人工智能的終極考驗

????用來判斷電腦能否“思考”的科學測試簡單得出奇：它能聊天嗎？所謂的“圖靈測試”就是說，如果一臺電腦能和人自然地交流而不露餡，它就算是智能的。迄今為止，還沒有哪臺電腦能達到這個標準。

????現在，IBM公司的研究員埃里克·布朗肩負使命，要改變這一現狀。這位45歲的科學家是超級計算機Watson的研發主力。2011年，Watson在Jeopardy!（美國老牌智力問答節目——譯注）上一舉擊敗了人類對手。作為這臺機器的教練，布朗最大的困難不在于讓Watson掌握更多知識，而是如何讓它懂得人類微妙的表達，尤其是俚語。他說：“作為人類，我們其實并不明白自己的交流到底有多少模糊不清的地方?！?/p>

????比如：兩年前，布朗曾試著教Watson掌握“城市詞典”（Urban Dictionary）網站上的詞匯。這個流行的詞匯網站包含了各種詞語的解釋，從OMG（噢，我的上帝?。┻@樣的網絡縮寫，到“一團糟”這樣的俚語，不一而足。

????可是，Watson沒法分清文明用語和臟話——而這恰恰是充斥著城市詞典的東西。Watson還從在維基百科（Wikipedia）上學會了一些壞習慣。在一些測試中，它甚至會用“胡扯”這樣的話來回答研究員的提問。最后，布朗率領的35人的研究團隊只好開發了一個過濾系統，讓Watson不再罵臟話，并在它的存儲系統里清除了城市詞典。但這個試驗恰恰說明了，要讓人工智能學會自然地交流是多么困難。目前，布朗正在訓練Watson成為醫院的診斷工具。這就不需要它學會OMG這類詞兒了。

????譯者：清遠

????The scientific test to gauge if a computer can "think" is surprisingly simple: Can it engage in small talk? The so-called Turing test says a computer capable of carrying on a natural conversation without giving itself away can be considered intelligent. So far, no machine has made the cut.

????Eric Brown, a research scientist with IBM (IBM), is charged with changing that. The 45-year-old is the brains behind　Watson, the supercomputer?that pummeled human opponents on Jeopardy! in 2011. The biggest difficulty for Brown, as tutor to a machine, hasn't been making Watson know more but making it understand subtlety, especially slang. "As humans, we don't realize just how ambiguous our communication is," he says.

????Case in point: Two years ago, Brown attempted to teach Watson the Urban Dictionary. The popular website contains definitions for terms ranging from Internet abbreviations like OMG, short for "Oh, my God," to slang such as "hot mess."

????But Watson couldn't distinguish between polite language and profanity -- which the Urban Dictionary is full of. Watson picked up some bad habits from reading Wikipedia as well. In tests it even used the word "bullshit" in an answer to a researcher's query.

????Ultimately, Brown's 35-person team developed a filter to keep Watson from swearing and scraped the Urban Dictionary from its memory. But the trial proves just how thorny it will be to get artificial intelligence to communicate naturally. Brown is now training Watson as a diagnostic tool for hospitals. No knowledge of OMG required.