paint-brush
AI vs. Sarcasm: Will It Ever Understand?by@Manish-sharma
889 reads
889 reads

AI vs. Sarcasm: Will It Ever Understand?

by Manish SharmaAugust 4th, 2023
Read on Terminal Reader
Read this story w/o Javascript
tldt arrow

Too Long; Didn't Read

AI program that's designed to detect sarcasm in online comments. AI's struggle to grasp human emotions, particularly sarcasm, stems from the inherent complexity of human interactions. Emotions are multi-dimensional and not solely visible on our faces. Achieving accuracy in analysing human emotions is difficult due to our subjective mindset.
featured image - AI vs. Sarcasm: Will It Ever Understand?
Manish Sharma HackerNoon profile picture



I saw this wild YouTube video recently that really delved into the dangers of AI in the future. It's about "iMom." So, in this futuristic world, they created this super advanced AI called iMom to be a mother substitute. But here's the thing, trouble starts when the iMom, unable to understand sarcasm or context, literally takes this as an instruction.


The real mom says, "Goodbye, my little chicken," referring to her infant daughter, and then tells the iMom to cook chicken. Guess what happens? The iMom ends up cooking the baby because it takes the mom's comment literally.


The video got me thinking, will AI ever truly understand our emotions, especially the nuances of sarcasm? I mean, sarcasm is such a human thing, relying not only on what's said but also on tone, body language, and facial expressions.


We humans often struggle to grasp sarcasm completely, so how can we expect AI to get it right? It's like we're asking machines to comprehend something inherently complex and nuanced in human interactions. I mean, the iMom took the mom's comment literally and cooked the baby! That's a scary thought because it shows how AI, despite its intelligence, can still miss the mark on understanding emotions and context.


Companies have already started an AI program that's designed to detect sarcasm in online comments. It's made by these scientists, Ivan Garibay and Ramya Akula, funded by DARPA, and the program sort of banks on sarcastic cues and phrases, memorizing them for future reference.


It's designed to help people, like business owners, sift through tons of comments and feedback. But, you know, it's one thing to identify text-based sarcasm and another to understand the subtle tones, facial expressions, and body language that often accompany it.




In a video featuring Will Smith, the Sentiment Analysis tool attempted real-time emotion reading.  The tool reads faces from videos and detects feelings based on this emotion wheel thing. It's got four directions: exciting, positive, calming, negative. But even there, sarcasm wasn't included as an emotion, highlighting the complexity of grasping subtle human expressions.


Also, there's so much wrong with the analysis. Emotions aren't just visible in our faces; sometimes, we can look angry when we're really just focused or be bubbling with rage while looking completely calm. AI's struggle to grasp human emotions, particularly sarcasm, stems from the inherent complexity of human interactions. While AI can detect text-based sarcasm to some extent, understanding subtle cues like tone, body language, and facial expressions remains challenging.



Emotions are multi-dimensional and not solely visible on our faces. Achieving 100% accuracy in analyzing human emotions is difficult due to our subjective mindset. Relying solely on AI for this may not be foolproof.


As the saying goes,"Emotions are a language that requires a human heart to truly understand."