Zodpovědný uživatel AI

• Design clear prompts with useful context and constraints.
• Review outputs for factual, privacy, bias, and copyright risks.
• Build a repeatable human-in-the-loop workflow.

Doporučuji nejprve: AI základy

1. 1

   Prompting with context

   Výsledek: Turn vague requests into testable instructions.

   Praxe: Rewrite a weak prompt with audience, context, constraints, examples, and success criteria.

   Prompt and context designMechanika jazykových modelů
   Prompt Engineering →ChatGPT a LLM →
2. 2

   Reviewing outputs

   Výsledek: Detect unsupported claims and decide what requires verification.

   Praxe: Annotate an AI answer as verified, attributed, uncertain, or unsupported.

   Output review and error detectionBias and fairnessOvěření zdroje a tvrzení
   AI halucinace →AI zaujatost →
3. 3

   Privacy and ownership

   Výsledek: Choose safe inputs and understand content rights.

   Praxe: Create a red/yellow/green data policy for an AI workflow.

   Privacy and data handlingCopyright and content provenance
   AI a soukromí →AI a autorská práva →
4. 4

   A dependable workflow

   Výsledek: Combine AI assistance with checkpoints and human ownership.

   Praxe: Map an existing task and mark where AI may draft, check, or never decide.

   AI-assisted workflow designTool selection and comparisonHranice automatizace a ochrany
   AI v každodenním životě →Nástroje pro psaní AI →