Artificial intelligence (AI) is the field devoted to giving machines abilities usually reserved for the human mind: to perceive, reason, learn, decide and create. Picture an apprentice who, instead of following written rules one by one, observes thousands of examples and works out for itself how to act — that is the spirit of modern AI.
From rules to learning
Early symbolic AI relied on explicit, hand-coded rules ("if fever and cough, then flu"). Today's dominant paradigm is machine learning: the system tunes its parameters from data in order to minimize its errors. Deep learning, built on deep neural networks, has driven the recent breakthroughs in vision, language and content generation.
At the heart of this learning lies the goal of minimizing a loss function that measures the gap between prediction and reality:
$$\theta^* = \arg\min_{\theta} \; \frac{1}{n} \sum_{i=1}^{n} \mathcal{L}\big(f_\theta(x_i),\, y_i\big)$$
Main families
| Type | Principle | Example |
|---|---|---|
| Narrow AI | Specialized in one task | Speech recognition, translation |
| Generative AI | Produces new content | Text, images, code |
| General AI (AGI) | Human-level versatility — still theoretical | Does not exist today |
Promise and caution
AI is accelerating medicine, industry and education, yet it raises serious issues of bias, transparency and data sovereignty. For the saMORAIs, mastering these tools is a lever for development — provided we keep humans at the center.
AI does not imitate intelligence: it reinvents a form of it, out of the data we entrust to it.