Tools like ChatGPT, DeepSeek and Claude took the world by storm. Powered by large-scale language models (or simply large language models, or LLMs) – massive artificial neural networks, trained to understand and generate language – these tools can address numerous tasks from text summarisation and revision, to writing computer code and even translation of texts unseen before with unprecedented performance. The capabilities of these models and the easy access have facilitated their adoption in academic institutions, in the public sector and in industry. However, the fast temps of adoption, the proprietary frameworks and models which dominate the LLM landscape, lead to a knowledge gap about the actual benefits and drawbacks of this technology as well as the impacts on individuals, society, economy and environment.

This round table presents insights on the capacities of these models and tools, their underlying reasoning and learning mechanisms and facilitates a deeper discussion on their current and potential future (positive and negative) impacts on a wider-scale.

Panel by Dimitar Sterionov