AI is utilized today in so many ways that it might be easier to list where it isn’t. In the tech industry, one of the most common uses of AI and machine learning is in photography, like touching up photos or even heavily modifying them, like the controversial FaceApp. Hopefully less damning, a new MIT research project called AI Portraits Ars takes those same principles to render selfies and photos in a stylistically correct masterpiece from the classics, the Renaissance, or even contemporary art.
Style transfer this is not. That is also a popular AI application that takes a photo and applies the painting styles and colors of another art style. Impressive as that may be, the final product doesn’t actually change the nature, form, or structure of the original phone.
AI Portraits Ars, in contrast, generates an image that looks like it was painted by an artist from that period. Lines are blurred, skin tones are modified, and even hairstyles are changed. And the actual art period it applies can vary depending on the presence or absence of elements in the original photo.
It’s definitely a very impressive demonstration of some of the capabilities of AI and machine learning but it could also be a vehicle for tangential education as well. For example, the researchers point out how the AI isn’t able to reproduce smiles in the final portrait. That’s not because of some programming error but because the source materials, which focused mostly on 15th-century European art styles, also shied away from portraying smiles for one reason or another.
Given the recent controversy surrounding the selfie-aging FaceApp app, it might have been poor timing to publicize this AI experiment, as impressive and whimsical as it may be. The site does have a privacy disclaimer that photos are immediately deleted after being processed by the AI but you’ll have to put your faith in a group of MIT Researchers to keep their word.