But what we’re saying is that the process is totally different - it’s only the result that is similar. The AI isn’t “misreading” the question - it understands that it’s comparing pounds of bricks to a distinct number of feathers. The issue is that when it searches its database for answers to questions similar to the one it was asked, and sees that the answer was “they’re the same,” and incorrectly assumes that the answer is the same for this question. It’s a fundamental problem with the way AI works, that can’t be solved with a simple correction about how it’s interpreting the question the way a human misreading the question could be.