For a long time, language was considered the last untouchable bastion of human specialness. We thought that weaving sentences, understanding nuance and telling stories were inextricably linked to consciousness and the mind. But with the advent of modern Large Language Models (LLMs), this assumption has fundamentally changed. These "strange new minds" - artificial neural networks - have learned not only to imitate us in our own domain, but often to surpass us. But how did an architecture of numbers and probabilities achieve this mastery?
This article traces the technological evolutionary path: from the early, wooden attempts at rule-based language processing to the "transformer moment" that enabled machines to capture context globally. It highlights the fact that AI does not "understand" the way a human understands, but rather understands language as a multidimensional mathematical space. We are dealing with a statistical mimicry that is so perfect that the boundary between tool and counterpart becomes blurred.
However, the central question is not just how they learned to speak, but what it means to live in a world in which language is generated by non-living beings. What happens to our perception of truth when texts are created without an author? How does it change human cognition when we begin to outsource our thoughts to "external minds"?
The lecture concludes with an analysis of the social consequences: We are in a phase of "demystification" of language. If the word no longer necessarily presupposes a mind, we must redefine what human intelligence is at its core. It is an invitation to see the strange new neighborhood of these digital entities not as a threat, but as a mirror that forces us to radically question our own concepts of thought and communication.
Admission: 5 Euro
