Is there certain reason why daz shop doesn't use a category based on ethnicity?
James
Posts: 1,025
Is there certain reason why daz shop doesn't use a category based on ethnicity for chars?
Post edited by Richard Haseltine on
Comments
While there would be some obvious choices, where do you draw the line? There are so many variations and mixes that it would be impractical, if not impossible, to maintain, let alone define.
Moved to the Commons as it is not a Daz Studio topic.
DAZ store search could and should be so much better, but that's one thing they got right... be it on purpose or by accident.
Yep, the problem is nobody agrees on what is a given ethnicity. For example, let's say you have a "hispanic" section. What fits into that category? Having worked with a fair number of Spanish people, Spain is not included (the Spanish think of themselves as another European country nestled by France and Germany, Portugal and Brazil are not included since they were not founded by the Spanish and don't speak spanish, while the Philippines would be included. "Hispanic" is really a linguistic New World term created by linguists and sociologists. Likewise with "Asian", do you include Israel and Turkey? This reminds me of how people think of species as a hard and fast term. A friend of mine (a physicist taking with chemists) described my discipline, biology, as a "squishy science" to a sociologist friend. Math, physics, and chemistry are thought of "hard" sciences while sociology is considered a "soft" science. Biology shares characteristics of both. Lions and tigers are different species and can't interbreed until they do (see liger and tigon). And any progeny are sterile until they aren't (see lliliger). Ethnicity is an even more vague concept.
Re: Categories - Does DAZ do anything other than white female aged 18 to 28 wearing high heels while visiting a seatless toilet at a bus-stop ?
*with a sword.
This is Anime female aged 18 to 28 wearing high heels while visiting a seatless toilet at a bus-stop with a sword-erasure!
They could add an AI search based on an upload of an image that represented whatever you wanted, whether it be a australasian, african, american, european, asian etc. This would avoid any confrontation while also allowing searches based on things like, indeed, bus stops, and toilets, which are important for some of us (though not after I finally downloaded the busstop freebie several years ago, toilets are still very important.)
That's the only way you're going to get that thing to flush.
That's the only way you're going to get that thing to flush.
Ethnicity is a floating point that people can get really touchy about. It's better not to have that as a filter. What would be a good filter to have IMO, would be one for poses and textures sets.
It's important to be specific about what we mean by "ethnicity", because it's a fairly amorphous concept.
That is the problem, unfortunately.... first, people don't necessarily look "American" (i.e. Taylor Swift and Beyonce both look American as they were both born in America) just as Naomi Campbell and Kate Moss both look English. Second, AI is notoriously awful at this sort of thing and tends to label thing based on the programmer's and source material's biases. If you ever look at old Tarzan movies, one strange and funny thing is that the "natives" are dressed in an odd mixture of African and New Guinea clothing and makeup. Now, these two peoples are the two most genetically and culturally distant people in all of humanity but the filmmakers, in their sloppiness, decided that they are basically the same.
To say nothing of the fact that technology can't help but inherit its programmers' biases, with one example being how early facial recognition technology frequently failed to recognize darker-skinned people as human beings.
Hemm... never thought ethnic category could be not simple.
Or maybe DAZ can just add a feature like google lens to find similar characters.
Let's say we throw an image, and it will find similar faces in certain range.
will that be more difficult?
Because I have an idea in my head of a certain character, but to find it, I have to browse the whole shop.
What does an "Asian" look like? Are Indians and Koreans both "Asian"? What about Arabs? Filipinos? Russians? What do "Europeans" look like, and how would we distinguish them from "Americans"? Are "Americans" or "Australasians" specifically the indigenous inhabitants, or the colonial majorities, or both? If the former, how do you tell the difference between an Inuit, a Quechua and a Maori? How you do distinguish any of them from "Asians"? If the latter, how do you distinguish a European settler of New Zealand from a European setter of America from a European? If both, what good is that classification anyway? Dwayne Johnson is an American of Australasian and African descent, so where would a character modeled after him be categorized?
Years ago they had a set that was various morph shapes and that had so much vitriol directed at it. "She doesn't look French!!" "OMG, she doesn't look _______" and they sort of stopped even trying to direct things after that in my opinion.
Seeing from many opinions, now I understand why DAZ has no ethnicity category, due to difficulty to put certain face to a certain ethnic category.
I guess that answer my inital question.
I have a Puerto Rican cousin who is blond with blue eyes.
Yeah. ;)
There is the remains of a body that was found in one of the caves in Cheddar Gorge in the UK dating back about 10000 years - so shortly after southern Britain became habitable after the last ice age. The DNA recovered from that person shows they had red hair, black skin and blue eyes. Hmm. Category twisted, maybe broken.
Regards,
Richard
the general idea of asking for something similar to an image is not bad, for the time being that lens will be the other people on the forum who are often willng to help with what they know
Can we one day acknowledge that AIs do not inherit the PROGRAMMERS biase but that of the fed information base?
Which have ZIL to do with the programmers skill, except the ability to learn from them at all.
Any working AI gets none of a programmers bias. If there is bias in the algorithms, its not really AI.
I maybe new to DAZ, but I am breeding AIs for some 28 years now and consider such statements simply uninformed (at best, which I assume here).
If there is bias or racism in an AI, change the training, not the underlying software and learning mechanisms. If your dog misbehaves you will train the next dog different, and not swap to elefants.
Sorry, had to vent this.
And I've seen Mexicans who look Japanese (of course, seeing as how those indivuals have Japanese ancestery, one has to wonder....)
Somewhat related to the topic, I've also come across more English speaking people in Mexico City than in Washington DC. Seem to be better at driving cars, too--or maybe they are just lucky.
I must gently disagree. I have a friend who is both a noted expert on AI and an AI artist (he creates algorithms and has his computer analyze his photographs to generate new images based on how the AI understands the image). He recently wrote in his substack about AI hallucinations where AI uses a data set set and gradually leaves reality as it builds model based on faulty data that are reiterated to produce even wobblier models as the errors multiply. The fault is the designer who should be curating the information going in. In a real world situation, there is that issue of AI confusing Black people with apes (https://www.nytimes.com/2023/05/22/technology/ai-photo-labels-google-apple.html) and (https://www.theguardian.com/technology/2018/jan/12/google-racism-ban-gorilla-black-people). The issue isn't that the data sets are tainted with racist imagery; the problem is that the programmers didn't consider that the data sets were contaminated with racist imagery. It is the programmer's duty to curate what is fed into their AI models. Unfortunately, programmers are human and flawed and their biases make them blind. There is an old question about Genesis in the bible; "what is the first creation?'. It is the Word (logos); the idea is that you can't describe something unless you can invision it. Our own biases blind us.
Er... there's ample evidence in the store that DAZ can't even tell the difference between a propeller driven plane and a jet... do you really want these same people to be messing with something that can be as sensitive a topic as labeling via race and ethnicity? :)
You can't just define flaws out of AI. AI isn't handed down from the heavens, it's created by people, and those people have biases that they may not even be aware of.
I rendered all my characters and put them in a huge image ordered by skin tone. This way I could find the texture I would like depending on character I wanted to create. If I wanted someone with a more red tone or more yellow tone. If I wanted a darker or lighter character. I can tell you it sounds easy, but it's almost impossible to do. There are so many ways a skin tone can differ. I would say it's almost an impossible job to give characters an ethnicity and get it right. There are way too many unclear cases.
There definitely is programmer involvement. Programmers are responsible for the granularity of the learning, learning what variations to adress and which to ignore, whether something outside the avarages found is an exception, a fluke, a temporary state or perhaps even something normal which was underrepresented in the learning data. At their core, AI programs are still a mere matter of working with statistics and applying functions to generate variations based on those statistics. The statistics may be entered by the person teaching the AI, but the functions and their methods of application are generated by the programmer. Both can be flawed, and inevitable are flawed.
That's actually a clever way, how you did that. And I agree, it's nearly impossible to get right anyway. There are too many ethnicities with similarities, and on top of that, there is no proper way to categorize any mixes of ethnicities. Someone who's just 5% Puerto Rican may, through both dominant and dormant genes, appear more Puerto Rican than someone who's more than half Puerto Rican. There's no proper way to put that into a category. "African" is also too broad a term, as there are distinct differences to find between people from the East, West, North and South of Africa. And I'm not going to touch on that LatinX misnomer with a ten feet pole.