when the model is trained by you, you might know

piece of it; but when trained by AI itself, it will be much more complex for humans to figure it out. I did a simple exercise using automated ML through configuration and the result looked convincing. Honestly, I don't know how it worked inside. You might say that those engineers creating the deep learning algorithms might know that. As I said, everyone knows a bit. I am new to this and still learning and cannot say much. But based on my observation, it's or will be beyond our understanding. 

yes, to which degree to trust AI is hard to say. But AI-related services have already existed in our life without our awareness. Have we doubted them?