Are literary translators redundant in an age of AI? Rita Kothari, Arunava Sinha answer

“Trust me, if translators could get help from AI for deep research, they would take it,” Sinha says. But, he warns, “the process that an AI tool agent uses is very different from the process of a human translator.”

The predictability model that AI agents follow, Sinha says, is “a highly volume-driven, mechanical approach, which works when the text is intended to do nothing but communicate, but when some thought has gone into writing, that is where the big gap appears.”

The Centre has run its own experiments with AI translation. “AI can be very competitive, but it is still awfully risky. Every now and then it will do something so out of the blue that you could end up having some really serious bloopers. So if the human eye is required to keep an eye, the human eye might as well work on the translation itself,” says Kothari.

Besides, while AI firms continue to scan unique books, a major data gap remains in India. “Especially for translations from Indian languages into English or between Indian languages, there are not enough texts for an AI agent to learn from. They need huge, humongous amounts of text… so that their mathematical probabilistic analysis gets sharper and better defined.”

The heydays of translation

Despite the pervading AI-anxiety, both professors are optimistic about the state of translation in India. “There is a lot more interest in translating. What we are witnessing today is absolutely the heydays of translation,” says Kothari.

If anything is missing, she says, it is connective infrastructure rather than volume. “The frameworks are missing, in the sense that there is so much out there. But what frameworks can we use to understand how it interacts with this other region? If Tukaram is available in English, what does Tukaram have to say to Kabir? I think that kind of cross-dialogue needs to take place… and that is not happening.”

“The only thing wrong, if at all, is that there is a hecticness in the activity of it. There is a lack of contemplation and a lack of building connections.”

“We do not have enough people who can translate between the non-English Indian languages,” says Sinha. “The biliterate people who existed in fairly large numbers, say 50-80 years ago, are not there anymore.”

Currently, most translation between Indian languages happens through a bridge language, usually English or Hindi. Asked whether something is lost in this process, Sinha says, while change is inevitable in translation, “two languages don’t map into each other perfectly”.

On letting a language die

On Nikhil Krishnan’s The New Yorker essay, ‘Sometimes It’s OK to Let a Language Die’, Sinha said the argument is “led by economic factors” and “a very utilitarian economic approach… you could equally then wipe out a lot of art and say it has no economic value.”

Kothari has always taken issue with the use of the word “dead”: “I find this use of the word death a bit dramatic. I do not know what it means to say a language is dying.” She pointed to her own language, Sindhi, which is shrinking in India, but not in Pakistan. “We need to think of languages whose roles have shrunk, rather than get into this drama about death.”

Asked about what she makes of grandchildren unable to speak the language of their ancestors, Kothari asks, “What stopped them from actually using some of their native language with their own children? To what extent, therefore, is an ecology of a language also the responsibility of people?” She refuses to call these regional tongues “dialects,” calling the word “a form of gaslighting.” “To me, they are all languages, and they constitute for those people an entire life world.”

The politics of coercion

If a language’s survival ultimately rests on human will and private choice, what happens when the state tries to force the issue? On mandatory language learning for people working in different states—most recently a Marathi mandate for Maharashtra’s auto and taxi drivers, and a longer-running Kannada requirement in Karnataka schools—Kothari says, “If somebody coerces me to learn a language, I will not like that very much. I may not take kindly to coercion.”

At the same time, however, she says she would learn the language of any region she lived in, not from political pressure but as “a mode of expressing love… to say I’m here, and I respect you for what you speak.”

Her objection is that this is only demanded of people with the least power to resist, like taxi drivers and daily-wage workers, while those most insulated from consequence opt out entirely, especially the English-speaking elite class, who believe “I don’t need to learn any other language.”

“That is the person we should be going after,” she says. “Not the taxi driver. He should not be harassed.”

For Sinha, the underlying issue isn’t linguistic at all. “It is about excluding people. Language is an easy marker to identify whom to exclude.”

The way forward

Asked what would actually help build a more multilingual India, the scholars circled back to the three-language formula in schools, meant to expose students to a language far removed from their own.

“Unfortunately it became a box to be checked,” Sinha says. Students, he says, quickly learned to pick whichever third language would get them the easiest marks, rather than genuinely engaging with something unfamiliar. “Its purpose, its objective, was not met. But also, maybe, because it was a kind of coercion in its own way.”

Kothari agrees the formula had real “structural possibility.” “If, for instance, I had grown up learning Tamil and some Sanskrit and maybe Hindi or French, it would have been fantastic,” she says.

Still, both scholars believe the goal is achievable with enough institutional will. For now, Kothari says, the aim of gatherings like Bhashavaad is more modest than policy change. “It is aimed at attitudes, perceptions. That’s a very big leap [to policy]. But the beginning is always the attitude.”