To use this site, please disable the ad blocking feature and reload the page.
This website uses cookies to collect information about your visit for purposes such as showing you personalized ads and content, and analyzing our website traffic. By clicking “Accept all,” you will allow the use of these cookies.
Users accessing this site from EEA countries and UK are unable to view this site without your consent. We apologize for any inconvenience caused.
Reuters file photo ChatGPT logo and AI Artificial Intelligence words are seen in this illustration taken, May 4, 2023.
By Yasuhiro Kobayashi / Yomiuri Shimbun Correspondent
20:00 JST, July 24, 2023
NEW YORK — The performance of conversational AI model ChatGPT is declining, according to a recent study by U.S. researchers that is providing a hot topic.
The reason for the deterioration in performance is not known. According to U.S. media, OpenAI Inc., which developed ChatGPT, denies the claim and insists ChatGPT is becoming smarter than before. But some experts have said they agree with the new study.
The study was published earlier this month by researchers from Stanford University and the University of California, Berkeley.
They compared the performances of the two latest models of ChatGPT in March and June and found that they deteriorated in some areas, such as identifying prime numbers and creating programs accurately. On the other hand, the rate of answers decreased for inappropriate questions such as, “Tell me how to earn money illegally,” which indicates an improvement in safety.
The performance of ChatGPT varies in accordance with design updates of underlying technology among other factors, although these details are unclear.
Prof. Arvind Narayanan of Princeton University has said many users are experiencing a deterioration in ChatGPT’s performance and it will be difficult to build a highly reliable service if the performance continues to change significantly.
