Please enable JavaScript.

Coggle requires JavaScript to display documents.

AI Augmented Engineer - Coggle Diagram

- - - - Gen AI
        
        Produce errors, biases, misleading information
        
        Make better decision
        
        Respond to user instructions to generate new text, image, and insights by analyzing patterns in data
  - - - Automate tasks
      - Personalize content
      - Analyze information
      - Make predictions
    - - Evaluate ideas
      - Broaden perspectives
      - Brainstorm
    - - S2: Determine how to measure results
      - S3: Plan how to observe and document results
      - S1: Develop a hypothesis
      - S4: Evaluate, reflect, refine
  - - - Provide context
        
        Constrains
        
        Background knowledge
        
        Framing detail
      - Assign a persona
        
        Identity
        
        Set of charactersics
        
        Roles
      - Clearly define the task
        
        Pinpoint specific goal
      - Give examples
    - - Ask smart follow-up questions
        
        Clarify vague points and redirect the tool's focus to needs
      - Stay alert to biases
        
        Seek balance by questioning one-sided answers
        
        Asking the tool what perspectives you may be missing
      - Check the logic
        
        Evaluate the reasoning in the output
      - Seek diverse perspectives
        
        Combine AI insights with multiple trusted sources
      - Verify claims -> Cross-references
        
        Statistics
        
        Facts
        
        Data points
    - - Explore instead of solve
        
        Creative and diverse range of possibilities
      - Reimage the problem
        
        Outside traditional frameworks
        
        Future-facing options
- - - - Specific
      - Open-endness
      - Clear
    - - Ambiguity
      - Over-complication
      - Overloading - too much (irrelevant) information
  - - - Audience - tailoring the content
      - Length - brevity or depth <Concise summary, In-depth analysis>
      - Style - define the framework <List, Describe, ...>
      - Tone - mood and feel <Formal, Casual, Playful anecdote>
    - - Accuracy
      - Relevance
      - Logical consistent
      - Factual correctness
    - - Bold -> Emphasize: **
      - Quotation marks "ABC"
      - Header: #
  - - - Prompt ChatGPT with discontinuous tasks
        
        Blending humor with unexpected topics
        
        Encourage creative leaps beyond its training data
    - - Output format
      - Context
      - Persona
      - Examples
      - Instructions
    - - One-shot
        
        Template
      - Few-shot
        
        Multiple examples
      - Zero-shot
        
        Reply on extensive pre-training
      - Chain of thought (COT) promoting
        
        Break down problems into simpler, understandable steps, leading to more accurate and detailed responses
- - - - Training Data Bias
      - Context tracking
      - Knowledge cutoff
      - Hallucination
      - Legal and ethical considerations
    - - Provide examples if necessary
      - Keep it consice
      - Be clear and specific
        
        Necessary information
      - ChatGPT
        
        Understand the prompt
        
        Generate response
        
        Identify the topic
    - - Training Data = Algorithms + Fine-Tuning