Hacker News new | past | comments | ask | show | jobs | submit login

[flagged]



[flagged]


It's a tool. Some people will do something interesting with it. Most people won't.


True. For most, it will be a tool to generate ugly pictures.


True, diffusion models do not cure Sturgeon's Law.

OTOH, there is nothing special about Sturgeon's Law applying to them as well as everything else.


This could be a fact if all you have in mind is kitsch, sure.


I mean, this defeats the point of democratizing art. What's the point of driving the cost of an image to zero and replacing a craft with automated art of that level? Where exactly is the progress here? Are we expected to be impressed by speed or quality? Where is the value? The level of kitsch here is overwhelming. Should we applaud for more kitsch in the world?

"Was einer möchte und nicht kann, wird Kitsch." —Jan Tschichold


> Should we applaud for more kitsch in the world?

Of course. It gets nauseating fast and makes worthwhile messages stand out more, causing the balance swinging to the other side. You're creating something for someone, what's the point if they don't value your carefully crafted message and want kitsch instead?

> Are we expected to be impressed by speed or quality?

Realtime generation in particular makes 3D rendering in the viewport and live painting possible. [1] This kind of tooling makes a lot of difference. What you make with it, either kitsch or something worthwhile, is up to you.

[1] https://www.youtube.com/watch?v=AF2VyqSApjA


> Where exactly is the progress here?

The progress here is that driving the inference phase of generating AI imagery down to a much smaller time/cost stops it from holding back the rest of the creative process using that toolset.




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

Search: