Comment by feoren
1 year ago
> We live in times were non-problems are turned into problems.
This is exactly what everyone who benefits from the status quo always says.
> Most software engineers and CEOs are white and male
55% of Software Engineers are white; 80% are male.[1] So somewhere around 44% of software engineers are white and male. That's not "most". You think it's perfectly fine if 100% of generated images for "Software Engineer" are white males, when ~56% are not in real life? What exactly is your definition of "truth" here?
An unregulated generative model trained on the entire Internet is not going to regurgitate facts, it's going to regurgitate existing beliefs, which is damaging to people who those existing beliefs harm, and to the people who are trying to change those beliefs to actually align better with facts. It is an amplifier of pre-existing perceptions and prejudices; facts have nothing to do with it, except for when they serendipitously line up with common belief. But common beliefs often don't align with the facts -- yes, even yours, as we discovered when you spouted off that "most software engineers are white male" misinformation as if it was some unarguable fact.
>55% of Software Engineers are white; 80% are male.[1] So somewhere around 44% of software engineers are white and male. That's not "most".
Actually, white women are less likely than women of other races to pursue engineering. So there could be closer to 50% white men. Obviously this is in the US. In China, 99.9% of software engineers would be Han Chinese lol. Would it be wrong to show them a group of Chinese engineers? How about showing them 100% non-Chinese when they explicitly ask for Chinese? That's how messed up Gemini is.
Anyway, this is all a stupid argument. Talking about numbers like that in a field as diverse as software engineering is a bad idea, because it has no bearing on the problem. Let the AI generate what it wants to by default, and let people fine-tune to get other ethnicities in there if they want to. If I ask for 5 people with one white, one asian, one black, one Mexican, and one albino, the AI should be able to do that. Focus on correctness and leave judgement to the people consuming the output. I think proportions are only a problem with Gemini because it produces 0% images of white people, even in contexts that demand at least some white presence to not be absurd.
I expect Gemini to still be biased against white people after it's fixed. It will just be more subtle.