Ola’s co-founder and CEO, Bhavish Aggarwal, has joined the artificial intelligence (AI) competition with the introduction of “Krutrim,” India’s first AI computing stack. This multilingual AI model aims to rival industry giants like OpenAI’s ChatGPT and Google’s Bard. The name ‘Krutrim’ translates to “artificial” in Sanskrit and includes two versions – Krutrim, the base model trained on 2 trillion tokens, and ‘Krutrim Pro,’ a more advanced model set to launch in early 2024 for advanced problem-solving.
During the launch event, Aggarwal showcased an AI chatbot powered by Krutrim, similar to ChatGPT and Bard, capable of understanding 22 Indian languages and generating text in 10 Indian languages. The Krutrim team, operating in both India and the Bay Area (US), claims to have trained the model in just three months, asserting its consistent outperformance of OpenAI’s GPT4 in Indic languages.
Aggarwal emphasized the importance of AI infrastructure, stating that AI is the soul, and silicon and infrastructure are the body. He highlighted the company’s effort to reduce energy consumption in data centers, claiming that Krutrim has a Power Usage Effectiveness (PUE) of 1.1, significantly more efficient than the average Indian data center’s PUE of 1.5.
Krutrim Pro, slated for release in February 2024, is touted as a multimodal trained model, capable of integrating various modalities as inputs for cross-modal training. The platform aims to cater to developers worldwide, with developer APIs set to go live in February 2024.
Ola group companies are reported to be utilizing Krutrim for internal workloads, and the company envisions releasing its AI-focused system in a package (SIP) by 2025, with prototypes expected by mid-2024. This aligns with Ola’s ambition to build a homegrown supercomputer tailored for AI in the coming years.
In summary, Ola’s Bhavish Aggarwal has entered the AI race with Krutrim, presenting a promising model designed to meet India’s linguistic diversity and compete with established players in the global AI landscape. The company’s focus on energy-efficient data centers and plans for an AI-focused SIP by 2025 further highlight their commitment to advancing AI technology in the Indian context.