Fuck AI

Yes it's sad that legitimate aid groups need to do this but that's the reality that capitalism and "AI" have imposed on us.

That's a fine general principle but some places around the world have very little, and no one locally to give rich-westerner levels of resources.

Giving not-locally can make way more of a difference in more people's lives.

Once you start making fake images of social evils, even with the best of intentions, you are compromised. After all, why not look at your analytics and juice the prompts to make the campaign work better? And soon the images diverge from the real problems you’re supposedly trying to solve and into the realm of lurid fantasy, presenting an exaggerated and simplified vision of the problem that pushes emotional buttons harder than realistic reportage would. From then on, the die is cast. If things improve, surely it’s better to keep making images of them being worse than ever, as after all, it’s good to have cash reserves just in case. And given how good a job you’ve done with that, surely you deserve a bonus. The computer-generated sad-eyed shantytown urchins wouldn’t want it any other way.

I think you make good points, and I am also somewhat conflicted on this.

Overall, I'm probably against this, due to my experience as a disabled person. I (and many other disabled people) hate the way that we are often depicted in media aimed at the wider world . It's often gross, whether it's the pitying "poor disabled people" stuff which is analogous to what you see in "poverty porn", or the superficially positive "inspiration porn" — they're both equally dehumanising. Less exploitation of our lived experience would definitely be great.

However, depictions of disabled people (whether in advertising or other charitable publication material, or the wider media) is pretty inauthentic to our actual lived experience, and this drives the stigma that disabled people face. AI models will have been trained on this same media ecosystem, so are likely to perpetuate harmful stereotypes by depicting things that are insensitive or inaccurate about disabilities. Even before AI, the self perpetuating cycle of bad disability representation in media was already a problem that caused real world harm to disabled people, so I'm not optimistic about the AI.

Overall, I think that we would be better served in figuring out how to depict things like poverty and disability without it being gross, exploitative and dehumanising. It's not an easy task, but it's a worthwhile one — not least of all because doing it properly would require involving people of the marginalised group you're advocating for (I have overwhelmingly found that charitable organisations that have disabled people working within them have significantly more nuanced and sensitive representations of disabled people in their advertising materials, for example).

More generally, I also worry that AI generated images could desensitize people to the things depicted. When I'm browsing the web and I see writing or images that seem to be AI generated, I often lose any interest that I might've had and just skim over them. It's an almost reflexive response, and I feel bad knowing that there are inevitably going to be false positives, in which I disregard something that wasn't actually made by AI. If the practice described in the OP becomes commonplace, I fear that it could cause people to tune out real images of human suffering. Perhaps some people might even use that as an excuse not to care; empathy is often uncomfortable, and the ignorance of "I don't need to feel emotionally affected by this, because it's not even real" could be an easy out when viewing even real images.

I normally beat the anti-AI drum anyway, but I don’t consider this an ethical use of AI. The way to avoid exploiting people by making poverty porn is to just not make the poverty porn to begin with, not to use image generators trained on all the poverty porn that’s been made in the past. Even without a specific victim, the underlying problem of stereotyped portrayal while removing the voices of the people actually affected remains unchanged.