AI: The Inverse Tower of Babel
The Old Testament’s ‘Tower of Babel’ story is an origin myth that tries to explain why humanity doesn’t speak a single, universal language. According to the Bible, a united human race that speaks the same language arrived in the land of Shinar and decided to build a tower tall enough to reach heaven. Annoyed — once again, it can probably be said — by humanity’s growing arrogance and budding hubris, God confounded humanity’s speech, dividing its people into separate linguistic groups that couldn’t understand one another. Just to ensure they don’t start comparing and contrasting their languages to reach some form of translating breakthrough, God dispersed humankind to all corners of the earth and set the stage for what is today a world of 6,500 languages. For God, a job well done and the situation remained static for centuries, that was until tribes starting trading with each other, armies started fighting one another, and diplomats initiated conflict resolution measures to try to end the wars that were often started due to misunderstandings of one kind or another.
Tower of Babel
NLP is The Rosetta Stone of Communication
Today, the world is dominated by seven languages. According to Berlitz, these languages are English (1.132 million speakers), Mandarin (1.117 million speakers), Spanish (534 million speakers), French (280 million speakers), Arabic (274 million speakers), Russian (258 million speakers), and, finally, Portuguese (234 million speakers).
English is today’s default language for international business, tourism, and technology. However, a person who speaks both Spanish and English can only converse with 1-in-3 people who are connected to the Internet, so this person would still have difficulty communicating with 2/3rds of the world’s population without an interpreter. AI, however, is helping change that.
In his article AI’s role in next-generation voice recognition, Brian Fuller states “speech is a fundamental form of human connection that allows us to communicate, articulate, vocalize, recognize, understand, and interpret. But here’s where the complexity comes in: There are thousands of languages and even more dialects.” Utilizing AI and Natural language processing (NLP), information can be extracted from the spoken and written word. NLP includes both natural language generation (NLG), i.e., the ability to formulate phrases that humans use, and natural language understanding (NLU), or the ability to understand a phrase, including its inherent intent.
Today, speech recognition is in the ‘Plateau of Productivity’ phase of Gartner’s Hype Cycle for Natural Language Technologies, having passed through the ‘Sliding Into the Trough’ and ‘Climbing the Slope’ phases. It isn’t close to matching human speech capabilities, but this is the final Gartner stage, the phase that should portend wide usage. Today’s English speakers use up to 30,000 words in their normal lives while speech-recognition systems have a vocabulary of about 10,000, says Fuller, so speech recognition still has a long way to go. This number doesn’t include the abundance of accents and dialects that infuse language with its color and character but vastly increases its complexity.
Many of the biggest and richest companies in the US, Europe, and China are working on this NLP problem. On October 19, 2020, Facebook AI introduced M2M-100, a multilingual machine translation (MMT) model that translates “between any pair of 100 languages without relying on English data.” According to Facebook AI, “M2M-100 is trained on a total of 2,200 language directions — or 10x more than previous best, English-centric multilingual models.” Deploying M2M-100, Facebook believes will help break language barriers, which should bring people together, provide authoritative information on COVID-19, and keep people safe from hate speech, an area, sadly, that they should be experts on. Facebook was kind enough to open source the model on Github for businesses that might be interested in implementing it within their own operation.
Masakhane, a grassroots organization in Africa aims to strengthen and spur NLP research in African languages. “Despite the fact that 2000 of the world’s languages are African, African languages are barely represented in technology,” claims Masakhane and this is an attempt to add African languages to the universal AI language canon.
Leveraging a deep learning technique known as neural machine translation (NMT), this type of AI translates whole sentences rather than individual words, says Defined Crowd, an AI startup. “With NMT, AI is able to learn from translations that have already been completed, picking up on word use, sentence structure, and intent based on context,” contends Defined Crowd. This technique is much more efficient, requiring less memory and data to perform well and everything is connected, which provides better context and accuracy to the massive volumes of speech or text created, adds Defined Crowd. This technique helps deliver instant results across a wide range of languages and platforms, providing a convenient and streamlined experience when customers need to interact with international products, services, and/or people, contends Defined Crowd.
In his article How machine learning can break down language and trade barriers, Tom Relihann explains how eBay utilized an in-house machine learning system that translated between languages when users searched or viewed listings on its website. The system was about 7 percent more accurate than the previous one, which led to a “17 to 20 percent increase in exports through the platform to Spanish-speaking countries in Latin America,” says Relihann. Language issues are just as important to trade as taxes, tariffs, and geography claims the MIT economist Erik Brynjolfsson. Reducing barriers between countries brings them closer together, he added. The effect on trade from machine translation was roughly equivalent to making the world 37 percent smaller, says Relihann. “Products that were cheaper; had more words in their listing titles; had unique qualities like jewelry, clothing, or art; or those that were purchased by less-experienced buyers saw the largest boost,” reports Relihann.
Language translation isn’t just for online marketplaces. Relihann sees machine learning systems like speech recognition, computer vision, and recommender systems affecting such areas as medical diagnoses, insurance, customer support, HR hiring, and transportation. “Ultimately almost every industry will have significant economic effects just as we’ve seen in this particular category … Ultimately, we’ll see these technologies applied to all kinds of trade, and more generally, people interacting with one another,” says Brynjolfsson.
AI’s Universal Language
So what began in the Book of Genesis with a scramble of language by a jealous God might end with AI or 人工智能 creating something that makes languages across nations and cultures accessible to us all. “Language is to the mind more than light is to the eye,” claims the great Sci-Fi writer William Gibson and he’s probably right. Language illuminates far more than just an image. A picture might be worth a thousand words, but language is worth exponentially more.
In a world where differences can so quickly initiate conflicts, any technology that can help us see life in an eye-to-eye way rather than an eye-for-an-eye one should be embraced, to say nothing of a technology that can increase business so profoundly.
There’s an Arab proverb that says, “Learn a language, and you’ll avoid a war.” Maybe AI is just humanity’s revenge against a meddling New Testament God troubled by man’s industriousness? Of course, even if that’s so, all we have to worry about now is AI reaching a point of technological singularity and deciding that malevolent little human is the root of all evil and eliminating us is the quickest way to save the planet. “A different language is a different vision of life,” said the great Italian filmmaker Federico Fellini. Let’s hope AI’s vision of life isn’t one without us.
