The concept of computers thinking like humans may be long-standing fodder for the pens of science fiction writers, but the era of cognitive computing is upon us — and it’s already changing the way we work and interact with technology.
Cognitive computing doesn’t look like a “travelling cyborg Arnold Schwarzenegger”, neither is it simply an extension of our current technology. Dale Potter, partner of Healthcare Transformation with IBM Watson Group, says that to truly understand what cognitive computing is, and how it will change the way we work and approach public policy, you need to understand where it comes from.
The history of cognitive computing
Back in the late 1990s, IBM set itself a challenge: could it teach a computer to apply strategy to solve a problem? It tested this hypothesis by challenging world champion chess grandmaster Gary Kasparov to a game — and won.[pullquote] “What we had built was a computer that was exclusively designed to play chess …” [/pullquote]
“The story of cognitive computing started in that space in that timeframe,” said Potter. “What we had built was a computer that was exclusively designed to play chess … it did very, very well.”
The next challenge was to create a computer that could emulate how a person thinks. So IBM set a new challenge: pitting its cognitive computer Watson against two grandmasters of the long-running American game show Jeopardy.
“The great conundrum of Jeopardy is that it is counterintuitive — they present the answer and you need to derive the question. Even for humans that is very difficult,” said Potter.
In an appearance that made worldwide headlines, and became a viral hit, Watson defeated two of the greatest-ever players of the game.
“We proved the hypothesis of ‘can a computer emulate the way humans think’,” said Potter. “That’s really the birth of processing cognitive computing. That’s really what cognitive computing is, having a computer that is able to think the same way as a human.”
Teaching a computer to think like a human
The first step towards a computer thinking like a human is to teach it to read — not just simply being able to recognise words, but to actually read literature and understand what it was reading.
“We sometimes take for granted the process of learning human language,” said Potter. “But if you think of things like feet can smell, and noses run, or you park in a driveway and drive on a parkway, there are plenty of intricacies in any language we take for granted.”
The next step is teaching the computer certain foundation principles, such as the passage of time, to reflect how a person also thinks. These “ground truths” are what allow the platform to be able to reason and apply logic to scenarios.
The third piece is teaching it a field of expertise. This is done through a process of annotation. Some medicines have a number of names, such as Aspirin, so Watson is taught this. It can also be taught how different medicines are used.
Potter explains that in the Memorial Sloan Kettering Cancer Center in New York, where Watson is used in the oncology department, oncologists and cancer physicians have spent more than 5000 hours teaching Watson — the same as they would teach a medical student or resident about cancer.
What makes it so different?
There are a lot of other tools that allow you to search for information, or do analytics on data, but cognitive computing is fundamentally different, says Potter.
“Any tools that search for things tend to be based on key words, say in the form of a question or sentence, and looks for sources of information that contain the key word,” he said. “Where Watson is different is that those sorts of solutions don’t understand the context of the data they’re bringing back.”
Most tools, such as search engines, only work with structured data, such as webpages, PDF files, spreadsheets and typically metadata stored in databases.
“But Watson has the ability to read and understand how to gather unstructured data in a meaningful way,” Potter said.
How cognitive computing is being used
There are already more than 300 instances of Watson up and running completing different applications.
“The field I’m working in is medicine,” Potter explained, “where Watson empowers physicians to be more efficient and effective around making decisions about their treatments.”
The Watson cognitive computer bases its responses on a “corpus”, a body of knowledge made available to it. “The one we use at Memorial Sloan Kettering currently has more than 12 million pages of textual information and more than 200 textbooks and information sources,” Potter said.[pullquote] “The one we use at Memorial Sloan Kettering currently has more than 12 million pages of textual information and more than 200 textbooks and information sources.” [/pullquote]
In a world where the amount of data and research is expanding at an exponential rate, cognitive computers have the ability to sift through all of this information at a rate no human could hope to accomplish.
“Watson can read millions of pages per second and understand what it’s reading,” Potter said. What cognitive computers do is enhance the ability of a professional to readily access all the relevant information that would otherwise be difficult or nearly impossible. It’s designed to extend a clinician’s ability to acquire greater insights into an individual patient’s condition and the best course of treatment. It serves to democratise and scale specialised treatment rather than generalise treatments.
Already the platform is being used in a diverse range of other industries to provide solutions: oil and gas engineering companies are using it to help map the complexities of large-scale plants, wealth management providers have taught Watson how to be a financial advisor attuned to the risk profile of each individual, and “ChefWatson” writes new recipes.
Cognitive computing is the future
“In the broadest terms,” said Potter, “there’s masses of information that could deliver some insights to solve a problem that are so massive that you couldn’t put enough people on the path to do it.”
IBM is working with a government client in the United States to try and identify people most at risk of obesity and type 2 diabetes. Deakin University in Melbourne is using Watson as a question-and-answer platform that provides students information that they may otherwise be unable to access.
Cognitive computing is also assisting people working in call centres to find the information most relevant to questions customers ask them. Public sector workers would be able to use cognitive computing to provide real-time answers on previous government policies. It can also be used for security management and emergency services by consuming lots of publicly available information and being able to rationalise the data to identify areas of potential risk.
Potter says the potential for cognitive computing to augment the skills and capabilities of workers is yet to truly be realised. “But with any of the domains we participate in, Watson is never meant to replace the human being,” he said. “What it’s meant to do is enhance their abilities to come to a good decision.
“The cognitive capability is to reason like a human being, and present a conclusion that is based on evidence. But a human can still say, ‘yes, but I’ll do things slightly differently’.
“That is what we believe is the future of computing.”
Written by Jacob Robinson