AI wrapper and search platform Perplexity AI’s CEO Aravind Srinivas not too long ago mentioned that Infosys co-founder Nandan Nilekani is improper in pushing Indians to disregard mannequin coaching abilities and deal with constructing on prime of current fashions. “Important to do each,” Srinivas argued in a submit on X.
“To be clear: Nandan Nilekani is superior, and he’s finished much more for India than any of us can think about via Infosys, UPI, and so forth. However he’s improper in pushing Indians to disregard mannequin coaching abilities…,” he additional mentioned, countering Nilekani’s imaginative and prescient of constructing India the ‘AI use case capital’ reasonably than investing in foundational mannequin coaching.
Nilekani’s argument is that India ought to prioritise AI purposes over competing with the US and China in creating large-scale fashions. “Our aim shouldn’t be to construct another LLM. Let the massive boys within the (Silicon) Valley do it, spending billions of {dollars},” he mentioned.
He’s appropriate, as India missed the generative AI bus lengthy again.
Notably, Srinivas is just too late within the dialog. Others like Ola CEO Bhavish Agarwal, Tech Mahindra CEO CP Gurnani, and HCL co-founder Ajai Chowdhry are, nevertheless, optimistic and imagine that one thing like NVIDIA might be born overseas.
Paras Chopra, founding father of Turing’s Dream, an AI analysis agency, mentioned, “Many AI labs in India deal with native issues, much like Indian firms. However the web has no boundaries, so why not purpose to be world-class?” He inspired founders to specialize in niches however try to be ‘SOTA’ (state-of-the-art) in these areas.
Tech Mahindra might be the one system integrator to have constructed its personal fashions from scratch. Different Indian IT companies are counting on current options to create tangible outcomes.
In addition to, whereas startups like Sarvam AI, TWO, and Krutrim are constructing merchandise, the influence that they’ve created is minuscule in comparison with one thing like OpenAI’s ChatGPT.
“We are able to’t all the time have the service mindset, and positively not when everyone seems to be chasing ASI,” KissanAI CEO Pratik Desai mentioned.
Equally, Soket AI Lab’s Abhishek Upperwal additionally believes that India ought to put money into deep tech startups. “Constructing LLMs isn’t nearly having one other AI mannequin available in the market. It’s about proudly owning the long run. Nations, organisations, and people who create these fashions will dictate industries, insurance policies, and progress.”
India is usually dubbed because the use case capital of the world in the case of most technological developments. The case in opposition to creating newer foundational fashions is actually because constructing LLMs from scratch is an costly, time-consuming, and capital-intensive activity.
Final 12 months, Nilkeni mentioned that India’s strategy to AI can be completely different from the worldwide development of constructing LLMs. As an alternative, India ought to deal with constructing AI use instances that can attain each citizen. Nilekani emphasised India’s benefit in its inhabitants and aspirations, urging the usage of current expertise to deal with challenges reasonably than ready for the following $10 trillion mannequin.
“The Indian path in AI is completely different. We’re not within the arms race to construct the following LLM…We’re right here to make a distinction, and our purpose is to place this expertise within the arms of individuals,” he mentioned.
Perplexity AI is a Wrapper
Commenting on the character of wrappers, Srinivas mentioned in an earlier podcast, “Wrappers are in any respect ranges; it’s simply that they’ve given you a lot worth that you don’t care.”
The time period ‘wrappers’ first got here into use in 2023 after the increase of GenAI startups, which had been constructed on current LLMs, inflicting a whole lot of uproar, particularly within the VC ecosystem. Srinivas not too long ago met Prime Minister Narendra Modi within the nationwide capital earlier this month.
Many founders and builders construct on current LLMs, creating worth on the software layer. Healthify is an attention-grabbing working example, which was featured in OpenAI’s latest DevDay. It demonstrated real-time AI conversations in Hindi utilizing OpenAI APIs. AIM has extensively coated how Indian AI firms, which had been constructed on Llama’s open-source platform in India, delivered nice worth.
Srinivas Turns to India to Construct Foundational Fashions
“I’m not ready to run a DeepSeek-like firm for India, however I’m pleased to assist anybody obsessed sufficient to do it and open-source the fashions,” Srinivas added. He confused that India ought to shift from reusing open-source fashions to constructing globally aggressive fashions excelling in each Indic languages and broader benchmarks.
This improvement comes in opposition to the backdrop of China’s DeepSeek lab, taking the web by storm with the launch of its open-source reasoning fashions – DeepSeek-R1 and DeepSeek-R1-Zero. These function options to proprietary methods like OpenAI’s o1.
“China is presently main a number of the greatest open-source fashions (Minimax and DeepSeek V3) and retains delivering nice video fashions as nicely (Hailuo and Kling), regardless of being constrained by US sanctions and can’t buy the most effective chip-making instruments from ASML (Netherlands) nor the most effective AI chips,” a person on X commented.
In distinction, Indian innovators face an absence of funding. AIM’s analysis discovered that Indian startups require solely $10 million to start foundational AI analysis, but funding stays scarce.
Mohandas Pai, head of Aarin Capital, highlighted this problem and mentioned, “Who will give $200 million to a startup in India to construct an LLM? Why is nothing like Mistral coming from India?”
Pai known as for a government-backed innovation fund, much like France’s $36 billion France Innovation Fund, which helps startups like Mistral. Such funding, he argued, might allow India to supply foundational fashions and compete on a worldwide scale.
The submit Perplexity AI CEO Says Nandan Nilekani is ‘Fallacious’ appeared first on Analytics India Journal.
