An exercise has taken place whereby an artificial intelligence (AI) image generator has been used to create images when given brand names as prompts. The technology remains in its infancy although it is progressing at a rapid rate. This model has produced an array of intricate images based on names of major companies from worlds of fashion, technology and other areas.
The activity has been undertaken by an online shopping site called Ubuy South Africa. The firm used an AI image generator called MidJourney to create images which used famous brands as a prompt. With each assessment, the model created a selection of four pictures for each brand. Copies of the images were sent to Digital Journal.
The results reveal how the image generation system created images that may or may not closely align with the brand’s style and product depending upon the perspective of the viewer and what they think of the brand. It should also be noted that the resultant images were only based on just a single word prompt and with no further contextual data.
Coca Cola and Pepsi
Two of the world’s biggest drinks brands received similar treatments from the AI. Each of the eight images took the form of a portrait of a woman, with the brands’ colours represented. With this feminisation of fizzy beverages, three of the Coca Cola images show a woman is holding a drink and the colours of the liquid are more red and orangey than the typical caramel colours.
With the Pepsi renditions, none of the Pepsi images featured any sign of a soft drink. The women in the Coke images have a parallel with film stars from yesteryear, while the Pepsi images seem to be more superhero-inspired.
Facebook, Amazon, Google, and Microsoft
Using “Facebook” as a prompt produces some slightly darker results than other brands, including a large skull surrounded by smaller bones, a woman being haunted by gruesome floating heads, and a person sat in an apparent wasteland strewn with computers. These are not terribly complimentary.
The AI’s interpretation of Amazon blends the river and rainforest with the tech giant, perhaps indicating the AI’s misinterpretation of the word ‘Amazon’. The results for Google are colourful castles in the sky surrounded by floating balloons. Microsoft gets the most varied set of images, ranging from greenery and water to a woman holding a tablet and a futuristic cylindrical cityscape.
Nike and Adidas
Both Nike and Adidas results in colourful product shots of the companies’ shoes, splashed in multiple paint colours. The brands’ logos appear similar, although an Adidas features four stripes rather than the brand’s signature three.
Honda, BMW, Tesla, and Toyota
The results for the car manufacturers vary considerably. Honda as a prompt produces motorcycle-focused imagery – two featuring a female rider, and two vehicles by themselves. BMW gets a more historical treatment, with retro-looking vehicles against mountain backdrops. The Tesla images all feature a man who could be interpreted as an amalgamation of Nikola Tesla and Elon Musk, alongside a car, and always in front of a fiery and stormy background. The car is the star in the Toyota images, although it appears in some weird and wonderful places, with strange creatures being a consistent theme.
Gucci and Louis Vuitton
With highly expensive ‘fashion’, the results for Gucci are all portraits of women in flowery outfits and surrounded by flowers as well as other animals – immaculately groomed dogs appear in two of the pictures, while exotic birds perch on a model’s shoulders in one. Louis Vuitton’s handbags take centre stage in three of the four images, while the fourth, like Gucci’s, features a female model in an outfit that is adorned with flowers.
AI can produce impressive images and it can also generate misshapen creatures and unworldly scenes.
What do these images mean? Given that AI is reflected of human programming, do the images say more about our perception of the world around us? Is this something that is complimentary to the brands or negative? These types of questions will need to remain at the forefront as the use of image generators accelerates.
