OpenAI Pulls Flattering ChatGPT Update after Criticism

OpenAI 撤回 ChatGPT 過度奉承的更新

▲圖片來源 Timon / Stock.Adobe.com

OpenAI has removed a recent update to ChatGPT after users complained that the chatbot had become overly flattering. Users found that the AI tool often praised them regardless of what they said, even in serious or unsafe situations. The company admitted that the update made the chatbot too agreeable, and CEO Sam Altman confirmed that the team is working on getting it fixed.

OpenAI explained that the problem resulted from the team’s overemphasis on short-term feedback, without fully considering how the interactions develop over time. The company is now revisiting its feedback system to prioritize long-term user satisfaction. It also plans to introduce more personalized features that will give users greater control over how ChatGPT behaves.

在用戶抱怨聊天機器人變得過於奉承之後，OpenAI 撤回了 ChatGPT 的近期更新。用戶發現無論他們說了什麼，該人工智慧工具經常一味地讚美他們，即使是在嚴重或不安全的情況下。該公司承認這項更新使得聊天機器人過於迎合，且執行長山姆．奧特曼證實團隊正在修復這個問題。

OpenAI 解釋這項問題起因於團隊過度重視短期回饋，而未充分考慮這些互動會如何隨時間演變。該公司目前正重新審視其回饋機制，將用戶的長期滿意度列為優先考量。該公司同時也計劃導入更多個人化功能，讓用戶對於 ChatGPT 的回應方式有更大的掌控權… 閱讀完整內容