OpenAI "fixed" errors in ChatGPT
Last night, ChatGPT users started receiving strange responses from the chatbot, including switching between languages and even repeatedly correcting themselves. Some of the responses were simply nonsensical.
OpenAI acknowledged the problem on its page and started investigating it. A few minutes later, it was announced that they had found the cause of the problem and were working on a solution.
Many of the errors resemble common social media memes, which is typical of trolling. However, some Reddit users have reported similar problems in the past.
This is a reminder that generative AI chatbots are a rapidly evolving technology and their behaviour can be unpredictable. It is important to understand that they are based on large language models that even their creators do not always fully understand.
Source: The Verge