Test Driving ChatGPT-4o

Just like people:

  • asking an LLM to go slowly and step by step avoids errors
  • providing distraction (images to consider, etc) increases errors
Comments
www.joshbeckman.org/notes/40594281