A team is developing a language model for two distinct tasks. For Task A, a medical information bot, the model must provide the most factually accurate and reliable information. For Task B, a creative story generator, the model needs to produce diverse and interesting narratives. The team implements a prediction rule for both tasks where the model always selects the single word with the highest calculated probability to be the next word in a sequence.

Analyze the suitability of this prediction rule for both Task A and Task B. Explain why this rule might be highly effective for one task but lead to undesirable results in the other.