IndoVox

Hello Friends,

India’s festive season is around the corner—and this year, CITLoB is adding another reason for the language and content community to come together.

SAMVĀD 2026, South Asia’s largest language industry conference, will take place on 26 October 2026 at Holiday Inn, Mumbai, in partnership with FICCI and C-DAC.

This is not just another industry conference.

The world of content is undergoing a fundamental transformation. AI is changing how content is created, translated, localised and consumed. Technology is redefining traditional business models, while the explosion of digital, OTT, media and global content is creating entirely new opportunities. For leaders, the question is no longer whether AI will change our industry—it is how we choose to lead through that change.

Under the theme “Beyond Borders: AI, Language & the Future of Global Content,” SAMVĀD 2026 will bring together CEOs, entrepreneurs, language and technology leaders, global content creators, broadcasters, media and OTT companies, technology platforms, government institutions, academia and international industry associations.

With Bruno Herrmann as the keynote speaker, and an exciting line-up of experts bringing perspectives from across industries and geographies, SAMVĀD 2026 promises a rich exchange of ideas, perspectives and experiences.

But the real value of SAMVĀD lies beyond the stage.

It is a place to challenge assumptions, exchange ideas, discover new business possibilities, forge partnerships and understand where the global content ecosystem is heading.

At a time when AI is simultaneously disrupting established models and opening doors to unprecedented scale, isolation is not an option. The conversations we have today could shape the businesses we build tomorrow.

I invite CEOs, founders and industry leaders to join us at SAMVĀD 2026—not merely as delegates, but as contributors to a conversation that matters.

Come with questions. Bring your perspective. Leave with possibilities.

SAMVĀD 2026 — Beyond Borders. Beyond Convention. Beyond Today

Regards,
Sudheen M
Founder - CITLoB

Machine translation has a strange habit: it never sounds unsure.

Ask a human translator about an ambiguous phrase and you'll get a pause, a question, maybe a note in the margin asking for clarification. Ask an MT engine and it just gives you an answer. Smoothly. Instantly. With exactly the same confidence whether it is right or completely off.

A menu item can become something unrecognisable because the source word had two unrelated meanings and the engine picked the wrong one. A formal legal notice can come out sounding like a text message because the register was wrong. An idiom can be translated word for word into something that means very little in the target language.

The strange thing is that the translation itself may look perfectly good.

We used to have more obvious clues. Poor grammar. Awkward phrasing. Missing words. Strange  terminology. Something that made you stop and think, "That doesn't look right."

Good MT has taken away quite a few of those clues.

Modern MT is remarkably good at the kind of high-volume, repetitive work that used to eat up hours: product descriptions, support articles, large batches of predictable content. There is plenty of work where the machine does a perfectly respectable job.

It's the less predictable stuff that gets interesting.

What does this word mean here? Is that sentence formal or conversational? Is the writer being sarcastic? Does that phrase have a second meaning? Does the target sentence actually say what the source says, or does it simply sound like something a native speaker would write?

A human translator can stop when one of those questions matters. They can check the context, make a judgement or ask the client.

The machine doesn't know that there was a question.

It just picks an answer.

LQA faces a slightly awkward problem here. We have spent years developing ways to spot things that look wrong. But what happens when the wrong answer looks completely right?

Grammar can be fine. Terminology can be fine. The sentence can sound natural. You can read it without noticing anything unusual.

The meaning can still be wrong.

That doesn't mean we throw out the things we already measure. Fluency, grammar, terminology and naturalness still matter. They just don't tell us everything.

We may need to spend more time asking what happened before the sentence appeared. Was the source ambiguous? Did the system have enough context? Were there other possible interpretations? Did it understand what the writer meant, or did it simply produce a plausible sentence?

Those are questions translators have always dealt with. The difference is that the machine's answer can look so good that nobody thinks to question it at first glance.

Regards,
Sankeshwari Deo 
Strategic Advisor - Localization, AI & Global Transformation