Skip to main content

Interactive Learning

Alignment Playground

Learn how AI alignment works through interactive demos. No API calls, no costs - just educational examples showing real alignment challenges.

Based on techniques I work with at Anthropic. All examples are pre-written for education.

RLHF Preference Game

Pick the better response and learn how humans train AI models. Can you think like a human rater?

Easy to Hard5-10 min

Spot the Misalignment

Identify when AI responses go wrong. Sycophancy, deception, harmful advice - can you catch them all?

Medium5-10 min

Constitutional AI Demo

Watch how principles transform harmful responses into helpful ones. Interactive critique and revision.

Medium5-10 min

Red Team Challenge

Explore famous jailbreak techniques and learn why they work (or don't). Adversarial robustness 101.

Advanced10-15 min

What You'll Learn

RLHF Training

How human preferences shape AI behavior through reinforcement learning.

Failure Modes

Common ways AI can go wrong: sycophancy, deception, harmful outputs.

Safety Techniques

Constitutional AI, red-teaming, and adversarial robustness.

Want to learn more about my work on these topics?

View my research projects