Empowering young minds: How 4 friends are teaching AI in low-income communities

Time of India

11-06-2025

Pune: "Why are firefighters always men? Why is a black, old, fat woman never the first image when we ask for a person?" These were some of the sharp questions posed by 11- to 14-year-old children learning about artificial intelligence (AI), its reasoning, and its biases.
As part of Pune-based THE Labs, a not-for-profit organisation founded by four friends, these children from low-income communities are not just learning how AI works but also how to challenge and reshape its inherent prejudices, how to train it, how to leverage it, and how to evaluate it.
Since June 2024, its first cohort of 20 students explored AI through image classification and identification, learning how machines perceive the world.
Now, they are gearing up to train large language models, equipping themselves with skills to shape AI's future. A new batch of 63 students has joined.
THE Labs is a non-profit after-school programme blending technology, humanities and entrepreneurship. It was founded by tech entrepreneurs Mayura Dolas and Mandar Kulkarni, AI engineer Kedar Marathe, and interdisciplinary artist Ruchita Bhujbal, who saw a gap — engineers lacked exposure to real-world issues, and educators had little understanding of technology.
"We first considered building a school, but the impact would have been limited. Besides, there were logistical hurdles," said Dolas, who is also a filmmaker.
Kulkarni's acceptance into The Circle's incubation programme two years ago provided 18 months of mentorship and resources to refine their vision.
In June 2024, THE Labs launched a pilot at a low-income English-medium school in Khadakwasla, training 20 students from standards VI-VIII (12 girls, 8 boys).
With no dedicated space, they conducted 1.5-hour morning sessions at the school. Students first learned about classifier AI — how AI identifies objects — and image generation AI, which creates visuals based on prompts.
Through hands-on practice, students discovered how AI's training data impacts accuracy and how biases emerge when datasets lack diversity. They experimented with prompts, analysed AI-generated images, and studied errors.
"We asked them to write prompts and replicate an image, and they did it perfectly. That is prompt engineering in action," Dolas said.
A key takeaway was AI bias. Students compared outputs from two AI models, identifying gaps — such as the underrepresentation of marginalised identities. "For example, children realised that a black, fat, older woman was rarely generated by AI. They saw firsthand how biases shape digital realities," Dolas added.
Parents and students are a happy lot too. Mohan Prasad, a construction worker, said he is not sure what his daughter is learning, but she is excited about AI and often discusses its importance at home.
Sarvesh, a standard VIII student, is thrilled that he trained an AI model to identify Hindu deities and noticed biases in AI searches — when prompted with "person", results mostly showed thin white men. "I love AI and want to learn more," he said. His father, Sohan Kolhe, has seen a surge in his son's interest in studies.
Anandkumar Raut, who works in the private sector, said his once-shy daughter, a standard VI student, now speaks confidently, does presentations, and is more outspoken since joining the programme.