Editing Chatbot (section)

== Limitations of chatbots ==
{{expand section|date=December 2024}}
Chatbots have difficulty managing non-linear conversations that must go back and forth on a topic with a user.<ref>{{citation |last1=Grudin |first1=Jonathan |title=Proceedings of the 2019 CHI Conference on Human Factors in Computing Systems – CHI '19 |pages=209–219 |year=2019 |chapter=Chatbots, Humbots, and the Quest for Artificial General Intelligence |publisher=ACM CHI 2020 |doi=10.1145/3290605.3300439 |isbn=978-1-4503-5970-2 |s2cid=140274744 |last2=Jacques |first2=Richard |author-link=Jonathan Grudin}}</ref>

[[Large language model]]s are more versatile, but require a large amount of conversational data to train. These models generate new responses word by word based on user input, are usually trained on a large dataset of natural-language phrases.<ref name="Caldarini-20223" /> They sometimes provide plausible-sounding but incorrect or nonsensical answers. They can make up names, dates, historical events, and even simple math problems.<ref>{{Cite journal |last=Stover |first=Dawn |date=2023-09-03 |title=Will AI make us crazy? |url=https://www.tandfonline.com/doi/full/10.1080/00963402.2023.2245247 |journal=Bulletin of the Atomic Scientists |language=en |volume=79 |issue=5 |pages=299–303 |doi=10.1080/00963402.2023.2245247 |bibcode=2023BuAtS..79e.299S |issn=0096-3402}}</ref> When large language models produce coherent-sounding but inaccurate or fabricated content, this is referred to as "[[Hallucination (artificial intelligence)|hallucinations]]". When humans use and apply chatbot content contaminated with hallucinations, this results in "botshit".<ref>{{Cite journal |last1=Hannigan |first1=Timothy R. |last2=McCarthy |first2=Ian P. |last3=Spicer |first3=André |date=2024-03-20 |title=Beware of botshit: How to manage the epistemic risks of generative chatbots |url=https://www.sciencedirect.com/science/article/pii/S0007681324000272 |journal=Business Horizons |volume=67 |issue=5 |pages=471–486 |doi=10.1016/j.bushor.2024.03.001 |issn=0007-6813}}</ref> Given the increasing adoption and use of chatbots for generating content, there are concerns that this technology will significantly reduce the cost it takes humans to generate [[misinformation]].<ref>{{Cite news |date=2023-01-06 |title=Transcript: Ezra Klein Interviews Gary Marcus |url=https://www.nytimes.com/2023/01/06/podcasts/transcript-ezra-klein-interviews-gary-marcus.html |access-date=2024-04-21 |work=The New York Times |language=en-US |issn=0362-4331}}</ref>