OpenAI's new tool called ChatGPT is a trained model that can converse and answer technical questions in almost human-like manner. Recently, I asked it to find open research questions in my field (quantum optics in space for gravitational wave detection) and it actually responded with some solid pathways that I had not considered. In this way, for me, ChatGPT could be a promising addition to my research supervisor for ideas and simple to read summaries of new fields of research. I want to know in the research community, have you had a go with ChatGPT and what kinds of potential use-cases do you foresee (if any).
See https://openai.com/blog/chatgpt/ for more details.
Not that much. What machines do is different from what humans do and it doesn’t make sense mixing them up.
I think the main issue with using text generated by tools such as chatGPT is that it is not guaranteed to make true statements. I can imagine (e.g.) using it to generate some synthetic introduction text, which might give you a useful stating point, but you would then (have to) check accuracy and relevance, and add appropriate citations. This might suit some work patterns, but I would think that the time saved using GPT would be balanced by the extra checks required (especially since GPT could make some quite plausible-sounding wrong statements that might be tricky to catch).
ChatGBT can be used as a starting point for your research and to further develop and refine the information through additional research, analysis, and critical thinking.
When ChatGPT generates responses that include citations or references, it is important to evaluate the quality and accuracy of the sources it provides before using them in your research. it may not provide reliable, accurate, or up-to-date sources.
However, if you use Google or ChatGBT, it is important to critically evaluate the sources you find. Even in Google, you must consider factors such as the author's credentials, the publication's reputation, and the quality of the evidence presented.
Ahmed Mounaf Mahdi, it is not just the "quality and accuracy of the sources" that GPT provides that are an issue. A source cited by GPT can be entirely ficticious, i.e. might not exist. And even if you challenge GPT about the accuracy, it may not reply with a true statement.
See e.g.
https://www.theregister.com/2023/03/02/chatgpt_considered_harmful/