No, it's not philsophical. Because if you optimize to predict, you are doing something different to optimizing for a reward. It's a different process - different objective function, different optimization, different set up.

[dead]