ContentDB - now with package translation support

[runs]

Wow!!! my mod Petz rocketed in ContentDB thanks that Spanish Youtuber.

[MCL]

Wow!!! my mod Petz rocketed in ContentDB thanks that Spanish Youtuber.

What about Juanchi Game?

[rubenwardy]

What about Juanchi Game?

Juanchi Game has the "gore" content warning, packages with content warnings are hidden on Android by default. So Juanchi didn't get many downloads that day.

Btw, all these stats are public. You can see Juanchi's stats at https://content.minetest.net/packages/r ... chi/stats/

[runs]

What about Juanchi Game?

Juanchi Game has the "gore" content warning, packages with content warnings are hidden on Android by default. So Juanchi didn't get many downloads that day.

Btw, all these stats are public. You can see Juanchi's stats at https://content.minetest.net/packages/r ... chi/stats/

I have to change the Juanchi LOGO Image, the game is not gore but the LOGO. I mistakenly thought that such a LOGO would attract more people for the sake of morbidity.

[runs]

rubenwardy, may you quit the "Gore" tag of Juanchi Game? I have replaced the LOGO with a cool AI LOGO.

juanchi_logo.jpeg (413.02 KiB) Viewed 4963 times

https://aqualxx.github.io/stable-ui/

prompt: minecraft character, forest, stairs, columns, cinematic, detailed, atmospheric, epic, concept art, Matte painting, background, mist, photo-realistic, concept art, volumetric light, cinematic epic + rule of thirds octane render, 8k, corona render, movie concept art, octane render, cinematic, trending on artstation, movie concept art, cinematic composition , ultra-detailed, realistic , hyper-realistic , volumetric lighting, 8k –ar 2:3 –test –uplight

[MCL]

Are binaries generated using technologies that utilize machine learning allowed on CDB?

I would assume yes, but there has been a deliberate misinformation campaign against binaries like that and many sites have banned them.

[snoopy]

Hi,

@MCL - THX however IMO some ambiguity and I wonder what the meaning behind this may be ...

Are binaries generated using technologies that utilize machine learning allowed on CDB?

I would assume yes, but there has been a deliberate misinformation campaign against binaries like that and many sites have banned them.

Let me ask this:

1. Does Lua 5.1 code already exist which truly & reasonably utilizes machine learning technology?
2. Which apparent deliberate misinformation campaign against binaries like that of the or around the MT community would you be pointing at here?

Any clarification would be highly appreciated.

Have fun with MT.

[Blockhead]

Are binaries generated using technologies that utilize machine learning allowed on CDB?

I would assume yes, but there has been a deliberate misinformation campaign against binaries like that and many sites have banned them.

I'm not sure what you mean by binaries, because Lua mods for Minetest are usually distributed in source form and interpreted as in ordinary "binary executables" for your computer (and "binary" as in "binary file" could vaguely refer to formats like PNG). Anyway it doesn't really matter, because machine-learning generated art assets are already present on ContentDB and I'm sure they'll they will probably continue to be. NathanSalapat in his post-Game Jam 2022 review stream noted how much use of AI-generated textures he made in his game. As for code: Rubenwardy, the administrator of the official ContentDB site, also wrote an article about generating Lua code for Minetest from GPT-3, so it's only fair to assume I think that he doesn't hate the technology and doesn't want to disallow generated code on ContentDB.. unless the code is so bad it doesn't pass the ContentDB reviewers' checks, since these AIs have been known to fill gaps in their knowledge with nonsense all the time :)

Edit: All of the above should be taken with a big grain of salt and only as evidence of the aforementioned peoples' actions. As Rubenwardy has said below, refer to the actual ContentDB policy page which has an entry regarding the topic.

[Corax]

"Not hating" a technology is not strongly coupled to "promoting or resisting" a technology.
Technology per se is neutral, so why would I love or hate it; but I can still find the use of it (or even just some specific use cases) unethical or otherwise detrimental even though I may want to experiment with it to learn about its possibilities and limitations, and/or because I find it generally interesting without considering possible ramifications.

I consider machine learning and automated generation of {code|art|literature|...} in itself interesting, that doesn't mean I have to love it. But when you look over the immediate horizon, there undeniably are legitimate concerns about ethicality, authorship, responsibility, scope and many more. That still does not mean that I have to hate the technology either.

[rubenwardy]

Please see the new copyright guide for our position on AI art: https://content.minetest.net/help/copyr ... used-an-ai

Errrr. This is a legally untested area, we highly recommend that you don't use AI art/code in packages.

For now, we haven't banned AI art/code from ContentDB. Make sure to clearly include it in your package's credit list (include the name of the AI tool used).

Check the tools terms and conditions to see if there are any constraints on use. It looks like AI-generated art and code isn't copyrightable by itself, but the tool's T&Cs may still impose conditions.

AI art/code may regurgitate copyrighted things. Make sure that you don't include the names of any copyrighted materials in your AI prompts, such as names of games or artists.

[MCL]

Are binaries generated using technologies that utilize machine learning allowed on CDB?

I would assume yes, but there has been a deliberate misinformation campaign against binaries like that and many sites have banned them.

I'm not sure what you mean by binaries, because Lua mods for Minetest are usually distributed in source form and interpreted as in ordinary "binary executables" for your computer (and "binary" as in "binary file" could vaguely refer to formats like PNG). Anyway it doesn't really matter, because machine-learning generated art assets are already present on ContentDB and I'm sure they'll they will probably continue to be. NathanSalapat in his post-Game Jam 2022 review stream noted how much use of AI-generated textures he made in his game. As for code: Rubenwardy, the administrator of the official ContentDB site, also wrote an article about generating Lua code for Minetest from GPT-3, so it's only fair to assume I think that he doesn't hate the technology and doesn't want to disallow generated code on ContentDB.. unless the code is so bad it doesn't pass the ContentDB reviewers' checks, since these AIs have been known to fill gaps in their knowledge with nonsense all the time :)

Edit: All of the above should be taken with a big grain of salt and only as evidence of the aforementioned peoples' actions. As Rubenwardy has said below, refer to the actual ContentDB policy page which has an entry regarding the topic.

When I said binaries, I clearly meant images. I didn't say anything about them being executable, so maybe you should learn to read before jumping to conclusions.

I'm not planning on using GPT-3 as it is corporate garbage with a piece-of-shit filter that prevents it from being used to its fullest extent.

GPT-2 is MIT/X and can be ran locally so I might use that in the future, but I don't know yet.

[rubenwardy]

When I said binaries, I clearly meant images. I didn't say anything about them being executable, so maybe you should learn to read before jumping to conclusions.

No, it wasn't clear, "binaries" usually means executable binaries in that context. Don't blame the reader when you communicated poorly. The correct thing to say here would have been "Oh sorry, I meant binary files, like images and sounds."

[debiankaios]

Does this include translations? I do most of the translating for DeepL, the name says it all, Deep Learning, and yes it is an AI. Even the text was translated with https://www.deepl.com. I wouldn't do so well without it, because my English skills are not good enough to translate mods, let alone whole books (the Minetest modding book in German wouldn't be finished without DeepL).

[rubenwardy]

Yeah, using machine translation is fine

[MCL]

For now, we haven't banned AI art/code from ContentDB. Make sure to clearly include it in your package's credit list (include the name of the AI tool used).

Check the tools terms and conditions to see if there are any constraints on use. It looks like AI-generated art and code isn't copyrightable by itself, but the tool's T&Cs may still impose conditions.

SD and many other models don't require you to attribute them; do you still need to credit them nonetheless even if you use a model that's freely or semi-freely licensed?

[Neuromancer]

In Minetest ContentDB if 2 different mods have the same name (for example "caverealms") and another mod depends.txt refers to that mod name (caverealms), which Caverealms mod gets downloaded? For example in this case would "Cave Realms Lite" by Shara or "Caverealms" by HeroOfTheWinds get downloaded? More importantly how do you force ContentDB to choose one over the other?

[rubenwardy]

In Minetest ContentDB if 2 different mods have the same name (for example caverealms) and another mod depends.txt refers to that mod name (caverealms), which Caverealms mod gets downloaded? For example in this case would "Cave Realms Lite" by Shara or "Caverealms" by HeroOfTheWinds get downloaded? More importantly how do you force ContentDB to choose one over the other?

If the user has neither installed, the one that has the highest score is downloaded. In this case, "Cave Realms Lite" by Shara. ContentDB strictly enforces right to a name, so both should work as a dependency.

To force ContentDB to use one, you need to download it first before installing the mod that depends on caverealms

For more information, see the dependency resolution docs or the package scoring docs

[Neuromancer]

In Minetest ContentDB if 2 different mods have the same name (for example caverealms) and another mod depends.txt refers to that mod name (caverealms), which Caverealms mod gets downloaded? For example in this case would "Cave Realms Lite" by Shara or "Caverealms" by HeroOfTheWinds get downloaded? More importantly how do you force ContentDB to choose one over the other?

If the user has neither installed, the one that has the highest score is downloaded. In this case, "Cave Realms Lite" by Shara. ContentDB strictly enforces right to a name, so both should work as a dependency.

To force ContentDB to use one, you need to download it first before installing the mod that depends on caverealms

For more information, see the dependency resolution docs or the package scoring docs

I meant moreso from the position of the mod developer (not the end user) being able to force which dependency mod gets downloaded by the user so that the group of dependencies work better together.

[rubenwardy]

I meant moreso from the position of the mod developer (not the end user) being able to force which dependency mod gets downloaded by the user so that the group of dependencies work better together.

No, there's no way to do this - ContentDB is based on the engine's dependency system, and the engine provides no way to depend on either caverealms mod specifically

[Neuromancer]

I meant moreso from the position of the mod developer (not the end user) being able to force which dependency mod gets downloaded by the user so that the group of dependencies work better together.

No, there's no way to do this - ContentDB is based on the engine's dependency system, and the engine provides no way to depend on either caverealms mod specifically

I thought as much. I think the answer is to get the mod creator to change the name of the mod in the mod.conf file so that it's unique.

[Mantar]

Might be a good idea to allow two names in that case. So like "caverealms_lite" is also "caverealms" for dependency purposes. Like an official name, and a compatibility name.

[Mantar]

Might be a good idea to allow two names in that case. So like "caverealms_lite" is also "caverealms" for dependency purposes. Like an official name, and a compatibility name.

[Neuromancer]

Might be a good idea to allow two names in that case. So like "caverealms_lite" is also "caverealms" for dependency purposes. Like an official name, and a compatibility name.

Yes this is going to become more of an issue over time as more and more forks are created of mods, and names are just duplicated, and then dependencies are built upon them. Another option would be for contentDb to have a unique name field that it manages and the engine uses that first.

[Blockhead]

It is my firm opinion that anyone who provides a mod on ContentDB with the same name as another has to provide compatibility with other mods of the same name, or else they should change their mod's name. ContentDB only has a policy of allowing forks and reimplementations that is meant to ensure this. So to me the only real source of problems is people who do a bad job of maintaining compatibility when forking mod or refuse to acknowledge that their fork should be renamed. And please people, contribute PRs back to mainstream unless you have irreconcilable differences!

[Mantar]

That doesn't solve the problem when a fork of "foo-mod" provides functionality that the original mod lacks, and another mod may depend upon the extended functionality. The second mod needs to be able to specify which exact fork of foo-mod it wants, like "depends = foo-mod++", while another mod may just want any compatible version of foo-mod, so it would "depends = foo-mod" and the ++ version should be able to satisfy that as well.
So if foo-mod++ could add a line that says something akin to "compatible_with = foo_mod" that'd be useful.