# AI Fluency: Framework & Foundations

- how to collaborate with AI system as trust partner. Finding possibility with AI.
- a practical bunch of skill in 4 keys:
  - effective
  - efficient
  - ethical
  - safe
- 4Ds
  - Delegation: when should human do the work and when should AI?
    - understand your goal and the problem that you are trying to solve.
    - know what AI systems can and can't do well
    - Decide how to divide the work between you and the AI
    - Generative AI fundamentals (FYI)
      - Traditional vs Generative AI
        - Traditional AI → analyzes patterns and solves specific tasks
        - Generative AI → trains on data, analyzes patterns, and generates human-like language and content
      - three pillars
        - algorithms (understand language in context)
          - Neural networks: inputs → predict what is the next token
          - transformers
        - data
          - articles and websites
          - code and multimodal content
        - computation (massive)
          - GPUs, TPUs, etc.
          - Computing clusters
      - AI context window:
        - prompts (your instructions)
        - AI responses
        - Any other info you've shared
      - What makes generative AI powerful
        - processes vast information during training and learns complex patterns.
        - adapts to new tasks through in-context learning
        - demonstrates emergent capabilities from scale
      - Capabilities and limitations
        - Knowledge cutoff date: Claude's training data extends through January 2025
        - Hallucination: generates plausible-sounding but inaccurate or fabricated information
          - Causes include: training data limitations, token prediction uncertainties, out-of-distribution prompts, and context window constraints
        - Non-deterministic output: produces varied responses even with identical prompts (due to sampling in the generation process)
        - Reasoning limitations: Extended thinking helps Claude tackle complex problems by spending more computational effort on reasoning before responding
      - Human vs AI comparison
        - Human strengths
          - critical thinking
          - judgment
          - creativity
          - ethical oversight
        - AI strengths
          - speed and scale
          - pattern recognition
          - processing vast information rapidly
  - Description: How do we communicate clearly with AI systems?
    - what you want the final output to be
    - how you want the AI to approach the task
    - How you want the AI to behave -- tone and style
    - Don't make AI guessing your mind; tell it step by step like a cookbook
    - Think of AI as your partner
    - Explain task, ask questions, provide context, and guide the interaction
    - Build a shared thinking environment
    - Effective Prompting Techniques
      - What we want (output goals)
      - How we want it done (approach and method)
      - How we want to interact with the AI (tone, style, format)
      - Prompt engineering is simply the practice of designing effective instructions for AI systems
      - tips:
        - context (background information)
        - example (show, don't just tell)
        - output constraints (format, length, structure)
        - breakdown (step-by-step reasoning)
        - space to think (allow extended thinking for complex problems)
        - roles/persona (assign a perspective for the AI to adopt)
  - Discernment: How do we evaluate what AI gives us?
    - Is the output useful and correct?
    - Is the AI taking the right approach?
    - Is the AI behaving as desired?
    - Need: Domain experts and understanding of how AI systems work and their typical limitations
    - checklist
      - for product quality
        - Is it factually accurate?
        - Is it appropriate to audience and purpose?
        - Is it coherent and well-structured?
        - Does it meet my requirements?
        - Does it add value?
      - for process quality
        - logical inconsistencies
        - lapses in attention
        - inappropriate steps
        - getting stuck on one small detail
        - getting trapped in circular reasoning
      - for interaction performance
        - Communicates information at the right level
        - Responsive to feedback
        - Interacts efficiently
      - feedback and correction
        - Specify the problem clearly
        - Clearly explain why it is a problem
        - Provide concrete suggestions for improvement
        - Revise the instructions or examples based on feedback
  - Diligence: How do we ensure our interaction with AI is responsible, transparent, and accountable?
    - Ensuring accuracy and taking responsibility for outputs
    - Honesty and transparency in AI limitations and capabilities
    - Ethical use and critical awareness
      - personal guidelines
      - organizational policies
      - professional standards
      - industry codes of conduct
      - legal and regulatory requirements
    - Fair, safe, and beneficial outcomes
- [Vocabulary](./AI_Fluency_vocabulary_cheat_sheet.pdf)
- Three ways to interact with AI
  - Automation: following fixed instructions → when you have a clear, well-defined outcome
  - Augmentation: collaboratively work together → AI acts as creative and solving partner
  - Agency: operates independently on your behalf → assign role and rough direction; AI acts with autonomy to handle unforeseen cases
  - implementation steps:
    - Set up vision and goals of project
    - Task breakdown and delegation analysis
    - Execute your project using description-discernment loops
      - Describe (communicate clearly)
      - Discern (evaluate output quality)
      - Refine (provide feedback)
      - Integrate (use the result and iterate)