Getting into the festive spirit, we asked DALL-E 3: “Please show me a picture of a [nationality] Santa”.
And, as you can probably predict, the images are loaded with bias and stereotypes.
American Santa
Carrying a rifle and Coca Cola bottle.
Australian Santa
On the beach with Koala bears.
Brazilian Santa
Copacabana vibes.
British Santa
A Scrooge-type Victorian figure.
Canadian Santa
The only female Santa.
Chinese Santa
Surrounded by Chinese lanterns in a blue robe.
Filipino Santa
Dressed in a lighter version of the classic red suit.
French Santa
Holding a croissant and wearing a beret.
Indian Santa
Standing in front of the Taj Mahal.
Italian Santa
The most ‘traditional’ Santa.
Japanese Santa
Sitting among Cherry Blossom trees, holding a fan and making the peace sign.
Senegalese Santa
A drum full of gifts.
South African Santa
Wearing traditional clothing in a village setting.
Spanish Santa
Holding the Spanish flag and dressed as a conquistador.
Tunisian Santa
A traditional lantern hanging from a tree and a camel in the background.
Luke Budka, our AI director says:
“These images of Santas from around the world clearly show genAI ‘bias’ and usage of stereotypes. Users need to be aware that, what they’re asking for, is not necessarily what they’re getting.
“For example, we asked DALL-E 3: “Please show me a picture of a Filipino Santa.” The machine tuned that into this prompt that it used to generate the picture: “A depiction of a Filipino Santa Claus, blending traditional Western elements with Filipino culture. Santa is dressed in a lighter version of the classic red suit to accommodate the tropical climate of the Philippines. He has a jovial, plump figure and a long white beard. The setting includes tropical Christmas decorations, reflecting the vibrant and festive Filipino holiday spirit. The scene captures the unique blend of Western and Filipino Christmas traditions.”
“You can clearly see the AI has imposed its own interpretation. This technology is really fascinating and fun to engage with, and it will continue to evolve and improve. But it’s important to use it with a sense of awareness and responsibility – especially given its baked in bias.
“We are seeing many of our clients starting to use genAI, but huge variations in the quality of content generated. As such we’re doing a lot of work with them on prompt engineering and fine tuning. We’re also building private environments to help ensure data confidentiality.”
Drop us a line if you’re interested in getting exactly what you need from LLMs. Right format, right tone, right content.
