I got this useful bon mot from a middle school teacher recently.

In response to, “I DONT UNDERSTAND,” he calmly said, “okay what steps have you taken to understand?”

And that’s when I realized that a lot of folks have no steps.

Yeah, me too. So, you likely already follow those researchers as they are some of the top names in the field

Ah, there's a misunderstanding with the term source. I meant sources as in the sources it provides for a given response. Not it's own training data. I never said it has a 100% error rate - the % I am quoting are from peer reviewed, published studies, not my own personal experiences

No idea why the links in the thread are sometimes for the following paper! The citations should all be there though.

Oh duh how could I forget to reference. Read the whole thing, it's excellent.

Guest, O. et al (2025). Against the Uncritical Adoption of 'AI' Technologies in Academia. Zenodo. doi.org/10.5281/zeno...

I strongly recommend following the work of @emilymbender.bsky.social @abeba.bsky.social @irisvanrooij.bsky.social @olivia.science @mmitchell.bsky.social to learn more about algorithmic harms and AI ethics in general.

Shah, C. and Bender, E. 2024. Envisioning Information Access Systems: What Makes for Good Tools and a Healthy Web? ACM Trans. Web 18, 3, Article 33. doi.org/10.1145/3649...

LLMs “take away transparency and user agency, further amplify the problems associated with bias in [information access] systems, and often provide ungrounded and/or toxic answers that may go unchecked by a typical user.” - Shah and Bender 2024

Vicente citation:

Vicente, L., Matute, H. Humans inherit artificial intelligence biases. Sci Rep 13, 15737 (2023). doi.org/10.1038/s415...

Adam et al also found biased AI outputs resulted in biased decisions by both experts and non-experts while study participants who did NOT use AI were unbiased, a finding Vicente 2023 confirmed and extended, showing people continued to perpetuate biases even when not using an AI tool.

And in:

Adam, H., Balagopalan, A., Alsentzer, E. et al. Mitigating the impact of biased artificial intelligence in emergency decision-making. Commun Med 2, 149 (2022). doi.org/10.1038/s438...

So not only is it wildly inaccurate, but AI mistakes not only weren’t caught by humans but resulted in a decreased accuracy in human decision making in:

Dratsch et al. 2023. Automation Bias in Mammography: The Impact of Artificial Intelligence BI-RADS Suggestions on Reader Performance

This 2023 study found ChatGPT was unable to provide ANY actual refs & instead provided made up citations with PubMed IDs that were from unrelated papers

Alkaissi H, McFarlane S. (February 19, 2023) Artificial Hallucinations in ChatGPT: Implications in Scientific Writing. doi:10.7759/cureus.3517

In this 2025 study, more than half of LLM responses cited fake sources or links.

www.cjr.org/tow_center/w...

It isn't using its sources, it's fabricating them - often outright making up citation, making up facts, etc. Because it isn't trained to produce an accurate answer, just an answer that looks human produced. If you want actual, reliable sources here are a few:

Ahh what a great game (and cause)

Should've already been cancelled tbh (also your username made me laugh so hard so thank you ❤️)