We all know the great mathematician, Srinivasa Ramanujan. But not many would know that before he became legendary, he struggled to be taken seriously by his peers. Even his first journal article was returned more than once. What held him back was not a lack of brilliance, but his inability to explain his work to peers.What helped others realise his greatness was input from S Narayana Iyer and others. They helped translate his body of work into a form others could understand. What they did not do was supply him with mathematics. That distinction matters.Cut to today, I ask myself: can AI be the next translator for the next Ramanujan? And if it did, would critics shame him for using AI or would they recognise his intellect?This echoes the outrage over Stanley Druckenmiller using AI to write his op-ed in The Wall Street Journal. Instead of focusing on his argument, users are outraged that he used AI to write. They overlook his expertise and his clarification that ‘There’s a reason I moved from an English major to economics...I write everything using AI like I use a calculator.’AI detectionThis debate is being held while we talk, in parallel, about AI digital twins and how they will think and talk like us. Truth be told, this narrative is being pushed by those with vested interests. Especially those whose revenue hinges on AI detection. That said, the threat of these tools aiding the spread of misinformation is well-founded.While this narrative does well to cook up a storm around AI slop, if overdone, it can cost this world its next Stephen Hawking. Had it not been for the speech synthesiser, the world would have lost the genius he shared with us. Must note, the speech synthesiser was not the source of his intellect, but only a medium for it to reach us.GenAI tools are doing the same at population scale. These tools are giving voice to those whose intellect runs faster than their minds can translate into language most of us understand. The outrage peaked when LinkedIn introduced a feature, “Seems like AI slop”.Whether this feature is itself slop is for another day, but it reminded me of when LinkedIn allowed everyone to author articles. The timeline was flooded with content that was anything but well-written. Thanks to GenAI, at least the timeline now reads smoother.Before you assume otherwise, I’m not against AI detectors, since they make an important point about disclosures. And most, if not all, writers have not been ethical about using AI. However, treating these detectors as the holy grail and using them to shame experts is in poor taste. The issue is not detection, but whether it is used to assess intellect fairly.This deserves deep introspection. Recently, Anthropic stirred a debate by introducing text watermarks. In its defence, it cited Article 50 of the European Union AI Act. However, it did not disclose that the article includes provisions to recognise human review and editorial responsibility.Whether text watermarking will be useful or cause more confusion, time will tell. What’s certain today is that any and every form of AI detection is being used to judge the author.But the same judgement needs to be cast on the tool creators who have shown little accountability for using, without permission, content from the internet.The recent lawsuit by WikiHow alleges OpenAI scraped more than 11,000 articles without permission. The latter is being accused of infringing at least 1,200 copyrights. As one would expect, OpenAI says the models have been trained on publicly available data grounded in fair use.If that’s not a double standard, I don’t know what is. The argument only works if accountability runs both ways.Glaring imbalanceOne can argue that the situations don’t stack up equally, but ethically, the imbalance is glaring. Regulations are still playing catch-up, and tool creators find their position easy to defend since laws on the topic are as dated as the colour TV. Recently, in India, the court refused to grant an interim injunction to Asian News International (ANI), which had appealed against OpenAI using its content to train models. The case was tried under an act drafted in 1957.Accountability aside, this topic needs to be seen through another lens that is central to the existence of the human race: cognitive development. Decades ago, I completed two master’s degrees, which also means two dissertations. The ability to think, reason, and solve complex problems had a profound impact and laid the foundation for my career.This is discussed in Bloom’s taxonomy (1956), which Anderson and Krathwohl revised in 2001. This taxonomy argues that writing can happen across levels, but creating original text sits at the highest level of the cognitive hierarchy. Just like Ramanujan and Druckenmiller, not everyone can operate at this level, but that doesn’t mean they aren’t great at analysis, evaluation, and application. This is where GenAI comes to the rescue. The world will benefit more if these mavericks spend more time in their field than learning how to write better.Extent of involvementBut to ask for disclosure from users is fair. What is also fair is a promise from the creators of these tools. Instead of finding new ways to target original content under the guise of compliance, the focus must shift to tagging the content. A great place to start is using “extent of involvement” as the core premise. The spectrum can range from AI-originated at one end to an authorised writing partner at the other, with human-reasoned, AI-drafted and human-authored, AI-edited in between. Such classifications are critical since AI-written content is now the new normal. In fact, LLM writing will soon merge with human writing, and the difference will be hard to tell.Bottom line: denying people the right to use these tools simply because they don’t speak a language you do is akin to how East never got a chance to put its history on a world map. Sit with that for a while.The writer is the Chief Analyst at Greyhound Research (part of The House Of Greyhound), a global, award-winning, technology research, advisory, consulting & education firmPublished on August 28, 2026