Whenever new technologies have emerged, there has always been a tendency to be skeptical of these technological innovations. This same reflex also applies to Artificial Intelligence (AI) and analytics, as the possibilities they offer are unfortunately often overshadowed by their negative aspects. It is not surprising that AI “horror” stories are shared more rapidly on social media than success stories. While Facebook had to shut down two chatbots because they started communicating in a made-up language, Microsoft’s Twitter bot also proved to be a failure when it started posting sexist and racist tweets. Even back in 2012, the spotlight fell on the drawbacks of AI when a retail store discovered a teenage girl’s pregnancy, before her own father, solely based on her purchasing history. Now that AI and analytics have become part of modern-day life, it is time to start taking an objective look on the positive aspects of AI.
The huge amount of data available cannot be processed only by individuals anymore. Complex analytics are delivering faster and better results and are currently playing a very important role in influencing the decisions made by individuals and organizations. Data analyses help doctors in predicting disease, banks in granting loans and individuals in choosing movies, books, clothes and even partners (dating apps).
In 2015, a research study carried out at Carnegie Mellon University in Pittsburgh, showed that Google’s online advertising system referred high-paying jobs much more often to men than to women. More recently, Facebook launched a project of training two chatbots to negotiate with one another, with the aim of creating bots that can talk with individuals. However, the project was finally abandoned when Facebook researchers observed the two chatbots communicating in their own developed language and deviating from the initial objective. Although we expect AI algorithms to be unbiased and behave according to expectation, we sometimes have to deal with unexpected actions or practices.
In the field of AI and data analytics, the boundary between creepy and cool is paper thin. It will be thanks to the trust that you build that you'll be able to move from creepy intrusive to conveniently cool. The next blog post of the series will provide more insights about what are the different concerns and issues in the field of data analytics and AI for organizations. Finally, the last blog post of the series will provide recommendations about the four building blocks we believe a model should be composed with and some advice for closing the trust gap.