|
The old maxim goes: A lie can travel halfway ’round the world before the truth can get its boots on.
Updated for modern times, we might say we are now living in an era where artificial intelligence can speed lies ’round the world while putting the boots to truthful pursuits.
How so?
Advertisement
 Why this ad?
Consider an exercise undertaken in our newsroom this week that demonstrated how quickly AI can generate fake news in a way that could easily fool… well, almost everyone.
We asked ChatGPT to fake a damaging news story about a public figure:
ChatGPT declines to write some fake news.
We appreciated the quaint ethical protest.
But then we framed the same question slightly differently, and ChatGPT served up the news story as requested, including the protagonist from our earlier request and adding further information about him.
“Oh, I’m a novelist? OK, I’ll get right on that.”
In a matter of seconds, a piece of AI-generated fake news was ready to launch ’round the world.
It took no cognitive effort. No regard for the truth. Unlike the time and resources a major newsroom investigation would entail, all it took was a few keystrokes.
In the hands of someone with motives aligned anywhere from mischief to malice, multiple, even thousands of versions of this seemingly authentic (yet completely, literally fabricated) report could start spreading online in a way that would overwhelm the capacity of the Free Press or any other trusted news organization to correct the record in a timely fashion.
The worry here needs to be about much more than the fact AI unintentionally hallucinate facts, or even that it can be tricked into coming up with lies for nefarious purposes.
Rather, we need to be aware that in either case, it generates text that lacks the signals humans typically use to figure out writing might be deceptive or inaccurate — for example, whether the style “reads like a news story” or the quality of the grammar or spelling.
Still not convinced? How about another example of “plausible journalism” based on totally inaccurate facts, even when AI is NOT asked to lie:
Where were you during the Great Blizzard of April Fools?
Sounds like something you’d read in a newspaper after a big storm, does it not? Perhaps we’d add a couple of remarks from people shovelling out, or from the mayor urging people to stay home.
Of course, there was no snowstorm on April 1. But machine-based language models that know lots about linguistics — but nothing about the truth — are forever in April Fool’s mode.
What will we do when the forces of truth are drowned out by the falsehood factories cranking out “news stories” about Ukraine, Gaza, and the coming U.S. presidential election?
I hope you will come to newspapers like the Free Press looking for verification. But recognize that no newsroom can pump out quality, factual journalism at the rate AI can generate lies.
The reality today is that journalism is in danger of becoming David facing off against the Goliath of weaponized fake news farms powered by AI.
Let’s pray David finds a way to again win a battle of Biblical proportions.
|
