人工智能搞翻译

agent124

The heat, fouled by exhaust fumes and unwashed by the slightest breeze, rose at her like the heat from a lift-shaft, but her Slavic features registered no complaint. She was neither dressed nor built for exertion on a hot day, being in stature very short indeed, and fat, so that she had to roll a little in order to get along. Her black dress, of ecclesiastical severity, possessed neither a waist nor any other relief except for a dash of white lace at the neck and a large metal cross, well fingered but of no intrinsic value, at the bosom. Her cracked shoes, which in walking tended outwards at the points, set a stern tattoo rattling between the shuttered houses. Her shabby bag, full since early morning, gave her a slight starboard list and told clearly that she was used to burdens. There was also fun in her, however. Her grey hair was gathered in a bun behind her, but there remained one sprightly forelock that flopped over her brow to the rhythm of her waddle.

这里引用的一大段，是为了帮助最后两句的理解：
There was also fun in her, however. Her grey hair was gathered in a bun behind her, but there remained one sprightly forelock that flopped over her brow to the rhythm of her waddle.

特别是这个however如何理解。因为只有对此理解透了，才可能有下面的译文：
不过，她身上也有一股活泼劲。她灰白的头发在脑后盘成一个发髻，一缕不听话的刘海搭在额头上，随着摇摆的步伐轻轻晃动。

将fun译成“活泼劲”，sprightly译成“不听话的”，都是人工智能gemini的手笔（其他是我自己译的），确实是大师级的水平。

试比较另外两种译文，可以看出gemini的高明之处：
1. 不过她也有一些有趣之处。她的灰发在脑后结成一个髻，蓬松的前发在她蹒跚前进时却披散在额前，（施华）
2. 然而，她也并非完全索然无趣。她的灰发在脑后绾成一个发髻，但仍留下一络活泼的刘海，随着行进的节奏在前额跃动。(李静宜）

应该说，sprightly译成“活泼”，也是很好的。不过这里因为前面用了“活泼劲”，后面如果再用活泼，就重复了。而且“不听话的”译法需要很大的胆识，更生动。至于fun，两位译者都没有深刻理解它的含义，所以只是简单地搬用词典的释义。这里的fun，对照上文，它的确切意思应该是：前面塑造出来的，是一个生活重压下的沉重形象，但这人也展现出一点生机。理解了这层意思，方能写出“活泼劲”这样的译法。

只有真正吃透了原文，深刻理解了，才能这样大胆地译。我是真心佩服。翻译最难的，就是大胆的意译，表面看和原文的形式或者词典上的意思有出入，细品却觉得丝丝入扣，十分贴切。这就是所谓的译品，或者叫翻译的艺术。这样的艺术只有对照原文，细细品味，方能欣赏。

难就难在意译的分寸很难把握。翻译水平不高的人（比如我）容易犯的毛病就是要么过于拘泥于原文形式，不敢越雷池一步，要么走得太远，而失去了忠实。关键就在于没有吃透全文，深刻理解含义，所以不敢大胆却又是准确地下笔。因此，根本原因还在于英语水平（中文水平当然也重要，不过，如果真正吃透了原文，找到合适的中文表达方式还是相对容易一些）。

总体来说，人工智能直接搞文学翻译还是不行的，译文很毛糙，生硬、不通之处颇多，还不如自己搞。但是，有些地方却又显示出翻译大家的水平。如果人工智能针对翻译，进行特别的训练，相信超过人类翻译家（比如董乐山这样的）水平，是完全可能的。

水壶转

人工智能只能机械的翻译。

丛林寻宝儿

AI只是个玩具，只能辅助做些事情目前