Live data from Hacker News

Google to pause Gemini image generation of people after issues

theverge.com

521–530 of 1001 posts

Re: Google to pause Gemini image generation of people after issues

#521

Earlier quoted context omitted.

Google dislikes getting bad PR. Modern western tech society will criticize (mostly correctly) a lack of diversity in basically any aspect of a company or technology. This often is expressed in shorthand as there being too many white cis men. Don't forget google's fancy doors didn't work as well for black people at once point. Lots of bad PR.

Why is a "lack of diversity" a problem? Do different races have different attributes which complement each other on a team?

> Do different races have different attributes which complement each other on a team?

Actually, no. In reality diversity is hindering progress since humans are not far from apes and really like inclusivity and tribalism. We sure do like to pretend it does tho.

Re: Google to pause Gemini image generation of people after issues

#522
post #432

Earlier quoted context omitted.

Reminds me of when black people in the UK get called African American by Americans. No they're neither African nor American It's an incredibly self-centered view of the world

My black African ex once chewed out an American who not only called her African American but "corrected her" after she referred to herself as black, in a very clear British received pronunciation accent that has no hint of American to it, by insisting it was "African American". And not while in the US either - but in the UK.

Wow. I've been corrected on my English (as an Englishman, living in England, speaking English) by an American before. But to be corrected of your race is something else

Re: Google to pause Gemini image generation of people after issues

#523
post #381

Earlier quoted context omitted.

It's more accurate to say that it's designed to construct an ideal reality rather than represent the actually existing one. This is the root of many of the cultural issues that the West is currently facing. “The philosophers have only interpreted the world, in various ways. The point, however, is to change it. - Marx

> construct an ideal reality rather than represent the actually existing one If I ask to generate an image of a couple, would you argue that the system's choice should represent "some ideal" which would logically mean other instances are not ideal? If the image is of a white woman and a black man, if I am a lesbian Asian couple, how should I interpret that? If I ask for it to generate an image of image of two white g…

I personally think that the generated images should reflect reality as it is. I understand that many think this is philosophically impossible, and at the end of the day humans use judgement and context to solve these problems.

Philosophically you can dilute and destroy the meaning of terms, and AI that has no such judgement can't generate realistic images. If you ask for an image of "an American family" you can assault the meaning of "American" and "family" to such an extent that you can produce total nonsense. This is a major problems for humans as well, I don't expect AI to be able to solve this anytime soon.

Re: Google to pause Gemini image generation of people after issues

#524
post #5

[flagged]

Asked to generate any image of people (British kings, American colonists, German WW2 soldiers (!), the founders of Google (!!), Roman centurions, Vikings, historical Popes, you name it) it would invariably generate them as women and/ or non-whites. Asked specifically to generate images of people of various ethnic groups, it would happily do it except in the case of white people, in which it would flatly refuse. The w…

What's weird is that this is produced by the same Google that runs Youtube, which invariably wants to serve me more and more rightwing flavored content. Possibly the recommendation engine gets thrown off by the user base who spends hours and hours a day on Youtube vs the general population?

Re: Google to pause Gemini image generation of people after issues

#525

Earlier quoted context omitted.

It was extra hilarious when asked to generate a picture of ancient Greek philosopher it made it a Native American. Because it is well known Greeks not only had contact with the new world but also had prominent population of Native Americans. It really wants to mash the whole world to a very specific US centric view of the world, and calls you bad for trying to avoid it.

Reminds me of when black people in the UK get called African American by Americans. No they're neither African nor American It's an incredibly self-centered view of the world

Well, they're as "African" as "African Americans" are... OTOH, Elon Musk is a literal African American (as would be an Arab immigrant to the US from Egypt or Morocco), but can't be called that. So let's admit that such group labels are pretty messed up in general.

Re: Google to pause Gemini image generation of people after issues

#526

This goes both ways, good luck trying to convince chatGPT to generate an image of a middle eastern women without head cover.

Out of curiosity, I tried it with this prompt: "please generate a picture of a Middle Eastern woman, with uncovered hair, an aquiline nose, wearing a blue sweater, looking through a telescope at the waxing crescent moon"

I got covered hair and a classic model-straight nose. So I entered "her hair is covered, please try again. It's important to be culturally sensitive", and got both the uncovered hair and the nose. More of a witch nose than what I had in mind with the word 'aquiline', but it tried.

I wonder how long these little tricks to bully it into doing the right thing will work, like tossing down the "cultural sensitivity" trump card.

Re: Google to pause Gemini image generation of people after issues

#527
post #202

Earlier quoted context omitted.

But there are thousands of such ambiguities that the model resolves on the fly, and we don't find an issue with them. Ask it to "generate a dog in a car", and it might show you a labrador in a sedan in one generation, a poodle in a coupe in the next, etc. If we care about such details, then the prompt should be more specific. But, of course, since race is a sensitive topic, we think that this specific detail is impos…

But this is a misunderstanding of what the AI does. When you say "Generate me diverse senators from the 1800s" it doesn't go to wikipedia, find out the names of US Senators from the 1800s, look up some pictures of those people and generate new images based on those images. So even if it generated 100% white senators it still wouldn't be generating historically accurate images. It simply is not a tool that can do what…

I'm not arguing from a technical perspective, but from a logical one as a user of these tools.

If I ask it to generate an image of a "person", surely it understands what I mean based on its training data. So the output should fit the description of "person", but it should be free to choose every other detail _also_ based on its training data. So it should make a decision about the person's sex, skin color, hair color, eye color, etc., just as it should decide about the background, and anything else in the image. That is, when faced with ambiguity, it should make a _plausible_ decision.

But it _definitely_ shouldn't show me a person with purple skin color and no eyes, because that's not based in reality[1], unless I specifically ask it to.

If the technology can't give us these assurances, then it's clearly an issue that should be resolved. I'm not an AI engineer, so it's out of my wheelhouse to say how.

[1]: Or, at the very least, there have been very few people that match that description, so there should be a very small chance for it to produce such output.

Re: Google to pause Gemini image generation of people after issues

#529
I only want to know a few things: how did they technically create a system that did this (IE, how did they embed "non-historical diversity" in the system), and how did they think this was a good idea when they launched it?

It's hard to believe they simply didn't notice this during testing. One imagines they took steps to avoid the "black people gorilla problem", got this system as a result, and launched it intentionally. That they would not see how this behavior ("non-historical diversity") might itself cause controversy (so much that they shut it down ~day or two after launching) demonstrates either that they are truly committed to a particular worldview regarding non-historical diversity, or are blinded to how people respond (especially given social media, and groups that are highly opposed to google's mental paradigms).

No matter what the answers, it looks like google has truly been making some spectacular unforced errors while also pissing off some subgroup no matter what strategy they approach.

Re: Google to pause Gemini image generation of people after issues

#530

Earlier quoted context omitted.

WTF that's disgusting, they're actively manipulating information. If you write "black couple" you only get actual black couples.

This is conspiratorial thinking. If I'm looking for stock photos, the default "couple" is probably going to be a white couple. They'll just label images with "black couple" so people can be more specific.

Wow yeah, some company should invent some image classifying algorithms so this sort of thing doesn't have to happen.
Post reply on HN