Always be very careful in using chatbots.

As a reminder, Natural Language Models on which they are based aim to make a realistic conversation…but there is no reason for the conversation to also have a correct content, which is instead key in science.

So never ever trust what chatGPT says.

However, we can use it to rephrase sentences in a better way, for example, we can literally ask “Can you rephrase with other words this sentence: XXXYYYZZZ”.

Another very decent use of ChatGPT (or co-pilot from OpenAI, which is built on the same language model) is to write code. Once we break a problem into parts, we can literally ask chatGPT to write a function (in C, Python, …) to do a specific task, and it usually does it very well. It is even more precise if we want to translate between languages, for example, if we have a Python code and want it written in C++, just put the code and ask to translate it in C++, it works extremely well.

Finally, for papers we might also want to use it to write very general introductions to some topic, however, I’d do it only for topics that we know very well, so that if there is any wrong information we can quickly correct it.