Now that I've had a chance to sit down and read through both the study and a few other things... you're correct, but it's not completely clear-cut, at least in my opinion, that the changes under consideration actually relate to reward-seeking, addict-like behaviour and aren't simply, say, a lack of sexual development due to being single.

They found a variety of different features in their test subjects (actual anatomical differences, differences in activity level within the caudate, differences in interconnectedness between pudamen and caudate...) and saw these were strongly correlated with level of pornography use, on the basis of addiction. However, there were some people in the study who used alcohol in a mildly problematic way. They showed only a r = ~0.25 (weak positive correlation) with porn usage. That strikes me as pretty inconsistent—if these are pathways strongly implicated in addictive behaviour, why didn't the drunks line up more neatly with their data? They don't mention alcoholism again in the discussion, except to draw parallels between porn usage and various forms of drug usage, and to suggest psychiatrists should ask about porn usage.

I guess you missed this article from two days ago, then? (The classic "mystery" in neural nets is how they distribute weights during learning, the answer to which is "rarely better than a human would, and according to the algorithm they train by.") I know you saw this one; you commented on it. Or perhaps you were talking about computer-automated proofs? Those aren't sophisticated, merely long-winded; the result of applying simple propositional logic over and over again.

If we had algorithms that were actually capable of exceeding human comprehension in a meaningful way—and not just outpacing or outlasting us at regression and tree search—the world would be a much different and more exciting place. It is very unlikely we will have AGI until the human brain is almost fully understood, simply because we don't know what direction we really need to be pushing in. Given that there's now evidence that the brain's neurons could be DNA computers, that definitely has a long way to go.