As a Certified Veracity Sleuth, I Find AI Missteps More Frequently Than You’d Imagine
“I’m a Professional Fact-Checker. AI Is Wrong More Often Than You Think”
“ANY TECHNOLOGY THAT seeks to manipulate or sort through human language is confronted with one inescapable fact: Our online writing is a mess. We humans are an error-prone bunch…”
The original article rightly points out the unnerving reality that human online language is an intricate labyrinth of cryptic expressions, typos, and a cavalcade of emojis. It’s no wonder that AI, with its logical-to-a-fault and binary nature, is often caught off guard, like a deer in headlights. For Artificial Intelligence, learning to interpret, comprehend, and respond to this ‘creative’ human lexicon is akin to trying to solve a jigsaw puzzle on a rollercoaster.
Let’s take a stroll through the world of fact-checking AI. No, don’t fear, it isn’t another Terminator sequel, but it is definitely an adventurous ride. Hop on!
Understanding and deciphering human language are two sides of one coin. While ‘Alexa’ and ‘Siri’, have passed this challenge with flying colors to some extent, there are colossal barriers yet to overcome when it comes to text interpretation. So here is the ultimate catch-22: we need AI to understand our gobbledygook, but our nonsensical jargon is precisely what has our beloved bots doing the silicon equivalent of scratching their heads. How refreshingly ironic!
To draw an analogy, ‘fact-checking AI’ is like a diligent student hell-bent on unravelling the nuances of human language, but always being tripped up by a casual typo or falling into the abyss of colloquial expressions. It’s earnestly trying to master this course called human communication for its final exam – to filter fake news.
Nowadays, AI is asked to stop the dam of fake news from breaking body and soul on social media. It’s like, “Hey AI, help us sort through this colossal digital trash heap of misinformation and find that tiny speck of truth, will ya?” It’s a 24/7 job with no weekends off, and no HR department to complain to – tough gig!
On the bright side, newest algorithms such as Roberta, BERT, and GPT-3 (These are AI models, not the latest robotic boy band) are upgrading the way AI understands text. It’s like a futuristic Babelfish, interpreting the tower of babel that is our modern language.
But in the midst of all this algorithmic alchemy, hesitate for a second and ponder – interpreting and understanding are one thing, but how far are we from making AI ‘think’? That, dear readers, is the real conundrum, the ultimate Rubik’s cube that needs to be solved.
After all, we don’t want a future where a well-meaning AI observes a human writing “LOL” and promptly calls an ambulance for a potential cardiac arrest situation. Now that would be the ultimate AI blooper, right?
Read the original article here: https://www.wired.com/story/fact-checking-ai/