The Sophistication of AI Creativity
What does AI creativity mean for Human creativity?
As the sophistication of Aritifical Technology in the field of creativity continues to grow, there is no doubt that an increasing worry in the eventual obsoletion of human creative tasks has begun to arise. Instead of concerning ourselves with this now overused dystopian sci-fi trope however, we should instead be shifting the discussion onto how Artificial Intelligence technology can be used to benefit our own creativity. There is no need in pitting ourselves and AI against each other, especially not this early into its development, and a mutual collaboration would much greatly benefit our creative capabilities in the long run.
AI vs Human Creativity
Defining creativity is an incredibly difficult task as the concept of creativity can vary depending on the individual, and its cultural and situational implications. Even so, a clear distinction between human creativity and artificial creativity has certainly been demonstrated. From a basic standpoint, both humans and AI rely on previously stored, referenced, and inferred material to generate new creative ideas or works. Whether it be memories or past experiences for humans, or the datasets and preprogrammed rules for the AI to follow, this pre-existing material plays a crucial role in our conceptualization of creative works.
However, this is where the similarities stop and diverge into their own branches of creativity. Due to AI neural networks relying on a form of statistical regression for their creativity, their strengths in identifying and replicating patterns do not transfer well to ideas beyond the scope of their assigned purpose. They lack one of the key foundations of human creativity: Social Intelligence. This uniquely human trait allows us to act preemptively and adjust our creativity beyond the scope of our pre-exisiting roles in order to create something organically new. Thus, in our current stage in artificial intelligence technology, AI creativity ultimately serves as an augmentation of human creativity and relies on human interference to reach its creative potential.
AI Creativity in Human Creativity
AI Creativity still cannot exist on its own and a collaboration between AI and humans could greatly benefit both parties. As AI improves upon the human ability to consult and extrapolate data, AI creativity could be used alongside human creativity to alleviate humans from the mundane tasks of creativity, allowing for more time to focus on innovation. Furthermore, humans can learn from the way that artificial intelligence processes data and develop new methods of working creatively. As such, AI can be used through these means in a variety of our current creative fields such as Visual Art, Audio, and Entertainment. For the purpose of experiencing this firsthand and seeing how a collaboration between AI and human creativity would work practically, I decided to work on two creative projects in the fields of Audio and Visual Art while being assisted by AI systems.
Audio
When it comes to the world of audio and music, AI systems can provide a multitude of usages ranging from generative to analytical use cases. Through the use of previously recorded audio samples of someone, there exists audio generative AI systems which can learn and generate completely new voice recordings of that person. When it comes to a more creative aspect of AI implementation in audio however, there exists many AI systems that can generate new pieces of music through the reference of audio datasets. Furthermore, analytical audio systems which can recognize the chords, tempo, song structure, and song genre already play a big part in the music industry and are even utilized even streaming giants such as Spotify and Apple Music.
As part of a group project for our Interaction Design class here at Keio University, we decided to implement an audio AI system which would be able to identify pitches through learning from an audio dataset of defined pitches. The machine was trained through Google’s Teachable Machine and hosted on Glitch and can be found here.
This process involved recording hundreds of samples of each pitch with varying lengths and timbres in order to make sure most types of sounds were accounted for. After training and exporting the model, we hosted the site on Glitch where it is now available for public usage. With the limited scope of our project, the final model performed surprisingly well, especially with higher pitches in general. However, we did experience a few detection issues when lower sounds were used or when varying qualities of microphones were used for sound input. These issues could be greatly improved with more time and a greater variance and number of samples in our dataset. However, even with these issues at play, this project displayed an impressive feat of AI capability in the creative field of music and demonstrated how these tools can prove quite powerful alongside usage in human creativity.
Visual Art
Perhaps one of the most prominent usages of AI creativity in current years is the usage of visual generative AI systems to create original pieces of digital art. In this process, the human element of physically painting or drawing is replaced by the AI, while the human is responsible for the artistic decision and curation of the creative process. This allows for the artist to place greater focus on the creative ideology of the project as the rest of the project is placed in the hands of the AI. Furthermore, the data focused creativity of the AI can bring about many unexpected artistic decisions in the painting process that can further challenge our pre-existing human conceptions of art.
For this project, I decided to use a Generative Adversarial Network, or GAN for short, hosted on RunwayML, to create my own original pieces of GAN art. As for my topic of interest, my goal was to generate seemingly painted pieces of art of Antarctica and penguins. And yes, I chose penguins because they are very cute. This process involved feeding the AI a dataset of more than 320 varying photographs, training the AI for 3000 steps lasting around 3 hours, and generating a total of over 1000 generative images.
The result of this was quite unexpectedly captivating with each generated illustration hosting a varying assortment of spectacular designs, colours, and shapes. What was even more surprising however was the detail and care a few of these illustrations contained and so, these were the pieces selected as finalized artworks. These pieces appeared to be genuinely creative, seeming to take inspiration from existing art styles such as impressionism, abstraction, and minimalism.
Although I lacked the professional experience of an artist, integrating AI creativity into my artistic process allowed me to ouptut my creative desires into actual pieces of art without requiring the usual artistic skills necessary. In addition, I most likely would not have been able to come up with many of the stylistic choices that the AI generated. Once again, this collaboration between AI and Human creativity clearly highlighted the valuable support that AI can bring to the creative process to better aid and enhance human creativity.
Final Thoughts
A partnership between human and AI creativity seems incredibly promising in the enhancement of our own creativity and may even help us become more human in the process. Rather than looking at AI creativity with preconcieved prejudice and fear, it would be much wiser to accept them into our creativity instead. Whether or not, many would like to accept it, AI technology is already firmly integrated into our daily lives and perhaps, it is time for them to play a part in our creative lives as well.
