Pre-Semester Frontier Post 50
(Facebook’s “Artificial Intelligence” Failing to Protect MCO Accounts)
Artificial intelligence” (AI) is another one of those contradiction in terms you will inevitably have the pure annoyance of running into these days. All social media (Facebook, Twitter, Instagram, YouTube, Google Search) use it to power these protocols that recommend what videos you should watch, rank the results of any search you make, etc.
Its only redeeming virtue, to me, is when you type “idiot” in the dialog box of an image search, what comes back are endless screenfuls of Donald Trump’s face. At least AI’s got that one right. Go ahead, try it yourself—it’s amusing.
Google must have fixed that somewhat after some Republicans roasted some Big Tech execs about it at a congressional hearing. Now the search results for “idiot” is still festooned with Trump icons but this time he’s joined by a bipartisan panoply of other Republican AND Democrat caricatures. Evidently, Google widened the AI word association making it more generic, from “idiot=Trump” to “idiot=politician.” I’d go along with that still.
But more recent personal experience has only upped my cynicism about AI to think that while to err is human, to totally crash the whole system you need AI.
Mia, a friend of mine who keeps the FB account “Mia Magdalena” (I totally recommend you follow her) woke up one morning to find a couple of fake FB accounts built around a stolen image of hers and a fusion of her name with some other unrelated persons’ names. The account is as fake as a three dollar bill, but the use of her avatar and first name combination produces a convincing-enough profile that some airheaded trolls are now using to proselytize some of her following. Worse, these trolls now post messages under her pseudonym channeling the mirror opposite of her public advocacies.
These are very short messages—20 words or so—which is the thought composition limit of your average troll. As expected ‘dey mispel werds’ and could never get subject and predicate to see eye-to-eye. You’d have no trouble telling these were not her thoughts.
I was warned about this last night so I quickly blocked the accounts. But all that did was to hide their nefarious activities from my view. These trolls still lurk in the murky cyberwaters of e-manipulation. It’s an annoyance more than anything else, like a bad TV commercial jingle on endless loop that just won’t leave your head.
According to FB’s advisory to me, they decided not to take down the accounts until the trolls have posted material that was actually harmful to the public interest. They’ve gotta be kidding—or at least they should know they’ve bitten off something larger than they can chew. I am no paragon of virtue myself but I would have a hard time making that judgment call of what’s harmful to public interest and what’s not. A computer doing the same task would face the same daunting challenge faced by a blind man in dark room looking for a black cat that isn’t there. Artificial intelligence, my foot.
I’d settle for plain old reference to registries. What is so difficult about pulling out Mia’s avatar—which is bitmapped to the last pixel—and seeing it matched to another name, no matter how uncannily similar, to conclude that the avatar had been hijacked? When is FB going to realize that the shortest path between two points is an unsophisticated straight line?
This is the problem with “artificial intelligence” and its cousin “machine learning.” They both cannot make the distinction between fact and fake except by using the metrics of user response. Perhaps if more people filed reports involving these same accounts, it might trigger a subroutine in the program code of whatever artificial “brain” these machines use. Finally it might make the mathematical inference “like < dislike = error 404”
A couple of months ago, I had a video of mine taken down by YouTube when I mentioned that some Catholics are leery about COVID vaccines because some vaccines against other diseases in the past utilized stemcells harvested from human placenta. Even that is largely a myth but is now being used by some conspiracy theorists to fuel vaccine hesitancy, the very thing my video was trying to fight.
But I got a “strike” notification from YouTube, warning me that I had violated YouTube’s Community Guidelines which prohibited uploading material that do not conform with WHO public health safety doctrines. I fired off a strongly-worded email (you can just imagine) to YouTube pointing out that they can’t just program their “machine learning” to stay on the lookout for suspect phrases and immediately jump to the conclusion that a video is “harmful to public interest.”
I said they have to bring back some human into the equation. Their AI machines are tack sharp on word- and phrase-detection but totally suck at CONTEXT analysis. Who was I kidding? I was talking to a machine!
The only “context” it knows is linking one website to another because the same Google ad appears in both. Context has nothing to do with it. I learned this the hard way—pun will become evident in a while—when I clicked on a link about Filipina weightlifter Hidilyn Diaz’s Olympic gold achievement. Done reading the main article I spied on an innocent-enough-looking link in one corner of the screen inviting me to “click to see more SPORTS.”
I should have known that disaster lurks just around the corner if I took a moment to notice that the word “sports” was in quotation marks. As soon as I clicked the link I was immediately funneled into the “bikini sports edition” of some lurid online pornsite. It featured women with voluptuous front ends that God would have taken one look at and said, “I didn’t make those!”
Unfortunately, once you are caught in the grip of these web chaining algorithmns, it’s impossible to “click out” The AI program code pulls you deeper and deeper into this hedonistic abyss of cyberorgasm. It took less than 2 or 3 more clicks of me struggling to come up for air before I had all these cucumber-shaped and other vibrating toys dancing on my screen, being offered on 50% discount—free-shipped in a discrete unmarked box.
Finally, I just had to power off my laptop—and I mean literally unplug the device in a cold reboot. Otherwise the darned website reloads itself!
It’s hard to estimate the profound damage that “artificial intelligence” wreaks on culture and society on so many levels. I can’t imagine what the solution is, except maybe to try to win the war against it one tiny battle at a time.
If you get tagged by this charlatan FB account called “Mia Somngi Licoben” go the extra mile to report it to FB anyway. And then look for the genuine item “Mia Magdalena” and follow THAT.
The author is a writer and lawyer based in Baguio City, Philippines. Former editor of the Gold Ore and Baguio City Digest, professor of journalism, political science and law at Baguio Colleges Foundation (BCF). He is a photographer and video documentarist. He has a YouTube channel called “Parables and Reason”