Читать книгу Societal Responsibility of Artificial Intelligence - Группа авторов - Страница 14

Artificial general intelligence (AGI) or strong AI has similar – and even superior – reasoning abilities to those of human beings. It is endowed with capabilities not limited to certain areas or tasks. It reproduces or aims to reproduce a mind, or even a consciousness, on a machine. That is to say, an evolutionary machine with its own reasoning and consciousness, capable in particular of independently elaborating strategies and/or decisions that go beyond human beings in order to understand them so as to help them (in the best of cases) or to deceive or even destroy them (in the worst of cases).

From a general point of view, AI can be illustrated as an algorithmic matrix that aims to “justly or coldly” optimize decisions. Naturally, the morality or fairness of this judgment is not predefined, but depends, on the one hand, on the way in which the rules are learned (the objective criterion that has been chosen), and, on the other hand, on the way in which the learning sample has been constructed. The choice of the mathematical rules used to create the model is crucial. Just like the human functioning that analyzes a situation before changing one’s behavior, AI allows the machine to learn from its own results to modify its programming. This technology already exists in many applications like on our smartphones, and should soon be extended to all areas of daily life: from medicine to the autonomous car, through artistic creation, mass distribution, or the fight against crime and terrorism. Machine learning not only offers the opportunity to automatically make use of large amounts of data and identify habits in consumer behavior. Now, we can also actuate these data.