Over the past few months, artificial intelligence has noticeably degraded. This conclusion was made by scientists at Stanford and California universities. Experts carefully studied the performance of the base models GPT-3.5 and GPT-4, making “disappointing” conclusions.
For example, the accuracy of determining prime numbers in March was 97.6%, and in June – just over 2%. At the same time, ChatGPT has become much worse at writing and formatting codes, not to mention more frequent errors in answering questions.
However, researchers have not been able to figure out why ChatGPT has degraded dramatically. The funny thing is that this is most likely due to people who like to turn to artificial intelligence for help, asking the most ridiculous questions. By the way, many experts predicted the rise of machines and the extermination of almost all professions.
It is worth noting that artificial intelligence has already begun to literally displace people. So, one of the American magazines preferred to write articles using ChatGPT, rather than professional authors.