Summary: A team of techies and engineering students have collaborated to build a Small Language Model (SLM) that is capable of generating focused and smaller amounts of content. The SLM, known as AI Chandamama Kathalu, was developed to preserve mythological stories and local folklore from the famous Chandamama magazine. In a four-hour hackathon, volunteers digitized and uploaded a dataset of over 40,000 pages of stories from Chandamama. The SLM is currently producing meaningful content and the team plans to release a Large Language Model (LLM) in March, which will have the ability to generate lengthy outputs. This project aims to build a Language Model for local Indian languages and create an AI solution for short stories.
The team behind this initiative includes Kiran Chandra, founder of Swecha and a Free Software Movement of India activist, Chaitanya, Chief Product Officer and Co-Founder of Ozonetel, and Gaurav Raina, a professor at IIT Madras. Chandra explained the importance of preserving the moral and ethical values embedded in ‘Chandamama Kathalu’ and the need for a language model that focuses specifically on storytelling.
The process of digitizing the stories was made efficient by enlisting volunteers from the Swecha community, who completed the task in just four hours. The entire collection of stories is now available on the internet for anyone to access, download, and improve upon.
Chandra compares this effort to the creation of the first Telugu Operating System 20 years ago and sees it as a step towards democratizing technology. The team plans to continue their work in this space and use their learnings to develop Generative AI solutions for other Indian languages.
This innovative AI model brings new life to beloved cultural stories, ensuring that future generations will have easy access to these cherished tales. With the ability to generate tailored content, the SLM opens up endless possibilities for preserving and reviving cultural narratives.
The source of the article is from the blog papodemusica.com