Wie „denken“ ChatGPT & Co.? Ein Blick hinter die Kulissen. (LNDW 2026)

How do ChatGPT, Gemini, and their peers manage to respond almost like humans? In this workshop, we demystify the chatbots that have become nearly impossible to avoid. Together, we explore how language models (LLMs) juggle probabilities, why they sometimes invent fantastical stories, and how skilful "prompting" can steer them. The workshop introduces the fundamental workings of large language models (LLMs). Its aim is to build an intuitive understanding of the concept of "next-token prediction," the prediction of the next fragment of a word, that lies at the heart of how these systems generate language.

• No prior knowledge is required.
• The event is open to anyone curious about science, suitable for teenagers and adults alike.
• The workshop will be held in German.

Event details

• Date: Saturday, 6 June 2026
• Time: 5:30 – 6:30 PM
• Moderation: Lisa Raithel (DFKI, BIFOLD, TU Berlin)
• Venue: Physikalisch-Technische Bundesanstalt (PTB), Elsa-Neumann-Bau, Room 104, 1st floor; Marchstraße 25 / corner of Guerickestraße, 10587 Berlin
• Language: German
• Admission: As part of the Lange Nacht der Wissenschaften Berlin
• Pre-registration is requested. Register here