Rahul Matthan: Does it really matter if something you liked was written by AI?

🔗 원문 전체 보기 → Livemint

요약

Modern AI writing is hard to tell apart from human-written work, but that needn't matter if the content holds value. Instead of resisting the inevitable march of technology, let's learn to work with it, immersing ourselves deeply in it so we can take advantag…

본문

This is a Mint Premium article gifted to you.
Subscribe to enjoy similar stories.

Earlier this year, the Supreme Court of Colombia dismissed an appeal after an artificial intelligence (AI) detector called Winston found that the filing contained “only 7% human content,” holding that it could not, for this among other reasons, be admitted as a validly submitted pleading. Shortly thereafter, a lawyer, Emmanuel Alessio Velasquez, ran the court’s own ruling through the same detector, which suggested that the judgment itself was 93% AI-generated.

X-raying documents for traces of AI seems to be all the rage these days. In September, an anonymous user posted Pangram results suggesting that Thélyson Orélien’s novel C’était ça ou mourir had been generated by AI. In response, the Goncourt jury took the extreme step of dropping the book from its prestigious longlist. And when a researcher, Linch Zhang, ran Magnifica Humanitas—Pope Leo XIV’s encyclical on humans in the age of AI—through Pangram, even though virtually the entire document was found to be human-written, the fact that Pangram flagged a handful of paragraphs as AI-generated was enough to shift some of the attention away from what it was saying to who wrote it.

Modern AI has become so good at generating text that, in many instances, it is impossible to tell its work apart from ours, which is why so many have turned to detectors. We have, as a result, become more obsessed with how content was created than with what it actually says. I have always believed that good writing is anything that is worth reading, regardless of how it was written or by whom. The articles I most enjoy are those that make me stop and think, force me to question my beliefs and encourage me to dive into new rabbit holes of discovery. I also like articles with a stylish turn of phrase and literary construction that make me step back and marvel at how they were put together. Every piece of writing that has some (or all) of these elements is, in my view, worth reading.

Almost from the day this column started, readers have asked me about my process. Where do I get my ideas from, how long do I spend on research, and how do I find the time to churn out an article every week for so long? I would be lying if I said that any of this was easy. Figuring out what to write each week—especially in the early days of the column—has always been a nerve-wracking exercise, and there have been weeks where my mind has remained stubbornly blank as the print deadline has drawn inexorably nearer.

Coping with this has meant developing a number of systems and processes over the years. I’ve kept notebooks to jot down ideas as they come to mind, used personal knowledge management systems to process drafts and a variety of writing apps to generate the final polished product. But nothing has helped me get from the germ of an idea to a published article with as much reliable certainty as my current AI-powered workflow. So much so that it is hard for me to say how much of the past 100 or so articles in this column have come out of my own mind and how much from AI.

Here is how it works. Every column starts from something I read or heard at a conference or a thought that occurred to me in the middle of a podcast. Today, the moment a new idea strikes, I register it in my germination bench—an “idea-store” that AI maintains, where every nascent idea is nurtured and processed until it has enough substance for a full-blown article. When the time comes to write, AI scans the germination bench to find something suitable—a new idea, one that extends what I have said before, or one that addresses an issue I have yet to cover. I then structure what I want to say using AI, working out the main claim of the piece and how it should be presented. We go back and forth on the narrative structure, working out what each paragraph needs to do until we settle on the spine of the piece to my satisfaction.

In the next stage, with the outline settled, AI does all the legwork. It researches the arguments, verifies the facts and finds the sources for every claim I intend to rely on. Its work is checked along the way—a separate reviewer questions the plan, and every fact is checked against its source—not just to limit hallucinations but also to make sure that the sources are sound and reputable. It is only then that it begins drafting, ensuring that every sentence it generates adheres as closely as possible to the “voice” of my columns. The draft that comes out of this AI workflow is far closer to something I would have written myself than if I had just asked it to write an article from scratch.

No theme has more consistently run through all my columns than the idea that, rather than resisting the inevitable march of technology, we need to learn to work with it, immersing ourselves deeply in it so we can take advantage of the new possibilities it presents. I have, in my own life, tried to practise what I preach, and as AI has become increasingly capable, I am constantly trying to figure out how to make it as useful to me as possible.

As with many of my recent columns, this, my 500th Ex Machina article, has been written entirely using my AI workflow. If you have read this far, it has passed the only test that matters—holding your attention all the way to the end. Some of you will doubtless run it through an AI detector to see how much of it is flagged as AI. But if you found it interesting, does it really matter whether AI wrote 5% of it or 95%?

The author is a partner at Trilegal and his latest book is ‘Ex Machina: Essays on a Decade of Tech Policy’. His X handle is @matthan.

About the Author

Rahul Matthan is a partner at Trilegal and author of ‘Ex Machina: Essays on a Decade of Tech Policy’. His X handle is @matthan

Catch all the Business News, Market News, Breaking News Events and Latest News Updates on Live Mint. Download The Mint News App to get Daily Market Updates.

Topics

• #artificial intelligence
• #technology