RE: Instead of Using AI for contenting, why not do this?
You are viewing a single comment's thread:
however if you do train a own model in local, spend a lot of time in refining, making loras and such then some monetization should be allowed
I don't really agree with this, it's the same as a dev setting up some kind of system for automated posts, lotteries/raffles, etc, and then expecting to get daily votes and rewards for a thing they spent some time on once/for a certain period. It defeats the consistent effort required by people putting in time, work and effort into creating posts that keep bringing users to consume them.
Sure one could say as long as posts are consumed it doesn't matter who/what created them, but just because you're now able to do this with less effort doesn't mean you should be valued/rewarded as much as you used to or compared to those who still spend a lot of time in creating posts of the same caliber/consumption in a purely non-AI way. It's proof of brain after all and it's important to keep author rewards focused on that, else we'll just end up having 10-100 accounts all taught by @tarazkp for instance to write like him and at that point we're rewarding one user 100 folds and less rewards are going towards other unique users which hurts decentralization through bad distribution.
Oh not spamming accounts, just yours... I say this because I made a lora myself for stable diffusion for images and it took me days to get a decent result, a lot of trial and error, I mean it was a lot of work... I guess one could burn part of the rewards?
In the end it's not very different from some posts I see now where a user posts a generated AI image and in the post he explains the prompt and the technology used, but it's the image the main content of such a post
Ah okay you were referring to images, thought you meant creating an AI agent that'd post "like you". Transparency would go a long way, if curators then still overreward you according to others they can then use their downvotes to adjust the rewards a bit and the fault is not on you because you've always been transparent about images or text's being AI generated.
For the text, I suppose one could train a model or lora, make it write a short post, and in the post write the prompt and some technical analysis + personal thoughts on the post generated so at least 50% of content is your own words, that could be a compromise I guess like with images ?
On a side note, I know which user the OP is talking about and on this the downvotes are correct, daily spam of obvious AI (not even a custom model) heavily over rewarded by autovotes, that's an other story... And I'm glad you helped to "fix" rewards there, 60-75 hive per post no way
I only just got to understand that you meant AI images. For sure, a lot of users do this. I think using in full content creation is where it becomes an issue.
Yeah I guess if you keep at least half of post your own words should be fine, regardless of Ai generated object, one could even make ai music and elaborate in the post the steps, the meaning and so on