ChatGPT's capabilities are definitely considerable, but it is also undeniably constrained by a few core weaknesses:
- Despite what some may believe based off the term "artificial intelligence," ChatGPT does not actually "think" when it replies to user prompts. It generates responses using probability to determine what word is most likely to come next based on the vast amount of data it has been trained on. In other words, ChatGPT is not consciously aware of the answers it provides - it just predicts and imitates.
- ChatGPT also has a cutoff date for the information it was trained upon - September 2021. Therefore, it has no accurate knowledge of current events since then.
As a result:
- ChatGPT frequently provides false information in its responses.
- In particular, it often provides citations for resources that don't exists when asked to provide resources for citations for a specific research topic - an issue that has already been experienced by faculty and librarians alike here at SHU.
You can check out the articles linked below to read more in-depth about some of ChatGPT's limitations.
FIND US ON