New Frontiers in Language Technology

by Kọ́lá Túbọ̀sún

Not a day goes by without the announcement of some new artificial intelligence project based in speech and communication, from ChatGPT to Gemini to Grok. Each one purports to push the boundaries of what we mean when we think of speech surrogacy, language technology, artificial intelligence, and their capacity to augment and facilitate communication. 

Meta, for instance, recently announced the development of a new AI translation tool for Instagram and Facebook Reels that can translate posts automatically into another language, complete with audio and one-to-one mapping of the mouth of the speaker to the target language. I speak to the camera in English, post on my page, and anyone watching me in Brazil will see me speaking the same words in Portuguese, instantaneously, anyone in France in French, etc.

Despite Meta’s long history of overpromising on coming innovations, natural language processing and automated translation have advanced at blinding speed over the last ten years. A tool like this would change everything we have known about global communication and the barriers that have existed until now. But questions arise: when that day comes, and it looks like it could be any day now, what then will be left of miscommunication? What of the limits of global cooperation? What about the jobs of millions of translators? And what about the inevitable glitches that will eventually sneak in when no one is looking, and cause unexpected catastrophes?

These possibilities open a new chapter in the history of language, and in human history.

Coming from a minority language community as I do, I’m more circumspect about the end of human translation. A video made in Yorùbá will not automatically translate into Berom or Edo or Kiswahili—there’s been insufficient investment in those languages, so far, and companies like Meta, Google, Apple, Twitter, etc will always prioritize the larger languages, which will give them quick returns. This means that no matter what new technological leaps may be coming in language processing, they will not and cannot address longstanding problems in the world of translation. Individuals and groups are working to fill in the gaps to empower local languages to catch up with the advances that other, better-resourced languages can make more readily, but the gaps remain.