双语 | 牛津大学预测“7年后机器翻译超越人类”，靠谱吗？

人工智能对翻译行业的影响究竟有多大？近日，英国牛津大学的一份研究报告指出，到2024年人工智能将比人类翻译得更好。

从2017年到2024年，短短7年的时间，译员就会面临人工智能翻译的威胁吗？

▲本文作者艾德•伯西（Ed Bussey）是内容制作公司Quill的创始人兼总裁

The latest scare story around the rise of robots is that within 120 years all human jobs will be automated. If that study from Oxford University is to be believed, we’re just 3 to 4 generations away from perpetual holiday. The report goes on to predict when AI will outperform humans and — more interestingly — how. Some aspects will be of genuine concern to certain industries: AI will be a better driver than human heavy goods vehicles drivers by 2027, AI will write better novels than we can by 2049, and, closest to today, AI will be better at translation by 2024.

最近关于机器人的兴起流传着这么一种骇人听闻的说法：120年之内，人类所有的工作都将实现自动化。如果这项牛津大学的研究真能实现的话，那么我们离永久歇业也只有三四代人的时间了。这份报告还预测了人工智能何时会比人类更好地胜任工作，以及更有意思的是，它如何能做到这一点。该报告真实考察了某些行业的某些方面：到2027年，人工智能将比人类重型货车驾驶员驾驶技术更好；到2049年，人工智能写出的小说将比我们的更加优秀；而离现在最近的目标是，到2024年人工智能将比人类翻译得更好。

AI has the potential to significantly reshape the translation sector, as it’s doing to many other industries already. However, given that the last time human translators were pitted against machine translation (in February) that 90 percent of the automated translation was judged “grammatically awkward,” that is a bold prediction. Korean is notorious for being a challenge for machine translation, of course, and the margin of victory would likely have been much smaller for more closely connected languages, such as English and Spanish.

正如其他行业一样，人工智能也可能会导致翻译行业大洗牌。然而，在上一次人机翻译大战（今年2月举行）中，90%的自动化翻译被评判为“有语法问题”【点击进入：双语 | 世界首次人机翻译大战昨日开战！结果你猜对了吗？】，基于这种情况，这个预测可以说是非常大胆了。韩国的机器翻译因此而声名狼藉，当然如果翻译的两种语言比较相近的话，比如英语和西班牙语，那么人类的胜利系数可能就会更小。

A child learning Spanish, for example, may struggle for weeks to learn the word for “spoon” before it sticks. Not so with AI. Neural network machines can even extract patterns and detect and predict trends within use of language. If 2024 is the year robots become better linguists, who knows when they’ll be able to predict what we want to say before we know ourselves, based on everything we’ve ever said before?

比如，一个学西班牙语的小孩可能要花费几周的时间才能记住“勺子”这个词的说法。而人工智能就不一样了。神经网络机器甚至能够抽取模式，识别并预测语言运用的趋势。如果2024年机器人能成为出色的语言学家，基于我们上述的种种可能，谁又会知道它们什么时候能比我们自己更先预测到我们自己想说什么？

The companies involved are some of the giants of the tech world, with Baidu, IBM, Google, and Microsoft all keen to make the breakthrough that puts automated translation above its human equivalents. As a sector, natural language processing (NLP) — the overarching umbrella term for the interaction between computers and human languages — is expected to grow from $7.63 billion last year to over $16 billion by 2021. And the pace of change is fierce. For instance, we’ve just seen the launch of an earpiece that can translate languages in more-or-less real time. Facebook has joined in too, by expanding AI-assisted recommendations into more languages. Within a few days of these announcements, Microsoft announced it’s launching a dictation app in 20 languages that could let you “ditch the keyboard” and provide real-time translation in 60 languages.

涉足这一领域的企业包括一些科技巨擘，百度、IBM、谷歌和微软均表现出极大兴趣，力图实现突破，使自动化翻译超越人工翻译。自然语言处理（NLP）是一个范围极广的概述型术语，指计算机与人类语言之间的互动，预计到2021年底，这一领域的规模将从去年的76.3亿美元（约合人民币516.6亿元）增长至160多亿美元（约合人民币1083.4亿元），其变化也可谓日新月异，例如最近刚推出的能进行接近实时翻译的耳机。Facebook也加入了进来，并在其人工智能助手中增加了更多语种，为用户提供推荐内容。这些消息发布后不久，微软便宣布将推出支持20种语言的听写应用程序，这种程序可以让你“告别键盘”，并提供60种语言的实时翻译。

So where does this leave the translation industry? From the wide array of AI solutions we’ve tested to date, machine translation still yields unacceptably poor quality content, especially for established brands that (rightly) set a very high bar for their content and brand tone of voice. But given the huge effort underway to vastly improve machine translation, it’ll likely redefine the role of humans in the translation process.

那么，这会把翻译行业置于何种境地呢？从目前尝试过的众多人工智能方案来看，机器的翻译质量依然粗劣不堪，对于著名品牌而言尤其如此，这些品牌（应当）对内容和语气要求更高。但鉴于当下人们为改善机器翻译做出的巨大努力，机器很有可能重新定义人类在翻译过程中的角色。

Instead of human translators, we’ll see humans in the loop. Machine translation will do the heavy lifting and people will edit the content. This form of augmented intelligence is likely to be prevalent long before we see standalone artificial intelligence.

我们可能不会看到人类译员，而会看到人类管理者。机器将完成翻译的主要任务，而人类将负责内容的编辑。在独立人工智能远未出现之前，这种形式的增强智能可能会盛行开来。

It’s hard to imagine robots completely replacing human translators, though. The most common questions we receive from our translators are related to cultural references, and it’s clear that an understanding of the source culture is as important as the source text. I’d question how soon AI can be taught to understand political, historical and cultural contexts in a way that would safeguard against awkward errors. Equally, it will be interesting to see how a machine contextualizes words that can’t be defined simply. For example, there’s a Czech word prozvonit, which means “to call a phone and let it ring once, to get the recipient to call you back.” Or the Indonesian word jayus, meaning “a joke told so badly it’s funny.”

虽然如此，让机器人完全取代人工译员还是难以想象的事情。译员反馈的最普遍问题是文化关联，了解源语的文化背景与理解源语本身的意思显然同等重要。不知还要多久才能训练人工智能去理解政治、历史和文化背景，以避免出现尴尬的错误。同样，让机器结合背景去分析难以简单理解的词，也是挺有意思的。例如捷克语里有一个单词prozvonit，意思是“在打电话时只让对方响一声，然后马上挂机等对方打过来”；还有一个印尼语单词jayus，意思是“讲得很烂却让人发笑的笑话”。

Away from all the bold predictions, where I do see AI making a measurable impact on the translation sector today is with things like workflow and automation.

抛开那些大胆的预测，我看到如今人工智能在翻译领域着实产生了重大影响，主要是处理工作流程和自动化之类的事情。

Translation memory is already able to facilitate faster human translation, providing translators with words and phrases that have already been translated. It’s clear that AI is increasingly underpinning many businesses, automating the repetitive, time-consuming work that few people enjoy doing.

翻译记忆库已经能够提高人工翻译的速度，为译员提供现成的词汇和短语译文。很明显人工智能正在为越来越多的业务提供支持，自动完成人们不愿意做的重复性耗时工作。

However, it’s not yet capable of the creativity, understanding, and personality that make for truly effective translation, localization, or transcreation. And there will always be difficulties with context, or when a word evolves in meaning. AI’s flaw in a field as emotive and personal as language is simply that it isn’t human. So, while we’ll see machine translation taking on the more basic parts of translation work, it’ll be many years yet before it fully replaces skilled humans.

然而人工智能依然不具备创造力、理解力和个性，通过这些才能实现真正有效的翻译、本地化或创译。语境也是经常遇到的难题，或者有些时候单词的意思会发生演变。在语言这样一个情感化和个性化的领域，人工智能的缺点就在于它不是人。因此，虽然我们看到机器翻译开始承担翻译工作中更为基础性的任务，但想要完全取代熟练的人工译员仍需时日。

文章原标题：Why AI-powered translation needs a lot of work？（为什么人工智能翻译还有很长的路要走？ ）

英文来源：VentureBeat

编译：阿狸、Janet、Lyla

推荐阅读