I would be interested in seeing if two neural network models with equivalent topology and weights could always be said to 'have' the same subject matter. (It's

I would be interested in seeing if two neural network models with equivalent topology and weights could always be said to 'have' the same subject matter. (It's a better example than for-loops, because I think both brains and computers can embody what it means to have a neuron in a close-enough fashion). I would expect the answer would be "no". You would have to know about the environment that trained them, which blows the door open for something which is irreducible without knowing about environmental conditions, and biology having too rich of an environment to be as precisely reducible as computers might ever be without them being also a computer system. In my other posts in this thread I emphasize the "in-principle" and "in-practice" distinction because I don't believe it is the "in-principle" part of reductionism that ought to be at stake. It's if we can actually do it in a way that doesn't invite approximations.