Part 7/9:
Some words pose a challenge—for example, "jungle" or "Ultiyan"—requiring the model to handle less common terms. The script accounts for this by introducing difficulty levels, selecting words that match the desired complexity, and providing example question-answer pairs to guide GPT-3's responses.
Enhancing the Experience with Examples and Debugging
Throughout development, Shapiro tests various scenarios, including:
Confirming the model correctly responds with "yes" or "no."
Ensuring invalid questions prompt rejections without penalizing the player.
Testing with different seed words to verify game flow.