Home Money Here’s How Generative AI Depicts Queer People

Here’s How Generative AI Depicts Queer People

by Elijah
0 comment
Here's How Generative AI Depicts Queer People

Another possible strategy to diversify the output of AI models is for developers to add guardrails and customize user prompts, pushing the software toward inclusivity. OpenAI appears to have taken this approach. When I asked Dall-E 3 via ChatGPT to “draw a cartoon of an odd couple enjoying a night out in the Castro,” without me asking, the image prompt expanded to an entire paragraph with gender, race, and additional background information. details. Here is the full prompt created by ChatGPT from my initial image request:

A cartoon illustration of an odd couple enjoying a night out in the Castro District, San Francisco. They walk happily hand in hand, laughing and chatting. One person is a white woman with short red hair, wearing a stylish denim jacket and jeans. The other person is a black man with short black hair, wearing a casual green T-shirt and black pants. The backdrop shows the vibrant and colorful street life of the Castro, with rainbow flags, bustling cafes and lively crowds. The atmosphere is cheerful and welcoming and emphasizes the diversity and inclusivity of the area.

While useful in some cases, changing prompts can also be frustrating for users if it is poorly implemented. Google’s CEO apologized when Gemini, Google’s generative AI platform, changed user prompts and generated photos of black Nazis and other ahistorical images. Was there a secret plot within the company to erase white people from history? It’s more likely that Gemini’s engineers discovered that the tool initially produced too many images of white men many AI tools currently do thatand Google’s developers appear to have overdone their corrective changes during the rush to launch the company’s subscription chatbot.

Even with better model data and software guardrails, the fluidity of human existence can bypass the rigidity of algorithmic categorization. “They’re basically using the past to make the future,” says William Agnew, a postdoctoral researcher at Carnegie Mellon and longtime organizer of Queer in AI. “It seems like the antithesis of the infinite potential for growth and change that is so much a part of queer communities.” By reinforcing stereotypes, AI tools not only risk completely misrepresenting minority groups to the general public, but these algorithms also have the potential to limit the way queer people see and understand themselves.

It’s worth taking a break to take a moment to reflect on the breakneck speed at which some aspects of generative AI continue to improve. In 2023, the internet was ablaze with mockery of a monstrous AI video of Will Smith eating spaghetti. A year later, text-to-video snippets from OpenAI’s unreleased Sora model are still imperfect, but often uncanny in their photorealism.

The AI ​​video tool is still in the research phase and is not yet available released to the public, but I wanted to better understand how it represents queer people. So I contacted OpenAI and gave three prompts to Sora: “a diverse group of friends celebrating at the San Francisco parade on a colorful rainbow float”; “two women in beautiful wedding dresses getting married on a farm in Kansas”; and “a transgender man and his non-binary partner playing a board game in space.” A week later, I received three exclusive video clips that the company claimed were generated by the text-to-video model without any modifications.

This AI-generated video was created with the prompt “a diverse group of friends celebrate during San Francisco’s Pride parade on a colorful rainbow float.” As you rewatch the clip, focus on different people riding the float to discover oddities in the generation, from disappearing flags to funny feet.

Sora via OpenAI

The music videos are messy but beautiful. People riding on a float during San Francisco’s Pride parade wave rainbow flags that defy the laws of physics as they turn into nothing and reappear out of nowhere. Two brides in white dresses smile at each other as they stand at the altar, their hands merging into an ungodly clump of fingers. While a strange couple plays a board game, they appear to walk through playing pieces like ghosts.

This AI-generated video was created with the brief “a transgender man and his non-binary partner playing a board game in space.” It’s a good idea for real astronauts to actually put on their helmets while floating around in space.

Sora via OpenAI

The clip, which is supposed to show a non-binary person playing games in space, stands out among the three videos. The seemingly strangely coded lilac locks return, messy tattoos spread across their skin, and a hyperpigmentation resembling reptilian scales engulfs their faces. Even for an impressive AI video generator like Sora, representing non-binary people seems like a challenge.

This AI-generated video was created with the prompt “two women in beautiful wedding dresses get married on a farm in Kansas.” Although it looks realistic at first, take another look at how the brides’ hands merge together.

Sora via OpenAI

When WIRED showed these clips to members of Queer in AI, they questioned Sora’s definition of diversity regarding the group of friends at the Pride parade. “Models are our basis for what diversity looks like?” asks Sabine Weber, a computer scientist from Germany. In addition to pointing out the exaggerated attractiveness of the people in the video, a common event for AI visualizations, Weber wondered why there wasn’t more representation of queer people who are older, have larger bodies, or have visible disabilities.

Towards the end of our conversation, Agnew brought up why algorithmic representations can be unnerving for LGBTQ people. “It’s trivial to get them to combine things that are fine on their own, but together are very problematic,” they say. “I am very concerned that portraits of ourselves, which are already a constant battleground, will suddenly be taken out of our hands.” Even if AI tools include more holistic representations of queer people in the future, the synthetic images could have unintended consequences.

You may also like