How exactly do ChatGPT and other large language models work? Do they possess intellectual abilities that already surpass those of humans? Or are they merely parrots that repeat what humans have taught them? Katharina Zweig, an award-winning computer science professor and Germany's leading AI educator, demonstrates what ChatGPT and similar models can really do and what they cannot. She explains whether these language models can actually summarize, compare, analyze, evaluate, or correct our mistakes – and how to use them most effectively. She also discusses the dangers of their new functions, such as AI agents – programs that autonomously perform tasks on the web for users. Because decisions must still be made by us humans, not by AI.