Leonardo da Vinci’s masterpiece, Mona Lisa in her new setting in the refurbished Salle des Etats … [+]
Why are AI hobbyists so bad at hyping generative AI?
On Twitter, the latest AI trend sees enthusiasts “fill in the gaps” of famous images, from the Mona Lisa to iconic memes, enabled by the launch of Adobe Firefly, which boasts a (genuinely impressive) Generative Fill feature.
This isn’t a new technological development, but it is vastly improved; now, users can easily replace unwanted clutter in photographs with imagery that suits the overall tone, or fill in gaps with just a few prompts.
Or, they can attempt to “improve” Leonardo da Vinci’s masterpiece, the Mona Lisa.
The trend was sparked by a viral Twitter thread showcasing an enlarged image of the Mona Lisa, floating in a craggy, alien landscape. The image was retweeted by critics, who noted that the edit was made in poor taste.
Twitter user @PlanetofFinks pointed out that “this is actually a great demonstration of what AI can do. In that it doesn’t know to give her legs and adds a second horizon above the horizon because it isn’t an intelligence and doesn’t know anything.”
Another viral thread showcased an image from Reservoir Dogs, filling in the space outside of the frame. The same thread used Generative Fill to “zoom out” of the close-ups from The Good, the Bad, and the Ugly, which horrified film buffs, who pointed out that the iconic shots were framed with great deliberation and skill.
These edits seems to have been made in earnest, being genuine efforts to showcase the potential of AI. It didn’t take long for Twitter users to poke fun at the trend.
Of course, Generative Fill is an extraordinarily useful tool, a timesaver for artists and Photoshop tinkerers who want to experiment with their images. From an artistic perspective, it is genuinely interesting to see how the AI interprets the prompts and original image, to see what the machine “thinks” should fill the gap.
The results are often quite revealing, the biases of the machines exposed via landscapes that resemble desktop backgrounds, and hypersexualized images of women with bulging breasts, contorted in provocative positions.
Of course, the images that AI enthusiasts choose to promote are just as revealing, and many noted their bleak, nihilistic attitude to art.
Online, AI enthusiasts are beginning to develop a bad reputation; many have even compared them to NFT guys, whose boundless enthusiasm for ugly art and energy-intensive technology was widely mocked.
The AI art explosion has proved that even powerful tools like Generative Fill can’t substitute skill; creating complex imagery has become faster and easier, sure, but intent still matters.