A recent BBC test of generative AI tools, including ChatGPT and others, showed they struggle with factual accuracy regarding current events. Out of 100 questions posed using BBC articles, over half of the answers were wrong, with 19% containing outright incorrect facts. Additionally, 13% of quoted material was altered. Examples include incorrect assertions about the UK's Prime Minister and the death date of a public figure. These findings underscore ongoing reliability issues in AI-generated responses.
More than half of the answers (51%) provided by generative AI tools during the BBC test were incorrect, revealing their struggles with current events.
Only 19% of the responses were based on accurate information, signaling significant reliability issues in the generative AI's grasp of factual data.
Collection
[
|
...
]