Indian AI Startups: Accel's Prayank Swaroop on Academia, Capital &'s Future
By Rediff Money Desk, NEWDELHI Mar 06, 2024 12:45
Accel partner Prayank Swaroop discusses the need for more Indian AI startups from academia, the capital intensity of building AI models, and the potential for India in the AI-native product space.
New Delhi, Mar 6 (PTI) Indian origin researchers are leading artificial intelligence and machine learning breakthroughs at companies such as Google, Meta and Amazon and yet not enough researchers based in India have opted to build AI startups, says Prayank Swaroop, partner at global venture capital firm Accel.
The capital-intensiveness of building AI foundational research models could be one reason behind this, he said, adding that India with its large talent pool can, instead, leverage high-quality AI models and create applications for the world.
"We need more Indian AI startups coming from academia," the senior executive at Accel, which counts Flipkart, Cure.fit, Ninjacart, Swiggy and BookMyShow among its portfolio companies, told PTI.
Indian developers are now dabbling with LLMs and building AI-native products, so it is only a matter of time before we see a truly disruptive AI-native product emerge from the country, said Swaroop, who is also leading the AI cohort of Accel's revamped early-stage accelerator programme, Atoms (now in its third edition).
Excerpts from the interview:
Q:Will the momentum on AI die down anytime soon? What's on the horizon for AI?
A: This is the year where we're seeing generative AI applications come into widespread use with a number of real-life use cases being implemented across customer support, sales agents, call centre support etc. Moreover, media applications of AI in copywriting, image creatives creation and music generation are starting to become much more widespread. I believe we are just starting on a multi-decade journey of AI becoming part of human productivity.
We are witnessing an ever-increasing pace of innovation in LLMs (large language model), LLMOps (large language model ops), ML models and GPU (graphics processing units) compute. Foundational models and AI and machine learning toolchains are undergoing rapid innovation as we speak, across text, voice and video.
A significant theme expected to emerge is the application of GenAI in vertical use cases or sectors such as healthcare, finance, education, entertainment, and security. In fact, the next generation of SaaS applications will all be AI-enabled, providing 10-times the value that they currently do.
GenAI also has the potential to significantly disrupt the service sector, which constitutes a majority of the GDP in developing countries. Service sector companies have not been able to adopt newer tech as quickly as other sectors, so have a lot more potential for disruption, especially in terms of how they operate.
Q: What is the outlook for startups building in AI in India as compared to, say, the Silicon Valley (US)?
A: Frankly, we are behind the curve on core foundational research in the generative AI space like LLMs, model fine-tuning methods, responsible AI etc. It is an irony, since the current way was set off by Indian origin researchers like Ashish Vaswani working at Google.
Today if you look at the Valley, heads of AI/ML at major players like Google, Meta or Amazon are Indian origin researchers. But unfortunately, not enough researchers based in India have come out and built startups - the only one that comes to mind is Sarvam.AI. Another example is Krutrim.ai, but some of their senior team members are based in the Valley. This is a big question we need to ask - why aren't there more Indian startups coming out of our academia, even if capital is available.
A part of the answer might be that the capital available in India is not in the same quantum as in the Valley. A USD 100 million cheque to an Indian researchers led startup is quite unlikely in India right now. And foundational research models today are very capital intensive to build. That is why we are seeing the big cloud providers -- Microsoft, Google and Amazon -- making major plays. To compete you need a lot of capital - like OpenAI has demonstrated.
But I'm hopeful there are still a large number of categories where Indian developers can win. India offers a vast talent pool that can leverage these high-quality models, and build applications for the world. We've seen that happen over the last decade as Indian SaaS companies are competing and winning against SaaS companies globally.
Indian companies can also focus on building AI tooling, companies that are the picks and shovels in the AI gold rush. By providing essential infrastructure and support systems for AI development, these startups play a crucial role in the ecosystem to enable faster AI adoption.
The rise of generative AI models increases the potential for India to bring AI-native solutions to market. GenAI does not require specialised AI knowledge to use, so India's massive developer population will be a big advantage for the future of AI development here - India has more than 13 million active developers on GitHub alone, second only to the US which has 20 million developers.
Indian developers are now dabbling with LLMs and building AI-native products, so it is only a matter of time before we see a truly disruptive AI-native product emerge from the country.
Q: How do you evaluate the potential of an AI company and what has made your AI investments successful? What are those startups doing differently or better?
A: We're in the very early innings of seeing the true implications of GenAI, so the playbook for what creates success in this space is yet to be written. But for building enduring companies, the one truth that is time tested is - what is the company doing that solves a customer pain point? Just because generative AI is the new technology does not mean that customers will buy anything.
Customers buy solutions to their problems, and AI startups need to build products that are calibrated to their customer needs. I do feel that a new AI startup can defeat existing large incumbent companies if they can truly weave AI into the DNA of their products versus just an add-on AI feature to their solutions.
Founders who are nimble, have the right mindset for navigating this highly unpredictable, dynamic and competitive AI environment, and stick to first-principle thinking while this market matures and changes, stand out to us.
More specifically, I can say that we're always excited to meet teams that have identified a problem statement and are building solutions offering 10X value to customers at one-tenth the price with AI. Teams that understand the ever-changing landscape and can build and deploy AI systems quickly, while also building systems that are secure and do not hallucinate.
Large, high-quality businesses are usually built by founders who have a strong bias for execution and remain customer-obsessed.
The capital-intensiveness of building AI foundational research models could be one reason behind this, he said, adding that India with its large talent pool can, instead, leverage high-quality AI models and create applications for the world.
"We need more Indian AI startups coming from academia," the senior executive at Accel, which counts Flipkart, Cure.fit, Ninjacart, Swiggy and BookMyShow among its portfolio companies, told PTI.
Indian developers are now dabbling with LLMs and building AI-native products, so it is only a matter of time before we see a truly disruptive AI-native product emerge from the country, said Swaroop, who is also leading the AI cohort of Accel's revamped early-stage accelerator programme, Atoms (now in its third edition).
Excerpts from the interview:
Q:Will the momentum on AI die down anytime soon? What's on the horizon for AI?
A: This is the year where we're seeing generative AI applications come into widespread use with a number of real-life use cases being implemented across customer support, sales agents, call centre support etc. Moreover, media applications of AI in copywriting, image creatives creation and music generation are starting to become much more widespread. I believe we are just starting on a multi-decade journey of AI becoming part of human productivity.
We are witnessing an ever-increasing pace of innovation in LLMs (large language model), LLMOps (large language model ops), ML models and GPU (graphics processing units) compute. Foundational models and AI and machine learning toolchains are undergoing rapid innovation as we speak, across text, voice and video.
A significant theme expected to emerge is the application of GenAI in vertical use cases or sectors such as healthcare, finance, education, entertainment, and security. In fact, the next generation of SaaS applications will all be AI-enabled, providing 10-times the value that they currently do.
GenAI also has the potential to significantly disrupt the service sector, which constitutes a majority of the GDP in developing countries. Service sector companies have not been able to adopt newer tech as quickly as other sectors, so have a lot more potential for disruption, especially in terms of how they operate.
Q: What is the outlook for startups building in AI in India as compared to, say, the Silicon Valley (US)?
A: Frankly, we are behind the curve on core foundational research in the generative AI space like LLMs, model fine-tuning methods, responsible AI etc. It is an irony, since the current way was set off by Indian origin researchers like Ashish Vaswani working at Google.
Today if you look at the Valley, heads of AI/ML at major players like Google, Meta or Amazon are Indian origin researchers. But unfortunately, not enough researchers based in India have come out and built startups - the only one that comes to mind is Sarvam.AI. Another example is Krutrim.ai, but some of their senior team members are based in the Valley. This is a big question we need to ask - why aren't there more Indian startups coming out of our academia, even if capital is available.
A part of the answer might be that the capital available in India is not in the same quantum as in the Valley. A USD 100 million cheque to an Indian researchers led startup is quite unlikely in India right now. And foundational research models today are very capital intensive to build. That is why we are seeing the big cloud providers -- Microsoft, Google and Amazon -- making major plays. To compete you need a lot of capital - like OpenAI has demonstrated.
But I'm hopeful there are still a large number of categories where Indian developers can win. India offers a vast talent pool that can leverage these high-quality models, and build applications for the world. We've seen that happen over the last decade as Indian SaaS companies are competing and winning against SaaS companies globally.
Indian companies can also focus on building AI tooling, companies that are the picks and shovels in the AI gold rush. By providing essential infrastructure and support systems for AI development, these startups play a crucial role in the ecosystem to enable faster AI adoption.
The rise of generative AI models increases the potential for India to bring AI-native solutions to market. GenAI does not require specialised AI knowledge to use, so India's massive developer population will be a big advantage for the future of AI development here - India has more than 13 million active developers on GitHub alone, second only to the US which has 20 million developers.
Indian developers are now dabbling with LLMs and building AI-native products, so it is only a matter of time before we see a truly disruptive AI-native product emerge from the country.
Q: How do you evaluate the potential of an AI company and what has made your AI investments successful? What are those startups doing differently or better?
A: We're in the very early innings of seeing the true implications of GenAI, so the playbook for what creates success in this space is yet to be written. But for building enduring companies, the one truth that is time tested is - what is the company doing that solves a customer pain point? Just because generative AI is the new technology does not mean that customers will buy anything.
Customers buy solutions to their problems, and AI startups need to build products that are calibrated to their customer needs. I do feel that a new AI startup can defeat existing large incumbent companies if they can truly weave AI into the DNA of their products versus just an add-on AI feature to their solutions.
Founders who are nimble, have the right mindset for navigating this highly unpredictable, dynamic and competitive AI environment, and stick to first-principle thinking while this market matures and changes, stand out to us.
More specifically, I can say that we're always excited to meet teams that have identified a problem statement and are building solutions offering 10X value to customers at one-tenth the price with AI. Teams that understand the ever-changing landscape and can build and deploy AI systems quickly, while also building systems that are secure and do not hallucinate.
Large, high-quality businesses are usually built by founders who have a strong bias for execution and remain customer-obsessed.
Read More On:
DISCLAIMER - This article is from a syndicated feed. The original source is responsible for accuracy, views & content ownership. Views expressed may not reflect those of rediff.com India Limited.
You May Like To Read
TODAY'S MOST TRADED COMPANIES
- Company Name
- Price
- Volume
- Srestha Finvest
- 0.89 ( 0.00)
- 69758135
- Standard Capital
- 1.26 ( -16.00)
- 55792425
- GTL Infrastructure
- 2.37 (+ 4.87)
- 50169700
- Vodafone Idea L
- 9.18 ( -1.50)
- 27130818
- Jaiprakash Power Ven
- 22.33 ( -2.45)
- 21183721
MORE NEWS
Amul Ice Cream Tampering Video Resurfaces:...
GCMMF has denied allegations of tampering with Amul ice cream after an old video...
Indo-European Relations: Convergence of Values...
FEBI President Remi Maillard highlights the strong Indo-European relationship driven by...
Srinivas Appointed CEO of GeM Public...
L Satya Srinivas, Additional Secretary in Department of Commerce, has been given the...