🔖 Tags!
Filter
Show Official Posts OnlyReplies
This is a pretty small update, and something that only effects the backend (for now!).
One of the major systems of organization that most sites have is tags. We've got "tags" on some areas of the site, including items, however those tags are not what I'd call complete, or even really tags. For example, items can have tags like "cash shop", but that doesn't actually put that item into the cash shop. We add it separately, and the tag is just to note to the staff that it can be found there.
I've now implemented a complete tagging system, that can be added to anything on the site, and grabbed very easily. We're going to be using this extensively, and the tags can be shared across "objects". For example, an Item and a Shared Outfit can share the tag "morostide". And, on a search page, we can show all things on the site tagged in a certain way.
We'll also be using tags on items to determine where they show up on the site, and make those searchable in a new and improved site-wide item search.
These kind of behind the scenes improvements are making it much easier to rewrite various features on the site, and hopefully soon will give you improved ways to tag your own content (shared avatars, forum posts, custom wearables, etc) and find it. We'll also be adding public tags as a searchable parameter for things like the vaults, or the wardrobe.
This is one place that I think generative AI might be useful, feeding in an item and it's description, image and uses and getting back a set of pre defined tags. I'm going to be testing with this, I'll let you know how it goes!
💖 ✨ 🤗
For clarity's sake, would you mind explaining the last paragraph a little better? Maybe explaining what type of generative AI you'd be looking at? I'm guessing this would be an outsourced AI since that's a lot for Subeta to produce on their own, so would you be disclosing the company you're going with in order for users to be able to look into?
I know Subeta already has a 'No AI art' stance, so I'm just trying to understand what type of AI you'd be using / etc. Thank you.
A functioning tag system sounds useful in many ways! Given the vast amount of things that can be tagged, maybe some sort of filtering could be applied? The big disadvantage of tags is that usually you can search for one tag at a time only. For example, many users searching for "morostide" might not only bring the servers down but will also bring up a not very helpful mass of results. So a filtering system by item type/usage/location etc. and/or several tags would make it even more useful.
is there a way we could get alerts for posts/updates in this forum, big or small? It's easy to miss when a new post has been made.
For clarity&;s sake, would you mind explaining the last paragraph a little better?
Maybe explaining what type of generative AI you&;d be looking at?
I&;m guessing this would be an outsourced AI since that&;s a lot for Subeta to produce on their own, so would you be disclosing the company you&;re going with in order for users to be able to look into?</p>
<p>I know Subeta already has a &;No AI art&; stance, so I&;m just trying to understand what type of AI you&;d be using / etc.
Thank you.
It's exactly what I said, we would feed this AI (ChatGPT as an example) an item image, name, and description and ask it from a pre-generated list of tags to say what tags it would apply to that item. We have tens of thousands of items, so being able to get a base set from something like that (what is actually called Machine Learning, but everything is AI now 🙄) by giving it out existing information would be very valuable for speed. It's not "generating" anything, it's using information it's been given to make suggestions across a large data set.
💖 ✨ 🤗
That answers part of my question, thank you! Any insight on the second part of it though? Specifically this part;
Honestly, the reason I am asking this information is because I'd be curious what other projects the company had, and whether the information you're feeding this program could possibly be scraped for other, potentially less-ethical projects of theirs. Additionally, artwork is not the only thing that is scraped off the internet without consent, of which I'm sure you already know. I think it's still a concern though. Again, thank you for considering my questions.
That answers part of my question, thank you!
Any insight on the second part of it though? Specifically this part;</p>
<p>
I&;m guessing this would be an outsourced AI since that&;s a lot for Subeta to produce on their own, so would you be disclosing the company you&;re going with in order for users to be able to look into?
Honestly, the reason I am asking this information is because I'd be curious what other projects the company had, and whether the information you're feeding this program could possibly be scraped for other, potentially less-ethical projects of theirs. Additionally, artwork is not the only thing that is scraped off the internet without consent, of which I'm sure you already know. I think it's still a concern though. Again, thank you for considering my questions.
Yup - we haven't made any decisions yet, but in the enterprise feature of almost all current AIs you can feed it information and not have it stored in it's model, or able to be referred to in the future.
I'm also a fan of tabby for my own personal uses, which is an on-device AI that uses a model of your choosing (one of the open source models) that does not send anything out to the broader internet to be stored in a model.
I'll be very upfront if we're using any Subeta data to train a model that is in use anywhere besides us :). Thanks for the questions!
💖 ✨ 🤗