Recently, Nieman Lab conducted an experiment to test the accuracy of ChatGPT in providing correct links to articles from news publications. They found that ChatGPT does not provide accurate links, instead, it confidently generates incorrect URLs. This behaviour is referred to as "hallucinating" in the AI industry, and it's comparable to a person confidently lying or making things up.
Nieman Lab's research
Nieman Lab's Andrew Deck requested the service to provide links to high-profile, exclusive stories published by 10 publishers that OpenAI has made deals worth millions of dollars with. These publishers included the Associated Press, The Wall Street Journal, the Financial Times, The Times (UK), Le Monde, El País, The Atlantic, The Verge, Vox, and Politico. Instead of providing accurate links, ChatGPT generated fake URLs that led to 404 error pages, as these links did not exist.
This behaviour was intentional as the system was designed to predict the most likely version of a story's URL, rather than providing the correct one. A similar experiment with Business Insider yielded the same result.
OpenAI’s response
An OpenAI spokesperson told Nieman Lab that the company is developing an experience that combines conversational capabilities with their latest news content. This enhanced experience is still in development and not yet available in ChatGPT, and the spokesperson did not provide further details regarding the fake URLs.
The timeline for when this new experience will be available and its reliability remains uncertain.
When ChatGPT generates URLs, it's essentially creating false information. This is because generative AI operates by attempting to predict the next word in a sequence, similar to an advanced version of autocomplete. It doesn't actually "understand" the text it's generating, even though it may appear that way.
Meanwhile, Chinese artificial intelligence (AI) companies are rapidly trying to appeal to users of OpenAI's technology. This follows reports that the U.S. company plans to limit access in China and other countries to its application programming interface (API), a platform that allows developers to incorporate its AI models into their products. OpenAI, the maker of ChatGPT, intends to block access to its technology for building AI products for organisations in China and some other countries, as per the Chinese state-owned newspaper Securities Times on Tuesday.
ALSO READ: ChatGPT is now available for all Mac users: Here's how to access it
