The article essentially claims that the more people there are the more likely it is that _someone_ will intervene. Which isn't the same as no one feeling less inclined to help since there are other people around.
That Slate dot com tier take became a theory, developed experiments, and discovered something that was obviously mathematically true: if not everybody in a group helps someone that they see in distress (only a subset do), but individuals always help when alone, then in groups, the average time before response for each individual is going to go up, and the average likelihood of any response at all from a particular individual is going to go down.
This would be true even if people were always helped faster by groups rather than individuals, and a subset of every group always intervened. The types of people that bystander experiments found to be likely interveners could just adhere to the stereotypes that people think of as physically authoritative and look to when something occurs. People look to them to intervene (large, male, athletic, maybe ethnic or rough looking in criminal/street situations) maybe those people intervene because there no one for them to pass the responsibility off to.