今年博鰲亞洲論壇首次嘗試使用AI機器同傳技術，引發關注，但由于很多翻譯呈現得并不理想，結果在業內引起爭議。一時間，有人批評譯文漏洞百出、準確率低，也有人力挺技術革新，稱這是翻譯技術勇敢邁出的第一步。
▲博鰲亞洲論壇2018年年會上使用AI同傳技術（圖片來源：環球科技）
那么，該如何看AI同傳技術首次“替代”人工譯員呢？這是AI替代人類的“第一槍”嗎？本期“天下·譯事”為大家提供一個觀察視角↓
Over the few past years, the demand for real-time interpreting services has increased considerably. One major contributing factor for this phenomenon is the globalization of business, as it has increased the number of opportunities for international trade and opened up new markets for businesses all around the world.
在過去幾年中，市場對實時口譯服務的需求大大增加。造成這一現象的一大原因是商業全球化，因為全球化增加了國際貿易的機會，并為世界各地的企業開辟了新的市場。
In order to stay competitive and meet this increase in demand for interpreting services, developers have been working on technological solutions to meet the requirements for high-quality simultaneous interpreting, but can technology really replace human beings with regards to interpreting services?
開發人員為了保持競爭力以及滿足口譯服務需求的增長，一直在研究技術解決方案，以滿足高品質同聲傳譯的要求，但技術真的能代替人類完成口譯服務嗎？
翻譯技術的進展
Real-time translation systems are available on the market which include applications that can be installed on smartphones, computers, or other devices connected to the Internet. The words of the speaker are transcribed by a computer server, which analyzes the content and selects the closest translation from a vast collection of phrase pairs in its database.
市面上已經有各種實時翻譯系統，其中一些應用程序可以安裝在智能電話、計算機或能夠連網的其它設備上。講話者的語句由計算機服務器轉錄，服務器對內容進行分析，并從其數據庫中的大量短語對中選擇最接近的翻譯。
There are already a few machine interpreting solutions on the market today. Despite their growing popularity, apps and services such as these have received criticism for their inability to accurately convey the meaning of what is being said. Humans often use context to determine the meaning of words, and consider how individual words interact with each other.
目前市場上已經有一些機器口譯解決方案。盡管這些應用程序和服務的使用范圍越來越廣，但由于其無法準確傳達講話內容的含義而遭到批評。人類經常根據語境來確定單詞的含義，并思考每個單詞之間是如何互相關聯的。
拿谷歌翻譯的一段法庭口譯內容舉例（西班牙語譯英語）：
Witness: Se me borró dónde quedó ese maldito cohete, pero si quieren saber, el pinche lana quedó en la cajuela del mueble.
Google Translate: I erased me where you got the damn rocket, but if you want to know, click the wool was in the trunk of furniture.（直譯為：我擦除了我你在拿到該死的火箭的地方，但如果你想知道的話，點擊羊毛就在家具的后備箱里。）
A real interpreter would have said something like: “I can’t remember where that damn piece ended up, but if you want to know, the damn dough ended up in the car trunk.”
人類口譯員可能會翻譯為：“我想不起來那該死的槍在哪兒，但是錢是在汽車后備箱里�！�
As you can see, communication broke down quickly once the witness used a little slang. The witness gave a perfectly coherent answer (and this type of answer is entirely common in court testimony), but Google’s rendering garbled it beyond recognition.
可以看到，證人稍微說點俚語，交流就無法進行下去了。證人的回答條理十分清晰（這種回答在法庭證詞中很常見），但谷歌的翻譯結果卻將其打亂，完全變了樣。
One problem is that Google cannot consistently give a dependable translation; a potentially greater problem is that it takes an interpreter to parse out which renditions are correct.
一方面谷歌無法始終給出可靠的譯文，另一方面或許更為關鍵，依然需要口譯員從語法上分析哪些譯文是正確的。
However, these combinations are in constant change owing to evolving human creativity.
不過，隨著人類創造力的演進，人機結合的情形也在不斷變化著。
所以說，技術能取代人工口譯員嗎？
Nevertheless, with all these technological advances, interpreter jobs will inevitably evolve, just as they have already over the years. The Nuremberg Trials are generally considered to have been the event that changed interpretation forever. Before The Nuremberg Trials, any kind of interpretation was done consecutively—talk first, and then wait for the interpreter to translate. In 1945, for the first time, interpretations were performed consecutively using a system of microphones and headsets to transmit the cacophony of languages.
各種技術在進步，口譯職業的提升在所難免，多年以來一直如此。一般認為紐倫堡審判長久改變了口譯行業，這場審判之前，所有口譯都是交傳形式，即講者先發言，然后等口譯員進行翻譯。1945年首次采用麥克風和耳麥進行同傳，盡管音質不佳。
In 2008, Livescribe launched its first “smart pen,” which featured an infrared camera just below the writing tip to record the pen’s movements, and a built-in microphone to pick up ambient sound. Handwritten notes are then synchronised with the sound recordings using a digital time signature for playback on demand. The Smartpen offers a ‘safety net’ of sound recording in consecutive interpretation settings where accuracy is key, such as healthcare and the justice system.
2008年Livescribe發布了首款“智能筆”，筆端下方的紅外攝像機記錄筆的動作，還有內置麥克風捕捉環境音，通過數字時值實現手寫文字與聲音記錄的同步，供需要時回放。智能筆為同傳的聲音記錄提供了安全保障，對同傳來說，準確是關鍵，如醫療和司法領域。
From a didactic standpoint, the decisions and ethical dilemmas interpreters face on a daily basis are countless and the potential for disagreement regarding those decisions is great. Technology Mediated Dispute Resolution (TMDR) processes can be particularly useful when misunderstandings and conflicts arise. It’s also thanks to tech that all work is documented and thus available for follow-up and review.
從教育的角度來看，口譯員每天面臨的決策和倫理困境是數不勝數的，在這些決策上存在分歧的可能性很大。當出現誤解和沖突時，技術介導的爭端解決進程特別有用。這也要歸功于技術，讓所有工作都被記錄下來，可供跟進和審查。
From our point of view, we believe that technology-assisted interpreting is more and more welcome. In its simplest application, smartphones, tablets, and online dictionaries are being put to good use, described by some as an “infallible information assistant” if personal knowledge comes up short. However, it should not be relied on completely, and human interpreters will always be required for certain nuances that may be almost impossible for even the most cutting edge technology to detect.
從我們的角度來看，我們認為技術輔助口譯越來越受歡迎。在其最簡單的應用中，智能手機、平板電腦和在線詞典正被充分利用，在知識欠缺的情況下，有人把它描述為“可靠的信息助手”。然而，人們不應該完全依賴這項技術，對于某些即使是最尖端的技術也幾乎不可能檢測到的細微差別來說，人類口譯員永遠是必需的。
Leading providers adopt technological solutions when the time is right in order to gain a competitive advantage. Of course, machine interpretation is still in its infancy, and who knows what the next wave of innovation will bring. For now, though, I think we can safely say that human interpreters are irreplaceable.
行業領導者在時機合適時采用技術解決方案，以獲得競爭優勢。當然，機器翻譯仍處于起步階段，誰知道下一波創新潮將會帶來什么？但現在，我想我們可以有把握地說，人類口譯員是不可替代的。
