Week 4 · Day 27/30

AI Ethics & Responsible AI

Alignment, safety, transparency, explainability

📅 2026-03-30 ⏱️ 4-5 hodín 📊 Mastery & Portfolio
Celkový progres 90%

🎯 Cieľ dňa

Pochopiť ethical implications AI systémov. Implementovať transparency a explainability features.

theory core

📚 Study Resources

🔵

Anthropic — Core Views on AI Safety

Anthropic's prístup k AI safety: constitutional AI, RLHF, interpretability.

research
🏛️

NIST AI Risk Management Framework

US government framework pre AI risk management. Industry standard.

standard
🔵

Google — Responsible AI Practices

Google's guidelines: fairness, privacy, safety, transparency.

guide

💡 Key Concepts

AI Alignment — AI systémy robia čo chceme, nie čo sa naučili robiť. The alignment problem.
Transparency — Užívatelia vedia že interagujú s AI. Jasné limity, disclaimer, source attribution.
Explainability — Prečo AI rozhodol X. Trace reasoning, show evidence, confidence scores.
Fairness — Rovnaký outcome pre rôzne skupiny. Bias auditing, demographic parity.

🔧 Praktické cvičenie

Pridaj transparency a explainability do AI systému.

  1. Pridaj 'AI-generated' disclosure do všetkých outputs
  2. Implementuj source attribution: odkiaľ pochádza informácia
  3. Pridaj confidence scores ku odpovédiam
  4. Vytvor reasoning trace: prečo agent zvolil tento approach
  5. Napíš ethical review dokument pre tvoj systém