綱要······················································································1 
方法論·············································································2 
主要發現··········································································2 
結論················································································4 

第一章 搭建舞臺：評估人工智能影響的方法 ································7 
測量人工智能能力和影響的已有研究 ··································· 10 
人工智能能力探索性測評的目標 ········································· 15 
本報告的結構·································································· 18 
參考文獻········································································ 18 
注釋·············································································· 20 

第二章 人類技能與人工智能能力的演變 ···································· 23 
技能供給的變化······························································· 25 
人工智能能力的最新發展 ·················································· 36 
測量人工智能能力的重要性 ··············································· 46 
參考文獻········································································ 48 
附錄2.A補充表格 ·························································· 52 
注釋·············································································· 62 

第三章基於“成人技能調查”的人工智能能力測評的方法論 ·········· 63 
“成人技能調查”概述 ························································ 64 
確定一組計算機科學家 ····················································· 70 
收集專家判斷·································································· 72 
問卷開發········································································ 74 
構建人工智能讀寫和計算能力表現的綜合度量 ······················· 76 
困難挑戰和經驗教訓 ························································ 77 
參考文獻········································································ 79 

第四章專家對人工智能讀寫和計算能力的測評 ··························· 81 
人工智能在讀寫領域的能力評估 ········································· 82 
人工智能在計算領域的能力評估 ········································· 96 
參考文獻·······································································111 
附錄4.A補充表格 ·························································112 

第五章2016—2021年人工智能讀寫和計算能力的變化 ················113 
人工智能讀寫能力隨時間推移的變化 ··································115 
人工智能計算能力隨時間推移的變化 ··································123 
參考文獻·······································································131 
附錄5.A補充圖表 ·························································131 
注釋·············································································135 

第六章人工智能能力的持續進步對就業和教育的影響 ·················137 
測評結果概要·································································138 
不斷發展的人工智能讀寫和計算能力對政策的影響 ················145 
測評人工智能的新方法 ····················································158 
參考文獻·······································································159 
注釋·············································································162 

譯後記 ················································································163

                                                                        前言
        近年來，人工智能（artificial intelligence，簡稱AI）發展迅速，給各行各業帶來了重大變革。隨著人工智能不斷演進，了解這項技術能做什麼、不能做什麼，變得愈發重要。經濟合作與發展組織（Organisation for Economic Co-operation and Development，簡稱OECD）認識到，有必要繫統地測評人工智能在掌握人類技能方面的能力，特別是掌握對人類就業和教育至關重要的技能方面的能力。這樣的測評可以幫助政策制定者和教育工作者更好地預測技術變革對勞動力的影響，使個體為應對未來的需求做好準備。
        本報告追蹤了早期試點研究，測評了最先進的人工智能在閱讀和數學方面
的能力。閱讀和數學是人類能力的兩個關鍵領域，對人們在工作和生活中眾多
領域的成功意義重大。報告不但展示了隨著時間的推移，人工智能在這些領域
的能力演變過程，而且顯示了與人類的閱讀和數學技能相比人工智能的表現情
況。對人工智能的測評是基於計算機科學家的評估。
       本研究隸屬OECD教育研究與創新中心（Centre for Educational Research and Innovation，簡稱CERI）正在開展的一個綜合項目，旨在測評人工智能的能力及其對工作和教育的影響。“人工智能與技能的未來”（the AI and Future of Skills，簡稱AIFS）項目旨在開發易於理解、全面周詳、可重復使用且對政策具有指導意義的人工智能能力度量。該項目通過使用包括專家評估在內的各種關於人工智能的信息來源，旨在為政策制定者提供決策所需知識，這些知識會影響對面向未來的教育和勞動力市場的塑造。
        報告顯示，自2016年以來，人工智能在閱讀方面的能力有了顯著提高，這反映了近年來自然語言處理（natural language processing，簡稱NLP）領域的進步。人工智能在解決數學問題方面的能力並沒有以同樣的速度提高，但專家預測，未來幾年內，人工智能研究和開發領域的投資將不斷增加，這將會推動人工智能在閱讀和數學方面取得顯著進展。
      報告還顯示，人工智能在數學和閱讀方面具備超越大部分人的潛力。這對人類就業和教育影響重大，因為在這些技能領域，未來的勞動者可能會面臨來自機器的日益激烈的競爭。這不但凸顯了提高勞動力基礎技能的必要性，而且強調了為勞動力在關鍵領域與人工智能合作做好準備的重要性。
     通過提供人工智能在人類兩個關鍵認知技能方面能力改進的示例，本研究強調了定期和繫統監測人工智能能力演變的重要性，也闡明了將人工智能的能力與人類技能進行比較的迫切性。這將幫助關注技術進步的政策制定者、教育工作者和研究人員進一步了解技術進步對人類就業和教育未來的影響。