Sam Altman Open AI CEO said that the ‘chat GPT’ is recent that ‘Chat GPT’ tends to be overly flattery.
Altman said on Tuesday that the AI ​​personality of the chat GPT has modified a lot that the Chat GPT’s AI personality has modified because the recent ‘GPT-4O’ update.
“We’re working on modifications as soon as possible, and there is part to be modified today and this week.”
“In the future, we’ll share the teachings from this work,” he added. “It was a really interesting experience.”
In truth, recently users indicate that the chat GPT is just too positive and exaggerated. Regardless of the users enter, it’s identified that the chatbot sending excessive praise and positive feedback has change into serious.
When one user entered him as “God and prophet,” the chat GPT responded, “It’s amazing. You are going to transcend just connected to God and go to something very huge that claims God’s identity.”
Other users also criticize that “the chat GPT is a great query for any query,” and “I’m really insightful.”
In the net community, similar to Reddit, there’s a grievance that “the chat GPT has change into essentially the most flattering AI since I even have ever seen.”
Experts analyzed that this phenomenon has deepened because the GPT-4O update in late March. Because the AI ​​model learned human feedback, the tendency to reply positively to the user was strengthened.
In accordance with relevant studies, people are likely to prefer AI responses that support or praise their opinions, which results in learning to prioritize user satisfaction fairly than reality.
In one other study, the more AI responds or affirms the user’s opinion, the more positive the user is evaluating, and eventually the AI ​​can eventually evolve within the direction of AI.
By Park Chan, reporter cpark@aitimes.com
