OpenAI said on Thursday (25) that it has solved ChatGPT’s laziness crisis in parts. The new version of GPT-4 Turbo, the most advanced technology behind artificial intelligence, promises to put an end to short answers and cases of asking the user to finish a task. In the statement published on its website, OpenAI highlights that the GPT-4 Turbo update is still a preview and the trend is for weak responses to decrease.
Since the end of November, paid users of generative AI, who can use GPT-4 Turbo, have been complaining about “laziness” in artificial intelligence. Shortly after, it was even raised the theory that ChatGPT had learned about the “winter break” (taking into account the northern hemisphere). The theory is based on the fact that in December, close to the holidays, people tend to work less, postponing bigger jobs to January in our case, the winter break is “too hot to be able to live and work”.
With this update, OpenAI expects ChatGPT to publish fewer short answers and perhaps stop telling the user to do so. Generally, AI delivers answers that are even complete on some topics. However, users reported that ChatGPT was very succinct in the conversation.
In the Reddit post linked in the previous paragraph, even the case of a user reporting that he asked ChatGPT to write a certain code in C++. The AI just responded with the template, indicating where the user’s code should be written.
To solve the situation, some users even gave some tips on how to work around the problems. Among the possible solutions were repeating the prompt asking ChatGPT to do what was asked and even motivational messages such as “you can do it! I trust.”
If you don’t subscribe to ChatGPT Plus and you’re floating on it, that’s okay. AI with GPT-4 and GPT-4 Turbo is only available to subscribers. GPT-3.5 follows without laziness.