Krutrim AI: Here's All You Need To Know | NDTV Profit

NDTV Profit
18 Dec 202301:41

TLDRKitum, an ambitious initiative by Ola founder Pavish Agarwal, aims to create a full-stack AI company tailored for India. The company plans to develop AI chips, extensive language models, and a user app, supporting all 22 scheduled Indian languages with responses in 10. Kitum is trained on 2 trillion tokens, ensuring a deep understanding of Indian languages and culture. Funding is primarily from Agarwal's personal resources, with some from friends and family. Data sources include proprietary datasets, partnerships, and open platforms. NDTV Prophet is closely monitoring Kitum's progress.

Takeaways

  • 🚀 Kitum is an initiative to create an AI company focused on India's specific needs.
  • 📈 Kitum aims to be a full-stack AI company, meaning it will be involved in every level of the AI system development.
  • 🔍 Kitum will design AI chips, build large language models, and develop a user-facing app.
  • 🗣️ The company supports all 22 Indian scheduled languages and can respond in about 10 of them.
  • 📚 Kitum is trained on 2 trillion tokens, ensuring a broad representation of Indian languages.
  • 🇮🇳 The AI developed by Kitum is claimed to have an Indian ethos and cultural sensibilities.
  • 💰 Funding for Kitum is primarily from Olaf founder Pavish Agarwal's personal resources, with some from friends and family.
  • 🤔 The exact funding figures and data sources have not been disclosed by Agarwal.
  • 🔍 Kitum's data sources include proprietary datasets, data partnerships, and open-source platforms.
  • 📢 NDTV Profit has conducted an interview with Pavish Agarwal and will continue to track Kitum's developments.

Q & A

  • What does 'full stack' mean in the context of Kitum AI?

    -In the context of Kitum AI, 'full stack' means that the company will be involved in every level of the artificial intelligence system, including designing AI chips, building large language models, and developing a user-facing app.

  • How many Indian scheduled languages does Kitum AI support?

    -Kitum AI supports all 22 Indian scheduled languages.

  • In how many Indian languages can Kitum AI respond?

    -Kitum AI can respond in about 10 Indian languages.

  • What is the significance of Kitum AI being trained on 2 trillion tokens?

    -Training on 2 trillion tokens indicates that Kitum AI has a large representation of Indian languages, which helps in understanding and processing text in those languages more effectively.

  • What does Pavish Agarwal mean by 'Indian ethos natively' in Kitum AI?

    -By 'Indian ethos natively,' Pavish Agarwal means that Kitum AI is inherently designed to understand and reflect the cultural sensibilities of India, making the text it generates more relatable to Indian users.

  • How is Kitum AI funded?

    -Kitum AI is primarily funded from Pavish Agarwal's personal resources, with some participation from friends and family.

  • What are the three major sources of data for Kitum AI?

    -The three major sources of data for Kitum AI are proprietary datasets, data partnerships, and other open source platforms.

  • What is the role of proprietary datasets in Kitum AI's development?

    -Proprietary datasets have been collected by Kitum over the past year, contributing to its understanding of Indian languages and cultural nuances, which is crucial for its AI models.

  • How does Kitum AI ensure its AI models are culturally sensitive to Indian users?

    -Kitum AI ensures cultural sensitivity by training its models on a large dataset that includes a significant representation of Indian languages and cultural contexts, allowing the AI to generate text with an innate sense of India's cultural sensibilities.

  • What is the significance of Kitum AI's user-facing app?

    -The user-facing app is an integral part of Kitum AI's full stack approach, allowing users to interact directly with the AI system, which is designed to cater to the specific needs and preferences of Indian users.

  • How does Kitum AI's focus on India differentiate it from other AI companies?

    -Kitum AI's focus on India differentiates it by ensuring that its AI models are tailored to understand and respond to the linguistic and cultural diversity of the country, which is a key aspect of its appeal and utility for Indian users.

Outlines

00:00

🤖 Ola's AI Initiative: Kitu's Vision and Capabilities

Ola, an Indian company, aims to establish a full-stack AI entity named Kitu, dedicated to India. This means Kitu will engage in every layer of the AI system, including designing AI chips, developing extensive language models, and creating a user-facing app. The company's India-centric approach ensures support for all 22 scheduled Indian languages, with the ability to respond in about 10 of them. Founder Pavish Agarwal highlights that Kitu is trained on a vast dataset of 2 trillion tokens, rich in Indian languages, and is imbued with an Indian ethos, reflecting the cultural nuances in its generated text.

Mindmap

Keywords

💡AI

Artificial Intelligence (AI) refers to the simulation of human intelligence in machines that are programmed to think and learn like humans. In the context of the video, AI is the core technology that Kitu, the company mentioned, is focusing on developing. The video discusses Kitu's involvement in various levels of AI system development, indicating their commitment to creating advanced AI solutions tailored for India.

💡Full Stack

In the tech industry, 'full stack' refers to the comprehensive involvement in every layer of development, from the front end (user interface) to the back end (server, database, and infrastructure). The video highlights that Kitu is a full stack AI company, meaning they are engaged in designing AI chips, building language models, and creating user-facing applications, covering the entire spectrum of AI development.

💡Kitu

Kitu is the name of the AI company that the video is discussing. It is an Indian-focused AI company aiming to develop AI solutions that cater specifically to the Indian market, including support for Indian languages and cultural nuances. The name Kitu is used to represent the company's efforts in creating a本土 (indigenous) AI ecosystem in India.

💡Indian Scheduled Languages

The Indian Scheduled Languages are those languages which are recognized in the Eighth Schedule of the Indian Constitution. The video mentions that Kitu supports all 22 of these languages, showcasing the company's commitment to inclusivity and catering to the linguistic diversity of India.

💡Language Models

Language models are a type of machine learning model used in natural language processing (NLP) that can generate or understand human language. The video indicates that Kitu is building a family of large language models, which suggests their focus on developing AI systems that can effectively process and interact with Indian languages.

💡User Facing App

A user-facing app is a software application that is designed to be used directly by the end-user. The video mentions that Kitu is developing a user-facing app, implying that they are creating an accessible interface for consumers to interact with their AI technology.

💡Indian Ethos

The term 'Indian ethos' refers to the underlying principles, values, and cultural norms that are characteristic of Indian society. The video claims that Kitu's AI has an Indian ethos, meaning that the AI is not only linguistically but also culturally attuned to the nuances of Indian culture, which is crucial for creating AI that resonates with the local audience.

💡Funding

Funding in a business context refers to the financial resources provided to support a company's operations, growth, and development. The video discusses how Kitu is being funded, primarily from the personal resources of its founder, Pavish Agarwal, with additional support from friends and family.

💡Data Sourcing

Data sourcing involves the collection and acquisition of data that is used to train and improve AI systems. The video mentions three sources of data for Kitu: proprietary datasets, data partnerships, and open source platforms. This indicates that Kitu is leveraging various data sources to enhance the capabilities of their AI systems.

💡Tokens

In the context of AI and language models, tokens are the basic units of text, such as words, phrases, or characters, that are used to train the model. The video states that Kitu's AI is trained on 2 trillion tokens, which suggests a vast amount of data used to ensure the AI's proficiency and understanding of language nuances.

💡Cultural Sensibilities

Cultural sensibilities refer to the awareness and understanding of cultural norms, values, and behaviors. The video emphasizes that the text generated by Kitu's AI has an innate sense of India's cultural sensibilities, meaning that the AI is designed to produce content that is culturally relevant and appropriate for Indian users.

Highlights

Ola's ambition to build India's own AI

Introduction of Kitu, a full-stack AI company focused on India

Full-stack AI involvement at every level of the system

Designing AI chips and building language models

User-facing app development

Support for all 22 Indian scheduled languages

Response capability in about 10 Indian languages

Kitu trained on 2 trillion tokens with a focus on Indian languages

Innate Indian ethos in the AI's cultural sensibilities

Funding for Kitu primarily from Ola founder Pavish Agarwal's personal funds

Funding also involves contributions from friends and family

Three major sources of data for Kitu: proprietary datasets, data partnerships, and open-source platforms

NDTV Profit's interview with Pavish Agarwal in Bangalore

NDTV Profit's commitment to tracking Kitu's developments