Large Language Models

Relying on LLMs as sources of truth is a dangerous shortcut: low-quality data, hallucinations, and structural compliance make a higher level of verification and oversight necessary compared to what applies to traditional media

LLMs excel at language, but without mathematical optimization they cannot make reliable decisions. The new frontier of AI lies in integrating language models with formal rigor

With generative AI for images, there is a shift to the creative act. This change lowers barriers but risks compressing complexity, diversity and visual authorship

LLMs are not just tools work assisting: they are becoming the environment through which we write, organize ideas and make decisions. Understanding how they work is the first step toward governing their effects