According to several Reddit users, ChatGPT gave unexpected answers that were labeled “complete nonsense.”
On Tuesday, the ChatGPT Reddit sub was flooded with reports from users about unexpected results from OpenAI’s ChatGPT AI assistant. According to some users, it looked like ChatGPT had suffered a “stroke.” OpenAI has now recognized the problem and is investigating the unusual behavior of the AI chatbot. Experts are currently speculating about various causes, including high temperatures or insects.
ChatGPT temporarily “crazy”
The reactions to some requests were initially completely normal, but then developed into complete nonsense. There are a few testimonials on Reddit that talk about ChatGPT “having a stroke,” “going insane,” or “wandering.”
No official statement has been made regarding ChatGPT’s unusual behavior, but experts are still speculating. For example, there is the possible explanation that ChatGPT’s temperature is set too high. Temperature is defined here as the extent by which the LLM deviates from the most likely outcome. As a result, the context from the past can suddenly be lost. Another speculation concerns OpenAI’s possible test for a new GPT-4 Turbo version, which contains unexpected bugs. Finally, another bug is also discussed, such as the recently mentioned “memory” function.