We’ve Reached the Point Where AI Has to Catch AI (The Internet Is Full of Slop)
· Vice
AI is working hard at making it hard for you to tell it’s AI at all. College essays, news articles, scam emails. AI-written prose masquerading as the final product of real, human sweat and tears (but probably no blood) is everywhere and a spreading plague. So how do you suss out what’s likely AI fakery and what’s not?
Relying upon tells like use of the em dash (the long “—”) and patterns of three is lazy. These useful cornerstones of writing have been used by authors for generations before AI’s arrival, and dismissing any tract that uses them as AI slop will sweep up a lot of innocent human-created work into condemnation.
Visit mwafrika.life for more information.
How do you determine whether a piece was written with AI or whether it’s just a blandly written piece of human creation, then? You can use AI detection. Take Pangram’s AI Detection Chrome extension, for example. It’s free to download; it operates on a freemium model, whereby you pay to access higher-level features. Turning AI against itself has a poetic quality.
An AI detective to unmask ai impersonators
(opens in a new window) PangramPangram AI Detection Chrome Extension (opens in a new window)
Available at Pangram Buy Now (opens in a new window)Garbage in, garbage out. AI’s lauded ability to write clean copy has always fallen flat to me, because by its very design it sucks up huge quantities of existing writing, including other AI-written content, and stews it down into a mixed-together, bland gruel. It doesn’t distinguish much between good writing and just, well, writing. It just gobbles it all up.
But that doesn’t make all that easy to spot. Who’s to say what was written in AI’s bland, lukewarm prose or written by a person who just writes in bland, lukewarm prose? AI “authors” have begun to pop up in all manner of places with the barest of disclosure (if any at all) that the portrait, name, and backstory may all be a ruse.
That last thing I’d want to do whenever a piece raised my suspicions would be to navigate to a website and jump through a bunch of hoops to see whether my hunch would be proven correct. It’s cool and all that such sites exist, but I’d click a button asking “Did AI secretly write this?” much more often if it just lived in my browser, such as with the Pangram AI Detection Chrome extension.
It’s right there one click away. Why should all of us have such a burden of searching out websites when it’s other people out there trying to fool us with AI? It’s still not fair that we have to do the work, but at least it’s made easier now that it can live right inside your Chrome browser.
It shows you how much of your feed is likely to be made by ai – credit: PangramPangram makes some heavy claims, saying that Pangram AI Detection has a “near-zero” false positive rate. A false positive is when a tool mistakenly says a piece of writing is AI-written when it was actually written by a (very unlucky, probably now ego-damaged) human. But then Pangram backs up its claims by saying that this accuracy was validated by third-party researchers at the University of Maryland and University of Chicago.
Even when a person asks an AI to “humanize” its writing by making it pull a few tricks to avoid telltale AI giveaways, Pangram says its AI Detection can figure it out. The company says their tool is already in use by thousands of teachers, publishers, and higher education institutions—you know, colleges and universities—to detect which students and authors are trying to skate by on AI’s merits rather than their own. That’s no separate product, but rather the same one available for free.
Pangram was founded by Max Spero, who worked previously for Google, and Bradley Emi, who’d been at Tesla. The two became acquaintances back in their freshman year of college, long before they joined to found Pangram Labs in 2023.
(opens in a new window) PangramPangram AI Detection Chrome Extension (opens in a new window)
Available at Pangram Buy Now (opens in a new window)That may seem like a fairly new company, and in the grand scheme of things it may be. But AI advancement is moving rapidly. We’re not dealing with a legacy tech timeline, as with TVs, and not even with something on the level of the smartphone industry, which has largely matured since its own gold-strike level of bursting enthusiasm years ago.
All of that explosive growth in AI has created opportunities (and lost some) for people, but more than that it’s created a breach in trust that penetrates from the images we see on social media to the news articles we read on major news sites. Trust is a foundational component of society.
AI-generated images are fine when they’re on the level and up front about being created or tweaked with AI. Too much of the time, though, people keep it hidden on purpose as a way to manipulate. Giving into constant suspicion isn’t a sustainable way forward. The worry had been whether scammers’ ability to use AI to fool us would always outrun AI’s ability to detect it.
Pangram’s ability to detect text written by ChatGPT, Claude, Gemini, and other popular AIs with 99.98 percent accuracy, as they themselves state, is a reprieve from just giving into a state of constant disbelief. If AI can fool us with the press of a button, it’s a boon to have a way to detect the foolery with the push of another button.
In a world of fake news, scams, and bogus product endorsements, who wants to risk being fooled? Your money, your worldview, and most importantly, your ability to fake your way through discussions of the latest Stranger Things episode are all at risk if you’re on the receiving end of bad info.
Being that it’s free to download to your Chrome browser or integrate directly into your Google Docs, there’s no commitment to try out the Pangram AI Detection Chrome extension. Install it, give it a whirl on the next bunch of articles you read, and see how it performs. It may provide you with your own Men in Black moment, where you realize that a lot of what’s out there presenting itself as human on the internet isn’t actually human beneath the skin at all.
pangram’s ai detection integrates with google docs – credit: PangramThe post We’ve Reached the Point Where AI Has to Catch AI (The Internet Is Full of Slop) appeared first on VICE.