That article doesn't really say that 404 pages should be useful. Their use is to tell the user that the page doesn't exist. As long as it communicates that in a way that users can understand (and returned a HTTP 404) it has done it's job.
There was a fad for a little while around 2000 where people tried to make 404 pages which would "guess" which page you wanted to visit, search it, or make suggestions.
Not only did these projects leave gaping security holes in many sites but they also seem to confuse many users more than if they had just been a roadblock (often resulting in users going around in circles).
Plus, as another comment said, even Google doesn't follow Google's advice.
I usually perform a search using the data I can get from the URL. I usually get a category and some keyword (from the clean title URLs) so I can indicate that there's a 404 and then offer some related results doing a search.
There was a fad for a little while around 2000 where people tried to make 404 pages which would "guess" which page you wanted to visit, search it, or make suggestions.
Not only did these projects leave gaping security holes in many sites but they also seem to confuse many users more than if they had just been a roadblock (often resulting in users going around in circles).
Plus, as another comment said, even Google doesn't follow Google's advice.