• Thu. Apr 18th, 2024

Do AI Models Actually Improve Diversity?

ByMaria Farsoon

May 16, 2023

London-based model Deba Hekmat recently shared a story on Instagram pointing out the eerie growth in dominance of AI-generated models of colour across social media and online shopping websites. She pointed to Shudu Gram, the first black digital model, as well as the fact that all of Vogue Brazil’s digital covers this March have exclusively featured artificially programmed models of colour, as opposed to “black [human] models that they could have paid”. 

One of the very first of the CGI celebrity cases was Lil Miquela, the figure that appeared on Instagram in 2016 and inspired the term ‘digital influencer’.

With 2.8 million followers, the robot’s creation has permeated social media by occupying the ‘influencer’ space, collaborating with brands and celebrities, releasing music, advertising fashionwear, and even becoming nominated for the Shorty Award for Best Celebrity.

Student Amber Quinn remembers a moment in time when the public was “so convinced that this was a real person”. The figure is ultimately a product of motion graphics but has successfully created an influential online platform by adapting to cultural phenomena, as technology itself does in order to thrive and attract users and consumers. Lil Miquela is not a human being. Understanding the inequalities provoked by the AI Model requires us not to ignore this fact. 

Shudu Gram, created in 2017 by Cameron-James Wilson, has been labelled by Lauren Michele Jackson in the New Yorker as “a white man’s digital projection of real-life black womanhood”.

AI models of colour have recently been generated by multiple fashion brands with the aim of representing diversity in an affordable, efficient, and innovative manner. This however raises the question of authenticity and equality, issues that permeate the discussion around the industry’s removal of opportunities for members of a marginalised group to not only express themselves, but to make a living.

History student Amirah Ahmed sees Shudu’s creation as just another example of the “typical” behaviours enacted by people in such positions of power whose absence of “respect for the work ethic of real people of colour” might perpetuate ignorance in the favour of capitalist ideals.

To elaborate, Ahmed agrees that it is especially problematic when AI models of colour are produced by someone not of that race. Shudu’s invention, as Jackson cites social theorist Patricia Hill Collins, is “an image ‘contrived by a white man who has noticed the movement of dark-skinned women.’”.

Is it truly sustainable and fair to justify attempts at diverse representations that are non-human, by replacing human expression with artificial technology? Does progress really mean creating artificial projections of underrepresented groups as opposed to giving the recognition and economic benefit to a real human in that group? It seems ironic and upsetting to utilise an advanced tool such as AI for the benefit of such regressive ideologies that in turn deprive the particular group from which one’s ‘inspiration’ is pulled. 

Though AI has been innovated in ways as a force for societal development, the opposite may also be said, with such misuse evidently prolonging the capitalisation off of under-credited ethnic groups.

Vogue” by mespechesmignons is licensed under CC BY 2.0.