AI and Creativity: An Efficient Partner for Creative People
Introduction
Nowadays, there are AIs everywhere. It’s in our phones, suggesting us interesting Tweets; our home, responding to our voices; or our cars, driving for us. It’s used to make our lives easier, and this could be true for creative activities, too. In the class Interaction Design, I learned about the current relation between AIs and creative activities, and explored ways AIs can be used creatively. Below, I will showcase what I created in the class, and then discuss what can be thought about AI and Creativity in general.
My Creation: Imaginary Pokédex
In the class, my friend and I made a Pokédex of an imaginary region, Gangener Region (GAN-generated). The Pokédex is made from Pokémon images generated by GAN, an image generation AI model. It has 19 Pokémon in total; 10 from drawings generated by AI, and 9 drawn by myself inspired by AI generated pixel images.
We have decided to create a Pokédex, since we guessed that AI could be a solution to a potentially upcoming problem of difficulty of coming up with new Pokémons, considering the growing number of Pokémons. There are currently 905 Pokemons, and more if region forms or other forms are included.
To create the Pokédex, we trained two GAN models; one which generates drawn Pokémon images, and one which generates pixelated Pokémon images. The drawing model was trained with 1113 images of official Pokémon drawings, and the pixel model was trained with 1112 images of pixel Pokémons.
After generating the Pokémon images, we picked ones to include in the Pokédex.
From the generated drawings, we picked ones which were Pokémon-like. From the generated pixel images, we picked ones which were at least recognizable as creature, and then I drew an image based off the pixel image. Each drawing took less than 20 minutes to create.
Then, from the images, we guessed each Pokémon’s names, types, and other details.
Once all the details were decided, we created the Pokédex.
Thoughts about AI and Creativity
AI can speed up and expand creative activity
Through the creation of a Pokédex using AI, I have experienced enhancement of creativity, especially in terms of speed and idea generation. As my friend and I have guessed, image generation by AI have greatly sped up the process of creating new Pokémon; it generated drawings, and became an inspiration for other drawings or the details of each Pokémon.
AI cannot create a completely new concept on its own, however. Since AI (especially the one I have used for the Pokédex) is trained by preexisting data, although they can be helpful when generating new versions of a preexisting concept, it cannot make something other than that. The AI I have used were able to create Pokémon, but not anything other than Pokémon. AI cannot, on its own, achieve the “transformational creativity”; one of the 3 types of creativity distinguished by Margaret Boden.
Further application and the future of AI
Some further application of AI in terms of creative activity could be to automate parts of a creative process. For example, when drawing, the coloring could be done by an AI. Since AI cannot create something completely new on its own, it is likely that in the future, AI will continue to be used as a helpful tool or partner for creators.
Implication on the future of human activity
In the future, it can be guessed that there will be an increase in quality of creative activities. This is because with the help of AI, many parts of various creative process can be automated, and become more efficient. There might also be an increase in new concepts, since AI can efficiently create new versions of existing concepts, and quickly deprive of new ideas potentially existing in the concept space.
Issues with data biases
Data biases could become a problem when working with AI. The data AI was trained with could potentially be skewed or contain problematic inclinations, such as with human gender or races. However, this also happens with human beings; the reason it might become a problem is because AI can achieve this very efficiently, which amplifies and makes it easier to notice of this problem. To avoid this, work done by AI could be reviewed by multiple human being, like how it is done with work done by human.
