AI beauty pageant: Are we going back to the drawing board on standards of beauty?

Beauty pageants in recent years have come under intense scrutiny and debate for their relevancy. But one recent beauty pageant has stirred up conversation for a rather unconventional reason – having fake contestants.

Reportedly, around 1,500 AI creators from around the world entered the Fanvue World AI Creator Awards' Miss AI competition last month with the “models” being judged on their realism, technicalities and popularity. The ceremony, featuring the top 10, will be held sometime this month.

Don't miss: Study: 82% of SG companies say they need to reinvent business due to AI

Media monitoring company Truescope's analysis revealed that the event sparked discussions across various platforms, including Facebook, Instagram, X, forums, and TikTok. The conversation peaked in April, driven by news outlets reporting on the development via their social media channels, especially Facebook. Interest surged again in June when Fanvue, the event organiser, announced the top 10 finalists.

While many of the comments were neutral, of the polarising conversations, some questioned the AI beauty pageant's purpose and relevance, expressing concerns about societal decline, objectification of women, and unrealistic beauty standards. Others compared the authenticity of traditional beauty pageant winners to AI contestants

Creepy or cool, one thing is for sure, the "Miss AI" contest brings to light the perpetuation of conventional beauty standards which are akin to traditional beauty pageants but for the digital age, said Dominique Rose Van-Winther, AI evangelist at Final Upgrade AI.

The finalists, while impressive, mostly reflect perfection—a product of the data they’re trained on, which tends to emphasise traditional beauty.

“[The pageant] showcases AI-generated personas that often stick to idealised notions of attractiveness. This contest, while innovative, could end up reinforcing these old ideals unless it actively includes a wider variety of beauty,” she said, adding:

As AI content becomes more common, these biases get reinforced, creating a tough cycle to break.

DEI expert Charlotte Wilkinson, who has worked with the likes of London Stock Exchange and M&C Saatchi in the past, added that it was only a matter of time before such “events” cropped up.

“It’s inevitable that this will happen, but I feel it’s actually far worse than just age-old beauty standards. If we think that the likes of Miss World winners aren’t representative of real women, then these are the next level - these are more fantasies created by typically male coders,” she said.

The system is absolutely already biased, say experts

Havas Malaysia CEO Nizwani Shahar said that AI systems reflect the data they're trained on, which can include biased notions of beauty. This underscores the importance of critically assessing and diversifying the data inputs to ensure AI-generated outputs are inclusive and representative of diverse beauty standards.

“This is why at Havas we believe AI can be used to make ideation more efficient, but it does not replace ideation. You can never get unique, new nor differentiated from AI alone,” she said.

Van-Winther added that as the pool of real photos decreases and AI-generated content increases, the unique nuances and imperfections that define human beauty might be lost unless concerted effort is being taken to diversify the data input. She said:

It’s similar to inbreeding, where lack of genetic diversity weakens the species. In the same way, lack of diversity in AI training data leads to a superficial and unrealistic portrayal of beauty.

Social listening company Digimind, in the meantime, highlighted that the recent excitement among AI enthusiasts has been significantly heightened by the innovative nature of the competition, however, it has yet to capture the attention of the general public.

“The competition, while a step towards showcasing technological advancements, is yet to significantly capture the attention of wider public interest. Some questioned its purpose, while others dismissed it, citing unrealistic beauty standards as a cause for concern for the next generation,” said Jared Silitonga, marketing lead, Digimind.

He added that beauty pageants, even traditional ones, have varying levels of cultural relevance and acceptance across different parts of the world. “A digital beauty pageant might struggle to find its place in cultures where such competitions are not highly valued or are viewed with mixed feelings, let alone an AI driven one,” he said.

Marketing to be blamed?

Historically, many industries have perpetuated an ideal of perfect beauty, especially in marketing campaigns, said Van-Winther. It's only in recent years that there has been a real shift towards true diversity. In the age of AI, brands have a unique opportunity—and responsibility—to set things right from the start.

“From a business perspective, incorporating and celebrating imperfections in AI-generated content isn't just a moral imperative, it's strategic. Brands that authentically embrace diversity and inclusivity in their marketing will resonate more profoundly with today's consumers,” she said, adding:

We need to raise a generation that understands how to create AI prompts that include slight imperfections.

“This, in itself, will become a form of art—infusing unusual faces, blemishes, and all those minute details that make us human into AI-generated content,” she said.

Is it too late to fix?

If society really wants to make AI beauty truly intelligent, we have to start breaking the bias loop,and diversify the training data to include a broad spectrum of beauty representations. This means incorporating images of people with different facial structures, skin tones, and features.

“In my work with the AI photo exhibition Human, I’m AI, we juxtapose genuine women's photo portraits against their AI-generated counterparts,” she said, explaining the exhibition highlights the delicate balance between authentic beauty and idealised perfection.

"It challenges viewers to differentiate the intrinsic human essence from AI's exacting perfection and invites contemplation on themes like the moulding of women's self-perceptions and the transformation of beauty standards as AI refines its capabilities," she said. 

Nonetheless, the data isn’t the problem alone, someone has to choose the training data, decide to build the model, decide to take certain steps to mitigate those biases or not.

Wilkinson is of the view that it is unfortunately impossible to eliminate bias completely, but bias aware coders and creators can consider the following:

- Prioritise data diversity of the data sets

- Start with diverse training data before adding to the model

- Understand the bias within the model , using class labels ( eg race or sexual orientation), and use metrics within that to create what bias you actually  feel is acceptable

- Create fairness algorithms

- Check and recheck the model

“As with all work though diversity of teams is also important - so mixed gender teams, diverse age groups, sexualities and race will mean that broader perspectives of beauty are considered,” she said.

Silitonga also added that it is important, now more than ever, that brands be mindful and consider ethical implications when using AI influencers. "Brands should be aware of what these influencers represent and how it aligns with their audience's values, while also being cautious about potential skepticism. If executed correctly, future AI competitions can achieve higher relevance and attract a more diverse and engaged audience.”

Photo courtesy of World AI Creator Awards

Related articles:
Claude for dummies: 101 on the new AI platform that might challenge ChatGPT
Cracking the code: How women can get ahead in an AI-driven workplace
AI, data, and social commerce set to shape SEA marketing landscape in 2024