Over the last several months, innovative online platforms like DALL-E2, Midjourney, and Stable Diffusion have erupted onto the market, allowing anybody with a smartphone or PC to produce highly polished art by entering in basic text instructions.
Sophisticated algorithms have learned to duplicate the precise styles, colors, and brushstrokes of famous painters, allowing users to instantly create their own unique renditions of masterpieces like Van Gogh, Dali, Turner, and Monet.
The technology can bring bizarre and strange things to life in stunning realism. If you search for Cookie Monster climbing the Shard, you’ll see the children’s television figure incongruously mounting the skyscraper. When you search for ‘Taylor Swift leading an army of the undead,’ a terrifying picture of the pop singer appears, as if created from the depths of hell itself.
The limitless possibilities have spawned an avalanche of memes on social media, bringing the concept of generative AI to the forefront and addressing some basic concerns in the process: if a computer creates art, is it actual art or merely the product of sophisticated calculations? And what does this entail for human artists working in video games, music, cinema, or television? Are their hard-earned creative abilities being undervalued, and their employment jeopardized?
The more I learned about how they’re made, what kind of data they not only use, but require, to generate results, the more hesitant I became, to the point where I can’t in good conscience or good faith recommend these tools to anyone I know, whether they’re a concept artist or an art director; nobody.
The idea of utilizing AI to create art may seem new, yet studies in training computers to emulate human creativity extend back decades. In the 2000s, innovation surged due to the advent of computer coding tools for artists, open-source initiatives, and the public availability of enormous datasets, such as ImageNet, that could be used to train algorithms to categorize photos and identify things.
It’s still early days for generative AI, and systems may fail to portray some aspects realistically, such as human or animal body parts or written information, which is often mangled. Nonetheless, they have already shown their capacity to compete with human art, fooling even seasoned art experts.