Halloween is approaching 🎃, and we want to share a real horror story from the IT world.
Imagine this: you’re an NLP (Natural Language Processing) specialist at a cool company. Your job is to teach an AI system to recognise ‘toxic’ or undesirable content. 🤖
One day, the system starts working… BUT INCORRECTLY! 🤯
It interprets completely innocent words, such as ‘night’ 🌑, ‘shadow’ 👻, and ‘white noise’, as reports of supernatural activity. It bans users who are simply talking about the weather. 🙅♀️ Panic ensues in the company chat: ‘Our AI sees ghosts! It’s cursed!’ 💀
What went wrong❓
It’s not a curse, it’s a BASIC NLP ERROR! 🐞 Someone forgot about deep contextual analysis. The system recognised words associated with ‘mysticism’ but did not understand their true role in the sentence. 😔