GPT chat does not work. Thousands of applications from users

Chat GPT, chatbot opracowany przez OpenAI

On Thursday, January 23, the popular GPT chat stopped working. The failure was reported by thousands of concerned Internet users.

The website providing online access to the popular GPT hut on Thursday, January 23, stopped working for a moment. The failure took place around 1pm of Polish time, which is clearly seen on the chart of applications on DownDector.com. Interestingly, at the same time Open AI websites still operated, which is responsible for GPT chat.

At the same time, problems with other Open AI services were also reported. GPT-4O and GPT-4O Mini models did not work. Some users pointed out that they do not open chatgpt.com and chat.com. Others noticed that GPT chat does not answer their questions. Applications created to support this model did not react.

This would not be the first GPT Chata failure. In the last few weeks there have been short breaks in the operation of the service. The largest took place in December, when there was a serious failure in the United States, also resulting in the mistakes of other Open AI services.

The great success of the GPT chat and later problems

Started prototype on November 30, 2020, Chatbot quickly gained great popularity and became a useful tool. Thanks to the huge data sets, he was able not only to talk, but also give advice from various areas of knowledge. After 5 days he had a million users, and after 2 months of operation he reached a ceiling of 100 million.

Over time, however, scientists noticed that GPT chat is “stupid”. In time, GPT-4 answered the questions asked worse and worse. For example, during the start of the tests in March 2023, ChatgPT recognized the first numbers with the accuracy of 97.6 percent. In June 2024, he failed the entire test – he recognized only 2.4 percent correctly. given numbers.

The authors of the study had no idea why Chatbot worsened its results so much. At X, it was suggested that this could be an issue of OpenAI strategy. Answers in the GPT-4 model are given much faster than in previous versions. Did the company choose speed beyond accuracy? We are currently not sure.

Similar Posts