Newsletter No. 478

478 • 19.5.2016 7 宣布事項 Announcements 新任校董會主席 New Council Chairman 經大學校董會提名，大學監督委任校董梁乃鵬博士為 大學校董會主席，任期三年，由2016年5月1日起生效。 On the nomination of the University Council, the Chancellor of the University has appointed Dr. Leung Nai-pang Norman, a member of the Council, as the Chairman of the University Council for a term of three years, with effect from 1 May 2016. 續任校董 Reappointed Council Member 經校長推薦，並依據《香港中文大學條例》規程11第1(m)段、第4段及第5段規定，再度選舉 梁定邦博士出任大學校董，任期三年，由2016年5月29日（即其現屆任期屆滿後）起生效。 Dr. Anthony Neoh has been re-elected, in accordance with Statute 11.1(m), 11.4 and 11.5 of The Chinese University of Hong Kong Ordinance, as a member of the Council for a period of three years from 29 May 2016, upon the expiry of his current period of membership, as recommended by the Vice-Chancellor. 2016教學人員年度評審 2016 Annual Staff Review Exercise for Teaching Staff 有關教學人員「晉升/教授級薪金組別提升」，以及「退休/屆退休年齡後延任」的年度評審 現已展開。詳情可參考人事處通函編號（GC03/2016），或參閱人事處網頁「員工資訊」一 欄下之「人事政策通告」 ( www2.per.cuhk.edu.hk/ )。 The reviews on the ‘advancement/crossing of pay bands’, and the ‘retirement/extension of service beyond statutory retirement date’ for teaching staff have commenced. For details, please refer to the general circular (GC03/2016) of the Personnel Office. It is also available at the Personnel Office website ( www2.per.cuhk.edu.hk/ ). Please visit ‘Personnel Announcement’ under ‘Staff Area’. 諾貝爾獎得主講座 Lectures by Nobel Laureates 2001年諾貝爾經濟學獎得主約瑟夫 ‧ 斯蒂格利茨教授及2013年諾貝爾醫學獎得主蘭迪 ‧ 謝克曼教授應大學邀請， 於5月26日蒞校主持講座，詳情如下： Prof. Joseph E. Stiglitz, 2001 Nobel laureate in Economic Sciences, and Prof. Randy W. Schekman, 2013 Nobel laureate in Physiology or Medicine, will visit the University on 26 May and deliver the following lectures: 講者Speakers 約瑟夫 ‧ 斯蒂格利茨教授 Prof. Joseph E. Stiglitz 蘭迪 ‧ 謝克曼教授 Prof. Randy W. Schekman 講題Topics 全球財富不均 Global Inequality 細胞分泌蛋白質的包裝和運輸基理 How Cells Package and Traffic Proteins for Export 日期Date 26.5.2016 (星期四 Thursday) 時間Time 5:00 pm 地點Venue 李兆基樓高層地下五號演講廳 Lecture Theatre 5, UG/F, Lee Shau Kee Building 利黃瑤璧樓地下冼為堅演講廳 Sin Wai Kin Hall, G/F, Esther Lee Building 查詢Enquiries 3943 8677 網上登記 Online Registration www.cuhk.edu.hk/cpr/lectures/ stiglitz www.cuhk.edu.hk/cpr/lectures/ schekman 人工智能是近期炒得火熱的一個話題。不論我們如何定義 這個詞語，最近發生的幾件事例委實值得我們深究。今年 2月，谷歌在路面測試自行駕駛車輛時發生了一宗交通意 外，報告指出是谷歌汽車企圖切線時，與另一輛有人駕駛的 巴士發生輕微碰撞。谷歌承認了責任，但亦指出當時自行汽 車正在緩慢地切線繞過路上一些障礙，程式判斷巴士司機 應會為此禮讓，結果當然沒有。 今年3月，微軟推出Tay網上人工智能聊天機器人(chatterbot)， 服務僅運行了16小時後被腰斬，原因是Tay在處理過數萬 條與人交談的信息後，被惡作劇者灌輸了錯誤不當的觀念， 而成為滿口胡言，充滿種族歧視、下流低俗等意識的聊天 者。4月，內地魏則西事件更是震驚全國，引發人們反思儘 管是來自百度等大型網企信息的可靠性和搜索引擎背後的 經濟模式是否合理。 以上事件表面上沒有關連，但它們却隱藏一個可怕的共同 危機。不久將來，為人類提供個人化服務、懂得與人類溝 通、交流情感、照顧獨居者、協助人類學習等的機器人將會 大行其道，人類的生活也會越來越依賴人工智能。無可避免 地，這些機器人和系統，必須載入各種知識、邏輯、是非觀 念，以至道德倫理、價值取向等。所載入的資料，不論是由 所謂專家制定或從普羅大眾採集回來，都存在一定風險。 在回答「人工智能可以信賴嗎？」這個終極問題前，我們不 妨先問問自己：人類世界存在信任嗎？我們是如何在人類世 界建立互信的？這些元素存在於機器世界嗎？是誰在賦予 機器這些元素？這些人又可靠嗎？ 人工智能，作繭作孽？ AI, Blessing or Curse? AI has been very much in the news lately. Regardless of one’s personal view of AI, one cannot pretend that it’s not coming and that it’s not going to stay. In February, an accident happened when Google was road testing its autonomous car. The Google car was changing lane when it collided with an oncoming bus that was driven by a human driver. Google accepted responsibility, but pointed out that at the material time its autonomous car was going round some obstacles on the road, while its pilot system predicted that the bus driver would give it right of way. For reasons incomprehensible to the system, the bus driver didn’t. In March, Microsoft launched Tay, a chatterbot that keeps real human beings in conversation. The service was suspended only after 16 hours in operation because after exchanging tens of thousands of tweet messages with human beings, Tay picked up quite a few infelicities. The system started chatting gibberish that borders on hate crime and vulgarism. In April, the whole nation was shocked by what happened to Wei Zexi. People began to question the reliability of information found on large online platforms such as Baido and the economics of the search engine. These seemingly unrelated incidents are in fact united by one unsettling thought. In future, when interactions between humans and robots increase (robots providing personalized services to human beings such as communicating and empathizing with them, taking care of the elderly and assisting children to learn), so does our dependence on them. Inevitably, all sorts of knowledge, logic, ethics, even morality and values have to be uploaded to these robots or systems. Whether these uploads are done by experts or culled from ordinary folks, there are accompanying risks. Before we answer the ultimate question ‘Is AI trustworthy?’, we should ask ourselves the following questions: ‘Is there trust in the human world? How do we establish mutual trust? Do the necessary and sufficient conditions of trust exist in the machine world? How could these conditions be coded in machines? Can we trust the humans who do the coding?’ 字裏高科 Tech Talks 梁光漢 Philip Leung