Using AI and Algorithms to Solve Complex Systems

It is not easy to understand what artificial intelligence is, and especially to understand the fields of its application. This task is not simplified by the fact that artificial intelligence groups together several different types of algorithms. Some of these algorithms are called deterministic, while others are more probalistic (or stochastic, if you prefer). In this article, I try modestly to give my point of view to explain the contributions of artificial intelligence compared to what we had with computer science in the past.

What Does AI Consist Of?

First of all, it is important to know what the different artificial intelligence algorithms consist of. For example, we have:

• Expert systems listing domain knowledge to reach conclusions.

• Neural systems imitating neural functioning to learn and infer conclusions.

• Stochastic algorithms doing probability and statistics.

• Deep learning modeling long chains of cause and effect.

• Machine learning learning based on statistics.

How Does AI Resolve Complex Systems? 

Compared to procedural algorithms that I would describe as classical, artificial intelligence algorithms seek not to solve a simple process but to solve what is called a complex system. Indeed, whereas a procedural algorithm is a simple succession of steps subject to simple conditions, artificial intelligence seeks to solve a problem containing many entities linked together in a complex way. And that is precisely the definition of a complex system. AI could, for example, study complex systems such as public opinion and the economy as well as other fields such as language and recognition of forms. All these domains are addressed by artificial intelligence algorithms. 

We Still Use Statistics

We could say that artificial intelligence solves complex systems, either by trying to embrace this complexity or by having a simple and procedural approach. One would thus have expert system algorithms and neural systems in the category of algorithms that embrace this complexity, whereas machine learning would be a more procedural method.

It is not at all shocking to have a procedural and statistical approach to address a complex system. One of the most telling examples could be, for example, agile methodologies. Agile methodologies were inspired by the same vision of a complex system of projects, and an attempt was made to resolve the complex nature of a project in a procedural manner, including all the associated comitology. And it is because agility understands the complexity of projects that it has succeeded in providing simple procedures to address this complexity. At the same time, agility is aware of its limitations by, for example, pushing Sprint review meetings and pushing to manage technical debt. It is indeed complicated, if not impossible, to solve most of the problems of a complex system without an improvement review procedure.

Then What? 

To say that artificial intelligence solves complex systems whereas previously procedures were solved makes it possible to measure the very strong impact that AI will have. Indeed, whereas before, we did processes and calculations, tomorrow, we will understand the complex system of our world. Obviously, not everything can be solved, and great progress remains to be made, for example, in biology, but with AI, we will be able to, for example, understand and predict economic crises. In short, AI brings answers to everyone's questions because we will have the answer even before asking the question. We can expect many changes, whether they're good or bad, but with a glimmer of hope for humanity.