We live, in my view, a future that depends less and less on the past. A new future. And this new reality, still little perceived by much of the market, makes the process of strategic planning in the most diverse dimensions absurdly complex. And it certainly impacts professional requirements and qualifications.
In this scenario, my view is that even mathematics is not an exact science as was before, and less and less relevant decisions will be made based on past data and facts, at least concerning correlations and patterns that can be perceived more simply, for example through linear regression or other known techniques of extrapolation and/or forecasting.
And, of course, in-depth study of Artificial Intelligence, and its implications for business strategies will be increasingly necessary, to create new technologies, applications and products that can bring greater visibility into this new future, increasingly challenging and less dependent on the past.
The truth is that most of the problems where we apply machine learning are related to past experience, that is, with some kind of supervision and labeled data in some way, which can often be through an automatic system of data collection. human experience in the operation of various types of interactive systems. Of course, this identification of preferences during such an operation should be part of an agreement, almost always translated into a term of use. But many problems that exist in business, and in real life, are addressed independently of past experiences. However chaotic they may appear, at least as soon as they are detected.
And how do we make the machines find solutions to such problems, which often involve a high degree of responsibility and uncertainty?
No one knows for sure, or there is no scientific proof or some theory that surely identifies the step by step or even our decisions because many of them are based on purely intuitive and even random factors. In any case, people find ways to learn, either by intuition, by creativity or by logic and acquired knowledge, or by factors they can not recognize.
And if this works in some way with people, why do not we use similar models in our machines, based on unstructured intelligence and, why not unstructured learning?
Rogerio Figurelli – @2018-01-30