Art — Is it Still Just a Human Thing?
The apprehension of AI, Terminator being the prime example, has left many wondering what humans can do better than machines. There aren’t many, given that as machines advance they will perform in ways that outdo human capacity. However, many pride themselves on what does make us unique in ways machines can never hope to accomplish: emotions and creativity. Both are central to human culture, but the medium through which they are particularly held high in regard is art. Art by definition is the expression of human creative skill or imagination with the intent of being appreciated for their beauty or emotional power.
Humans and only humans have long been the sole possessors of art, but is this still the case? Is art still just a human thing?
Well, the answer to that question is not so simple any more because today machines can create art. As artificial intelligence progresses, it is being integrated into more and more aspects of society, and one of its many amazing applications happens to be the product of artists finally getting their hands on AI.
As I said, machines can create art. But perhaps it’s better described as machines generating art because AI art is trained on many thousands of images. These large sets of data are extensively analyzed, reconfigured, manipulated to bring into effect aesthetically pleasing images.
Art such as this cannot be created so easily as to type a few lines or a prompt on DALL•E, which is all the rage lately when it comes to AI art. Instead, artists of the like design art by write code that incorporates machine learning technology. Algorithms are utilized to discard a strict set of rules, and instead “learn a specific aesthetic” and style by analyzing thousands of images.
The process entails:
- Pre-curation — choosing specific images that complement the “visual outcome in mind” (large collection of data is still needed!)
- Utilization of algorithms called “generative adversarial networks”. “Adversarial” because there are two sides: one that generates random images, and the other that is taught to judge the images based on the input (pre-curation)
- Post-curation — selectively choose images to keep or use
There’s a lot of other things going on behind the scenes as well, and interestingly enough Natural Language Processing works together with AI-driven art in order to ensure accuracy in DALL•E and other AI art tools, which ask for a prompt through their user interface in order to get a general, or perhaps more specific, direction of the generated images. WIRED highlights a significant contribution a few years back in which NLP researchers found a way to satisfy the need for large datasets by scraping text from novels and the internet, highly valued because statistical patterns could be analyzed from large amounts of structured text.
AI-art generators still have their limitations, however, a major example being how they can reflect societal biases. For example, doctors are generally rendered as male while flight attendants as female. WIRED interestingly attributes this as well as nonsensical or glitch images to “the systems’ lack of understanding of the world” and “[replication] the [of] web-sourced images they were trained on”.
Users and people in general had an overall positive response to DALL•E and other such platforms. In fact, Benjamin Von Wang described them as “wonderful tool[s] for someone like me who cannot draw”, expressing the satisfaction this can give to people that are not the best at art, but nonetheless enjoy it. Surprisingly, even artists have found AI-art to be a source of inspiration or an “explosion of creativity”, perhaps through the meshing of many different images into one. The implications have even convinced some that image-generating AI will “permanently change humanity’s visual movement”, as what was initially hard to imagine can become an easy-to-understand visual that can impact efficiency, time, and expenditure on projects.
But can AI really be considered an artist of its own making? After all, it doesn’t make anything on its own. Instead, it’s required to have an input, a pre-curation, a pre-defined direction of what the art should be like. In the end, AI-art is just a tool, nothing more. Can it come ever come close to match the depth of the Mona Lisa? To this day, the controversiality of whether the Mona Lisa depicts happiness or sadness continues. Moreover, is AI-art in the end truly original? Can it derive or establish its own meaning in an attempt to express itself?
How much longer will creativity and imagination be associated only with humans? Of course, in accordance with reality there is doubt about whether or not AI has the potential to become originally creative or an independent thinker. They believe the limitations of AI are bound, or should be bound (again Terminator 🤣), but what is the future if not a place to be unrealistic, if not a place to entrust our hopes and dreams, even the ones we don’t ever plan on seeing?
Sources:
- https://www.wired.com/story/when-ai-makes-art/
- https://www.smithsonianmag.com/innovation/with-ai-art-process-is-more-important-than-product-180970559/