"Artificial intelligence plays a key role in transforming modern society, opens new horizons, and poses serious challenges before us."
Congratulations! You just read an entire sentence and learned absolutely nothing.
This phenomenon even has a name — orbital argumentation. Researcher John Gallagher describes how a neural network can circle a topic endlessly without ever getting to the point. Pretty metaphors, confident phrasing, generic reasoning — and almost zero actual information.
Signs of orbital argumentation
- **Manufactured gravitas.** Everything is "key," "landmark," "fundamental," and "opens new horizons."
- **Favorite constructions.** "This isn't just a technology, it's a new philosophy," "from small startups to global corporations." And always three fancy descriptors in a row.
- **Synonyms instead of repetition.** The model carefully avoids repeating the same word, but has no problem repeating the same thought several times, phrased slightly differently each time.
- **Authority out of thin air.** "Scientists have proven," "experts note," "studies show." Which scientists? Which studies? Who knows.
How to actually tell if AI wrote it
But here's the problem. None of these signs actually prove a machine wrote the text. Humans are perfectly capable of writing meaningless drivel without any help from ChatGPT. And neural networks are getting better and better at imitating human style.
I've already written about Pangram — a service that tries to detect AI authorship in text. But even these detectors get it wrong.
So maybe the more useful question isn't "who wrote this text?" It's "what exactly did I learn from it?"
If the answer is nothing, then does it really matter whether a human wrote it or a neural network?
