Reference News Network, February 22**Dao AFP, February 21**Dao, said that the famous generative artificial intelligence platform ChatGPT (chat generation pre-trained converter) appeared nonsense for a few hours on the 20th, giving some beginningless sentences when answering user questions. It's also a reminder that such systems are still in their infancy.
The American Open Artificial Intelligence Research Center (OpenAI), a start-up that launched ChatGPT at the end of 2022, said on its ** on the morning of the 21st that ChatGPT has returned to "normal" operation. On the afternoon of the 20th (San Francisco time), the company announced that it was "investigating the characteristics related to some unexpected answers to ChatGPT." A few minutes later, OpenAI announced that "the issue has been identified and is being resolved."
According to the report, many users put screenshots online, showing some errors or incomprehensible answers given by ChatGPT.
One user regretted on the forum: "It generated some words that didn't exist at all, omitted some words and created a short set of keywords that I couldn't understand at all." There are also a number of other anomalies. Another user complained that conversations with ChatGPT "quickly became incomprehensible for three hours." (Compiled by Lu Longjun).