The Trouble With AI is It Ain’t That Smart

Friday The New York Times dropped a story about satire and Facebook’s bots inability to recognize it:

Since 2013, Matt Bors has made a living as a left-leaning cartoonist on the internet. His site, The Nib, runs cartoons from him and other contributors that regularly skewer right-wing movements and conservatives with political commentary steeped in irony.

One cartoon in December took aim at the Proud Boys, a far-right extremist group. With tongue planted firmly in cheek, Mr. Bors titled it “Boys Will Be Boys” and depicted a recruitment where new Proud Boys were trained to be “stabby guys” and to “yell slurs at teenagers” while playing video games.

Days later, Facebook sent Mr. Bors a message saying that it had removed “Boys Will Be Boys” from his Facebook page for “advocating violence” and that he was on probation for violating its content policies.


© Matt Bors

Facebook has had trouble identifying the slipperiest and subtlest of political content: satire. While satire and irony are common in everyday speech, the company’s artificial intelligence systems — and even its human moderators — can have difficulty distinguishing them.

Matt Bors and The Nib is the main cartoonist of interest,
though Ed Hall and Adam Zyglis and Ruben Bolling also get name-checked.

As Matt explains the problem:

While his main source of income is paid memberships to The Nib and book sales on his personal site, he gets most of his traffic and new readership through Facebook and Instagram.

Mr. Bors said losing his Facebook page would cost him 60 percent of his readership.

The takedowns, which have resulted in “strikes” against his Facebook page, could upend that. If he accumulates more strikes, his page could be erased, something that Mr. Bors said would cut 60 percent of his readership.

“Removing someone from social media can end their career these days, so you need a process that distinguishes incitement of violence from a satire of these very groups doing the incitement,” he said.

The New York Times story can be read here.

 

Alan Dershowitz, at conservative Newsmax, agrees with The Times, but notes:

Although the humans who program the robots eventually permitted some of the cartoons to go on line, by the time they did so, the contemporaneous impact was lost.

With censorship as with humor, timing is everything.

and

The other side of the coin is that some really offensive and/or dangerous material evades robot censors, because humans have figured out how the algorithms work and how to circumvent their censorship.

So the end result is that robots both over censor and under censor.

 

On a related note …

The day before the N.Y. Times story The Patriot Post complained
that a Gary Varvel cartoon had been fact-checked:

On March 8, 2021, we posted a cartoon from one of our syndicated cartoonists, Gary Varvel. Two days later, it was labeled as “Mostly False” by independent “fact-checker” PolitiFact and our page was punished (again) for sharing “false information.” The cartoon was “missing context” and did not technically toe the line for the leftist version of the truth about the Democrats’ tyrannical HR 1 legislation. The sentiment surrounding the well-known push by Democrats to manufacture new voting blocs (among other things) was essentially ignored. The whole point of the hyperbolically humorous cartoon was lost on the humorless wokescolds.


© Gary Varvel

But it doesn’t matter that, à la Charlie Hebdo, editorial cartoons are meant to be satirical commentary on politics and society. Satirical cartoons, once hailed as sacred by the Left, have now become the next prime candidates for cancellation. The Left will stop at nothing to ruin humor.

Well, conservative humor anyway.

The full Patriot Post column is here.

 

Back to some problems with The N.Y. Times article.

Right wing sites had trouble with this portion of The Times article:

Facebook often dealt with far-right misinformation sites that used “satire” claims to protect their presence on the platform, Mr. Brooking said. For example, The Babylon Bee, a right-leaning site, frequently trafficked in misinformation under the guise of satire.

From Clay Waters at NewsBusters:

Wait, Isaac cites as a “far-right misinformation site” a “right-leaning site”? Some copy editor needs a stern e-mail.

It’s not the first time the paper has made the “misinformation” claim, as if the Babylon Bee was some kind of think tank. Liberal journalists quickly lose their sense of humor when their ox is gored.

And from Reclaim The Net:

That means a recent NYT piece focused negatively on a particular kind of satire sites that its writers think should be “dealt” with – those of conservative provenance like the satirical site Babylon Bee – but not those in the same genre but from the other end of the ideological spectrum, like the Onion.

 

Ruben Bolling, as he explains at his blog, also had problems with The New York Times as the article used his art without asking or getting his permission, or attributing the art to Matt, or paying him for the use of his cartoons. The original article as posted included none of the above. An amended version carried attribution:

 

March 24 update:

Ruben isn’t the only having trouble with the N. Y. Times article.

The CEO of The Babylon Bee says he is considering taking legal action against The New York Times for claiming the Christian satirical news site publishes “misinformation.”

In an email to site subscribers, Seth Dillon wrote, “The New York Times points to The Babylon Bee as an example of a ‘far-right misinformation site’ that ‘sometimes trafficked in misinformation under the guise of satire.’ They said we dishonestly ‘claim’ to be satire to protect our presence on the platform. This is false and defamatory.”

“Notably, the words ‘trafficked in misinformation’ are hyperlinked, presumably to a supportive source,” Dillon continued. “But the link they point to is another NY Times piece that actually refutes the claim that we traffic in misinformation by describing us as a legitimate satire site.

“The reason they don’t link to a supportive source is because they don’t have one,” he added. “So what can be done about this? Well, we’re talking with our lawyers again about the best path forward. Damaging defamation is serious.”

The conservative BizPac Review carries this story.

 

One thought on “The Trouble With AI is It Ain’t That Smart

  1. I’ve been meaning to subscribe to The Nib, having leeched off the free stuff for so long – this may be the impetus I need to actually open my wallet.

Comments are closed.

Top