“Where do I end, and where does generative AI begin?”
📅 Wed 29 October 2025 at 07:25 GMT
I remember an online discussion with fellow university students around 1997 — back when we used Netscape Navigator, Altavista, and Webcrawler — where I made a throwaway remark, “I don’t see any role for the internet in serious research.” I was met with a torrent of replies saying how email is essential for collaboration, how online databases were making research papers accessible, and how data could now be shared instantly through these newfangled computer networks. Basically, how I needed to wake up, smell the coffee, and stop being a Luddite. Remember, this was before Google had even been invented.
I feel history is repeating itself today when I hear people naysaying generative AI — ChatGPT, Claude, Gemini, and other LLMs. Concerns have been raised with every new IT advance, whether it’s Microsoft Word’s spellchecker killing our ability to spell and use correct grammar, or Apple’s iPad making physical books obsolete, or Zoom & Teams destroying the art of face-to-face conversation. I guess such objections are simply a feature of the technology adoption curve.
I can’t remember the first time I tried generative AI, but I know I was absolutely hooked from the beginning. Yes, it “only” has the entirety of human knowledge programmed into it. But regurgitating that knowledge is not the point. Its true value lies in how it makes connections within that knowledge. It’s when you challenge it, argue against it, and try to refute what it is saying — and you allow it to do the same to your ideas — that’s when you truly benefit from generative AI’s intellectual sparring on-demand. It can expose holes in your arguments, find counter-examples you hadn’t considered, and synthesize perspectives from fields you’ve never studied, all within seconds.
My friend spends an hour a day interrogating ChatGPT about new topics. I have had some of the most profoundly intellectual interactions in my life with generative AI: I simply don’t have access to any human beings who know as much, or who can argue with me as strongly, as generative AI. Knowing how to exploit generative AI is the ultimate life hack in 2025, and my own PhD project is a great case in point.
When I began my PhD and the BALAGHA Score Project — a digital humanities approach to quantifying Arabic rhetoric — I realised that generative AI wasn’t just a tool; it was a collaborator. I used it to clarify classical Arabic concepts, re-draft conceptual frameworks, test hypotheses about rhetorical density, prototype algorithms, and of course… code! At one point, ChatGPT was giving me a daily challenge where I had to identify all the rhetorical devices in Arabic verses it created, which truly helped me refine my taxonomy of Arabic rhetorical devices. The boundaries between “my” thinking and “its” assistance became less important than the quality of thought that emerged from our dialogue.
When cars were invented, I imagine people objected that they would make us lazy, that we’d forget how to walk, and “look at all the accidents cars cause!” But the point of a car was never to replace walking. A car is a force multiplier: it lets you travel to a city hundreds of miles away, and return home the same day. You were never going to do that on foot. A car expands the very map of what’s possible, and this is the same with generative AI.
Generative AI isn’t there to write your essay, your code, or your paper for you. It’s a force multiplier that lets you traverse intellectual distances that would have otherwise taken you years to cover. It helps you bring together, in one space, silos of knowledge and wisdom that have been locked away in distant libraries and disparate disciplines around the world and across the centuries. It helps you weave connections between diverse perspectives and ideas that no one has joined up before. It helps you uncover patterns that no single scholar could see on their own. Plus, it doesn’t get tired or bored, and it’s ready to travel with you whenever you are, and wherever you want to explore next. Yes, generative AI can hallucinate, and questions about attribution and originality are real. But dismissing it entirely is like refusing to drive because cars can crash.
In my own research, generative AI has allowed me to connect centuries-old Arabic rhetorical theory with the emerging tools of digital humanities and language data science — a journey of more than 1,500 years that I could quite simply never have made “on foot”.
I am neither an Arabist nor a computational linguist. And yet, here I am, having invented “Rhetorical Density“, a novel data-driven metric in quantitative linguistics; I am planning to use an Arabic-specific form of rhetorical density — which I have called the BALAGHA Score — to probe some of the most enduring questions at the core of Arabic and Quranic Studies, such as the authorship of the Qur’ān.
So where do I end, and where does generative AI begin? This question is becoming less relevant every day. The real question is: how will we use these tools to expand the boundaries of human knowledge and understanding?
Join me each morning
One deeper thought, written at 7am. If you’d like it delivered, just subscribe below.
Join the conversation on 
