”[H]ow hard is it to have a comments editor and software that ensures the comments are free of just purely hatefull content? Isn’t computer technology responsible for all kinds of incredible things?”
You would be surprised. The article alludes to the workload a human editor/moderator has, but if you want to have a better idea then try managing such a site yourself. The mods here at Fluther have it easy since, compared to many other sites, Fluther is relatively low-traffic and the people here are generally civilized. Yet even moderating a small site of well-behaved people is work.
If you want an idea of what moderators are up against, imagine trying to keep discipline someplace like /b. (That link actually is to /b, so click at your own risk; you have been warned!) If you don’t know what /b is well enough to grasp the relevance, then think of some other wretched hive of scum and villainy where rules are either weakly enforced or non-existent.
Could you spend 943 hours a day staring at the worst humanity has to offer? It really is that much work; Facebook has an unbelievable number of full-time employees on their payroll that do nothing else in order to split the load enough to keep each mod working a mere 8–12 hours per shift.
So no, having people do it is not feasible. The process would have to be automated… leading to other issues.
Computers may do things you see as “incredible”, but they have very sharply defined limits on things they just cannot do as well as capacity limits that we humans lack. A computer may be able to memorize the entire Library of Congress with ease, but it can’t understand a word of it. Computer are only really able to do four things, though they can do those things better than most humans; memorize large amounts of information, recall that information quickly and accurately (given proper search parameters), multitask, and perform numerical computations.
They lack the ability to make judgment calls, as anyone who has ever used autocorrect can attest. They also lack the ability to do anything that they are not explicitly designed to do through proper programming and use of peripherals. For instance, a computer without some form of camera cannot do anything visual like recognize faces or read facial expressions. True, something like a CMM that has no eyes but a very sensitive probe analogous to the pressure sensing nerves in our finger tips can touch an object in multiple places and “see” the same way a blind person can from mental images by touching something, but even that requires specialized hardware and software.
That software is quite complex too as the computer lacks the ability to extrapolate the way humans can, though proper programming can mitigate that to an extent by saying basically, “If data is incomplete, then assume [whatever].”. Humans don’t need to be told that, so we are FAR superior at dealing with unforseen things like a part being placed on the table more than a few microns out of position. The only reason I never slammed a tool into the table of a CNC mill is because I programmed it correctly. Incorrect programming, including parameters such at Tool Length Offset and work coordinate offsets relative to machine zero will crash the machine, and the machine is too stupid to know better so it would be my fault if it crashed in that manner. Some fancy CNC machines have probes like a CMM in order to do some things like check the position of the part/fixture and adjust offsets accordingly, but those machines have hardware and software that grants them that ability. Furthermore, those programs don’t automatically run (they must be explicitly called up and executed) and aren’t smart enough to handle situations outside their programming.
Now figure, if a multi-million dollar machine cannot think for itself well enough to perform just one task (cut material into a specific shape) that deals with purely unambiguous stuff like “Spindle speed is 3,000 RPM” without massive human oversight and supervision, what do you think it would take to get a computer that could do a wider variety of more complex tasks that involve ambiguity and uncertainty that would force the computer to make judgment calls? Given that computers are so literal, the programming required would be inordinately complex; complex enough that even the large number of people who have tried for decades still haven’t come up with anything much better than Siri. And that software would need some major-league hardware to run; something like a Tianhe-2 that costs $390,000,000 might not even be up for the task.
Sure, moderating the comments section is a smaller task than being a full-fledged human consciousness, but the main thing required for moderation is the hardest thing to get a computer to do; read comments, understand them with consideration given to context (basically disambiguation), and make a judgment call. Getting computers to understand natural language is a task that still eludes us. Even relatively smart computers require humans to alter their syntax; we have to meet the computer halfway. Oh, and gawd forbid that a speech recognition program has to deal with someone speaking Cockney!
So we can’t have a computer do it, at least not any better than rolling dice to decide whether any post is deleted regardless of content. While a computer may see nothing wrong with the phrase, “Your Mother.”, if you asked someone what they did last night and that was their response, context would make it offensive. It is not technologically feasible to expect every site with a comments section to plop down tens of millions of dollars on something that cannot do the job.