In 2017, the artificial Mongolia Email List of Google AlphaZero beat the holder of the title of world chess champion for the first time: Stockfish – who is not a human but an algorithm as well. Stockfish is the heir to a long series of strictly “superhuman” chess software since Deepblue, the first supercomputer to beat a world champion (Kasparov) in 1997. In the example of the AlphaZero vs. Stockfish, we can assimilate Alphazero to Machine Learning and Stockfish to Mathematical Optimization: AlphaZero is a Deep Learning algorithm, that is, it has learned itself (by unsupervised training by playing against itself) to play better than a grandmaster at chess. Stockfish is based on the working principles of mathematical optimization; it owes its

level to chess to its ability to calculate for each move millions of possible alternatives, each move being judged more or less good according to a largely human model of the game (strength of positions, strength of pieces, etc. ) Does this ML victory over OM then call into question the interest of mathematical optimization in the business world? Businesses are not chess boards Business models aren’t quite like a game for two reasons: They are not circumscribed : in chess, the pieces cannot leave the chessboard and each piece has a precise and immutable role; in companies, the ramifications of stakeholders (employees, suppliers,

The Tech-savvy Shepherd

customers) on the value chain are often infinitely complex. They are dependent on the outside world : in chess, I won’t win a game by flicking my opponent’s king down; companies, for their part, are dependent on the market, the availability of resources, contingencies of all kinds … In my opinion, these two peculiarities of companies still prevent machine learning from being all-powerful as it is in chess. Likewise, Kasparov’s loss to Stockfish in 1997 did not signify the impending obsolescence of human skills in the world of work. The value of ML and that of OM in our companies are nevertheless very real, but they are different from each


other (just like that of Men!) Artificial intelligence is popular! A few months ago, I was working for a client in the industrial sector on a “factory 4.0” oriented mission. You don’t necessarily need to go into the details of the mission to share a finding: it would have been difficult to convince our interlocutors of the relevance of the solutions we were proposing without talking about IoT, IA, Big Data, etc. In general, these technologies are enjoying a fashion effect, they are currently trends! While in some cases they can positively and profoundly transform organizations, misunderstanding them can also lead to failure.. In this sense, new

To Hell With The Sheep, Concrete Applications

technologies are not the answer to everything and often prove to be fully effective in complementarity with other tools or in specific organizational methods. Machine Learning is no exception and where we could see in this technology an improved version of mathematical optimization (OM), we should rather see a compatible and complementary solution with it: in our business organizations which are not circumscribed and dependent on the outside world, the ML remains a tool of prediction and gives visibility while the OM arms its users for decision-making in the face of these predicted situations. The tech-savvy shepherd Let us illustrate the complementarity Machine Learning – Mathematical Optimization by an example, that of the

technophile shepherd. This shepherd has two concerns: he seeks to reduce the risks in the management of his herd (loss of sheep for example) he tries to minimize his efforts and that of his dog to manage his herd (walk less) His taste for new technologies pushed him to equip himself with a mathematical solver and a Machine Learning algorithm through which he modeled his organization. Here is how the two tools work in complementarity to enable the Shepherd to solve his problems: In this example, Machine Learning is used to identify “patterns”, that is to say patterns that repeat themselves formed by the organization of the sheep

Leave a Reply

Your email address will not be published.