The Artistic Abilities of AI
Art has been a form of spiritual and personal representation for human beings since the palaeolithic era. Cavemen designed figurines in cave walls for the satisfaction in engendering interesting depictions. Back then, it was rudimentary and unsophisticated. Today, art has become employed in every aspect of the world. Architecture, new technological innovations, and mechanisms are all based on artistic formulas that help facilitate the process of innovation. AI is fostering a vital role in this process, albeit with little steps. As of recent contributions, there are AI prototypes that can sketch and draw freely, in the way us humans did as cavemen.
The credit for this development goes to researchers who employed generative adversarial networks, or GANs - a class of 2 neural networks which both produce designs and evaluate them. These GANs were employed to create two GAN prototypes that had artistic purposes: SkeGAN and VASkeGAN. Their potentials were observed in a competition to envision which prototype holds greater strengths for art and which has greater efficiency. SkeGAN was created using 50% publically available material and data; whereas VASkeGAN was designed with serious auto encoders - enforcing a bright dichotomy between the two models. It would be interesting to see which "is the better artist."
The challenge was to be taken place with Google's Quick Draw Dataset in order to stimulate a training ground for the designing of simple creations: cats, bugs, stick figures and vehicles, as modelled in the picture. The prototypes were then trained, to become acquainted with the patterns in drawing these materials, and to see which one could not only reproduce these patterns - but twist them into creating unique and personable material, with creativity that rivals that of human beings. A material labelled "ske-score" was incorporated in order to analytically evaluate which prototype worked most efficiently and accurately. It was defined as "the ratio of the number of times the pen is lifted to the number of times it touches the paper while the sketch is being drawn." A good ske-score would be a 5, with the worst being a 1.
Eventually, the results indicated that SkeGAN was the champion. SkeGAN received a unanimous ske-score of 4 in both clarity and beauty and a 5 in efficiency. Its art, which you can see in the picture below, exceeded VASkeGANs almost indubitably. The art was more pronounced, sophisticated, and mature for a clunk of machinery. Furthermore, the art was not only qualitatively better, but SkeGAN also held that natural "creativity" aspect, in which it could memorise the patterns and then twist the processes into developing unplanned designs and models.
With these steps being taken to bolster the role AI plays in art, it is inevitable that eventually, our designs for innovations such as rockets, spaceships, high-tech vehicles and structures will be heavily aided by AI. The creativity that humans innately hold can be transmitted to AI with practice and clarity. Yet, it is doubtful that SkeGAN will be the sole artistic AI prototype in this endeavour. As the researchers designed VASkeGAN, it is probable that other companies will design their own artistic AI models that perhaps not only rival VASkeGAN - but SkeGAN itself. The question of which-AI-can-draw-the-best will be challenged for the years to come. Perhaps, even the fields of art and design will grow stronger and larger - as AI can cooperate with these artists efficiently and productively in order to generate newfound designs and structures that truly impact the world. The possibilities are endless.