I think trying to understand the mind must be in one of his main motivations. If it wasn't for that, he would have not contributed to machine learning to begin with. So going off the deep end is a side effect of whatever it is that made him a great researcher.