AI-powered creativity
The recent development of OpenAI, including the famous image generation model like DALL-E and GAN have quickly changed the concept of art and the implication of AI on human creativity. For some, AI like DALL-E and GAN force them to reconsider the exclusive value that “human artists” can offer. For others, image generating AI paved a way for new forms of art. Now that AI has become an integral part of our everyday life and it is prospected that AI would become even more essential parts of our lives in the next decades, we must take a few moments to question ourselves “How can AI help our creativity?” “How would our relationship with AI change in the next decades?” and “What would be considerable risks of AI-based human society?”
AI based image generation
DALL-E is an AI system that turns texts into images. As one enters descriptive phrases, DALL-E can come up with highly creative image outputs. When a text prompt is given to DALL-E, text encoder maps the prompt to match text inputs to corersponding image outputs. Image decoder then generates an image that represents the semantic information of text inputs. The below example shows images generated by DALL-E with text prompt: “”a bowl of soup that is a portal to another dimension as digital art”.
Another technique that have been widely used to generate images artificially is Generative Adversarial Networks (GANs). GANs are systems that consist of neural network models that attempt to generate “fake” data that is extremely close to the readily available sets of data. The idea behind GANs is that a “generator” tries to create fake data in order to fool the “discriminator” whose role is to distinguish between a real and fake samples. This way, GANs can create images that look somewhat like real data samples that, in reality, do not exist.
How can we use these AI-based techniques to enhance our creativity? How can we use them for Transformative Creativity?
While some claim that AI may replace humans’ creative jobs, there are strong possibilities that AI-based techniques actually enhance our creativity. One way in which that is possible is that AI becomes a source of inspiration for artists. One of the prominent examples of this a work by Refik Anadol who calls himself a “technology-driven artist”. As a prominent example, Anadol uses AI based image generation technique to create thousands of images of a landscape (The statue of Liberty for instance) to create a dynamic, shifting image. Put together, these images look as if images are moving.
Another way in which AI based techniques can enhance our creativity is the use of AI powered art by non-artists. Currently, the ability to draw or visually represent one’s idea instantaneously is a skill limited to artists and artistic people. Thus, it is often challenging for non-artists to share ideas in ways that are visually stimulating. As AI-based techniques are developed to capture semantic information from text-inputs and suggest the best visual representations of them, non-artists are likely to get access to communication method previously unavailable for them.
There is yet another way in which AI-based techniques help enhance our creativity: AI can significantly advance our transformative creativity. Transformative creativity refers to the idea of expanding the usage of one object into areas previously unimagined. Examples include applying Miura-ori, a complicated origami folding technique to a satellite folding. AI aids us with transformative creativity as systems like DALL-E significantly supports us visualize multiple objects that are currently disconnected. Essentially, there exists many ideas that have not been made into reality only because of the difficulty creating physical models. However, as systems like DALL-E can quickly model a visual representation of a seemingly bizarre ideas, we are able to obtain first impressions of mind-blowing ideas. This is significant as oftentimes, getting a generalized image of the final product greatly motivates artists and innovators to proceed to the next step, which is modeling or prototyping.
Any further applications? How do you imagine people will use these technologies 20 years from now?
Whilst the aforementioned analysis mainly focused on AI-based art, the use of AI and its influence on our lives are certainly not limited to the field of art. Already, we see that AI is ubiquitous. However, how would the applications of AI evolve 20 years from now and what would it mean to us all?
Potentially the biggest contribution of AI in today’s world is the rapid advancement of the recommendation and matching algorithm. From online ads to channels we follow on SNS, numerous meta data and qualitative data are collected so that the contents that grab are attention are constantly shown. However, beyond the existing system, it is reasonable to assume that much more advanced matching system becomes possible. For instance, AI may link disconnected bodies of research in different academia to recommend a government policy. As a hypothetical example, during covid pandemic, policy makers have to collect hundreds of specialists in various range of fields and arrange dozens of meetings everyday to solve issues at hand one by one. However, if situations similiar to covid pandemic takes place again 20 years from now, AI may instantly come up with possible policies that takes into account public health as well as philosophical or ethical aspects of a given situation.
What is the implication of these models on the future of life in general?
To this date, the question of whether AI can bring about positive changes to the human society remains a matter of considerable debate; frankly, we are likely only know how all this plays out many years from now. Nevertheless, there is one thing that can be said with confidence: AI is becoming an integral part of the society and it would become even more ubiquitous going forward. As such, it is worthwhile considering the implication of the AI-driven society on our ways of life.
Perhaps one of the most visible implication of AI models on the ways we live would be felt in the field of education. At the present moment, AI and its internal mechanism remains a “blackbox” to many users. In other words, the operations of AI (particularly machine learning models) are left to “specialists” and the general public are frequently unaware of what data is fed into a model, how exactly a model learns data, and potential loopholes or weakness of a model. However, as AI becomes more pervasive and humans become even more reliant on AI based technologies, understanding of AI mechanisms would likely become an “essential knowledge for all”. That way, the society as a whole would better understand what an AI technology “can” and “cannot do”. More importantly, through thorougher education accessible to as many population as possible, we would become aware of the unique intangible assets that humans can provide that AI is perhaps not strong at.
