Computational Modelling and Artificial Intelligence Improve Virtual Testing for Better Real Solutions

Bitcoins, self-driving cars and industry 4.0 are among the topical subjects of debate and discussion in the business community: artificial intelligence is the fil-rouge connecting them. Within artificial intelligence, computational modelling is becoming increasingly important. As a sign of our times, on February 20, the Government Office for Science of the United Kingdom published a Blackett review on Computational Modelling (https://www.gov.uk/government/publications/computational-modelling-blackett-review). The review recognizes the importance that computational models and simulators are now gaining in the business environment to help decision-making and increase competitiveness: Computational models can help us to deal with an increasingly complex world that is changing quickly, often in unexpected ways. Increasing computational power and the greater availability of data has enabled the development of new kinds of computational model. These allow us to do virtual 'what if?' experiments about our world before we try things out for real. This presents huge new opportunities, which we must strive to grasp [p. 12].

The statements of the Blackett review indeed take a fresh look at ideas that have been maturing in the industry for the last fifty years at least. For instance, the following quote is attributed to Frederick W. Smith, the well known founder of FedeX: By modeling various alternatives for future system design, Federal Express has, in effect, made its mistakes on paper. Computer modeling works; it allows us to examine many different alternatives and it forces the examination of the entire problem.

Not only, but as is well known, computational models are crucial in helping policy-makers in deciding about important societal issues such as the identification of policies to mitigate the effects of climate changes.

Sometimes, models are the only way in which we can understand complex relationships in large datasets. The Stanford Encyclopedia of Philosophy at the Chapter devoted to Models in Science, evidences that two types of models are mainly developed: models of phenomena and models of data. In the case of models of phenomena, the model is one of the three vertices of the scientific triangle: one vertex is a scientific theory, a second vertex is a mathematical model whose equations descend from the theory and the third vertex is the real world phenomenon which id explained by the theory and described by the mathematical model. Sometimes, however, the theory vertex might be missing or even impossible to build. Then, models of data come into play. Models of data are essential in nowadays business. We have huge quantity of data whose fast and proper utilization is a key factor for the competitiveness and sometimes even the survival of a corporation, large or small. Data need to be interpreted via computational models to provide useful information. Starting from the data of a large fashion brand, a brilliant Italian data science firm has created a linear optimization model (technically a linear program) to support the logistics of a well-known fashion brand. Why the model? The company management wish to consider the problem in its entirety; given the several locations and goods to be transported, they realize that they have over one million strategies among which to choose: no human would be able to identify the best solution in a few seconds.

The type of decision support model in question is an optimization model and needs mathematical tools to efficiently find the minimum cost or maximum profit. This optimization model supports directly a business decision. Optimization problems are becoming so complex that scientists are thinking of using artificial intelligence tools (such as artificial neural networks) to help finding the optimal solution. At the same time, several problems in artificial intelligence, for instance training an artificial neural network, are in fact optimization problems. That is, we are now assisting to a cross-fertilization of scientific disciplines that appeared distant some years ago. On the business side? The implications are numerous and we are only at the initial stage, but the development pace is fast and the cross-contamination will make it even faster. Then, it becomes increasingly important for managers to be exposed to this evolution, to get to know the new tools. As John Little, Professor Emeritus at MIT, was writing over the years, the manager does not need to be a specialist of the technical detail. However, the manager needs to be able to know which algorithm can help her solving the business problem at hand.