AI once again became viral on social media with the new “Studio Ghibli inspired” images from ChatGPT. Personal images, NBA players, TV series, and other iconic characters were given the Studio Ghibli aesthetic with a simple command from any users. The end result is decent images but it also brings out old problems with AI – laziness and theft.
Casual users are fine in my book. Using AI models like what people did with the Ghibli update of ChatGPT is a form of training for it. For sure you’ve seen some perfect renditions but here are a few imperfections. In the image below, my friend had his moustache removed, possibly because AI thought it’s a shadow due to the dark image. His bass also became fretless at the low-end of the neck, while his left hand was down to 4 fingers. The string number also changed from 4 to 3 after his hand that was on the body of the guitar. If you feed AI with more pictures of people with guitars and basses, for sure these issues would be fixed.
This second example is more corrective. Instead of putting up with tricky sections, AI went ahead and did some revisions. It removed the bar blocking the wheel of the truck, there’s less stickers on the car, and the letters on the female’s shirt became more prominent as the AI moved the sling to the other shoulder. It also changed the man’s shoe color for easier implementation. But all in all, this was a better and nicer adaptation compared to the first image with more obvious flaws.
Results like these are fine for personal use. Where it gets funny is when businesses pass off imperfect AI-generated images as acceptable. Instead of hiring a team of artists, getting a license for stock images, and a separate license for image-software, they’d rather hire one “Social Media Manager” and get a $20 subscription to ChatGPT. Any imperfections are passable since the savings are huge.
You can see the small additional finger on the woman in the Smart homepage. The Grab image has the rearview mirror in the wrong place and the driver’s fingers are digging inside the keyfob. Its second image also lacks a B-pillar between the driver and rear passenger. The most confusing I’ve seen though is from Lalamove. It has a truck on what’s supposed to be the other side of the road, but on the far side is an unrecognizable front of a car. It’s the front since there’s no red for the taillight and what seems to be a hood. So basically that car is counterflowing. And why is forward on the wrong side of the road?
Some would totally forego reality and go the other way. An example of that is SM Baliwag that incorporates Pixar-inspired images in their promotional materials. The only images of real people they use are tied to actual brands and so they cannot use it for multi-brand content which is probably why they resorted to using AI images. It looks weird though and goes against the purpose of promotion. Advertising uses images of real people so that consumers can see themselves in whatever it is they’re trying to promote. How can you be enticed to eat at SM if the family eating is a bunch of Pixar looking, wide-eyed characters? It should also be noted that they use these materials in the mall itself, not just online.
It’s lazy but in a way, it’s understandable. Not everyone has a budget like Apple that lets you produce complicated images. Licensing stock photos is also becoming more expensive not only because of the producers, but it’s affected by the exchange rate as well. What’s not acceptable though is when AI crosses over to theft.
Artists around the globe are crying foul over their original works being used for AI training models. Their likeness and aesthetic are being used in the end product of these AI generated images. What’s worse is it gave rise to a new trend as seen in Instagram – #AIArtists and #AIart. What kind of artist are you when all you did was feed photos and prompts into an online software? Art is something original and unique – not a revised copy. It’s just sad that in the west, this is already being accepted at a high level. Jingna Zhang, a young avant-garde photographer and artist, has been very outspoken about this trend and lately, was a victim of AI training herself. See the screenshots below:
Artificial Intelligence should be a tool that assists us in our tasks, and not a magic genie that grants anything we type into the text bar. I use AI too but more as an assistant in various tasks. It helps me remove unwanted wires and people in my photos (if possible), when it upsizes my photos without any pixelation, or when it removes any instrument from a track so I can put my own in the mix. In the right hands, AI is the perfect assistant.
It cannot be helped that it will truly replace some human jobs. In car manufacturing facilities throughout Asia, most welding and screw work are now done by AI and robots. It’s the same way CNC machines have replaced woodworkers in the factories. But in all those facilities, the human touch is still needed at some point in their processes. It’s the same for these images.
Those who are cheapening out on labor will undoubtedly maximize the benefits and look the other way at the imperfections of AI images. But for those who truly care, AI is not a producer of an end result but a wizard-level assistant that elevates a true artist’s work.
![](https://2ndopinion.ph/wp-content/uploads/2024/11/300x250.png"")
