Can I use AI to help find sources?
Note: The following is based on our current understanding of, and experience with, the free versions of the Large Language Model (LLM) AI, ChatGPT.
In your research papers you will need to use verifiable sources generated by human scholarship. Unfortunately, ChatGPT, and other LLM AI systems, will sometimes respond with incorrect information and have a tendency to create fake citations.This may improve in future iterations of LLMs, but it continues to be a real problem at the moment. Trying to track down these fake citations and fact-checking AI output is an enormous waste of your time.
Articles:
- Brown, N. & CNN. (2023, December 29). Michael Cohen says he unwittingly sent attorney non-existent case citations generated by AI | CNN Politics. CNN.
- Emsley, R. (2023, September). ChatGPT: these are not hallucinations – they’re fabrications and falsifications. Schizophr 9, 52 (2023).
- News, iSchools. (2023, September 18). News Feature: Citation Accuracy in ChatGPT. iSchools Inc.
- OpenAI. (n.d.). OpenAI. Does Chat GPT Tell the Truth? Note: This blog post is continuously updated and it's fascinating to see what's changed and what hasn't.
Books:
- Broussard, Meredith. Artificial Unintelligence: How Computers Misunderstand the World. Cambridge, Massachusetts: The MIT Press, 2018
Although based on AI systems before the rollout of ChatGPT, it is very helpful for understanding the data that underlies all machine learning. - McQuillan, Dan. Resisting AI: an Anti-Fascist Approach to Artificial Intelligence. Bristol: Bristol University Press, 2022
Begins with a chapter explaining neural networks, and discusses the social and environmental impacts of AI.
Podcasts: