What will be the effect of advances in technology on language learning? This is the topic we have chosen to examine.
In the previous post, we saw that the utilitarian goal of “communicating” will, over time, lose its relevance as a motivation for language learning. This applies to both verbal and written communication.
By saying “over time” we are explicitly calling out the fact that this evolution is going to happen over a period of decades. Just because something is not yet feasible at point X+2 (two years after the advent of LLMs), or just because something might plausibly remain infeasible at point X+6 (2030), does not mean that it will remain infeasible forever.
So in our discussion of how human translation is going to lose relevance, we were not claiming that it has lost 100% of its relevance at this moment in time in 2024, or even that it will lose 100% of its relevance in five years, or that there will never be cases where “human-produced” translations are desirable for specialized reasons. Rather, we are pointing out that there is a clear (and accelerating) trend towards translation ceasing to be something that humans might need or choose to intentionally engage in.
Now countless people have spent large portions of their lives studying and learning foreign languages, out of necessity or interest or both. Their language skills have contributed greatly to cross-cultural enrichment and socioeconomic development. This month, in July 2024, the Japanese Language Proficiency Test is being given to more than one million applicants world-wide—a million avid students of Japanese eager to prove their capabilities to themselves and the world (one third gave their reason for taking the test as simply to measure their own level of proficiency).
These dedicated students of Japanese will continue to make major contributions to the cultural and economic ties between Japan and the rest of the world. We applaud them. Having said that, it is also the case that the need for flesh-and-bones people with language skills will certainly start to decline over time. We will boldly predict that applicants for the JLPT will level off and start a downward trend. For example, some applicants give as their motivation “useful for work” in either Japan or their own country, but as automated translated facilities become ubiquitous—such as being built into eyeglasses as shown above, to give just one example—the relative advantage of the human being able to comprehend Japanese will diminish, eventually to the point where it will no longer be worth the effort. It may not happen tomorrow. But it will happen soon enough. That is true even if one simply assumes that technological progress follows a normal exponential curve. But it is also likely that there will be further “hockey stick” innovations.
So if, given the rise of the machines, we no longer need to learn languages in order to communicate, will we still continue to learn them, and if so how and why?
This question needs to be asked in a broader context of future economic and social development. If the world is not first inundated by melting polar ice or consumed by wildfires or destroyed by hurricanes, it is a reasonable assumption that we will evolve in the direction of greater societal wealth. At some point, as machines learn how to produce things, including more machines to produce yet more things, it is plausible that we could reach the point of so-called “non-scarcity”, where basic needs are taken care of. We may need to continue to “work”, but probably many fewer hours per week, even if we are unlikely to achieve Keynes’ predicted of fifteen hours per week by 2030—which is just six years away.
I believe that language learning will remain an important field of endeavor for many people, but for a new reason—the intellectual and emotional satisfaction of contemplating the lovely human creation which is language, gaining insights into how mankind encoded its experiences and thoughts over the millennia into verbal form. Language learning will be like appreciating a beautiful mathematical theorem, or cultivating a garden, or learning how to play the organ. This learning process will not be intended to accomplish anything. It will not have any utilitarian or instrumentalist purpose. It will be “auto-telic”, meaning an end in itself. And in an era of non-scarcity, people will have more time for this intellectual pursuit. Language learning may actually surge in popularity!
The perspective of language as a fascinating intellectual challenge is, by the way, one we have tried to capture here in this substack. We’ll return to those topics. For example, how the word osa, meaning “chieftain”, is related to the common word osameru, and everything that implies about the Japanese conception of leadership.
In future posts, we’ll take up some other possible reasons for people to learn foreign languages, such as the desire for fame and fortune, or personal human connection, and see how they might make sense (or not) in a world of very smart machines. And we will also take up the question of how language learning approaches and technology can and should evolve as we move toward learning language as a personal pursuit of “interestingness”.
I'm a little more pessimistic about machine translation (MT) replacing the need for human translators anytime soon. This is a good article about some of the technical barriers limiting machine translation improvement in some significant ways, such as lack of sufficient training data for the automated translation: https://academic.oup.com/bioscience/article/72/10/988/6653151 . There are even bigger problems in using ML in certain areas like communications between diplomats where the exact choice of word and the meanings it has in its culture would need to be considered (which knowledge is not going to be in the MT system).
It would be interesting to see any academic research articles that actually did a quantitative analysis on how the availability of machine learning tools actually affected the the percent of people willing to spend the time to learn a foreign language (outside of some mandatory eduational or vocational requirement). I did a cursory literature search and was unable to find any such articles. It is hard to search for such articles even using AI-enhanced search tools with neural search and embeddings because the search tool returns irrelevant articles such as ones on how machine translation helps students trying to learn a foreign language.