Technology and Language

May 2024

The fixation of artificial intelligence is always on what it will replace, and never on what it will complement. One of its many predicted victims is the skill of being a linguist; specifically, those who work as translators and those who make the effort to learn foreign languages. AI alone will kill off the former but not the latter. 

As a paid service, translation is an entirely transactional process. It normally takes place at one-off meetings or talks. There is often no expectation of future communication between the two parties and no prospect of a genuine relationship. Perfect real-time translation (RTT) will eventually replace translators and their associated burdens. Unlike their human counterparts, AI RTT will not need to be paid dozens or hundreds of dollars per hour, it will not require the hassle of scheduling, nor will it carry the risk of last-minute cancellations. It will drive obvious efficiency gains and the human job of a professional translator will inevitably disappear, probably by the end of the decade. 

It begs the question of why, in an age of flawless translation technology, anyone would bother to spend hundreds or thousands of hours learning a foreign tongue. But people will continue to learn languages irrespective of how brilliant AI translation becomes. That is because the incentives to learn languages will remain. 

No level of RTT will ever impress people. When abroad, the shop owner, waiter, or passer-by that you converse with via RTT will duly answer back , but they won’t form a connection with you if they are speaking to a machine. As far as they know, you need not have made the slightest effort to learn a single word of their language, nor one aspect of their culture. Your grandmother who lives far away in your parents’ country of birth will not be impressed if you cannot converse with her naturally. You are unlikely to marry, or found a business, or become friends with someone if you don’t speak the same language even if you can, technically, communicate. 

Much of learning a language is not just to facilitate the immediate transfer of information, but to act as a signal on a meta level. For those in an immigrant household, speaking your mother tongue will remain a signal of commitment to one’s heritage. For those aspiring to learn a language, their linguistic ability will remain a signal of intelligence, hard work, and of cultural and social openness. The opportunities granted by these signals are extensive indeed, which is why people who learn a language to a high level will never discourage someone from doing likewise despite the required effort.  

AI language capabilities will probably see languages whittled down from school syllabi around the world. That will have less of an impact than one might think. Firstly, many people never seriously progress with a language they learn at school. If they do, it is helped by their effortless consumption of TV series and YouTube videos they watch, or video games they play (i.e. how most young people globally are familiar with English). Secondly, a reduction in people being forced to learn a language will further increase the incentives for learning of one’s own accord. In a world where no one is compulsorily taught languages, those who learn will be given even more praise for their efforts and initiative. 

Certainly AI will play a great role in language learning. It will be both an omniscient, personalised teacher and a fluent language buddy. But no amount of LLM parameters will substitute the personal commitment needed by the person to learn a new language. It is unavoidable that learning a language naturally takes significant effort. 

Beyond AI, technology will eventually threaten all language learning though. We will eventually be able to download fluency in a language onto a brain-implanted chip. Because you could not tell if a person gained fluency naturally or through downloading their linguistic ability, this advancement would remove any incentives for language learning. There would be zero advantage to learning a language naturally, with no differentiated upside but requiring enormous relative effort. The study of languages would perhaps hold out as a niche but endangered academic discipline.

The rise of implant-augmented intelligence will threaten the very concept of spoken language (which transmits information at ~40 bits/second) and give way to ‘telepathic’ wireless transmission between humans’ digital faculties (on electromagnetic signals of at least 10^8 bits/second). Our descendants will perhaps ‘speak’ in bits, not words, and in bytes, not sentences. AI will not kill language learning, but technology may kill spoken language.