Back in November 2022, OpenAI launched ChatGPT to the public. By January 2023, it became the fastest-growing consumer software application, gaining over 100 million users and contributing to the growth of OpenAI’s valuation to $29 billion. It fueled the boom in public interest in the field of Artificial Intelligence and further intensified the AI revolution.
In late 2023, Indian entrepreneur Bhavish Aggarwal (co-founder of Ola Cabs and founder of Ola Electric) announced Krutrim, India’s own AI; And in early 2024 launched the Beta version of Krutrim AI chatbot to the public. As mentioned on their official website, Krutrim AI has been trained with over 2 Trillion tokens and supports all 22 official languages of India.
Being someone interested in new technologies/products and a frequent user of ChatGPT, the facts about Krutrim AI compelled me to give it a try. So I have tried various aspects of India’s own AI and below are my observations.
Before I start, please keep in mind that Krutrim AI is still in Beta phase and might not match your expectation if compared to ChatGPT. It is continuously learning and evolving.
Observation-1
In initial days the answers provided by Krutrim AI were little vague and not very useful. But with time It has improved a lot.
Below you can see the answer it provided to same question over 1 week.
This clearly shows that the AI model is continuously learning and generating better answers, but it still needs to improve its answers with more detailing and information.
Fun Fact: It also once provided the below answer to the same question.
Observation-2
Next, I tried asking Krutrim AI to give me a typescript code to reverse a string. And below is the code snippet it provided.
The logic provided in the code snippet is correct, except for declaring the variable reversed as a const. Const variables cannot be reassigned a new value. The variable should be declared as let or var.
Also the Output mentioned for the input string is wrong (it should have been “olleH”). Below you can find the correct code snippet and output.
Observation-3
Next, I tried asking the same question in multiple languages (11 languages I tried) it supports and below are the results.
English: Able to understand and generate related answer.
Hindi: Able to understand and generate related answers.
Odia: Able to understand and generate related answers.
Tamil: Able to understand and generate related answers.
Telugu: Able to understand and generate related answers.
Kannada: Able to understand and generate related answers.
Malayalam: Able to understand and generate related answers.
Marathi: Able to understand and generate related answers.
Bangla: Able to understand and generate related answers.
Assamese: Able to understand and generate related answers.
Gujarati: Able to understand and generate related answers.
It clearly understands and is able to answer in various languages as claimed on their website. This is really great for India, where we have people speaking diverse languages.
If we compare the answers, they were nearly the same, with around 20% difference. And I found it fascinating that the answer provided in Odia language is more structured.
Conclusion:
From the little I have explored till date, I think in future Krutrim AI is going to become the next big thing in the Indian tech industry and surely going to give tough competition to global leaders in the AI revolution.
But to achieve this, the AI model should be trained and refined more. Also, the response time could be reduced to give a better user experience.
Considering the fact that this is just the Beta version, I am somewhat satisfied with its performance. But going to production needs more improvement and testing.
In coming days, I will be exploring more on Krutrim AI and will be sharing my findings. Stay tuned to learn more about Krutrim, India’s own AI.