Background

I recently watched a video of Prof. Terry Tao’s talk at ICM 2026 that prompted me to break out of my cocoon of thoughts and pen down my own experiences with generative AI and what advancement of generative AI means for all of us.

Disclaimer

Before I say any further, I would like to put up a disclosure that the content herein is an output of my mind and no help from generative AI tools was sought while writing this up. While my thoughts would and must have been influenced by a variety of factors including but not limited to generative AI, the write up itself is my own work.

My tryst with AI

When I was doing my bachelors, not very long ago, RNNs and LSTMs were all the rage. AI, was still a buzzword, as much as it is now, back then. I somehow felt privileged to learn the math behind all these fancy algorithms that powered these “language models”. What was even interesting back then is the paradigm shift that the professors talked about, from symbolic AI, to frequentist models, to ML, and to more powerful general purpose language models. What my naive self could not realize back then was that the paradigm was actively shifting but in a much accelerated manner. Fast forward to my master’s (which again was not really long ago!), I came across this paper that said something about “Transformers” with a catchy title (it will be too clichéd to write the full title of the paper as you know what it is :D). This one sounded way more complicated than RNNs but thanks to the wonderful professors at the university I could comprehend a thing or two about it. In all the time that passed by, advancements in computing coupled with some really fancy systems engineering, brought AI into the market for human trials (as I see it). Like every major invention it started getting commoditized. In a matter of few months the models kept getting better at almost everything what was previously cherished and coveted as a product of human endeavor: art, prose, code, and now math (I could have categorized everything into art though). What sounded fantastical at first became the norm in no matter of time. Much to my chagrin, I find it really annoying to read messages and emails sent by “humans” that are generated by AI these days.

Destination vs journey

Now that is a segway to what I really want to talk about: what does it mean to be human in the world of AI? We instinctively know whether something is written by AI. Our brains are wired and trained enough to recognize patterns that are specific to human creation. While AI does a fantastic job (kudos!) at generating fluent essays on answering a simple question such as, “How is the weather today?, it gives itself off by its sycophantic style of writing. Maybe we are wired to be willing to hear a yes or no answer sometimes while expecting to willingly hear a long rant (such as this one) when there is no absolute reason to! These imperfections and whimsical tastes of the human mind is probably what differentiates humans from AI. We are bad, no, rather horrible, at optimization. Our journey to a destination, which may be a literal destination like the sunny shores of California, or the breezy monsoons of Dharwad, or a decision in life, probably makes a profound impact on our minds more than the destination itself. The way we get to those destinations, things we see, hear, learn, and adapt along the way, is what shapes us, and our narrative. AI can at best read through millions of such documented narratives and provide a glorious account of its own journey but it for sure cannot experience one. Do note many of these journies are not perfect. Some hurt us, some make us laugh, some make us feel contended and grateful, while others leave us yearning for more. These imperfections, or perturbations from optimality, are what I consider the hallmarks of human nature, which get reflected in the works created by humans. Now clearly, enumerating all such possible perturbations or journeys and making AI read through those is an impossible task, not because AI cannot do it, but because humans cannot write everything down or coherently describe their experiences in the first place. So I theorize this is possibly why we can differentiate text generated by AI from what a human writes, as our brains are “tuned” to infer information from these imperfections and “expect” such in human generated prose.

Skills (not skills.md xp)

This brings me to question: among the list of skills I have garnered and worked upon over time, what is the subset of those skills that would be “more” important than the rest, in this world flooded with news of AI? While I have deliberately tried to make this statement sound precise, in rather mathematical-sounding ’lingo’, that specific wording carries meaning. Whatever we do in life carries some meaning, whether or not explicit.

First comes critical thinking. The way we approach a problem. I always could not help but wonder, how on earth did people even think of something as crazy as quantum mechanics or general relativity, “picture” those ideas, and coherently write those down for other people to understand, when there (possibly) was no computer or a “thinking machine” to assist? (okay, I watched Dune but didn’t read Frank Herbert). When I first heard about “s, p, d, f orbitals” I could barely understand what they meant. But over time, with considerable effort and practice, I came to appreciate the beauty of what they represented. When I came across textbooks and videos with richer visualizations, it was then, I really felt awed by the insane amount of “imagination” that people had to come up with, in order to theorize and conceptualize such other-wordly topics! Clearly, the advent of such vivid visualizations and imagery helped better my understanding of arcane concepts in physics and chemistry, and to some extent, math (and machine learning), but I always wonder: what if I were presented with these visualizations first and then asked to imagine? Would I still appreciate these as much I would as I do now? Is this process of appreciation commutative? (okay, I took an “Elementary algebra and number theory” class during bachelors which I believe I barely passed. There was nothing elementary about it :D).

Second of those skills would be math. Math of all sorts and kind. Math that made sense to me sometime in my past but no longer does (I know how to take the derivative of a univariate exponential function, yayy! :D). Math that scared me and thralled me. Math that I think I know and talk about sometimes, but math that I know nothing about and am aware that I entered this phase of decadence which many professors had alluded to in the past. If you have been in similar shoes as I am you would get what I am referring to but nevertheless it suffices to say that math is something is deeply respect but secretly sad that I don’t get to do it with same rigor these days.

Third, of the “more” important skills, is the ability to convey my thoughts clearly, whether verbally or in writing. This, by far, has been the more difficult one for me (like you could guess :D). While a plethora of thoughts run through my mind at a much faster “bit-rate”, it is not an easy job for me to slow that process down enough to document and convey those thoughts in a coherent and easy manner for someone to hear or read them out. It is like having this super powerful GPU and an ultra-fast SSD, only to be bottlenecked by a mediocre CPU, and a horrible NIC, that slows you down to a standstill. (okay, I analyze systems performance for a living xp) It is frustrating yet important learning curve, I feel. The fact that I am able to write this long, trying to maintain my “chain of thought”, is something I feel truly grateful about. Today, every major social platform hinges upon how long it can make a user stick to its platform, in other words, how long they can capture the attention of a user. Having been thrown to interface these platforms everyday, as a means of communication at work, or by personal choice for entertainment, the art of having to endure by reading through long passages of writing, or writing them in the first place, is possibly becoming a lesser needed skill. I feel what demands my attention is not those platforms but these works that require me to stay focused on one topic and make the most of the information it has to offer. After all these years being a student and early-in-career professional, I feel multi-tasking may not be a boon after all. To drive the point home, I think I want to say that writing and speaking coherently, listening with earnest and having an attention for detail, is what I feel is really important. It makes the life a bit difficult for you, do doubt about it, but it makes others lives a bit easier as the signal to noise ratio is high in your information and their brain gets to work a little lesser when listening to you. You also score a bonus point for having come across as approachable and not a weirdo to others, if you are clear and coherent in communication! The other, rather paradoxical benefit of this, is it helps make AI understand you better as well. It requires relatively “lesser” computation to figure out what you are talking about, and spares you from the frustration of having to go through multiple iterations of prompting to making AI understand what you truly meant, and save you some dollars as well (AI is getting costly!). This being said, I would like to view AI as that friend who does not lose interest in what you have to say midway but anyways forgets what you said, eventually.

An AI prompt a day keeps your brain at the bay

With such “more” important skills listed out and commented upon, I wonder, how good am I at those skills? In general, how good am I at art: art of speaking, art of writing, art of thinking, art of “math-ing”? How good is AI doing that? How good are others doing that? While this is probably not a quantifiable measure, I could at the best try answering those questions myself, based on social cues. I am pretty sure no one likes to go through an essay to figure out the crux of what an AI is trying to say. At the same time, people don’t like spending a lot of time analyzing others work (I hear “reviewers” making sad noises: research work reviewers, code reviewers,..). Maybe the first one is justifiable but the second one is not, in the world flooded by the news of AI. When we read or listen to something or someone, we always subconciously analyze it, some may call it a form of bias whereas the others may call it an evolutionary trait. But the lesser we analyze (or review) and take things on their face value, the more prone we become to being “sellable” to the popular narrative, whether or not it may be true. Ironically, I feel we are developing faith in AI, in some respects. While being paranoid is not good either, our discretionary antennae for differentiating “right” from “wrong” are becoming rusty. A possible reason for this is our brains do not want to spend a lot of time on something that has all of a sudden become a norm in our daily lives and spending significant portion of energy on this review of information is something our brain is not willing to do. It takes short-cuts, which sometimes work for our good, and fail spectacularly in other times (thinking of PR reviews by AI and for AI!).

I’m not a doomsayer

So, AI is here to stay, like were the steam engines (replaced by IC engines), the mills, the computer, the internet, and other such life-changing innovations. Like people say, “you do not get replaced by <x> rather you get replaced by the person who knows to better use <x> than you do”. You either get really good at it or pivot ways to find something that interests you with a hope you get really good at it and not get shot down by another AI-like innovation.