Intelligent machines for good?

Machine learning and modern Artificial Intelligence (AI) systems are influencing several aspects of our human lives. Many of these algorithms, based on Artificial Neural Networks (ANNs), have been empowered to make decisions and take actions, based on the well-known notions of efficiency and speed. The aura of objectivity and infallibility of such algorithms, nonetheless, have been already put into question (e.g., refer to the debate about the recent tragic car crashes that have involved self-driving cars). In this setting, our intuition identifies a key issue around the problem of AI errors and bias into the insufficient or inaccurate (human) activity of comprehension and codification of the context where the ANNs will have to operate. We present here a simple cognification ANN-based case study, in an underwater scenario, where we recovered from a situation of partial failure, by including additional contextual factors that were initially disregarded. Our final reflection is that a nuanced consideration of a complex context, and subsequent technical actions, should be always kept in mind before an AI-based system takes its final shape. Because machines have still no context for what they are doing, it is a human duty and responsibility to codify it.