technologies. But it does not follow a logic-based set of rules as the supercomputers of the past did. Instead, it decomposes questions from natural language to understand the context of what is being asked. Then it analyzes the corpus of available information in research and articles and comes up with candidate answers. It is not deterministic; it is probabilistic -- producing a set of best answers with ranking and supporting evidence. For example, a simple question about the color of the sky depends on the circumstances: It could be blue, gray or white. Watson tries to comprehend the question and then uses thousands of algorithms to score the answer.
Consumers, too, are ready for intelligent systems. Conditioned by the prevalent use of intelligent personal assistant application, like Siri on Apple's iPhone, people expect computers to recognize natural language and respond to complex questions.
"What IBM is doing at the high end, as well as companies like Google and Apple that are working natural language understanding into machine learning, we, as humans, are being conditioned to feel more comfortable talking to some form of artificial intelligence," Miller said.
Miller expects there will be a ripple effect that will eventually bring high-end Watson-powered apps to the masses. "But it doesn't take a Watson-like investment to get an interface that is conversational and human-like in nature," he said.
More on cognitive computing
Podcast: A crash course on cognitive computing
Cognitive computing may serve in governance, risk and compliance
IBM's $1b dollar investment in cognitive computing
Watson isn't simple or inexpensive. While Bigham wouldn't disclose WellPoint's financial arrangement with IBM, the process of training Watson for use by the insurer includes reviewing the wording on every medical policy with IBM engineers, who define keywords to help Watson draw relationships between data. The nursing staff together with IBM engineers must keep feeding cases to Watson until it gets it. Teaching Watson about nasal surgery, for example, means going through policies and inputting definitions specific to the nose and conditions that affect it. Test cases then need to be created with all of the variations of what could happen and fed to Watson.
And things change, so it is an ongoing process. Bigham said the company can now teach Watson new things over a period of several weeks. It is a significant time and money investment, but WellPoint is bearing the brunt of the work to develop an affordable commercial app that it can license to other health insurance companies.
This painstaking process of training Watson is most likely the reason cognitive technology is not catching on like wildfire, Bigham said. According to Miller, there are subtler things at work.
"Generally, solutions like Watson are put in the 'emerging technologies' category, and the processes involved with building a business plan to make an investment in Watson are just now being defined," Miller said. So though cutting-edge technologies like cognitive aren't typically associated with objectives like return on investment, they are starting to be.
Similarly, other companies are working on their own strategies with different forms of cognitive technology. Enterra Solutions, based in Newtown, Pa., has developed a cognitive reasoning platform that combines big data and artificial intelligence to find insights that can improve performance in areas like the supply chain and consumer marketing. Google, Facebook and Yahoo have all recently hired AI researchers or acquired startup vendors to lead machine learning development efforts. And NuLogix Labs, recently relocated from Princeton, N.J., to San Jose, Calif., is using a complex event processing technology, called the Cognitive Information Management Shell, to develop an agro-intelligence platform designed to help increase food supplies -- starting in India.
The CIM Shell, developed at Louisiana State University, can drill down into complex events and activities to adapt rapidly to evolving situations. The work being done with a university in India is focused on increasing specific crop productivity by using sensors to collect data on ground activity and a synthesis program to dynamically reconfigure in real time to adjust to environmental changes. The agriculture app provides actionable intelligence to scientists who use it to direct experiments and inform decision making.
The technology can be put to use in other industries as well. Oil and gas company BP gave LSU a $250,000 grant to develop a prototype of the technology that could be used to help prevent future oil spills.
The CIM Shell's distributed intelligent agents fuse disparate streaming data, like text and video, to create an interactive sensing, inspection and visualization system that provides real-time monitoring and analysis. If there are any changes in data patterns -- temperature or pressure of equipment, for example -- it sends an alert, noted Dr. S.S. Iyengar, a computer science professor at Florida International University and chief scientist at NuLogix who co-invented the technology. It's not the first technology that detects changes in conditions -- abnormal situation management applications can, but they only flag things out of the ordinary. The CIM Shell not only sends an alert but reconfigures on the fly in order to isolate a critical event and fix the failure.
"The goal of CIM is that nobody should have to write the program," said co-inventor Supratik Mukhopadhyay, an assistant professor in the department of computer science at LSU. "You tell the computer what it needs to do and it writes a program itself that will solve the problem in real time."
The human factor
The CIM Shell and Watson take different approaches to understanding complex events, but they both are built to respond, learn and continue processing, just like the human brain.
What these cognitive systems can't do is analyze the risk that might not be represented anywhere in the unstructured data. That includes factoring in cultures, environments, people and accountability.
"You have to be able to analyze risk," said Jose Bravo, a chief scientist at oil company Shell Global. Shell is looking at big data systems and is considering a variety of artificial intelligence products, but, according to Bravo, there are still limitations to what a deep learning machine can do. For example, if a predictive model says buy oil in the Middle East, but a leader in the region is at risk of being deposed by a revolution, it must be factored into the decision. "If you could predict how the future will develop, that would be great, but you can't," Bravo said. "And you can't hold a machine accountable if it makes a disastrous decision."
But that's why there will always be a human element in the cognitive machine mix. At WellPoint, the staff ultimately chooses whether to accept Watson's recommendations. The value of Watson is the speed, efficiency and consistency for responding to a doctor's request and for complying with medical policies and guidelines, Bigham said. For the doctor's office, users are typing a natural language question into a browser on demand. There is no calling and waiting to submit a request. And every day Watson gets smarter, drawing connections between concepts based on things it's already learned.
As time goes on, Opus Research's Miller predicted, cognitive technology will evolve for the masses, getting less expensive and easier to use. So Watson and other artificial intelligence systems won't fade away like the cognitive fads of the past, Bigham said. "In my opinion, this is one of the next big things."
STEPHANIE NEIL is a freelance writer and a correspondent for Business Information. Email her at email@example.com.