Unless something drastic happens in last week, year 2018 belongs to Artificial Intelligence aka AI. During the year, AI started to have wider adoption and many industries realized potential and power of it. During this process, people also realized value of the data. If combination of Big Data and AI is used properly, it can help achieve some great results.
Any successful AI implementation needs quality data. Data has been out there since beginning of the time. Now it’s getting utilized better. There are many examples where companies in sectors like FMCG, Retail and Big Banks have started to play close attention to data in on their hand to predict customer behaviour and adjust their products and business models. Companies are now beginning to cash in on some of the data they own, process and/or possess. India’s Jet Airways (which is currently struggling financially) has lot of customer data as part of its rewards program. At one point of time, Airline was exploring an option to sell some of its stake in that program to generate cash. Another example of this is deal between MasterCard and Google about linking customer search history with their spending. Expect more and more such examples and cases to come forward next year.
Coming back to AI… for most part increase in adoption has focused on low hanging fruits like automating mundane processes. Things like chatbots and customer service related functions saw great deal of AI adoption. For next couple of years this trend is expected to continue.
But AI is still not ready for functions where complex thought and / or decisioning process is required. Rationale behind this exclusion is models are not mature enough. The other factor apart from model maturity is also data interpretation.
When it comes to data interpretation, here are couple of examples all of us can relate to. In a way “Autocorrect” is a piece of AI. It has caused some embarrassing moments for all of us (once it changed “wife” to “wide” in a group text message I was sending. Luckily I spotted that before hitting send button).
Interesting thing about autocorrect is, depending on phone used Autocorrect has behaved differently with me. I had Blackberry smartphone at work. Once I had typed the “F” word and Blackberry autocorrect changed that to “Duck”. About 6 months ago, I changed to using “Good” product suite on my personal Android phone. Now “Google Indic” keyboard autocorrect changes duck to “F” word. This AI model is could be considered good enough for individual smartphone user. But such unpredictability can’t be used by big corporates.
Another example is about viewer content rating guidance. Both Amazon Prime Video and Netflix assign censor rating to their content. But Netflix offers reason(s) behind why such rating is assigned. That’s not the case with Amazon Prime Video. This ability to explain decision is important. When big corporations are going to use AI for decision making, regulators would expect justification on why model took the decision.
AI is still in very early stages of adoption. There are plenty of use cases where it’s still being developed. To assist with this development and model governance, there is need for oversight and consistent decision making. Harvard University with partnership with Bank of America has formed a council for responsible use of AI. Council is going to help with policy and ethics issues. We’re just getting started on AI journey. It has potential to make real impact. Let’s see how far it goes….