Global AI Conclave: Sarvam.ai to release more LLMs, build platform for building Gen AI applications

Bala Parthasarathy of People + AI (l) in conversation with Vivek Raghavan of Sarvam AI at the CNBC TV18- Moneycontrol Global AI Conclave in Bengaluru on December 16

Homegrown generative artificial intelligence startup, Sarvam AI, which recently released the first Hindi large language model (LLM) Open Hathi, and raised $41 million, plans to release more such models and a platform where developers can use such models for their work, said Vivek Raghavan, Founder of Sarvam AI at the CNBC-TV18-Moneycontrol Global AI Conclave in Bengaluru on December 16.

During a fireside conversation with Bala Parthasarathy of People + AI at the event, Raghavan said, "One of the important things that we are doing is we're not just building models. We are also going to build a platform for developers where one can use a combination of various kinds of models some which are from us, some which are open source, some which may not be open source, and figure out how to deploy generative AI applications at scale, and understand and evaluate their performance efficiently."

Raghavan said that he expects this platform to be available to developers in the next couple of months. Apart from that, Raghavan also expects to release more models like the recently-released Hindi LLM Open Hathi, in various languages and sizes.

Earlier this month, Sarvam AI raised $41 million in its Series A funding round led by Lightspeed Ventures with participation from Peak XV Partners and Khosla Ventures.

Raghavan, who was also instrumental in building digital public infrastructure (DPI) like Aadhaar, said that the funding will be used to hire talent and increase computational capabilities.

"We have an amazing team, but we believe that it is talent that drives this kind of thing. And so it (funding) is it is to get talent. Of course, the other thing is compute. This is extremely expensive compute-wise..," he added.

Raghavan also made three predictions on what he expects from the AI ecosystem within the next year. First, he would prefer to talk to an automated customer service rather than an actual person by next year. "I think that there will come a time and I'm predicting it is sooner than later that you will get better responses from the bot than what the human representative," he said.

Second,  he predicted that in contrast to the excessive demand for graphic processor units (GPUs) currently, in the future, there will be a GPU glut.

"There is a tremendous shortage right now. I think that shortage will ease because that is how the cycles of things go. The fact that there was such a severe shortage last year, basically caused several different players to ramp up in various kinds of forms. And I think that that that will always go in a cycle...," he added.

Last, Raghavan made an ominous prediction that many companies will go out of business next year.

"AI is a tool, right? And you have to use that within your business process. Right... Organisations that leverage AI in fundamentally their core business processes will be more effective than those that don't. And I think that's the thing, and you won't know the difference until one day it becomes too obvious. It will be too late," he added.