Tech Mahindra, one of India’s leading IT companies, recently unveiled a groundbreaking project called IndicChat. This project aims to develop the largest Indic Language Model (LLM) to date, specifically designed to comprehend and generate content in Indic languages. The primary goal of IndicChat is to provide a more efficient and culturally sensitive AI communication platform for the non-English speaking population in India.
While existing language models like OpenAI’s GPT models have revolutionized AI communication, their proficiency in Indic languages is limited due to their predominantly English training datasets. This drawback prompted Tech Mahindra to take the initiative in developing an Indic-based foundational model for Indian languages. According to CP Gurnani, the Chief of Tech Mahindra, IndicChat has the potential to cater to a significant portion of the world’s population, approximately 25%.
The first phase of IndicChat will focus on developing a 7 billion parameter LLM capable of supporting 40 different Hindi dialects. Gradually, more languages and dialects will be added to broaden its reach. The versatility of IndicChat holds immense potential for various industries, such as healthcare, retail, and tourism. The model’s capability to perform tasks like Q&A and fill-in-the-blanks using the same engine makes it an attractive solution for specialized industries.
Moreover, the cost of generating content in Indic languages is significantly higher using the existing GPT models. IndicChat aims to provide a more cost-effective alternative, eliminating the token pricing constraints associated with Indic languages. Additionally, this initiative contributes towards the preservation of unrepresented languages and allows Tech Mahindra to leverage the model’s techniques to benefit their customers.
FAQ:
Q: What is IndicChat?
A: IndicChat is Tech Mahindra’s revolutionary project that aims to develop the largest Indic Language Model (LLM) capable of comprehending and generating content in Indic languages.
Q: Which languages will IndicChat initially support?
A: IndicChat will initially support 40 different Hindi dialects. More languages and dialects will be added in subsequent phases.
Q: How will IndicChat benefit non-English speakers in India?
A: IndicChat will cater to a wider section of non-English speakers in India, democratizing AI technology and providing a culturally-sensitive communication platform.
Q: What are the industries that can benefit from IndicChat?
A: IndicChat’s versatility makes it suitable for specialized industries like healthcare, retail, and tourism, where it can perform tasks such as Q&A and fill-in-the-blanks.
Q: How will IndicChat address biases in datasets?
A: Tech Mahindra will implement guardrails during the dataset collection process to ensure there are no biases. Human annotation and automatic techniques will be utilized to mitigate racial, ethnic, and gender biases.
Sources:
– [Tech Mahindra](https://www.techmahindra.com)