Ok, so let's put our neural network inside a robot with sensors, so it can see, hear, smell, taste and touch its environment. Let the robot wonder around and learn how to do various tasks (imitation learning, self-supervised learning, reinforcement learning, etc). Let's say after a few years of such training it starts to behave like it was sentient. How do you know what structures have formed in its multi-trillion parameter space? How do you know if those structures are fundamentally different from human brain structures responsible for sentient behavior?