人類仍做出85.5%的AI模型決策，尽管AI代理承擔更多工作
研究團隊發現，在AI模型開發中，人類做出了85.5%的關於方法和參數的決策，而AI代理處理了更多的任務，但人類仍然控制了目標設定和最終選擇。
a team from fudan university studied their development of atria dawn preview, a 744 billion parameter ai model. they analyzed 700+ task logs from 56 participants and found ai agents were used in 96.5% of tasks. humans made 85.5% of decisions on methods and parameters, while ai made 9.2%. humans controlled goal setting and final choices in 93.4% of cases. the study also found a third of completed ai-assisted tasks wouldn't have been attempted without ai.