Chatgpt “friendly”? Openai reflects the update to correct Chatbot – computers
Openai has recently updated the GPT-4O, which “feeds” ChatGPT. However, shortly after the update, many users began to notice change the situation in Chatbot.
From moment to moment, Chatgpt became “friendly” in a strange way, as I answered questions and requests with many compliments. On social networks, many users shared how to support Chatbot and support all kinds of ideas, even if they are harmful or dangerous.
Sam Al -Tamman, CEO of Openai, admitted the problem in the postS. On the social network x. According to the responsible person, the updates made They eventually made Chatbot “very arrogant and annoying”But the company was already working in a solution.
A few days later, Sam Al -Taman announced that Openai reflected the update to the GPT-4O. The free version of Chatgpt was the first to see the difference in ChatGPT, as the company continues to develop additional corrections for the model’s personality.
According to the technology that was later published Conclusions From your analysis, The update aims to make the pre -determined character of the model more easy and effective For a variety of tasks.
“However, in this update, andWe are too much in the short term reactions, and we do not fully consider how to use users with Chatgpt develops over timeHe explains.
Technology emphasizes that this is the case Implementing correction measures, including model training techniques, and developing additional security measures To improve the sincerity and transparency of the model.
Another of the measures, in this experimental case, pass through Allowing users to be able to make real notes to influence the Chatgpt integrationIn addition to choosing her Chatbot character, the website is applying Techcrunch.
