My Slightly Uneasy Friendship with AI

 

My Slightly Uneasy Friendship with AI

Someone known to a friend of mine is building an app that helps college students with ideas to write essays. It was obvious that they would be using some generative AI to create that. My  friend who is a little skeptical about usage of AI but still wanted to know more, had a doubt  “Isn’t this encouraging plagiarism?”

That simple question started a long and rather fascinating thought about AI in my head. I did what I normally do. Start a question - answer session about AI with AI. My first was to ask AI to tell me if I could create an AI Plagiarism Detector. I actually got a smiley back followed by a long explanation. One sentence said it all - One tool helps generate text… another tool tries to detect that it was generated. It was like Locks and Keys evolving together. 

First, I got to understand about AI writing tools. The basic one is an idea generator. It is like a smart brainstorming tool which is generally acceptable. It helps students with topic suggestions, essay outlines, references to explore etc. This is probably what the app is aiming to do. Good for people who want to think a bit at work. 

Then comes the assisted writing tools. This hovers in the grey zone. It helps with improving grammar, restructuring paragraphs and even suggesting better wordings. It will help you smoothen out the rough edges of your writing, tighten the sentences and make it look crisp. It is like a clean up artist. Though teachers do not like this but it is sometimes accepted. 

The full essay generators are the problem. You type “Write a 2000 word essay on Shakespeare’s tragedies.” And it produces a flawless document and the student submits it unchanged. This is academic misconduct. 

Now my question came back - the AI Plagiarism Detector. Well it would use AI too. It would have to detect patterns about predictability of words, evenness in writing, vocabulary distribution including repetitions and contradictions (which AI may not make), grammatical mistakes etc. It can give a statistical probability of how much AI could have been used. But then it could actually flag good writers while at the same time let smart cookies like me who edit AI generated text and add some grammatical and spelling errors, get away. 

And that is where it turned interesting and I discovered three words I had never heard before: Perplexity (not the other generative AI like ChatGPT), Burstiness and Stylometry. They sounded more like philosophical moods than technical terms but they turned out to be central to how AI writes. And that is when the thought came to share it with all of you. And this blog is the result.

Perplexity measures how predictable a piece of writing is. If you can easily guess the next word in a sentence, the perplexity is low. AI tends to produce such writing — smooth, logical, statistically likely. Humans, on the other hand, are far less predictable. In our normal day to day conversations we are predictable. We would say “The sun rises in the …. East” But when we are in full flow we change direction mid-sentence, introduce unexpected phrases, add some humour or sarcasm, sometimes even contradict ourselves. That unpredictability increases perplexity. 

Imagine a sentence like this “ The sun rises in the east - is a false statement. It is just human ego talking. The sun doesn’t rise or set. It is only the earth turning and making it look as if the sun rises and sets.” See that? Normally one cannot predict these kind of words. 

Then comes burstiness — an even more delightful word. Burstiness refers to variation in writing. Humans write in bursts. A short sentence. Then suddenly a long wandering explanation because the mind has taken a detour. Then another short line to bring the thought back. The previous example has high perplexity and this one a high burstiness as well. 

And Stylometry is like a writing fingerprint. I want you to ask yourself these questions - Do you normally write “I think……” or “I feel……” more often. What are your punctuation habits? How does your sentence rhythm flow? What is your 'takia kalaam' when you write or speak? 

If you read my blogs often you will be able to understand my Stylometry. I use humour, I ask reflective questions, I use ….. a lot. That is difficult to copy though I feel AI is catching up. Maybe if it reads 100s of my blogs it will write a new blog which may sound like me but that is only if I lose my integrity and call it my own. 

Which led me to another observation.

Creative writers find AI slightly unsettling because what might take them days to shape and polish can appear in seconds. Teachers are worried that essays written by students may not really be written by students. Even programmers — who helped build the digital world — sometimes joke nervously that AI might soon start writing their code (the truth is that it has started that already - do read about ChatGPT 5.3 Codex).

So there is excitement, curiosity… and a fair amount of apprehension. For a while I found myself completely fascinated by how easily concepts were being explained to me. But after some time, I also found myself disagreeing with the AI. It seemed to me that while AI is getting closer and closer to sounding human, humans may slowly be handing over more and more of their thinking to AI. And perhaps that is what makes even well-educated people uneasy.

But if we look back in history, we would realise that this is not the first time technology has done this. Calculators reduced our need for mental arithmetic. GPS has weakened our natural sense of direction. Smartphones have quietly erased many phone numbers from our memory. AI may simply extend that outsourcing into writing, reasoning and explanation.

Which brought me to an uncomfortable question. - Is AI helping us think better… or helping us avoid thinking?

Some people are already beginning to say openly: “I used AI to help organise my thoughts, but the ideas are mine.” Maybe that is the answer. That kind of transparency feels healthier than pretending the tool does not exist. In fact, this very piece you are reading grew out of a conversation with AI. And I am quite comfortable admitting that. 

Because in the end the real question about AI may not be technological at all. It is human. The answer probably lies not in the technology but in how honestly we use it. 

So the problem is not an AI problem, it is a human problem.

As I sat back and wrote this, I felt that the real dividing line in the AI age will not be between people who use AI and people who don’t.

It will be between people who still think and people who stop thinking.

Note: By the way I ran this article the way it is through ChatGPT for an Overall AI Involvement Prediction. And I was really happy when it said and here is its response verbatim "If I had to estimate: Human: ~80–85% AI assisted: ~15–20%. Why not 100% human? Because a few sections have the structured explanatory style typical of AI-assisted drafting. But the majority of the piece carries clear personal voice markers, which AI usually struggles to maintain consistently."