Perplexity AI CEO Challenges India's AI Strategy, Offers Investment for Foundational Model Development

AI start-up founder says Nandan Nilekani is 'awesome' but also 'wrong'; here's why he said so

Aravind Srinivas, the Chief Executive Officer of California-based Perplexity AI, voiced his disagreement with Infosys' non-executive chairman Nandan Nilekani's stance on artificial intelligence (AI). Srinivas called Nilekani 'awesome' but also criticized the tech veteran's suggestion that Indian startups should avoid building large language models (LLMs) and instead focus on creating practical AI applications. While he acknowledged Nilekani's significant contributions to India's technological and economic progress through initiatives like Infosys and the Unified Payments Interface (UPI), Srinivas raised concerns about the strategy of sidelining foundational AI model training. "Nandan Nilekhani is awesome, and he's done far more for India than any of us can imagine through Infosys, UPI, etc. But he's wrong on pushing Indians to ignore model training skills and just focus on building on top of existing models. Essential to do both," Srinivas stated in a post on social media platform X. Srinivas stressed on the the importance of fostering both model training expertise and application development to enable India to fully benefit from AI advancements. Srinivas argued that India is making a strategic error by assuming that building AI foundation models is prohibitively expensive. "Re India training its foundation models debate: I feel like India fell into the same trap I did while running Perplexity. Thinking models are going to cost a shit ton of money to train," he said. He wrote about India's potential to excel in AI, drawing parallels to the nation's achievements in the space sector. Citing Indian Space Research Organisation's (ISRO) ability to achieve significant milestones with limited budgets, he noted that India could apply similar principles to AI development. "Elon Musk appreciated ISRO (not even Blue Origin) because he respects when people can get stuff done by not spending a lot. That's how he operates," Srinivas pointed out. Srinivas urged India to shift its focus from merely leveraging open-source AI models to developing its own foundation models that can compete globally. "So, I hope India changes its stance from wanting to reuse models from open-source and instead trying to build muscle to train their models that are not just good for Indic languages but are globally competitive on all benchmarks," he said. He also highlighted the progress of the Chinese AI firm DeepSeek as an example of what could be achieved with a determined approach to AI development. While he expressed his inability to lead such an initiative himself, Srinivas offered support to anyone committed to the cause, adding, "I'm happy to help anyone obsessed enough to do it and open-source the models."

Perplexity CEO is 'Ready to Invest $1 Million, and 5 Hours/Week' to Make India Great Again in AI

Perplexity CEO Aravind Srinivas, in a post on X, expressed his support for the development of AI in India. "I am ready to invest $1 mn personally and 5 hours per week of my time into the most qualified group of people that can do this right now to make India great again in the context of AI," he said. Srinivas was referring to a previous X post by him in which he assured help to anyone willing to run a "DeepSeek-like company in India and open-source the models". Notably, he means it with all seriousness. "Consider this as a commitment that cannot be backtracked," he affirmed. In order to make the cut, he said the team must be "cracked and obsessed" like the Chinese research lab DeepSeek, whose recent models have shaken up the AI world with unprecedented efficiency. Moreover, he added that he would invest an additional $10 million if the team aims to beat the latest DeepSeek-R1 model on all benchmarks with rigour. Srinivas believes India should build a foundation AI model of their own instead of building on top of existing open-source models. "India must show the world that it's capable of ISRO-like feat for AI," he earlier said. Expressing a disagreement with Infosys co-founder Nandan Nilekani, Srinivas said the former is "wrong in pushing Indians to ignore model training skills and focus on building on top of existing models". "Essential to do both," Srinivas argued in a post on X. Srinivas is an Indian citizen and is an alumnus of the Indian Institute of Technology, Madras. Recently, he also announced plans to recruit talent to expand the company's presence in India. "I am looking to recruit someone to work together on growing Perplexity in India. It will be fun and intense. You must be based in India and willing to travel and meet with strategic partners and institutions - scrappy startup mode," Srinivas said in a LinkedIn post. Last year in December, Srinivas met with Prime Minister Narendra Modi in Delhi to discuss AI's potential in India. Following the meeting, Srinivas posted on X, "We had a great conversation about the potential for AI adoption in India and across the world. Really inspired by Modi ji's dedication to staying updated on the topic and his remarkable vision for the future." "Was great to meet you and discuss AI, its uses and its evolution. Good to see you doing great work with Perplexity AI. Wish you all the best for your future endeavours," PM Modi responded.

India On Wrong Path, Should Build Foundational Models: Perplexity CEO

Currently, majority of the Indian AI companies are pushing for a product finetuned on open sourced foundational models AI search engine Perplexity founder and CEO Aravind Srinivas believes India is on the wrong path when it comes to building AI models. Taking to X, Srinivas said, "Re India training its foundation models debate: I feel like India fell into the same trap I did while running Perplexity." Notably, Perplexity AI uses almost all of the major foundational models to provide real-time answers to users' queries on the platform. Srinivas believes that instead of finetuning a foundational model, Indian companies should focus on training their models from scratch. He said that while the thinking models are going to be costly to train, India "must show the world that it's capable of ISRO-like feet (sic) for AI". "I think that's possible for AI (to train models frugally), given the recent achievements of DeepSeek. So, I hope India changes its stance from wanting to reuse models from open-source and instead trying to build muscle to train their models that are not just good for Indic languages but are globally competitive on all benchmarks," he added. DeepSeek is a China-based AI company that develops large language models (LLM). On (January 20), the company launched its latest reasoning models, DeepSeek-R1 and DeepSeek-R1-Zero, to take on platforms like OpenAI-o1. DeepSeek, which has raised a little over $4 Mn, is being touted as a competitor to OpenAI. "I'm not in a position to run a DeepSeek-like company for India, but I'm happy to help anyone obsessed enough to do it and open-source the models," said Srinivas. Currently, majority of the Indian AI companies are pushing for a product finetuned on open sourced foundational models. However, few of the notable exceptions include Ola's Krutim AI and the Indian-government backed BharatGen. It is pertinent to note that Infosys cofounder Nandan Nilekhani, in October last year, said that Indians shouldn't be focusing on building foundational models from scratch. Instead, they should create synthetic data and smaller models quickly. "We will use it (open-source foundational models) to create synthetic data, build small language models quickly, and train them using appropriate data," he said at the time. One of the major AI companies in India is probably following his advice - Sarvam AI. The Lightspeed-backed Bengaluru based company is focusing on training its AI models using synthetic data. However, Srinivas doesn't agree with Nilekani. "... He's (Nilekani) wrong in pushing Indians to ignore model training skills and just focus on building on top of existing models. (It is) essential to do both," Srinivas wrote in a separate post. Industry veterans like Google India Research Director Manish Gupta also shared similar opinions previously, saying that India will benefit from building foundational models.

Perplexity AI CEO Says Nandan Nilekani is 'Wrong'

"To be clear: Nandan Nilekani is awesome, and he's done far more for India than any of us can imagine through Infosys, UPI, etc. But he's wrong in pushing Indians to ignore model training skills...," he further said, countering Nilekani's vision of making India the 'AI use case capital' rather than investing in foundational model training. Nilekani's argument is that India should prioritise AI applications over competing with the US and China in developing large-scale models. "Our goal should not be to build one more LLM. Let the big boys in the (Silicon) Valley do it, spending billions of dollars," he said. He is correct, as India missed the generative AI bus long back. Notably, Srinivas is too late in the conversation. He said that this approach could make India overly reliant on foreign AI advancements. Others like Ola CEO Bhavish Agarwal, Tech Mahindra CEO CP Gurnani, and HCL co-founder Ajai Chowdhry are, however, optimistic and believe that something like NVIDIA can be born out of the country. Paras Chopra, founder of Turing's Dream, an AI research firm, said, "Many AI labs in India focus on local problems, similar to Indian companies. But the internet has no boundaries, so why not aim to be world-class?" He encouraged founders to specialise in niches but strive to be 'SOTA' (state-of-the-art) in those areas. "We can't always have the service mindset, and definitely not when everyone is chasing ASI," KissanAI CEO Pratik Desai said. Similarly, Soket AI Lab's Abhishek Upperwal also believes that India should invest in deep tech startups. "Building LLMs isn't just about having another AI model in the market. It's about owning the future. Nations, organisations, and individuals who create these models will dictate industries, policies, and progress." India is often dubbed as the use case capital of the world when it comes to most technological advancements. The case against creating newer foundational models is usually because building LLMs from scratch is an expensive, time-consuming, and capital-intensive task. Last year, Nilkeni said that India's approach to AI will be different from the global trend of building LLMs. Instead, India should focus on building AI use cases that will reach every citizen. Nilekani emphasised India's advantage in its population and aspirations, urging the use of existing technology to address challenges rather than waiting for the next $10 trillion model. "The Indian path in AI is different. We are not in the arms race to build the next LLM...We are here to make a difference, and our aim is to put this technology in the hands of people," he said. Commenting on the nature of wrappers, Srinivas said in an earlier podcast, "Wrappers are at all levels; it's just that they have given you so much value that you do not care." The term 'wrappers' first came into use in 2023 after the boom of GenAI startups, which were built on existing LLMs, causing a lot of uproar, especially in the VC ecosystem. Srinivas recently met Prime Minister Narendra Modi in the national capital earlier this month. Many founders and developers build on existing LLMs, creating value at the application layer. Healthify is an interesting case in point, which was featured in OpenAI's recent DevDay. It demonstrated real-time AI conversations in Hindi using OpenAI APIs. AIM has extensively covered how Indian AI companies, which were built on Llama's open-source platform in India, delivered great value. "I'm not in a position to run a DeepSeek-like company for India, but I'm happy to help anyone obsessed enough to do it and open-source the models," Srinivas added. He stressed that India should shift from reusing open-source models to building globally competitive models excelling in both Indic languages and broader benchmarks. This development comes against the backdrop of China's DeepSeek lab, taking the internet by storm with the launch of its open-source reasoning models - DeepSeek-R1 and DeepSeek-R1-Zero. These serve as alternatives to proprietary systems like OpenAI's o1. "China is currently leading some of the best open-source models (Minimax and DeepSeek V3) and keeps delivering great video models as well (Hailuo and Kling), despite being constrained by US sanctions and cannot purchase the best chip-making tools from ASML (Netherlands) nor the best AI chips," a user on X commented. In contrast, Indian innovators face a lack of investment. AIM's research found that Indian startups require only $10 million to begin foundational AI research, yet funding remains scarce. Mohandas Pai, head of Aarin Capital, highlighted this issue and said, "Who will give $200 million to a startup in India to build an LLM? Why is nothing like Mistral coming from India?" Pai called for a government-backed innovation fund, similar to France's $36 billion France Innovation Fund, which supports startups like Mistral. Such funding, he argued, could enable India to produce foundational models and compete on a global scale.

Perplexity Rolls Out Free Subscription To IIT Madras Students

This comes a day after Srinivas stirred up a storm in the India AI ecosystem after he said that Indian companies focus on training their models from scratch, rather than finetuning LLMs Perplexity AI cofounder and CEO Aravind Srinivas has said that the AI search engine has provided its premium subscription plan for free to students and faculty members of his alma mater IIT Madras. "We've given free Perplexity Pro to all students and faculty and staff of IIT Madras, where I did my undergrad. Super excited to start there as we begin our expansion for Indian campuses," said Srinivas in a post on X. This aligns with Srinivas' comments in December 2024, when he said he was open to "figuring out" an economic structure with Prime Minister Narendra Modi to offer Perplexity Pro to Indian students, faculty and researchers. Last month, he also met PM Modi in New Delhi and discussed the potential for AI adoption in India and across the world. In a separate post on X on Wednesday (January 22), Srinivas said that he is ready to invest $1 Mn personally and 5 hours every week towards a "group of people" that would make India "great in the context of AI". "I am ready to invest a $1mm (sic) personally and 5 hours/week of my time into the most qualified group of people that can do this right now for making India great again in the context of AI. Consider this as a commitment that cannot be backtracked. The team has to be cracked and obsessed like (the) DeepSeek team and has to open source the models with MIT license," said Srinivas. The Perplexity cofounder also said that he will invest $10 Mn more in the Indian company than can beat DeepSeek R1 on all benchmarks with "rigour". DeepSeek is a China-based AI company that has developed large language models (LLMs), and is being touted as a major competitor to giant OpenAI. With less than $4 Mn in funding, DeepSeek launched its latest reasoning models, DeepSeek-R1 and DeepSeek-R1-Zero, earlier this week to take on platforms like OpenAI-o1. Reacting to the Perplexity cofounder's post, Indic-focussed LLM maker SarvamAI cofounder Pratyush Kumar pitched his own startup. "Aravind, at @SarvamAI, we are building sovereign models that combine deep reasoning and Indic language skills. Would love to have you join this mission!" Kumar said. This comes a day after Srinivas stirred up a storm in the Indian AI ecosystem after he said that Indian companies should focus on training their models from scratch, rather than finetuning existing foundational models. While noting that "thinking models" are costly to train, Srinivas called on the Indian entrepreneurs to "show the world that it's capable of ISRO-like feet (sic) for AI". He was referring to the cost-effective and frugal approach of the Indian space agency.