E-Paper
Of AI hallucinations and brain rot
On what AI is doing to our thinking and our focus and what we need to do to avoid its easy temptations so we can continue to nurture the intellect
There was a time before smartphones when I used to read one or even two books a week. Then the damned device charmed its way into my life and things have never been the same. Don’t get me wrong – I don’t hate it. I love it; that’s the problem. From being a reader who could read two books a week, I am now down to being someone who struggles to read even one book in two months. Every time I open a book and, with the best of intentions, start reading, my phone conspires against me. There’s some notification or other. Once I pick it up, it’s game over. The temptation to just take a quick peek at mail or social media is too great. One quick look turns into an hour of scrolling. The time set aside for reading vanishes.
I am not the only one suffering from this ailment. With smartphones in practically every hand, it’s just a part of the human condition now.
What this was doing to our brains was already a serious problem before artificial intelligence got into the mix. Now the smartphone has not just WhatsApp, Facebook, Instagram and X. It also has ChatGPT, Gemini and Claude. Where until recently you might have searched for an answer to a question on Google, now you’d probably do it on AI. Instead of a list of links that you have to sift through, AI will give you a neat, well written and even thoughtful answer.
No need to read or even scan through a few pages of text to get the answer you were looking for. You’ll get it in a few handy paras.
It can do the trick with big books too.
Want to know what’s there in a book without spending the time and effort to read the whole thing? Just ask AI; it’ll give you the summary in seconds, in text or audio.
It’s all just so convenient, so easy.
The question is, what is all this convenience doing to us?
I’ve been an avid reader since I was in Class 2. I’ve written three books. And I now struggle to finish books.
I’m afraid that soon, everything beyond one para will evoke a single reaction from most of us, whether we express it or not: tl;dr. Too long; didn’t read.
We as individuals and even collectively as a species are losing our ability to read long pieces of text, let alone write them.
Considering the central role these abilities – reading and writing – have played in the growth of human civilisation, this is a serious threat for all of us.
Fake cases, fake research
ChatGPT was launched on November 30, 2022. It is not yet four years since that day. Already, AI is everywhere. People are using it for all kinds of things, sometimes with alarming or comic consequences.
In July, the Supreme Court of India raised the alarm about non-existent AI-generated precedents being cited in judgments and court cases. The court was hearing a case in which it found that the National Company Law Tribunal had passed an order relying on fictitious AI-generated case laws.
The problem is hardly unique to India. Other countries are also grappling with a surge in lawyers citing non-existent AI-hallucinated cases. A US court, for example, even chastised the country’s Justice Department after one of its lawyers was caught citing a judgment in a case that didn’t exist.
Even researchers, including medical researchers, are not immune to the easy temptations of AI.
A groundbreaking study published recently in one of the top medical journals in the world, The Lancet, found 4,046 completely fabricated references spread across 2,810 published research papers. The rate of fake citations has exploded since the advent of AI, rising 12-fold from four in every 10,000 papers in 2023 to nearly 57 in every 10,000 papers in early 2026.
There are plenty more such examples, across fields, but my favourite is the one of South Africa’s AI policy. The country’s draft National Artificial Intelligence Policy had to be withdrawn after it was discovered that it used references to academic journals that did not exist, and papers that had never been published, all generated by the marvels of AI.
AI hallucinations
There is a common tendency among people, including those drafting AI policy, to somehow feel the machine knows best. It’s always well written, and always plausible – and that’s exactly what it is designed to be.
The problem with the artificial intelligence we have is that it’s not really intelligence at all. There are different types of AI but the top ones which most of us use for answers to questions, such as ChatGPT, Gemini, Claude and DeepSeek are all Large Language Models.
The way they work is by first converting every word into numbers and representing them mathematically. Based on the prompt the user enters, and its own internal training data, it then generates a probabilistic score to determine the first word of the answer it will generate. Once this first word is out, the maths functions will come up with the next word, and the next, and so on. So, the AI is in fact a very sophisticated predictive text system.
The implication of this is that it is not generating an answer that is grounded in fact. Nor is it giving you a piece of text that is creative. It is in fact optimising for the most predictable sequence of words in response to a given prompt. What you get is a plausible sequence of words on, for instance, AI policy, or legal cases, or medical research – which may contain references to cases that don’t exist or research that was never done.
Brain rot
The potential fallout of AI hallucinations is only one part of the problem.
With everyone from scientists to lawyers to policymakers outsourcing research and writing to AI, the need for deep reading – what we used to call study – is reduced. But reading a book with concentration is about more than what one learns from it. It also affects our brains in a way that is very different from what digital devices do.
In her book Reader, Come Home: The Reading Brain in a Digital World, author and academic Maryanne Wolf wrote about the shrinking attention spans of adults, among other things. We don’t read any more; we skim. There are always a myriad of things competing for our attention, so staying focused has become a challenge.
These changes are internal. They are happening in our brains. The way our eyes move on a screen is different from the way they move on a piece of paper. With time, the habits of skimming and browsing become ingrained. We simply lose the ability to focus. We are perennially distracted.
Professor Wolf highlighted the critical relationship between the quality of reading and the quality of thought, which is influenced heavily by changes in attention spans and what she calls “cognitive patience”. She mentioned letters she had received from professors who complained that their students simply did not have the patience to read dense texts.
The loss of cognitive patience leads inevitably to what is being termed cognitive surrender. People just outsource their thinking to AI, treating it as an oracle for all occasions.
Naturally, the temptation to do so is especially strong when one is confronted with a writing task and a deadline. No wonder that everyone from undergraduate students to judges in tribunals are getting caught out for AI generated content.
Writing is thinking
A key feature of the human brain is neuroplasticity, the brain’s ability to keep strengthening, forming or pruning neural connections through life. When we stop using certain neural circuits, those circuits break.
That’s not very different from how the rest of our body works. We have to exercise our muscles if we want them to be strong; someone who doesn’t do any kind of exercise and eats only junk food will inevitably be unfit.
These days, people tend to be very careful about what they eat. However, while they watch what they put into their bodies, they are less careful about what they put into their minds. A diet of endless doomscrolling on X or Facebook is hardly conducive to good mental health.
The reasons for why we should take care of our brains as we do of our muscles are many. The way to do it is simple.
We have to protect our reading habits.
We no longer need to walk or jog any more to get from place A to place B. The car, bus or motorcycle can transport us. But walking remains important for our own health.
Similarly, we don’t really need to read a book to get the gist of something; there are quicker and easier ways to get information. But we still need to read deeply to build and sustain our own cognitive patience and attention spans.
AI can write our emails, reports and papers. The thing is, writing is also simultaneously an act of thinking. We have to understand our materials and organise our thoughts before we can write them down in a cogent and coherent manner.
When we outsource all that to AI, we can copy-paste the result, but we won’t know the material, and it won’t be our own thinking.
Over time, the dictum of “use it or lose it” that applies to our bodies and minds will kick in, and we – and our children - will simply lose the ability to do our own research and thinking.
The growth of artificial intelligence really will be accompanied by the growth of natural stupidity if we let go of reading and writing. Reading and writing are the gyms for our minds.
Samrat Choudhury teaches at the Writing Centre, Shiv Nadar Institution of Eminence.
