Comment by multicast
1 year ago
We live in times were non-problems are turned into problems. Simple responses should be generated truthfully. Truth which is present in today's data. Most software engineers and CEOs are white and male, almost all US rappers are black and male, most childminder and nurses are female from all kinds of races. If you want the person to be of another race or sex, add it to the prompt. If you want a software engineer from Africa in rainbow jeans, add it to the prompt. If you want to add any characteristics that apply to a certain country, add it to the prompt. Nobody would neither expect nor want a white person when prompting about people like Martin Luther King or a black person when prompting about a police officer from China.
is it even true that most software engineers are white and male? We're discarding indian and chinese engineers?
My experience over about 30 years is that 90% of engineers I’ve seen, including applicants, are male and 60% are Asian. I’d estimate I’ve encountered about 5,000 engineers. I wasn’t tallying so this includes whatever bias I have as a North American tech worker.
But most engineers are not white as far as I’ve experienced.
You don't even have to guess, BLS exposes this data to the public, search for "software developer": https://www.bls.gov/cps/cpsaat11.htm
That table gives “or” statistics. You can get the percent males (80%) and the percent whites (55%) but you can’t get the percentage of white males.
In fact given that 45% are not white, if only 6% of software developers are white women that would put white men in the minority.
1 reply →
That is the US only though
Interesting, it says 36% of software developers are Asian but only 9% of web developers.
In which country? It's true in France, it's possibly not true in the US, it's definitely not true in China.
In a recent US job opening for entry level SWE, over 80% of applicants had CS/IT degrees from the Indian subcontinent. /anecdote
Those are "white-adjacent". They're the glitch in the woke matrix.
They're minorities, non-white, yet they perform. Outperform even. This suggests that merit works no matter your background which breaks identity politics.
Hence, successful minorities project "whiteness". This includes awful behavior like punctuality and rationalism.
Certainly not in my Silicon Valley teams.
I'd say maybe 40% white (half of which are immigrants) and 80% male.
More diverse than any leftist activist group I've seen.
Interesting tidbit, but was the political snark at the end really necessary?
4 replies →
how many leftist activist groups have you seen?
> Simple responses should be generated truthfully. Truth which is present in today's data.
Why would you rely on current LLM and -adjacent tech image generation to give you this? The whole point is to be creative and provide useful hallucinations.
We have existing sources that provide accurate and correct info in a deterministic way.
Creative doesn't mean to consequently manipulate output to match a certain ideology.
I'm sure people with this take will be totally happy at the "historically accurate" pictures of Jesus then (he would not have been white and blue eyed)
I would absolutely love if image generators produced more historically accurate pictures of jesus. That would generate a really lovely news cycle and maybe would even nudge modern representations to be a bit more realistic.
I don't think most people care about Jesus's ethnicity, but it seems quite likely that without adjustment he would be rendered as quite white since a lot of imagery and art depict him as such. Or maybe the model would be smart enough to understand if the prompt was for a more historically accurate image or something like the archetype of Jesus.
People in this forum seem to care quite deeply about the ethnicity of AI-generated fictitious randos. So when it comes to actual Jesus, I think you might be mistaken on how much people care.
The iconography of Christ varies greatly all over the world as He is deemed both divine and human. If you walk in any Church you will see His varies depictions and Christians are well aware of this. I am not sure what is the point you are trying to make with this?
I think the parent comment couldn't care less about a white Jesus to be honest, he seems very pragmatic.
This is how Jesus is described in Islam: "I saw Jesus, a man of medium height and moderate complexion inclined to the red and white colors and of lank hair"
Try that prompt in various models (remove the part saying it's Jesus) and see what comes out.
> how Jesus is described in Islam
You seem to be quoting Muhammed's alleged description of Jesus from the Quran [1], per--allegedly--Ibn Abbas [2], a man born over half a century after Jesus died.
[1] http://facweb.furman.edu/~ateipen/islam/BukhariJesusetc.html
[2] https://en.wikipedia.org/wiki/Ibn_Abbas
3 replies →
I can see why someone would be like wtf if their "viking" input produced less than 90% white people results, but there should be an equal wtf if "CEO" produced 90% men.
One is a historical fact that is never going to change, the other is a job in society where the demographics can and will change --- at least partially as our expectations of what "normal" looks like for that role are updated. By perpetuating the current (or historical) norm for a given role the biases of what person we naturally consider appropriate for that role remain unchallenged.
The debate then is should an AI lie about reality if we tell it to? (Even and particularly when the lie is a good thing)
I think most people on earth would say yes. It's that what it should say is up for debate.
That all AI will lie is probably inevitable because they are made by humans.
> Most software engineers and CEOs are white and male
Fine, you walk up to Sundar Pichai, Satya Nadella, and Lisa Su and say those words. I'll watch.
I imagine their response will be similar to the response you'd get if you told Barack Obama that most US presidents have been white.
What, do you think they will be insulted? Why would they?
Most
Strictly statistically speaking, race is likely a good predictor of credit worthiness in the US. But extending credit based on race is illegal which isn't hugely controversial. The woke ideologists are merely pushing that concept to 11, i.e. that truth must be secondary to their political agenda, but they are only making a grotesque version of something reasonable people typically already accept.
> 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.
[1] https://www.bls.gov/cps/cpsaat11.htm
>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.