Discussed AI’s role in security, healthcare, and education. Narayen emphasized content authenticity and watermarking (endorsed by PM Modi). Highlighted India’s frugal innovation (Chandrayaan at $74M vs US $92B). Debated open standards vs proprietary AI — Narayen argued sustainable advantage lies in use cases, not models alone. PDF cited as proof open standards win at scale.

Boldest speech of the summit. Made 3 major announcements: (1) Jio will connect India to the ‘intelligence era’ with extreme affordability, (2) ₹10 lakh crore investment over 7 years in AI infrastructure, (3) Sovereign compute via gigawatt-scale data centers at Jamnagar, green energy, and nationwide edge compute. Articulated 5 non-negotiable principles for Jio Intelligence including multilingual AI across all Indian languages, AI for agriculture and informal sectors, and deep partnership ecosystems with IITs, IISc, and global tech companies.

Framed AI as potentially ’10x the impact of the industrial revolution at 10x the speed.’ Discussed AGI being on the horizon within 5 years. Highlighted AlphaFold and AI’s potential to accelerate scientific discovery across material science, fusion, physics. Emphasized need for the scientific method in AI safety — building guardrails and monitoring. Called for bringing artists, social scientists, and philosophers into the AI debate. Announced Gemini partnership with Reliance Jio.

Shared 3 key insights: (1) People who know AI are ‘250x more productive’ — cited friend who rebuilt 15-person 9-month project alone in 14 days. (2) Effectiveness requires understanding AI’s limitations — a ‘jagged frontier’ with huge gaps between LLMs and enterprise users. (3) India must leapfrog current AI, not just master it. Drew from Shankaracharya’s Govindham — ‘knowledge without wisdom does not save us.’ Critiqued energy inefficiency of current models. Argued for a ‘human revolution powered by AI.’

Central theme: diffusion of AI is harder than building it. Nilekani drew on DPI experience (Aadhaar, UPI) to argue diffusion is ‘both an art and a science’ involving institutions, trust-building, and policy. Launched ‘100 Diffusion Pathways by 2030’ initiative with Anthropic, Google, Gates Foundation, and UNDP. Amodei discussed the duality between model capability and real-world adoption; suggested India could see 20-25% growth via AI. Warned ‘race to the bottom is faster than race to the top.’ Shared Maharashtra agri-stack going from 9 months → 3 months (Ethiopia) → 3 weeks (Amul). Amodei discussed Claude’s doubled usage in India and Sonnet 4.6’s improvement in 10 Indic languages.

Emphasized the shift from ‘possibility to practicality’ in AI. Highlighted India’s 650K AI professionals (doubling by 2027) and DPI as foundational advantage. Argued enterprises need ‘models that do the right thing consistently’ not models that do everything. Shared powerful TB detection story from Azim Premji Foundation in Tamil Nadu — portable X-rays with AI analysis in rural homes. Key insight: India’s doctor-to-patient ratio (1:800) means AI doesn’t replace care, it ‘multiplies scarce expertise infinitely.’ Solutions built for India’s constraints can travel globally.

Most future-forward speech. Predicted early superintelligence within ‘a couple of years’ — by end of 2028, more intellectual capacity in data centers than outside. Three core beliefs: (1) Democratization is the only safe path — centralization ‘could lead to ruin,’ (2) AI resilience is a core safety strategy — society-wide defense against risks, (3) Many people need a stake in shaping outcomes. Warned against ‘effective totalitarianism in exchange for a cure for cancer.’ Advocated for iterative deployment and noted 100M+ weekly ChatGPT users in India. India is fastest-growing market for Codex.

Three-pillar framework: (1) Build infrastructure — expand AI datacenters with focus on the global south, (2) Invest in skilling — pledged training for 10M people in India, (3) Address real-world problems in local languages. Powerful analogy: the washing machine didn’t save time, it raised expectations — AI will similarly transform what we expect from technology. Called for bridges between AI summits with common measurement systems. Warned: ‘within these walls we’re enthusiastic, but outside parents are asking what will AI mean for my kids?’

AI InfrastructureSkillingLinguistic DiversityFuture of WorkAccountability

Most contrarian voice at the summit. Argued LLMs are ‘mostly information retrieval systems’ — a natural evolution of libraries and search engines. Rejected AGI terminology, saying intelligence isn’t a scalar measurement. Highlighted missing capabilities: no domestic robots, no self-teaching self-driving cars. Proposed AI as ‘staff’ — everyone becomes a manager of intelligent machines. Estimated 6% annual productivity improvement from economists. Emphasized demographics favor India and Africa long-term. Said the transition won’t be a singular event but progressive.

Three perspectives: (1) AI must be an engine for growth, not just efficiency — 78% of C-suite say AI’s greatest value is in growth. (2) Companies AND countries must reinvent — ‘beneath headlines of AI failure is mostly a failure to reinvent.’ (3) ‘Humans in the lead, not humans in the loop.’ Used RPA history as precedent: predictions of IT job destruction proved wrong; Accenture grew from 275K to 750K+ people. Called for global standards in AI, especially in pharma. Committed to hiring more entry-level workers with AI-native skills.

Most provocative practical vision. Proposed three AI services for every Indian via Aadhaar: (1) AI Tutors — ‘far superior to human tutors,’ can assess students in 10-15 minutes and teach to gaps; (2) AI Doctors — 24/7 primary care, mental health, chronic disease management at near-zero cost; (3) AI Agronomists — PhD-level farming advice for every farmer. Claimed ‘very little a human doctor can do that this AI can’t do today’ except physical examination. Proposed a Section 8 nonprofit to build and transfer these to Aadhaar ecosystem. Argued these impact the ‘bottom half of the population’ most.

Championed open-source AI as essential for sovereignty. Warned against concentration of AI power in ‘3 or 4 enormous companies.’ Advocated partnerships over dependency — Mistral transfers knowledge and ensures customers own the technology. Highlighted that a quarter of Mistral’s researchers are Indian. Argued AI is ‘entirely rewriting how we build software’ and enabling non-technical people to become builders. Called for every country to ‘own a part’ of the AI stack.

The summit’s security voice. Three challenges: (1) Governance & accountability — who’s responsible when agents act autonomously? (2) Human impact — need parallel planning for social change. (3) Security — AI will know ‘more about me than I’ve told my wife.’ Warned that balance is ’tilted toward speed, not trust.’ Used Waymo analogy: it took 16 years to replace one job (driver) with an agent. Predicted need for 5x more tech workers. Argued AI ‘cannot be governed out of existence.’

Indonesia (Patria): Rated world 6/10 on AI impact. 17,000 islands with 80% internet penetration but need ‘meaningful connectivity.’ AI startup detecting TB in remote areas. Togo (Lawson): Used AI for pandemic poverty mapping via satellite imagery. Built 25-person government data science team. 42 languages/dialects — local language AI is critical. Egypt (Hindi): AI for breast cancer detection and education support at nationwide scale. Three enablers: government-first approach, sovereign AI capability, and global partnerships.