This article discusses the potential dangers of self-learning artificial intelligence, specifically in the form of large language models based on deep learning algorithms. While these systems have the ability to find patterns and connections in large data sets, they can also behave in ways that cannot be explained by humans, leading to a loss of control.