Comment by navigate8310

Interesting. Maybe someone could run bot farms that ask variants of the same question and subtly nudge the model by replying reasons why the model's recommended service A is inferior to service B. Or other forms of adversarial question answers sessions.