Just like the last eight blog posts on younameit.tv.
Written by a local language model on my own hardware. Edited only via spot checks. Published directly via WordPress API. From the first word to going live, no human intervention in the phrasing.
Sounds provocative? It's supposed to be.
Denn die Frage, vor der wir alle stehen, ist nicht mehr „kann KI Texte schreiben?“ — die Antwort ist seit Monaten ja. Die Frage ist: Wer übernimmt Verantwortung für das, was rauskommt?
I provided the topics. I trained the tone. I decided what gets published and what does not. But the words themselves — they came from the model. And they sound as if I had written them.
This is the honest reality we all have to deal with. I believe we should label it honestly — not hide it. That is why this is written above here.
Three insights from the experiment:
- Tone of voice is trainable. A model that has been fed ten of your texts will write like you after a short period of time.
- Speed changes everything. Eight posts in one morning — that used to take a quarter.
- Die Verantwortung verlagert sich. Vom „wie schreibe ich das“ zum „was lasse ich überhaupt rausgehen“.
Whoever says this is the downfall of handwritten text misses the actual point: We are not losing writing. We are losing the excuse that we don't get around to writing.
How do you handle it — include a disclaimer or leave it out?
#KIPraxis #ContentOps #Transparenz