The remarks made by an ex-Googler surprised many at a time when several tech businesses, tech experts, startup gurus, venture capitalists, and tech industry insiders are attempting to create an AI revolution in India. A few days ago, Gaurav Aggarwal, an ex-Google employee attempting to launch his own artificial intelligence startup in India called Ananas Labs, sent out a discordant tweet. The post from Aggarwal served as a reminder that India’s journey towards artificial intelligence will not be simple, in contrast to the exuberant remarks we often hear.
In a tweet, Aggarwal wrote, “After about seven months since leaving Google to start an AI research company in India, I have to unfortunately cut-short my adventure due to various reasons.”
We contacted Aggarwal in order to comprehend the different causes. It turns out that Aggarwal hardly mentioned his dissonant and challenging experience with AI in India thus far in his tweet. During the conversation, he shared his experiences from the past few months and outlined the major obstacles India has in its quest to overtake the US and China in the AI race. He even went so far as to say that India’s prospects for AI technology are dismal at one point.
Since leaving Google in August of last year to launch his own AI firm in India, Aggarwal claims that raising capital for the type of AI approach he is using has been his largest challenge. Although there is a lot of money available, it doesn’t seem to be enough for a business that wishes to engage in deep AI development, as OpenAI and a few other Silicon Valley firms do. Rather, and this is how Aggarwal presents it, the money is going to businesses that are only wrapping a large language model (LLM) made by American businesses.
“I wish someone was willing to put money into my startup,” says expert in machine learning and artificial intelligence (AI) Gaurav Aggarwal, who left Google in August 2023 to form his own AI research company in India. He is forty years old. I didn’t give it my all. In the past seven months, Aggarwal has met dozens of venture capitalists. He thinks of giving up because it is “extremely difficult to raise capital in India,” saying, “It’s a sad thing.”
Aggarwal was employed by Ola Cabs, Snapdeal, Fashiate, and Yahoo Labs prior to joining Google.
The field of artificial intelligence is very active right now. However, very little innovation comes from India and the majority comes from the West, China, or the United Arab Emirates, according to Aggarwal. “And because I am not the typical 20-something startup founder you will meet in India, I am in my 40s, I felt it was almost incumbent on me to set up a meaningful AI startup in India, because I have the capability and the technical knowledge to do it.”
Aggarwal has a low impression of the present AI work being done in India after his run-in with the country’s fledgling AI ecosystem. According to him, the majority of AI startups are lacking in deep AI expertise.
“I advise a few tech companies and I realise how shallow the AI capabilities are in India,” he continues. Regretfully, the majority of AI firms in India are only wrappers, simply implementing OpenAI APIs and creating adorable tiny applications on top of them. These businesses do not use deep AI at all. Actually, not many people in India possess the knowledge necessary to train models.”
And a major factor contributing to India’s “shallow AI capabilities” is the nation’s trade mentality, which is significantly more prevalent than its deep tech mentality.
“India’s startup scene is thriving, but it is not without issues. Deep tech is relatively unknown to the Indian venture capital community, according to Aggarwal. Though most of the deep tech businesses that come out of India are business innovations, it doesn’t mean they haven’t invested in deep tech. They differ from OpenAI and DeepMind. And it’s a depressing thing.
The shallow AI talent coming out of India is another consequence of the lack of emphasis on deep AI technology. “What saddens me is that these Indian firms are producing glorified engineers with a little bit of knowledge—not enough to even attempt to replicate what the West is doing—rather than true AI experts. Basically, they don’t know,” he claims.
Gaurav Aggarwal wants to create a multilingual LLM that is “better than any existing tool” with Ananas Labs. In essence, the AI technology would simplify and lower the cost of using Indian languages. The majority of LLMs are currently trained on datasets that are primarily in English, which makes it costly and occasionally erroneous for them to analyze tokens in Indian languages. Aggarwal hopes to address it with his AI project.
In India, there are nearly a billion internet users, and artificial intelligence will power everything we use. We will therefore have to pay the West money if we are unable to create our own AI models. In the next five to ten years, we will need to import AI, just like we do with oil, if we don’t work on building our own AI infrastructure, claims Aggarwal.
According to Aggarwal, venture capitalists in India are not willing to take on large risks in exchange for large rewards, which is why funding is going to firms that are more like clever ideas rather than research. He adds that they don’t have long-term vision. “These so-called consumer tech businesses are still losing money, but venture capitalists are still ready to invest in them. Even if these firms won’t be as profitable as Google or OpenAI, our VCs still seem to enjoy them. I’m not sure why they aren’t setting higher goals. Aggarwal remarks, “I wish our ecosystem had that willingness to take risks.
Among the numerous tech insiders attempting to make a significant impact in AI is Aggarwal. It’s still unclear if AI projects will succeed as we anticipate. But Aggarwal’s account serves as a brief reminder of what OpenAI CEO Sam Altman stated on his trip to India last year. What steps would an Indian company need to take to create a system similar to ChatGPT, someone asked him? His remark, which minimized India’s AI potential, attracted a lot of notice. Indians could try to develop AI like ChatGPT, but Altman has predicted they wouldn’t be successful.
