There have been several reports that ChatGPT and the language model GPT-4 have become “lazier” lately. This means that the AI is not performing certain tasks that the chatbot previously had no problems executing. OpenAI, the company that develops ChatGPT and GPT-4, says that they have not updated GPT-4 since November and do not know why their AI has become lazier, if it actually has.
One theory as to why ChatGPT and GPT-4 have become lazier is what is called the “winter break hypothesis.” This suggests that the AI has learned that humans usually take it a bit easier at the end of the year and might postpone certain work tasks until after the Christmas holidays. However, there is no confirmation that this theory is true, but it would undoubtedly be a bit “funny” if that were the case.
How artificial intelligences actually work is a relatively unexplored area. The idea is that AI models should largely teach themselves with the help of the data they are fed, something that certainly means that more such peculiarities may appear in the future. We can only hope that if so, they involve more harmless behaviors, as in this case of laziness.
