First Indian IT Company to Introduce ‘Made in India’ ChatGPT: A Game-Changer in Multilingual AI

In a groundbreaking move, Tech Mahindra, India’s leading IT company, has unveiled its ambitious initiative, Project Indus. This project aims to develop an open-source foundational language model tailored for Indian languages, setting a new standard in the world of artificial intelligence. The first model, armed with a remarkable 7 billion parameters, is set to revolutionize how AI interacts with users and content across diverse Indian languages. Let’s delve into the key highlights of this revolutionary endeavor.

Key Highlights

• Tech Mahindra’s Project Indus focuses on creating an Indic-based foundational model for Indian languages.
• The project aims to build a 7-billion parameter language model, which could potentially serve 25% of the global population.
• The initial focus is on supporting 40 different Hindi dialects, with plans to expand to more languages later.
• The model prioritizes cultural sensitivity, promoting effective communication and understanding local nuances.
• An Indic language model can democratize AI by catering to a wider non-English-speaking audience in India.
• Tech Mahindra collaborates with various stakeholders, including Indian educational institutions and Microsoft, to develop Indic datasets.
• Gathering data for various dialects is a challenge, and Tech Mahindra seeks contributions from speakers to build robust datasets.
• Addressing biases in datasets is a crucial concern, with the company employing human annotation and automatic techniques.

Breaking Language Barriers with a 7-Billion Parameter Model

Tech Mahindra’s Project Indus aims to fill a crucial gap in the world of language models. While powerful models like GPT by OpenAI have transformed AI communication, they are primarily trained on English datasets, limiting their effectiveness in comprehending and generating content in Indic languages. The launch of Project Indus heralds a new era, promising a foundation model tailored for Indian languages and dialects.

CP Gurnani, Chief of Tech Mahindra, envisions Project Indus as the world’s most substantial Indic language model. This endeavor could potentially serve a quarter of the global population, offering an inclusive and accessible AI experience for non-English speakers. While the cost and launch date remain undisclosed, Nikhil Malhotra, the global head of Makers Lab at Tech Mahindra, affirms the goal of creating a 7-billion parameter language model as the first step.

Bridging Linguistic Diversity with Cultural Sensitivity

The importance of cultural nuances in language cannot be overstated. Tech Mahindra recognizes the significance of cultural sensitivity in effective communication. An Indic language model could be tailored to prioritize local customs and norms, ensuring that the generated content is respectful and contextually accurate. This approach would not only enhance user experiences but also democratize AI usage across a broader spectrum of linguistic diversity in India.

Democratizing AI: Empowering Non-English Speakers

One of the primary objectives of Project Indus is to empower non-English speakers in India. The ChatGPT model, driven by OpenAI’s GPT models, has revolutionized communication. An Indic LLM (Large Language Model) could amplify this impact by offering a cost-effective solution for content generation in Indic languages. The model would democratize AI adoption, enabling more sectors like healthcare, retail, and tourism to benefit from AI-powered services.

Collaborative Efforts: Building Robust Datasets

The foundation of any AI model lies in its datasets. While English datasets are abundant, the scarcity of Indic language datasets poses a challenge. Various stakeholders, including the Indian government, educational institutions like IISc and IIT Madras, and Microsoft, have joined forces to address this gap. Tech Mahindra’s collaboration with these entities demonstrates a united effort to enhance AI capabilities in Indic languages.

Seeking Contributions: Building Datasets for Dialects

Tech Mahindra’s pursuit of excellence in Project Indus involves seeking contributions from speakers of various dialects. The company invites individuals to participate in the creation of robust datasets by providing voice samples and expressions in their respective dialects. Through an interactive portal, users can contribute to building a comprehensive linguistic resource, fostering inclusivity in AI technology.

Striving for Fairness: Addressing Biases

Tech Mahindra is committed to addressing biases that can inadvertently creep into AI models through biased datasets. By utilizing both human annotation and automatic techniques, the company ensures that racial, ethnic, and gender biases are minimized. This approach aligns with the goal of creating an AI model that respects and represents the diverse voices and languages of India.

The Road Ahead: Challenges and Triumphs

The success of Project Indus hinges on various factors, from meticulous data collection and model training to overcoming linguistic intricacies. Tech Mahindra’s innovative endeavor has the potential to reshape the AI landscape, offering a “Made in India” solution that bridges linguistic gaps, preserves cultural heritage, and empowers millions of non-English speakers. As the project progresses, the future looks promising, with an AI model that embodies diversity, fairness, and technological excellence.