Beyond black‐box AI: Comparing ChatGPT‐4 interpretability and accuracy to CNNs in melanocytic lesions diagnosis

Key Points

• The objective is to compare the interpretability and accuracy of ChatGPT-4 and CNN models in diagnosing melanocytic lesions.
• Compared ChatGPT-4 and CNN models for melanoma and nevus classification.
• Analyzed the accuracy of classifications without annotations.
• Evaluated the descriptive capabilities regarding dermoscopic features.
• ChatGPT-4 surpassed CNN models in accuracy for melanoma versus nevus classification.
• ChatGPT-4 showed limitations in describing dermoscopic features accurately.