Did AI researchers let AI hallucinations into scientific papers?
Investigating a claim that 100 made-up claims were found in AI papers.
AI can make mistakes – and AI chatbots like ChatGPT warn you about that whenever you ask them anything.
These mistakes sometimes involve making up entirely fictitious, factually false statements known as “hallucinations”.
Whether these hallucinations matter depends on what you’re using AI for, and whether they are spotted and corrected.
The team on More or Less were slightly surprised to read a headline in Fortune magazine, claiming that a top academic AI conference accepted research papers which contained 100 AI-hallucinated citations.
You might think that the top AI researchers in the world would be careful about using AI to write their research papers.
Alex Cui, CTO and co-founder of GPTZero – whose company discovered the hallucinations – explains what’s going on.
CREDITS:
Presenter and producer: Tom Colls
Sound mix: James Beard
Production co-ordinator: Brenda Brown
Editor: Richard Vadon
Last on
Broadcasts
- Sat 21 Feb 202605:50GMTBBC World Service except Australasia, East and Southern Africa, East Asia & South Asia
- Sun 22 Feb 202605:50GMTBBC World Service East and Southern Africa
- Sun 22 Feb 202609:50GMTBBC World Service Europe and the Middle East & West and Central Africa only
- Sun 22 Feb 202611:50GMTBBC World Service except East and Southern Africa, Europe and the Middle East & West and Central Africa
Download this programme
Subscribe to this programme or download individual episodes.
