Father has knocked about the road
Holidays a game resemble
When the uncle’s granny gets an idea from play
The trip cannot go wrong
Sauna the granny is happy on the way
Her dog Wille is happy when granny arrives at the sauna *
This is a poem co-created a couple of years ago by a primary school pupil and a computer. It was part of a study where schoolchildren wrote poems together with the Poetry Machine, a computer program designed at the University of Helsinki.
The study focused on the co-creativity of humans and computers, a new field of study examining the ways in which computers are able to produce something considered creative.
“Such products can be fine arts, music as well as stories, poems or other linguistic creations,” says computer scientist Anna Kantosalo.
Kantosalo, one of the developers of the Poetry Machine, discussed the development of the system and related user experiences among primary school children in her doctoral dissertation examined in August.
Creativity stems from artificial intelligence
The creativity of a computer is based on various systems of artificial intelligence. Traditionally, artificial intelligence applications have been assigned a certain problem, such as interpreting X-ray images, whereas creative artificial intelligence looks for all possible solutions within a given dataset.
“A computer can be told to draw a duck, but you can’t define where the duck is or what it is doing. This is up to the computer’s interpretation,” Kantosalo explains.
The term computational co-creativity is used when a computer creatively interacts with a human. In such instances, humans and computer systems together bring about something neither of them would have been able to produce by themselves.
In the case of the Poetry Machine, the system created word options, from which humans chose the ones they thought best suited to a poem. On the basis of the human input, the machine offered more options, eventually resulting a finished text considered a poem.
From machines to genuine co-workers
The Poetry Machine was developed for research use and is currently not in commission, but there are examples of co-creative systems from around the world ranging from the video game industry to interior design. A computational system can, for example, propose a certain roof shape to an architect or help interior designers come up with ideas.
“The human designer tells the system what kind of interior elements they would like to have in, say, a hotel lobby. The computer starts compiling images from the mass of images fed to it, picking suggestions for wallpaper and carpet patterns. Then the human chooses their favourite options and the system adjusts the image. This way, the human and the computer are able to process alternatives that are more numerous and divergent than those the interior designer would be able to conceive by themselves,” Kantosalo says.
For the time being, such systems are based on the human component leading and the machine component following. Kantosalo believes that, in the future, humans and machines will increasingly work simultaneously.
“In the future, computers will be able to choose what to communicate to humans as well as the manner of communication. In other words, the system will be able to pick the right moment to interrupt the human in their work and present its proposal in good time, much like a human colleague would do. For machines to become increasingly full-fledged co-workers, they have to be able to also contribute to framing tasks and discussing work methods.”
Public research provides protection
Computational creativity can also be used to cause harm, for example, through the distribution of fake news and disinformation. Indeed, Kantosalo finds it important that these technologies are investigated at universities with the aid of public funding.
“This enables us to provide decision-makers with information on systems in general, as well as to identify propaganda where such systems have been utilised. After all, someone somewhere is developing systems that churn out fake news. It’s also important to make knowledge publicly available instead of it being the sole possession of big businesses,” Kantosalo says.
*Source (poem translated from the original Finnish): Kantosalo, A., & Riihiaho, S. (2019). Experience evaluations for human–computer co-creative processes – planning and conducting an evaluation in practice. Connection Science, 31(1), pp 60-81.