who-03-hi

From Human Intelligence to WHO-India’s Needs

Slide 1: Human Intelligence

(The way we’ve always done it. The way they live it.)

# Human Intelligence
- Conclusion: We need a better way.
- Limitations: Time. Bias. Scale. Burnout.
- Results: 1 report. 1 officer. 1 deadline missed.
- Methods: PubMed → Embase → 10,000 abstracts → 3 months
- Background: What we know, what we don’t

“This is your day. This is your burden. This is what we’re here to change.”


Slide 2: AI Powered

(Not magic. Not replacement. Augmentation—by type.)

# AI Powered
- Five types of AI → Five types of tools
  5. Embodied AI → Ultimate goal (no human input needed at any point, Full automation)
  4. Generative AI → Disciplined branches (solving problems in data collection, analysis, reporting, ethics) 
  3. Agentic AI → Pulsing trunk (speed of delivery of results, Human-in-loop)
  2. Perception AI → Smarter roots (rigorous analysis)
  1. World AI → Faster soil (data access, Hybrid intelligence)

“We don’t replace you. We grow with you. One layer at a time.”


Slide 3: WHO-India

(Not global. Not generic. Theirs.)

# WHO-India
- Conclusions: Helpful to have precise language 
    → Vocabulary beyond “AI tools”
    → Understand features of various types of AI
    → The industry's goal is a fully-stack embodied AI (the only form AGI can take)
- Limitations: Invariant aspects of science & public health
    → Bias must be named (by experts, human standards)
    → Gaps must be acted on (collecting, finding, accessing data)
    → Evidence must be found (knowledge gaps)
- Results: Tailored to WHO-India 
    → policy speed 
    → PRISMA, equity 
    → 700+ officers
- Journal & language agnostic
    → Hindi, Tamil, etc (eg OpenEvidence only accesses NEJM, JAMA: Journals in English)
    → Grey lit 
    → Field notes
- Literature most relevant to India 
    → NCDs
    → TB
    → AYUSH, UHC

“This isn’t AI for AI’s sake.
This is your evidence, your language, your impact—made faster, fairer, scalable.”