AI has the same racial bias as old film formulas did
Earlier film formulas were engineered with light skin tones in mind. Any image of dark-skinned individuals in light settings results in lost facial detail. In a sense, film was racist. I've heard people dismiss this concept without actually considering it, but that's not the discussion I want to have now.
AI has the same bias.
In Japan, AI-generated imagery is quite commonplace and even somewhat popular. Many restaurants have resorted to garish menus and depictions of their own food that don't resemble what they actually serve. This is another topic I don't feel like getting into right now.
What's really irking me is all the AI-generated thumbnail images flooding the MeetUp app. In Japan, language and cultural exchange events are the most prominent international social events. In a city like Osaka, you can find numerous events daily, from cafe chats to hiking. Up until a couple of years ago, cover images for these events were either photos of past events or a flyer cooked up on Canva, good enough to get a sense of what you were joining.
But now most 'images' depict a group of individuals creepily smiling at each other, or all reacting to a joke that no one seems to have said, or even jogging while holding paper cups of coffee. Yes, all slop. But what annoys me here is the races depicted in every image. It's usually a mix of very generic Japanese people (all long, straight black hair), and seemingly White Americans ripped out of a Target flyer from 20 years ago. But this isn't representative of who actually attends the event. Half the participants are Japanese (usually much older, but that's another issue), but the foreign contingent is usually from all over the map. Besides Westerners, there are several participants from Mexico, Peru, India, Nepal, Taiwan, China, Poland, and... You get the picture. Everywhere!
But that's not what AI-generated images depict. If the prompt is something like "Give me an image of a Japanese/English language exchange meetup", and that's the image produced, the AI has a racial bias. What's more upsetting, though (maybe?), is that if the user doesn't reply with "No, make it more diverse. It's an international group.", then the user also has a bias. Even more, now the AI, which everyone has put too much faith in, strengthens this bias.
Yeah, I can point it out to the organizer, which isn't easy. They can claim ignorance, which I'm willing to give the benefit of the doubt, to an extent... Or people can say I'm being too sensitive; that AI can't be racist, because it's not a person. At that point, I'm done with the conversation, sadly.