Ola’s Krutrim, India’s most talked about large language model (LLM), arrived with much fanfare (attained unicorn status without even MVP). However, the reality fell far short of expectations.
Krutrim’s shortcomings – nonsensical responses, lack of nuance, and inability to handle basic conversation – have left many feeling disappointed.
Even though a buggy product is fine for any early stage product, but the fact that Ola decided to integrate Krutrim inside the main app is IMO, a bad move. While it ensures distribution to a much wider audience who are probably experiencing AI for the first time, the product is far from even being perfect – and might leave a bad taste for the audience when it comes to experimenting with Indian AI Startups.
And all of this, when we are in the middle of ‘What’s the Indian AI regulation story’ , this might just hurt the AI ecosystem badly.
But here’s the silver lining: Krutrim’s misstep could be the spark that ignites a firestorm of innovation in the Indian AI landscape and can inspire a generation of geeks to build better.
Krutrim’s Shortcomings: A Learning Opportunity
Krutrim’s issues stem from a combination of factors. Right from a rushed development cycle, resulting in a poorly trained model to limited access to diverse training data that hinders its ability to understand the intricacies of Indian languages and cultural nuances.
Additionally, a lack of transparency around capabilities and limitations breeds frustration. These issues serve as valuable lessons for future endeavors.
Building robust AI requires time, meticulous training on high-quality data, and a deep understanding of the target audience.
You can’t hurry love (LLM). You just have to wait
Phil Collins
The Indian AI Opportunity: A Playground for Geeks
Krutrim’s failings highlight a crucial gap: the need for AI solutions tailored specifically for the Indian audience. India boasts a massive and diverse population, with a rich tapestry of languages and cultural contexts. This presents a unique challenge and a phenomenal opportunity for Indian geeks.
Imagine AI assistants that understand Hinglish, navigate complex social situations, and seamlessly integrate into daily life. Think of educational tools that personalize learning for students across diverse backgrounds. Consider AI-powered healthcare solutions that bridge the language gap between doctors and patients. The possibilities are endless.
Building the Future: A Call to Action
Krutrim may not have lived up to the hype, but it has served a purpose. It has once again proven that funding doesn’t equal great product (rather, LLM). Here’s how the Indian AI community can respond:
- Open Collaboration: Foster a spirit of open collaboration among researchers, developers, and entrepreneurs. Sharing knowledge and resources will accelerate progress.
- Data is King: Focus on acquiring and curating high-quality, diverse data that reflects the Indian context. This includes regional languages, dialects, and cultural references.
- Focus on User Needs: Don’t get caught up in the hype of “big models”. Instead, prioritize user needs and build solutions that solve real-world problems for the Indian audience (instead of going all-in)
- Think different formats: My belief is that Indic LLMs will be voice-first, instead of the standard chatty experience.
Conclusion
Krutrim may be a cautionary tale, but it doesn’t have to be the end of the story. Let’s turn disappointment into motivation. Let Krutrim be the catalyst that propels India to the forefront of the AI revolution. Let’s build AI solutions that inspire, empower, and truly reflect the vibrancy of our nation.
The future of Indian AI is bright, and it’s up to the ProductGeeks to shape it.
