The program “Breitband” on Deutschlandfunk Kultur regularly addresses current topics related to digitalization at the intersection of technology, society, and politics. The latest episode, “Ärger um OpenAI – ChatGPT polarisiert mit neuer Version” (EN: Controversy surrounding OpenAI – ChatGPT polarizes with new version), examines the new version of OpenAI’s AI-powered chatbot, ChatGPT-5, and the partially controversial reactions to the software. Among the interviewees is Hendrik Heuer, research professor and head of the research program “Design of trustworthy artificial intelligence” at CAIS.
Hendrik Heuer explains why standardized tests for measuring performance in AI models are only of limited significance and why transparent and independent evaluations of language models are necessary. In addition, he criticizes the lack of responsibility on the part of AI companies with regard to users’ emotional attachments to chatbots, which have been given too little consideration to date.
Despite over 700 million users worldwide and high expectations for the successor model, the verdict is sobering or even critical: Compared to its predecessor, GPT-4o, ChatGPT-5 is perceived as less friendly, less “personal,” and less emotional overall. Those who use the chatbot in everyday life as a conversation partner, friend, or for therapeutic help are particularly critical of the change.
The complete episode from August 16, 2025, is available online (in German):
https://www.deutschlandfunkkultur.de/mensch-maschine-warum-polarisiert-die-neue-version-von-chatgpt-100.html