Writing columns with AI

Why shouldn't Artificial Intelligence do our intellectual work for us? Maybe it can't. At least not all at once.

The thought came to me like a suddenly shot arrow that hits the target and turns on the light. Epiphany. I was sitting at the beachside bar, in the shade, enjoying a cocktail and the view of the sea in front of me, while a stubborn lazy streak was not giving up inside me. I will not write a column this week. I will ask AI to write it for me. Hehehe. I will send it to the editorial office and I will take the soft, sweet money and stay lazy like this for another week or two by the beach. I will also make the waiter an apprentice with tips.

I rubbed my hands and opened my laptop. Klaudin. I told him to find all my columns and, based on that, to write me a column because I'm really lazy.

He asked me what I would like him to write about. I told him it didn't matter, whatever he found interesting about everyday life in Kosovo.

A couple of minutes later the column came out. Read it. here.

I read the column. The machine provided accurate data that I could also verify, made some spelling mistakes, had some logical omissions (ugh, ugh, an Albanian in Germany did not know the situation with the institutions of Kosovo) and invented Xhavit. It was indeed somewhat close to my style. However, the way I write is the result of the semantic memory of readings since childhood and of experiences and experiences transformed into habits, worldviews and emotional burdens. The machine has programming according to how it has been fed with data and cattle. According to that programming and my specific requests, it had written the column with notes that may be accurate, but it had given the non-existent example of Xhavit, because it does not exist. Xhavit does not have an uncle, cousin, or neighbor. He does not exist either as an experience, or as an emotion.  

The AI ​​column therefore did not have my position on this issue. Of course, I could have told him to change his position this way and that. Improve his writing here and there. Maybe he would have done it someday, but it was more work for me than writing it myself from scratch. I gave up on the epiphany, oh damn, ideas come to you on the beach. I sat down to write without a cocktail and without a view of the sea on the beach. Phew.

I first encountered AI in university exams shortly after the release of Chat GPT. A master's student came with a seminar paper that Artificial Intelligence had written for her. She handed me the paper with a look: I've won the game this time! At first glance, it seemed to have the right structure, there were no spelling mistakes, but as soon as I started reading on the second page, I noticed some sentences and quotes that were not only inaccuracies, but were such inconsistencies that even if students love it, they don't know how to make. I immediately went to the bibliography and, oops, 6 out of 8 books that were cited didn't exist at all and some of the authors didn't exist. One was particularly interesting: Rexhep Selimi. 2005. "Gramatikë e għuħa sħ sommet shqipe", Rilindja, Prishtinë.

Before Chat GPT came out, I was intrigued by a lecture on YouTube that asked the question: Can machines write poetry? The author had done some Turing tests to prove the result. The Turing test (according to Alan Turing) says that if you are communicating with a machine and it does not recognize that it is a machine, then it should be called thinking. So, if a computer can write poetry and a human cannot recognize that it is written by a machine, then, according to the authors, yes, a machine can write poetry. 

In an essay for the film “Her,” just six years ago, I challenged this. How valid is the Turing test for poetry? How many people can tell the difference between a good poem, a column, or an original painting? However, the main discussion lies elsewhere. The difference between a human and a machine is that a machine has no cognition. This is the most frequent statement I hear whenever Artificial Intelligence is discussed. Don’t get it! So what is cognition? Do we know it? Have we as humanity managed to decipher exactly what it is? No, you guys. 

Six years ago, the question seemed like a philosophical conundrum. Now we know that machines write columns, seminars, poetry, but the fundamental question has not yet been solved. The question is no longer whether machines can? They can. The question is what do we do that machines can only faintly imitate? If we don't know the answer, then another question arises: what kept us convinced all these years that we are different?

Machines make mistakes and stupid things and goof up, slandering from time to time. Just like humans. As humans, we don't know much more than we know. We make mistakes, stupid things and slander. However, as a machine, AI helps us do calculations faster, it can help us edit something technically. For many technical professions, I've been told that it has the ability to help a lot. It's progressing. It's getting better and better at some things: including translation, we're getting it. It's an assistive device, we're getting it, like dad, when we're little and we ask him about the world: why is the sea blue, how much is one plus one, is it better to be a doctor or an engineer? And dad sometimes gets it right, sometimes he's wrong and sometimes he goes completely crazy.  

The student was standing in front of me, caught in a trap. I had prepared to say something about academic honesty, about work, about the name that is built sentence after sentence and book after book read. Nothing seemed appropriate to me at that moment. 

I didn't know what to say, just as I don't know what to say about the car. Six years after the essay on Artificial Intelligence, even recent experiences with Chat GPT haven't given us a complete answer.

We're just learning to live with it. Without knowing what it is. After all, that's how we've always lived with ourselves.

In other words: Klaud wrote me an "in other words", as quickly