Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

If there's a reason Google doesn't want to talk about, it's that the biggest legal risk of AI image generation is that it can be used for revenge porn and CSAM.

I don't think it's problematic to be worried about that!



The biggest legal risk, empirically, is massive copyright lawsuits. The AI art models are rapidly eliminating most artist/photography jobs.

Revenge porn and CSAM are trivially solved by community moderation methods like Midjourney, simply force every generation to be tied to a paid account, and made in the public eye (posted on discord channels).

So in reality, its not Google's risk conservatism, its Google's utter lack of imagination and creativity, and the lack of organisational incentive for anything other than ad money.


> The AI art models are rapidly eliminating most artist/photography jobs.

Automation is associated with increased employment.

https://en.wikipedia.org/wiki/Jevons_paradox

> Revenge porn and CSAM are trivially solved by community moderation methods like Midjourney

Please don't go around saying things like entire safety departments being trivially solved by anything. Community moderation isn't a legal solution to CSAM, it's arguable the community users are committing crimes by looking at it. (Extra difficult because this depends on the country.) And they're certainly not getting healthcare benefits for it, so they could probably sue you.

This is the reason they're on Discord though, it's because Discord handles the legal compliance for these things.


You realize stable diffusion has been out for 6 months and is more powerful than midjourney?


My god, sorry for the dumb post, but SD is only months old? It already feels like it has been around forever. This space is insane.


Was, when it came out, since then MJ have released 3 new models (plus variants) and most users assess the MJv5 model as more powerful. Of course new SD models are also on the horizon…


> Of course new SD models are also on the horizon…

SDXL is available at https://beta.dreamstudio.ai/ though they say they're going to release more variants.

I think ControlNet is a lot more interesting than just "better tuned models"; it means there's no line between creating something yourself and asking an AI to do it anymore.


Maybe in some technical sense but in practice the majority of quality (subjective) work I see shared on Twitter is Midjourney backed. Which is crazy to me because I don’t think I could design a more frustrating interface than Discord.


Midjourney gives you good images with near zero work. SD makes you work for the output but gives you FAR greater control but also needing to be technically skilled, and have pricey hardware. Given this, it is easy to understand why the majority of what you see is done in midjourney.


They could just filter it like Midjourney or DALL-E.


How do they know they can "just" do something that came out a year after they invented Imagen?

Especially when getting it wrong once might mean Europe makes AI illegal. Making things illegal is their favorite hobby.


Midjourney and DALL-E were both announced and in closed beta a month before Imagen was announced, with filters already in place.




Consider applying for YC's Winter 2026 batch! Applications are open till Nov 10

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: