[{"data":1,"prerenderedAt":47},["ShallowReactive",2],{"q-fintech-22-037":3},{"exam":4,"examName":5,"chapterSlug":6,"question":7,"related":22},"fintech","金融科技力知識檢定","c4",{"id":8,"exam":4,"session":9,"qno":10,"question":11,"options":12,"answer":17,"chapter":18,"freq":19,"explanation":20,"difficulty":21},"fintech-22-037","22",37,"若以生成式 AI 協助合約審閱而需降低誤判與強化可稽核性，系統輸出應如何設計，才能讓法遵人員快速追溯依據？",[13,14,15,16],"僅輸出摘要不附來源","允許模型自行編造條款","以結構化解析（條款切片）+ 引用原文段落","不保留審閱紀錄",2,"人工智慧",1,"生成式 AI 用於合約審閱時,最大風險是誤判與「幻覺」(編造不存在的內容),因此輸出應採結構化解析:先把合約切分成一條條條款,再讓每項判斷都附上所引用的原文段落,法遵人員即可逐條回溯依據、快速驗證,也留下可稽核的軌跡,故 (C) 正確。(A) 只輸出摘要不附來源,無法查證判斷依據;(B) 允許模型自行編造條款會直接造成錯誤與法律風險;(D) 不保留審閱紀錄,事後無從稽核與究責,三者都與降低誤判、強化可稽核性的目標相反。","easy",[23,27,31,35,39,43],{"id":24,"question":25,"qno":26},"fintech-19-006","1980 年代人工智慧發展出的專家系統(Expert System)把取自專家的資訊彙編成一組規則，這些規則往往以「若...則...」結構呈現，稱為下列何種規則？",6,{"id":28,"question":29,"qno":30},"fintech-19-007","下列何種技術可以讓電腦模仿人類大腦的運算，讓電腦像人一樣思考？",7,{"id":32,"question":33,"qno":34},"fintech-19-008","有關機器學習訓練與預測的敘述，下列何者錯誤？",8,{"id":36,"question":37,"qno":38},"fintech-19-036","下列何種機器學習是所有資料都沒有標註，機器透過尋找資料的特徵，自己進行分類？",36,{"id":40,"question":41,"qno":42},"fintech-19-040","人工智慧使用運算晶片如 GPU，未來運用在特定人工智慧特定應用領域的專用晶片，稱為下列何者？",40,{"id":44,"question":45,"qno":46},"fintech-19-044","有關 2023 年推出新一代的語言模型 GPT-4 屬性之敘述，下列何者正確？",44,1793070566273]