Utpal Kaul, Global Head New Product Incubation, Carlson Wagonlit Travel
By definition, Cognitive Technology is a discipline within computer science which aims to mimic some higher level functions of human brain, for example image and pattern recognition, natural language understanding / processing and data mining. Although we are still light years away from a machine being able to think, feel and act autonomously like a human, substantial progress has been made in several narrow fields wherein a machine can compare favorably with a human being.
The evolution of natural language processing or NLP programs is one such area. For a machine to understand the context of a dialog is remarkably complex. Let's take an example, "this is a book" and "book me a f light", both sentences use the word "book", however in the first sentence, "book" is a noun and in the latter sentence it is a verb. The ability to discern this difference comes quite naturally to humans but machines need to be trained to understand this distinction. Let's consider another more nuanced example; we can use the word "really" as a question such as "really?" or as an expression of sarcasm "Really". In both cases the context of the dialog determines its meaning and it is this cognitive ability to gauge the context which makes NLP hard yet attractive to companies. It has several advantages over humans, for starters, it can be accessed anytime, anywhere and costs less to deploy than hiring and training humans. Chatbots, which use this technology, have become hugely popular not just in industries which are consumer direct, like retail, but also for internal use within companies for information sharing and knowledge management. By the year 2020, Gartner Research predicts that 25 percent of the customer service and support operations will have a chatbot deployed. That is up from less than 2 percent, just 2 years ago.
Image Recognition is another fascinating cognitive technology which has gained wide attention over the past decade.
Image Recognition has huge implications for diverse set of industries, from self-driving cars, to identity validation to potentially spotting a specific individual in a crowd
This has huge implications for diverse set of industries, from self-driving cars, to identity validation to potentially spotting a specific individual in a crowd. In fact facial recognition for identity verification has already been piloted by Delta airlines and Jetblue airlines at several airports. Companies such as Uber and Waymo are busy testing and perfecting driverless, autonomous cars. A little known fact is that we already have driverless trucks and vans ferrying people and goods on a few limited routes. You may well question why is driverless better? For one, humans are more prone to error than machines because the mundane but common limitations which apply to humans don’t apply to machines, like lack of sleep or driving inebriated.
There is an active argument in the innovation community on whether car ownership will become a thing of the past in the not so distant future. The argument has merit. Cars and other means of private transport have perhaps the least asset utilization of any high value durable good. According to Fortune Magazine, cars get utilized less than 5 percent of the time. The other 95 percent of the time we are paying for an unutilized asset. It is quite plausible that intelligent software will be able to predict where and when you need to get picked up and dropped off and a smart vehicle will come and ferry you just like a private car without owning one. You just pay for the time you use it.
Perhaps the most significant value of these deep learning and predictive technologies is in the world of healthcare. We can now diagnose and treat medical conditions with a very high degree of accuracy. By crunching an enormous amount of patient data like age, ethnicity, past personal and family medical history, gene markers, it is now possible to design treatments which are highly personalized to each individual patient. This promises to have a quantum improvement in treatment outcomes and quality of life.
There is however an area which is yet uncharted. This revolves around ethics and governance oversight. A commonly cited example is what if an accident is inevitable and an autonomous vehicle has to choose between an old person and a child. Whom should it collide with? Or if there are two patients waiting for an organ transplant, one has low likelihood of long-term survival despite the transplant but is first in line and the other has high chance of long-term survival but is behind in line. How do you choose? Real life is actually a lot more complicated than these straightforward examples. Humans for most part are highly accomplished in making these judgement calls. How will intelligent machines make these decisions, is anybody's guess.
The concern around governance is equally stark. Consider another emerging technology called Blockchain. In simplest terms, Blockchain is a distributed ledger with multiple nodes distributed across the globe. Every transaction between two or more parties gets instantly recorded in all the nodes, which could be in hundreds, or thousands or even millions. This effectively eliminates the possibility of fraud or dispute because a million copies of the transaction exist on all the nodes. So far so good. What if two or more nodes conduct a transaction which is unlawful? Are you culpable by retaining a copy of this unlawful transaction? It may seem far-fetched but there is actually a huge debate raging on this issue in the academic and legal community. This is where governance oversight and policy planning is crucial so they can plan ahead for such exigencies.