Re:Stable Diffusion 1.5
by blackoil on Friday October 21, 2022 @01:29AM
(#33284818)
I am more hopeful. Unlike popcon/napster these models aren't directly impacting existing bottom line of any company/organization. Most of the models are trained on opensource / public datasets, so you won't find any company to sponsor the fight against these models. The cost of these models is an issue right now, but Mr. Moore has always handeled that well.
[ Parent
| on HN ]
Ah., but SD does compete with OpenAI's Dalle2 no?
I am however not sure if that will cause too much trouble though.
Unfortunately, you’re right. These models are beneficial to large corporations, and they do the most harm to the small individual artists who created the content that made them possible in the first place, so it’s unlikely there will be any serious legal challenges.
Re:Stable Diffusion 1.5
by capitalsigma on Friday October 21, 2022 @08:41AM
(#33287381)
I don't think that Moore's law is what's driving down ML compute costs in particular, where there seems to be a lot of innovation going on in terms of hardware architecture and compilers (much of which is proprietary). Even just thinking about memory bandwidth, which historically has scaled much slower than compute: the $/second required to push 10+ TB of training data into some piece of hardware that can do useful work on it isn't going to fall by 100x in a decade.
Moore's law is not driving anything, it's just an observation. The effet might have multiple causes. So Moore's law also applies here