Companies that are building artificial intelligence (AI) solutions tailored to India’s domestic market can develop the scale, trust, safety, security, and enterprise readiness. These are useful traits when these firms inevitably go global, Alex Haskell, General Counsel and Head of Global Affairs at US-based conversational AI firm ElevenLabs, told Soumyarendra Barik and Bijin Jose in an interview.
Many believe the company, which was recently valued at $11 billion, is the current leader in voice AI. It has found use cases, particularly in automated customer support services deployed by an increasing number of companies around the world, including in India. Edited excerpts:
Some Indian startups are trying to build a niche by offering support for various Indian languages in their models. Do you view that as a prudent approach? What are the pitfalls of focusing specifically on Indian languages? Can a global business be built on this premise?
Alex Haskell: To answer the question directly, yeah, I think so. India is a very important country in the world for business, and for the sheer number of people. So I think focusing on India can allow companies to build the scale, to build the trust, the safety, the security, and the enterprise readiness that’s needed to serve other companies and other people all over the world. At ElevenLabs, we want to be both the best at Indian languages and also the best at trust, safety, scalability, and all those other things.
How expansive is your coverage of Indian languages right now, and how do you plan to expand that? What specific steps will you take for it?
We are very strong in 11 languages already. We know that’s not enough, given the diversity of languages spoken in India, so we’re focusing our research efforts on those other languages, getting the data that we need to power models in other Indian languages. Given that the Indian market is our second biggest enterprise market, it’s a place where we’re hyper-focused. We know that to succeed here demands expanding beyond the 11 languages we’re already good at.
What are some insights you can share about the user base in India on the customer side? How many Indian companies are using your services, and how many customers have they served?
Story continues below this ad
What I can give you are some of our marquee Indian customers, including Cars24, TVS, Nykaa, and quite a few others that we’re continuing to expand our work with.
Can you talk us through the process of building for Indian languages? How does your model achieve it, and how do you account for various dialects? Does it take a lot of data annotation work in India? And where is that kind of work happening for you at this point?
We were the first company to create a truly natural-sounding voice model with emotion and the right intonation, pace, and prosody. That was several years ago, and of course, it was in English. But the team—our research team—I can humbly say, is the best audio research team in the world. They’re capable of teaching models to speak in any language with a much more limited dataset than others, because of just how good they are at research and how frequently they have research breakthroughs. We’re now up to almost 100 languages globally. The reason that we’re able to serve customers well in all those languages is the architecture and other research-related breakthroughs, in addition to the data, that allow the model to work across languages.
Can global AI companies be built from India, especially in the space that you are operating in, like customer service?
Story continues below this ad
The short answer is yes. We know that to serve the Indian market, which is such a unique market, we have to build here. We have to have lots of people here and offices here and ensure that we’re meeting the unique needs of this unique market. So we are a global company. We’ve been global since day one, but India is such an enormous focus, and we are already building here and fully intend to keep aggressively building here.
Regarding models that are currently being built in India, which are focusing on Indian history, culture, and context—would they remain serviceable largely domestically, or can training a model for these particular aspects also have some benefits if they want to sell their service globally?
So you’re talking about other companies? Yes. At the heart of artificial intelligence innovation is research. It’s actually not something I fully understood when I first started working at the company—why we have this group of people that were called researchers. But what they’re doing is research in an academic sense, and picking up tons of learnings along the way that help in foreseen and unforeseen ways to deliver results to people all over the world.
The process of doing that research here in India—the companies that you’re talking about—they’re going to learn so much about how models work and how models can be best tailored for specific purposes, and those learnings will be valuable everywhere.
Disclaimer: We do not own any of the content, ideas, images, or text presented here. All rights belong to their respective owners. For more information and to view the original source, please visit the following link: