Newsletter No. 536

吸 引數千師生往邵逸夫堂漂走萬卷書的博群書節漂 書園才落幕兩天，同一地點在3月20日再度水泄不 通，全因被譽為「人工智能界先知」、社交平台擁有 五千萬追隨者的 李開復 博士大駕光臨，主講邵逸夫爵士傑 出訪問學人講座。 曾在蘋果、微軟、谷歌身居要職，後於北京開辦「創新工場」 的他面對滿堂逾千名師生和科技迷，繪聲繪色的講解人工智 能（AI）的快速崛起，分析中美間的AI角力，預言勢將遭科 技淘汰的工種，以及描畫人與機械和諧共融的藍圖。 李博士首先指出，在過去六十年的AI發展史上，最大突破是 「深度學習」技術的發明。這技術能吸收大量數據，並以超 人的準確度學會辨識、預測和合成。例如，當我們向深度學 習網絡輸入大量的食物圖片，網絡就能學習辨認食物，認 出是「熱狗」還是「非熱狗」。當輸入大量在公路上駕車所 攝的圖片、視頻和感應器收集的數據，網絡就能學會如人一 般駕駛。 這項科技躍進引發了李博士所稱的「人工智能四浪潮」。  「互聯網AI」始於1998年，令亞馬遜得以預測用戶的購物 意欲，臉書得以猜出用戶喜歡跟蹤的專頁。「商業AI」始於 2004年，數據成為銀行、保險公司、醫院等大機構的寶庫， 令決策更精準、效益最大化。「感知AI」始於2011年，AI自 此獲得看、聽、說的能力。2015年起，「自主AI」能對環境作 出反應，靈活移動，並能像人一般輕易操控物件。 AI的大發現時代一直由美國及其頂尖研發人員主導，但時 至今日已進入AI的實踐時代。李博士說：「在這新時代，講 求的並非誰能寫出更好的論文，而是誰能更快的運用大量 數據，應用技術。」中國的強項由此大派用場。 在這場AI競賽中，中國相較美國有數項獨特優勢。大批中 國年輕工程師進軍AI業，日以繼夜推動中國產品創新，漸漸 已趕上美國步伐。十年前，中國公司只是一味模仿美國產品 （例如百度模仿谷歌），現已發展至引領原創產品（例如阿 里巴巴旗下的科技金融公司螞蟻金服）。 中國市場龐大，賦予創業家更多用戶、更多營業額、更多投 資，最重要的是更多數據，因為數據是AI前進的最重要燃 料。「如果說數據是新石油，中國就是新的石油輸出國。」 李博士比喻道。 AI創造巨大價值的同時，人類文明也受到挑戰。「AI勢將取 代單調重複的工作。不只是藍領工作，還有許多白領工作。」  李博士預測，未來十五年內，世上將有四成工作被科技取 替，包括貨車司機、電話營銷員、保安員甚至放射科醫生。 創造型和複雜型工作，例如科學家和藝術家、經濟學家和 公司總裁等，飯碗可保，因為「AI只能優化，不能創造。」另 一樣AI無能為力的是愛。「機械人可以提醒老人吃藥或防止 The era of AI discovery has been led by the US and its visionary researchers, but now we have entered the era of AI implementation where, as Dr. Lee put it, ‘what really matters is not who can write a better scientific paper, but who can quickly build applications on a large amount of data.’ That is where China’s strengths come into play. China has a number of distinct advantages over the US in the AI race. A massive number of young Chinese engineers are marching into the AI industry. They are relentlessly driving Chinese product innovation which has been catching up with that of the US. Over the past 10 years, Chinese companies have leapfrogged from copying from the US (e.g., Baidu copied from Google) into leading brand-new innovations (e.g., the Alibaba-owned fintech company Ant Financial). China is a large market, which gives entrepreneurs more users, more revenue, more investment and, most importantly, more data which are the rocket fuel for the AI engine. ‘If data were the new oil, then China was the new OPEC,’ Dr. Lee quipped. AI is creating immense value, but it doesn’t come without challenges to humanity. ‘AI will increasingly replace repetitive jobs. Not just blue-collar work but a lot of white-collar work,’ Dr. Lee explained, predicting that 40% of jobs in the world, including truck drivers’, telesales people’s, security guards’ and even radiologists’, will become displaceable by technology over the next 15 years. The creative and the complex, such as those done by scientists and artists, economists and CEOs, will be protected, because ‘AI can optimize but not create.’ Another thing that AI cannot do is love. ‘It’s OK to build robots to remind older people to take their medicine and prevent them from falling, but robots don’t show love.’ Therefore, jobs that may not be particularly creative but require compassion, empathy and trust, such as tutors and elderly caregivers, will also be safe. At the end, Dr. Lee shared his vision of how humans should embrace the positive changes brought about by AI. ‘In the future, we will thank AI as something that comes to liberate us from having to do the routine jobs, to push us into doing things we love and are good at, and to remind us what it is that makes us human.’ After all, AI is merely a tool that we, as humans, control and run. We are the ones who have free will and it is up to us to direct the plots of the AI story. Christine N. 李開復預言 的未來 聽先知的話 Hear the Oracle Speak Lee Kai-fu forecasts the future of 他們跌倒，但無法展示愛。」因此，某些工種雖然創意不足， 但需要同情心、同理心和信任，例如家庭教師和老人看護， 同樣無需擔心被取代。 演講尾聲，李博士展望人類應樂觀迎接人工智能的正面效 應。「終有一天，我們會感激AI將我們從枯燥乏味的工作解 放出來，從事我們所愛和擅長之事，並明白生而為人的真正 意義。」畢竟，AI只是供人類控制和運用的工具。擁有自由 意志的人類才是這個AI腳本的總編導。 J ust two days after the conclusion of the ‘Book Crossing Pavilion’ held by the I·CARE Book Festival in which tens of thousands of used books were snatched up by throngs of CUHK members, the foyer of the Sir Run Run Shaw Hall was once again swarming with people who came to hear Dr. Lee Kai-fu , hailed as ‘the Oracle of AI’ and followed by over 50 million fans on social media, deliver the Sir Run Run Shaw Distinguished Visiting Scholar Lecture on 20 March. Addressing a packed house of over one thousand students, faculty and tech enthusiasts, Dr. Lee, a former Apple, Microsoft and Google executive who now chairs Sinovation Ventures, offered a visionary and compelling presentation on the rapid rise of AI, the Sino-American battle for AI supremacy, the inevitable automation of routine jobs, and a blueprint for coexistence between humans and robots. At the beginning of the lecture, Dr. Lee pointed out that throughout artificial intelligence’s 60-year history, the biggest breakthrough that happened in the field is deep learning. It is a technology that can take a huge amount of data and learn to recognize, predict and synthesize at superhuman accuracy. If we show the deep learning network a massive number of food photos, for example, it can learn to recognize food and tell ‘hot dog’ from ‘no hot dog’. Or if we show the network many pictures, videos and sensory data from driving on the highway, it can learn to drive a car as well as a human being does. Such technological leap is leading to what Dr. Lee called ‘the four waves of artificial intelligence’. Internet AI, which started in 1998, enables Amazon to predict what its users might want to buy and powers Facebook to suggest what its users might want to follow. In Business AI starting in 2004, data have become a goldmine for banks, insurance companies and hospitals to improve decision-making and maximize productivity. Perception AI started in 2011 when we gave AI the ability to see, hear and speak. In the fourth wave of Autonomous AI starting in 2015, AI is given the ability to respond to the world around it, to move intuitively, and to manipulate objects as easily as a human can. 04 # 5 3 6 | 1 9 . 0 4 . 2 0 1 9