but doesn't the AI node in BP utilizes the same GPTs under the hood ? why would these results differ from each other ?