扎克伯格的鋼鐵俠之夢能否成真

Technology leaders such as Elon Musk of Tesla Motors and venture capitalist Peter Thiel have warned that rogue AI, or artificial intelligence, could one day threaten the human race.

科技行業領袖，如特斯拉汽車公司(Tesla Motors)的埃隆氠斯克(Elon Musk)和風險投資家彼得泰爾(Peter Thiel)，都曾警告稱，失控的人工智能(AI)有一天可能威脅人類。

But Mark Zuckerberg has set himself the challenge this year of building his own personal AI assistant, describing fears of a super-intelligence running amok as “far-fetched”.

但是，馬克丠克伯格(Mark Zuckerberg)今年設定的個人挑戰恰恰是構建自己的AI個人助理，他表示，對超級智能失控的恐懼是“缺乏依據的”。

In a post on his company’s website on Sunday, the Facebook founder said he had decided to build a smart machine to control functions in his house and help him work, akin to a disembodied butler. He compared it with Jarvis, an intelligent computer in the Iron Man films.

這位Facebook創始人週日在Facbook上發帖稱，他已決定造一臺智能機器來控制自家住宅的各項功能，並且幫助他工作，類似於一個無形的管家。他將其與《鋼鐵俠》(Iron Man)電影中的智能電腦賈維斯(Jarvis)相提並論。

Mr Zuckerberg regularly sets himself personal challenges at the start of the year. Past ones have included meeting a new person every day, reading two books a month and learning Mandarin.

扎克伯格每年初都爲自己設定個人挑戰。往年的挑戰包括每天結識一個新人，每個月讀兩本書，以及學習中文普通話。

But this year’s resolution is the first to involve a software coding test with such a direct bearing on Facebook’s future. The company already has two teams working on artificial intelligence — one that focuses on pure research and another that looks at ways to apply AI to the social network.

但是，他今年的新年決心首次涉及一場軟件編碼測試，直接關係到Facebook的未來。該公司已有兩個團隊研究人工智能：一個專注於純理論研究，另一個研究如何將AI應用於社交網絡。

The challenge also highlights the AI race that has broken out among the leading internet companies as they try to invent new ways for people to interact with machines. Facebook has set up three artificial intelligence labs, in Silicon Valley, New York and Paris, and its high-profile hires in the field include Yann LeCun, a New York University professor who was one of the pioneers in deep learning — a form of AI that tries to mimic the functioning of layers of neurons in the human brain.

這項挑戰也凸顯出，領先的互聯網公司之間已經展開一場AI競賽，試圖發明人與機器互動的新方式。Facebook已經在硅谷、紐約和巴黎建立了三個人工智能實驗室，其在該領域挖來的知名人才包括紐約大學(New York University)教授、深度學習的先驅之一揚勒坎(Yann LeCun)；深度學習是人工智能的一種形式，試圖模仿人類大腦神經元層的功能。

Mr Zuckerberg said his work at Facebook often involved getting into deep technical issues with the company’s engineers, and that building his own intelligent machine would bring a different perspective to that work.

扎克伯格表示，他在Facebook的工作往往涉及與公司的工程技術人員深入探討技術問題，構建自己的智能機器將爲這方面的工作帶來一個不同的視角。

He described his personal project as “a simple AI”, taking an off-the-shelf system and teaching it to recognise his voice to “control everything in our home — music, lights, temperature and so on”.

他把自己的個人項目稱爲“一個簡單的AI”，涉及採用一臺現成系統，然後教會它聽懂他的語音，以便“控制我們家中的一切——音樂、燈光、溫度等等”。

He also said he was interested in using voice and face recognition so the machine could identify who was in a room, for instance adjusting the temperature to reflect the fact that he prefers colder rooms compared with his wife. Mr Zuckerberg added that he intended to teach the system to turn data into visual representations that he could see with a virtual reality headset.

他還表示，他有興趣使用語音和臉部識別，使機器能夠識別是誰在房間裏，然後據此發揮功能，例如調節室內溫度，顧及他與太太相比偏愛較低室溫的事實。扎克伯格補充說，他打算教會該系統把數據轉化爲可視形式，讓他能夠用一臺虛擬現實頭盔觀看。

Hermann Hauser, a Cambridge-based computer scientist and entrepreneur, said that recent advances in machine learning algorithms, computing power and availability of huge amounts of data, have combined to allow computer engineers to create AI systems.

劍橋計算機科學家和企業家赫爾曼樠澤(Hermann Hauser)表示，機器學習算法、電腦運算能力和海量數據可用性方面的最新進展相結合，讓電腦工程師能創建各種AI系統。

“[Mr Zuckerberg’s] is a perfectly realistic aim,” he said. But Dr Hauser was concerned that only Silicon Valley’s largest companies could access the level of information necessary to create functioning AI.

“（扎克伯格的目標）是一個完全可實現的目標，”他說。但豪澤博士擔憂，只有硅谷大牌企業才能獲取創建實用AI所需的那種水平的信息。