Inclusive AI Starts with People, Not Just Algorithms

Contents

Executive Summary

This panel discussion at an India AI Summit emphasizes that inclusive AI development requires centering human potential, diversity, and equitable access rather than focusing solely on technical advancement. The speakers—representing investment, entrepreneurship, enterprise AI, chip manufacturing, and grassroots education—argue that India's competitive advantage lies in rapidly applying AI across sectors and including historically excluded populations in AI development and benefits, rather than chasing foundational model development.

Key Takeaways

Inclusive AI is Faster, Better AI: Bringing diverse talent—women, youth, people from low-income backgrounds, underrepresented geographies—isn't just ethical; it's a competitive strategy. India's advantage lies in rapid execution and broad participation, not foundational research.
Apply Before You Build: Focus on domain-specific, high-impact applications (agriculture, healthcare, automotive) using fine-tuned models rather than chasing the foundational LLM race. This creates economic value faster and builds trust in AI.
Start Now, Unlearn Later: Training young people and newcomers in AI is more effective than retraining industry veterans because they lack ingrained assumptions. Grassroots, rapid-access training programs work and can transform lives within weeks.
Build Wisdom, Not Just Information Access: In an age of abundant AI-generated data, the rare skill is asking the right questions, thinking critically, showing resilience, and understanding emotional/social intelligence. Parents and educators should emphasize these over rote knowledge.
Geography & Decentralization Matter: Avoiding concentration of AI talent in metros and instead building centers of excellence in Tier 2/3 cities prevents inequality, distributes opportunity, and leverages region-specific expertise (e.g., tailoring backgrounds → computer vision excellence).

Key Topics Covered

Inclusive AI & Diversity: Women's representation in tech; building 50/50 gender balance in AI; reaching youth and underprivileged populations
AI Kiran Initiative: A community of 10,000+ women in AI across India; scaling from 10 women in ChatGPT results to 10,000 self-organizing members
Geographic Decentralization: Establishing AI centers of excellence in Tier 2 and Tier 3 cities (Kolkata, Visakhapatnam, Coimbatore, Shillong, Hubbli) rather than concentrating talent in metros
Applied AI vs. Model Race: Focus on domain-specific applications (precision agriculture, healthcare AI, automotive AI) rather than competing in foundational LLM development
Workforce Transformation: Reskilling and upskilling workers; AI's effect on employment; enabling rapid job transitions (6-week training leading to $120K jobs)
Education Reimagining: K–12 and higher education disruption; EQ (emotional quotient) vs. IQ; teaching resilience, critical thinking, and the power of asking good questions
Child Development & Parenting: Five senses and nine emotions (Navarasa); building curiosity; failed fast and learn; protection from AI harms while enabling AI literacy
Infrastructure & Compute: Balancing investment in technology, infrastructure, and human intelligence; ensuring compute access is not a limiting factor for ambitious teams
Trust & Provenance: Building trust in AI-generated information; how to verify and curate AI outputs in an age of AI-generated content
Digital Gap & Grassroots Training: Addressing pre-digital literacy barriers before teaching AI; starting from foundational technology skills

Key Points & Insights

Diversity as Competitive Advantage: Radha Basu's iMerit employs 53% women and 10,000+ people globally, with 3,500 in India. She explicitly states: "If anybody asks you how do you run a company with 50/50 women, look them straight in the eye and say, 'Have you seen the world lately?'" Gender balance is achievable and necessary.
Rapid Scaling Through Community: AI Kiran went from documenting 10 women in ChatGPT results (2023) to 10,000+ self-organizing women and youth in ~6 months. This demonstrates that visibility and community platforms can catalyze exponential growth in participation.
Geography as Leverage: India's advantage is not in building foundational models (competing with OpenAI, Google DeepMind) but in applying AI across diverse sectors and regions. Establishing centers of excellence in smaller cities (Kolkata for generative AI, Visakhapatnam for healthcare AI, Coimbatore for automotive AI) creates distributed talent and prevents brain drain.
The "Unlearning Problem": Radha Basu notes it's harder to train experienced industry professionals in AI because they must unlearn old paradigms; training youth from low-income backgrounds or rural areas is more effective because they approach AI without baggage or assumptions.
Compute ≠ Limiting Factor: Asha George (AMD) counters the assumption that chip access will concentrate AI power. The real limiting factor is "ambition you carry" and execution speed. India's track record (moving faster than competitors in services) suggests those with bold vision and speed will succeed regardless of compute constraints.
Applied AI Creates Societal Impact: Small vision models and language models fine-tuned for precision agriculture (crop failure detection) and healthcare (breast cancer screening for Indian women) drive real economic value and societal benefit faster than pursuing general-purpose large models.
Emotional & Relational Skills Over IQ: Multiple speakers (Anurag, Radha) emphasize that in an AI-augmented world, EQ (emotional quotient), resilience, curiosity, and the ability to ask meaningful questions matter more than raw IQ. Arts, music, and creative disciplines are essential to human development.
Rapid Reskilling is Possible: Mihir Shukla (Automation Anywhere) cites examples: 500 women in Africa trained in 6 weeks, 500 employed within a week; Mississippi Delta residents earning $12/hour flipping burgers now earn $120K/year after 6-week AI training. This technology democratizes upward mobility.
Youth as Wisdom Source: Kids and young adults instinctively understand the future better than adults; they recognize when questions are biased toward AI and ask for "questions worthy of humans." Parents and educators should learn from them, not lecture them.
Trust & Provenance Crisis: As AI systems generate content at scale, traditional curation by people is disappearing. Building provenance, verifying sources, and creating guardrails against AI harms—while preserving access to AI's power—is a critical unsolved challenge.

Notable Quotes or Statements

Kitika Chopra (AI Kiran): "What would you do if you weren't afraid?" — framing the core question for taking risks and pivoting to new opportunities.
Lakshmi Venu: "Innovation is happening everywhere... how do you find them, connect them and get them to tell their stories?"
Radha Basu: "If anybody asks you how do you run a company with 50/50 women, look them straight in the eye and say, 'Have you seen the world lately?'"
Radha Basu: "It's harder to take somebody from industry and train them in AI because they have to unlearn first than to take a young person from Odisha from a village and skill them in AI."
Asha George (AMD): "I don't think the limiting factor is access to chips. I think the limiting factor is the ambition you carry in this space."
Anurag (Manzel Mystics): "It's less about AI, it's more about HI—heart intelligence—because what makes us human is a heart pumping and making us alive."
Mihir Shukla (Automation Anywhere): "You don't have to invent a technology. The success lies in applying that technology in every aspect of economy. That is India's superpower."
Youth Panelist (via Mihir's example): "Ask me something that is worthy of human" — challenging adults to pose meaningful, non-trivial questions that AI cannot easily answer.
Radha Basu (on young talent): "Average age of our company is 24.5. The sassiest people in my company would say if it was not for you it would be 23."

Speakers & Organizations Mentioned

Name	Role / Organization	Key Context
Kitika Chopra	Founder, AI Kiran; Co-Founder, Optimize Jio	Led AI Kiran from 10 women to 10,000+
Lakshmi Venu	Pioneer, AI Kiran	15+ years connecting innovators and storytellers; worked at Intel, VC, philanthropy
Radha Basu	CEO/Founder, iMerit	10-year-old AI company; 53% women; 10,000+ employees globally; centers in Tier 2/3 cities
Mihir Shukla	CEO/Chairman, Automation Anywhere	~500M digital workers on platform; authored "A 5-Year Century"; focus on applied AI and inclusion
Asha George	Representative, AMD	Emphasis on compute infrastructure and execution speed over chip access barriers
Anurag	Founder, Manzel Mystics; AI Kiran Fellow	Mobile music school; 60,000+ children across 900 schools; teaches IP and creativity through music
TCS (Tata Consultancy Services)	IT Services	Ran hackathon with 1,200 non-native English-speaking women; successful rapid upskilling model
John Deere	Industrial Partner	Partner with iMerit on precision agriculture and crop failure detection
Anthropic	AI Research (mentioned speaker: Daria)	Discussed AI application landscape
Google DeepMind	AI Research (mentioned speaker: Sun)	Discussed AI applications
OpenAI	AI Model Provider	ChatGPT mentioned as baseline for comparing women-in-AI representation
Intel	Semiconductor / Corporate (Lakshmi's prior role)	Historical context
HP (Hewlett-Packard)	Technology / David Packard's founding role	Radha's early career; historical example of Silicon Valley creation
I Support	Early unicorn software company (Radha's prior)	Context for Radha's track record
Automation Anywhere	RPA/AI Automation Platform	Platform with global reach; 90+ countries
ServiceNow	AI Kiran member organization	Mentioned as partner
Women for Ethical AI	Movement	Mentioned in relation to building trust in internet

Technical Concepts & Resources

Concept / Tool	Definition / Context
Small Vision Models & Small Language Models	Fine-tuned, domain-specific models derived from large foundation models; more efficient and targeted than general-purpose LLMs
Fine-tuning	Adapting pre-trained models to specific domains (e.g., crop failure detection, breast cancer screening)
Reinforced Learning with Human Feedback (RLHF)	Training technique mentioned for improving model behavior; involves human evaluation and iteration
Red Teaming / Model Tormenting	Adversarial testing of AI models to find failure modes; Radha describes it as "torment it till you break it"
Dataset Creation	Critical step in applying AI; requires domain experts (radiologists, cardiologists, agronomists, etc.) to label and validate data
Computer Vision	Applied focus area at iMerit; used for precision agriculture and healthcare imaging
Natural Language Processing / Generative AI	Centered in Kolkata and Shillong in iMerit's geography strategy
Autonomous Mobility & Robotics	Largest business area at iMerit; centers in Kolkata and Mettur Bruce
Precision Agriculture	AI application: crop failure detection; yield optimization via John Deere partnership
Healthcare AI	Visakhapatnam center of excellence; breast cancer screening tailored for Indian/Asian women parameters
Automotive AI	Coimbatore center of excellence; global competitiveness in automotive AI models
ChatGPT, Perplexity	Consumer AI tools referenced as sources of consumer/business decision-making
Digital Worker / RPA (Robotic Process Automation)	Automation Anywhere's platform; 500M digital workers, ratio 1:20 (1 human to 20 digital workers)
GEO (Generative Engine Optimization)	Kitika's startup Optimize Jio; helping brands rank in AI-driven search (ChatGPT, Perplexity, etc.)
Navarasa (Nine Emotions)	Classical Indian aesthetic/emotional framework; Anurag advocates integrating into child development
Five Senses	Foundational to human development; Anurag emphasizes teaching alongside AI literacy

Implicit Policy & Systemic Themes

While not formalized policy recommendations, the panel implies support for:

Regional AI investment: Government/private funding for Tier 2/3 city AI centers
Inclusive education: Reforming K–12 and higher ed to emphasize EQ, critical thinking, resilience
Rapid reskilling programs: Public/private partnership on short-cycle AI training (6 weeks)
Women & youth in AI: Policy incentives or targets for representation
Guardrails for AI safety: Addressing harms while enabling access (acknowledged challenge, not yet solved)
Digital infrastructure: Ensuring connectivity and compute availability beyond metros

Document Quality Note: The transcript contains significant repetition and some garbled audio (noted by repeated phrases like "you you you," "and and and"). This summary prioritizes clear, substantive content over parsing every artifact, but the core themes and insights are preserved and accurately represented.