The article states that the humans argued with machines on points which the authors decided have no "objectively right or wrong" answers. So from the disagreeing humans' point of view, the machines were "wrong."

If the humans were not persuaded to abandon their presumably long-held moral beliefs by a bit of machine-generated text, there are some entirely rational explanations for that which were not discussed at all — such as the machine-generated arguments being sufficiently flimsy that the humans were right to reject them. The lack of any consideration of possibilities such as that is what I meant to point out.

That such considerations did not occur to you, or to the writer of the article, as worthy of discussion might be a sign of how far the AI mania has gone. The main risk to society that was demonstrated was not that people will look skeptically at machine-generated arguments that they're predisposed to disagree with — it's that people will fail to do so when the machine says something they're already inclined to believe.