237
OpenAI says it’s “impossible” to create useful AI models without copyrighted material
(arstechnica.com)
this post was submitted on 11 Jan 2024
237 points (100.0% liked)
Technology
37801 readers
333 users here now
A nice place to discuss rumors, happenings, innovations, and challenges in the technology sphere. We also welcome discussions on the intersections of technology and society. If it’s technological news or discussion of technology, it probably belongs here.
Remember the overriding ethos on Beehaw: Be(e) Nice. Each user you encounter here is a person, and should be treated with kindness (even if they’re wrong, or use a Linux distro you don’t like). Personal attacks will not be tolerated.
Subcommunities on Beehaw:
This community's icon was made by Aaron Schneider, under the CC-BY-NC-SA 4.0 license.
founded 2 years ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
It is an open question. As others have pointed out, a human taking inspiration from the work of others is totally fine. My issue is that AI are not human.
A human's production of work is limited. A human can only produce so fast for so long. An AI could theoretically be scaled infinitely and produce indefinitely. I don't want to live in a world where FAANGCORP's OmniAI is responsible for 90% of all art, media, and music because humans can't keep pace with it.
A lot of this can be traced back to the invention of photography, which is a fun point of reference, if one goes to dig up the debate at the time.
In any case, the idea that humans can only produce so fast for so long and somehow that cleans the channel just doesn't track. We are flooded by low quality content enabled by social media. There's seven billion of us two or three billion of those are on social platforms and a whole bunch of the content being shared in channels is created by using corporate tools to make stuff by pointing phones at it. I guarantee that people will still go to museums to look at art regardless of how much cookie cutter AI stuff gets shared.
However, I absolutely wouldn't want a handful of corporations to have the ability to empower their employed artists with tools to run 10x faster than freelance artists. That is a horrifying proposition. Art is art. The difficulty isn't in making the thing technically (say hello, Marcel Duchamp, I bet you thought you had already litgated this). Artists are gonna art, but it's important that nobody has a monopoly on the tools to make art.
"It's too fast" is a really really dumb argument against AI
It's like the classic "objection!" "On what grounds?" "It's devastating to my case!" Scenario.
Throughout history technology has repeatedly been developed that lets people do things faster than the people currently doing it. That's usually the point of technological progress. Of course the people left behind by that will complain, but that alone is no reason to limit the rest of us who would benefit from the advance.
who benefit from automating creative process
Anyone who enjoys creative things, since they now have access to a lot more of it a lot more easily.
is low quality
Then it's not a threat to professional creatives, is it?
because average person always prioritize quality over low cost
How dare people have different priorities than you.
Mass produced garbage is still mass produced garbage. As you point out AIs aren't human and while that removes the limitations of the flesh (including limitations that we might want there - no human ever says oops, I made a child porn), it imposes limitations of the machine. AI output isn't that good at anything practical. It writes garbage code that even if you manage to get it working, the business manager or whoever isn't capable of seeing the flaws in it. The art is devoid of any sort of soul and almost always has glaring flaws that require actual humans to identify and fix.
We are about to be inundated with AI produced garbage, sure, but that only proves the lie that shady internet sites and social media have always been a cesspool of shitty, unreliable content, and connecting with hundreds of thousands of faceless strangers was never a good idea. Hopefully we'll come up with (or go back to) solutions that don't treat the problem as simply one of volume.
It's not right to say that ML output isn't good at practical tasks. It is and it's already in use and has been for ages. The conversation about these is guided by the relatively anecdotal fact that chatbots and image generation got good so this stuff went viral, but ML models are being used for a bunch of practical uses, from speeding up repetitive, time consuming tasks (e.g. cleaning up motion capture, facial modelling or lip animation in games and movies) or specialized tasks (so much science research is using ML tools these days).
Now, a lot of those are done using fully owned datasets, but not all, and the ramifications there are also important. People dramatically overestimate the impact of trash product flooding channels (which is already the case, as you say) and dramatically underestimate the applications of the underlying tech beyond the couple of viral apps they only got access to recently.