Explore the Human-AI Perception Gap with Hot or Slop
Introducing Hot or Slop, a visual Turing test blending gaming with cutting-edge research. Swipe through images and guess if they’re AI-generated or real!
Key Insights:
- Player Performance: Average accuracy is 65%, with top players achieving 85-90%.
- Model Challenges: FLUX and Imagen models deceive users over 80% of the time.
- Critical Findings: Speed counts! Rushed guesses drop accuracy by ~15%.
Every interaction contributes to a dynamic dataset tracking which AI models are trickiest to identify, revealing how our abilities evolve as AI progresses.
Tech Highlights:
- Built with React, Express, and SQLite, processing ~15 metadata points per guess.
- Current scaling challenges prompt a shift to better-sqlite3 or Postgres for enhanced performance.
Join the fun! Test your skills at Hot or Slop — it’s strangely addictive. Let us know your accuracy rates and detection strategies in the comments. Share your experience and insights! 🌟
