No time right now?
The marketing industry has been discussing the question of whether machines can write good texts for ten years. In the last year it has become clear: it works – and there is much more to it than converting data into body text.
“Jane, don’t you understand? It is amazing. We could just reach out and touch the city or the sky. See how far we could go We can control this world, but we cannot control ourselves. “
These lines come from a digital dime novel. As a reminder to everyone under 30: The penny was the slang term for the ten pfennig piece. For this you got chewing gum and sweets at the kiosk or a thin booklet with trivial love stories or crime novels at the station kiosk. It’s the second paragraph in a story called Breakups. They were written by Ella from Cologne. Ella is not a pseudonym. Ella is an artificial intelligence.
“At the beginning we had the idea of using an AI tool to create tons of dime novels,” explains Michael Keusgen, one of Ella’s two fathers. “It’s a huge market and the stories are always very similar,” he explains. The former film and television producer changed genre two years ago. He was burned out, wanted to create something new and not argue with TV editors and scriptwriters. This is how Ella was born. The Rhenish AI has long produced much more than trash novels. In two large projects that Keusgen is not allowed to talk about, Ella is used to communicate with people. There are marketing projects of two internationally known large corporations whose orders ensure that Ella survives financially. The marketers from both companies believe that Ella can write creative, funny and touching texts.
Ella writes word for word herself, something Keusgen is particularly proud of. At the beginning, it first produced 179 scrap texts before a readable work was included. In the meantime, every seventh text is so good that it is written on the specially designed for your short stories Online platform Frogs42 can be published.
The machine writing of texts is usually based on one of two very different approaches: Either the machine generates a kind of probability calculation for the next best word in the text – like Ella – or it makes use of a huge learned library and uses existing text fragments and Sentence building blocks. Whether one method is better than the other is as much an academic question as is whether either form of machine writing is really creative. “Better” is a relative term that is primarily oriented towards the task at hand. If you need simple product descriptions, it is no problem if text fragments have already appeared elsewhere. This automated form of texting may be faster and cheaper.