This means that each example in the training set has an input (the data) and a known output (the label or expected outcome). The goal is for the model to learn to associate the inputs with the correct outputs, allowing it to make predictions about new data. For example, supervised learning can be used to predict whether an email is ‘spam’ or ‘not spam’ based on a set of previously classified emails.
In unsupervised learning, on the other hand, the data is not labelled. The model is responsible for identifying patterns or structures in the data independently. This approach is useful for tasks such as clustering or dimensionality reduction, where the goal is to organise the data in a more understandable way. An example would be grouping customers into different segments according to their buying behaviour, without prior information about existing categories.
Reinforcement learning, on the other hand, works differently. Here, the model learns by interacting with a dynamic environment, performing actions and receiving rewards or penalties based on the results of these actions. The goal is to maximise reward over time. This type of learning is widely used in games, robotics and control systems. For example, a robot can be taught to walk by adjusting its movements to achieve balance, learning from the rewards obtained.
These three approaches are applied depending on the problem and the available data, and can often complement each other.
What are the advances and limitations of artificial intelligence?
Despite all the advances, artificial intelligence still has clear limitations. It cannot demonstrate genuine creativity, as everything it creates is based on pre-existing patterns and data, without the ability to innovate or be inspired like a human being. Moreover, AI does not understand emotions and complex contexts in an authentic way; it can interpret emotional signals, but lacks real empathy.
Another major limitation is its inability to make ethical decisions, as AI has no intrinsic values or morality, relying solely on programmed rules and patterns. In completely new or unpredictable situations, it also faces difficulties, as it relies on previous data to learn and adapt. Finally, AI does not engage in intuitive or critical reasoning, processing information logically but without the depth or abstraction of human thought.
These limitations reinforce that artificial intelligence should be seen as a complementary tool, amplifying human capabilities but not replacing them.
