The Beauty of Imperfections
Human intelligence is imperfect, while AI is sometimes too perfect. We should embrace our imperfections - and disclose it when our "perfect" output is the result of AI
At first glance, it can be difficult to tell the difference between natural and artificial. But as you look at them more closely, you notice that they’re different. The one is rare and authentic, the other is fake and ubiquitous. And in a world of identical ubiquity, the imperfect and authentic stands out.
It’s not me making this observation this time, but Al Cook, the CEO of De Beers, the world’s leading diamond company. He and I caught up this week on the occasion of De Beers publishing a book on the history of the century-old company and its iconic “A Diamond is Forever” marketing. You can watch the video below.
I appreciated Al’s insights, as I see the similarity between his world and ours: in the diamond industry, it is now possible to make a perfect artificial diamond in a few days in a factory, which has understandably caused some disruption in the diamond industry. But as with artificial intelligence, an artificial diamond is perhaps too perfect.
Every natural diamond has hundreds of natural “birthmarks”, as Al put it, small imperfections which make each diamond unique as it is formed in its billion year journey in Earth. And as I see AI permeate the internet, and writing, and even LinkedIn comments or posts, I’m starting to believe that imperfections in writing are as valuable for “human intelligence” as these birthmarks are for natural diamonds.
It’s why I’m often now intentionally not correcting some grammatical or vocabulary errors in my posts. At the very least, I figure, people will immediately know it is me, a human, who wrote this, and not some generative AI program posting on my behalf.
More broadly, though, we’ll need a more agreed upon way to deal with the ubiquity of AI. It is now all around us, and most of us use it in our daily lives. Just in the past week, I used AI to send an email to the local police in French; dictated questions to an AI program while on the ski slopes, and asked it to reformulate them; and of course, used AI programs to inform my own thinking, knowledge and research, in the way I would use Google in times past.
So far, I’ve not needed a guidebook to keep my own authenticity. But the near misses are getting more and more common, as are the instances in which I get frustrated when others use AI-generated emails, messages, and questions to interact with me. It feels offensive, and disrespectful. And surely, I’ve offended others, such as perhaps the local policeman, when I used an AI program to write my email to him.
What should those guidelines look like? I’ve found these ones, from my book publisher Wiley about the use of AI in research, to be a good start:
Seek input from AI, then implement changes in the work yourself
Make final decisions about what to include in your work
Disclose your use of AI transparently in your manuscript so readers can understand your process
And for the disclosures, the recommendations they made include these:
When to disclose the use of AI? When AI generates or substantially rephrases text for the manuscript; creates first drafts of standard sections
When not to disclose the use of AI? AI corrects grammar and spelling; suggests word choices; rephrases awkward sentences
It may feel a bit awkward at first to disclose the use of AI, especially for those who made writing, or public speaking, or professional communication their trademark. But the alternative is worse. Using AI and not acknowledging may cause frustration with those on the receiving end. And, at some point, it may lead to public shame or dismissal, especially when the “truth” of your use of AI is revealed.
On the bright side, the smart use - and disclosure - of AI is likely to lead to better outputs, whether in Pulitzer prize winning journalism, or business analytics. I for one will try and test disclosures in this newsletter going forward. Although this edition, rather exceptionally, was written without the help of AI. As Al Cook jokingly told me, in the case of our interview, it was A-L, not A-I, who provided the insights on natural vs. artificial intelligence and diamonds. And the same went for me.
Stay safe, and until next week!
PS: De Beers’ coffee table history book is called “A Diamond is Forever: The Making of a Cultural Icon, 1926-2026”. You can order it from Assouline here. (I was not paid for this promotion!)
