The sophistication of AI-generated content may give us priceless content like Joe Biden playing Call of Duty or the Pope in a puffer jacket, but its potential for believable fakery could be an open goal for propagandists.
On Monday, Twitter users would doubtless have had their mouths agape in droves when they saw an image shared of the Pentagon, a hotbed for the USA’s strategy, with smoke pouring out of it.
It came from a post from what appeared to be a Bloomberg News account called “Bloomberg Feed”, which was quickly debunked as a complete fake and taken down.
Usually, as the old adage goes, where there’s smoke there’s a fire. In this case, ‘where there’s smoke there’s an AI-shaped smokescreen’, which falls decidedly less trippingly from the tongue.
The Pentagon’s official communications team were quick to deny any disaster whatsoever. Their account said: “There is NO explosion or incident taking place at or near the Pentagon reservation, and there is no immediate danger or hazards to the public,”.
This announcement was not quick enough to prevent others from taking the bait and running the story. RT, a Russian state media outlet, tweeted the story, though not long after took it down with its tail between its legs.
The post caused a hit to the stock market too, not just the Twitter-sphere, with a brief dip in S&P 500 stock index, which went from up 0.02% at 10:06 a.m. to down 0.15% just three minutes later.
On the one hand, it’s easy to find fault or a sceptical angle upon seeing the image. For one, the Pentagon is dangerously bad architecturally from an aesthetic point of view, but that’s beside the point.
It’s unlikely and a tad ironic that the literal Department for Defence wouldn’t be able to prevent an attack, or less likely, wouldn’t remember a garlic bread left in the staff kitchen oven.
On a more serious note, US Twitter consumers will be particularly sensitive to this kind of content and forgivably susceptible to believing it entirely from the offset. The country did of course experience the horror of the September 11 attacks, which actually included a plane collision with the Pentagon itself.
While one of the images may be demonstrably faker than the other, they are both believable and serious in nature to the naked and uninformed eye. That’s what makes this instance of fake news potentially way more dangerous than an easily fact-checkable word from Trump or Nigel Farage’s mouth.
The momentary meltdown could be perceived as a watershed moment in AI’s ability to create fake political narratives, fusing information and entertainment in a bloody terrifying way. Let’s enjoy US presidents debating video game tier lists on Discord servers while we still can.
If you’re struggling to tell fact from fiction, take solace in the fact AI is already bettering the critical facilities of the Punching Up team, as we try, an increasingly fail, to tell the difference in our podcast’s ‘two truths and AI’ round. Listen and lose hope folks.