71–80 of 88 posts
This post was originally made in 2022. In early 2023, I spent a lot of effort training a model with a new approach that made the generation quality extremely good... Then the ChatGPT API came out and made all my effort obsolete. In one hour, I was able to create a Notebook ( https://colab.research.google.com/github/minimaxir/chatgpt_a... ) that was able to create mechanically valid and relatively balanced cards given…
This post was originally made in 2022. In early 2023, I spent a lot of effort training a model with a new approach that made the generation quality extremely good... Then the ChatGPT API came out and made all my effort obsolete. In one hour, I was able to create a Notebook ( https://colab.research.google.com/github/minimaxir/chatgpt_a... ) that was able to create mechanically valid and relatively balanced cards given…
Yes. Very scary. This is a very good example of the speed of AI advancement. You used AI in 2022 to make a cool thing, then by 2023 GPT could do it better faster. Everyone is saying the AI craze is hype and already dying away. Meanwhile I can just talk to GPT4 conversationally and it will do a project for me in minutes that would have taken days. It is not hype, and change is still rapidly occuring.
I've been working with generating AI Magic cards since RoboRosewater was popular.
I’m stunned WOTC hasn’t sued them over this yet. I undertook a related project in ‘21 [0] and had two different lawyers tell me Wizards are notoriously litigious and that they’ve even come after creators of fan art in the past. [0] https://hardwork.party/#/entropy/ … recursively-generated MTG-style cards generated by GPT-3 (not 3.5t) and VQGAN+CLIP. The recursion brings out the crazy! I ended up scrubbing for every n…
I’m stunned WOTC hasn’t sued them over this yet. I undertook a related project in ‘21 [0] and had two different lawyers tell me Wizards are notoriously litigious and that they’ve even come after creators of fan art in the past. [0] https://hardwork.party/#/entropy/ … recursively-generated MTG-style cards generated by GPT-3 (not 3.5t) and VQGAN+CLIP. The recursion brings out the crazy! I ended up scrubbing for every n…
Intellectually, and as a creator, I love AI image generation. As a casual user of the internet, I'm growing more and more annoyed with it, simply because of all the times I lean in to examine an image that my visual cortex is stumbling on some part of, only to realize after a couple seconds, "oh, it's AI", and sit back up straight.
Earlier quoted context omitted.
Humans aren't exactly great at this either; cards needing to be banned has only gotten more frequent over time, although it looks like this year has been the first one in a while not to require a lot[1] [1]: https://mtg.fandom.com/wiki/Banned_and_restricted_cards/Time...
There has been a paper on using genetic algorithms for deckbuilding in magic[0] I'm working on at least partially reproducing and improving upon the paper but the reality is that MTG as a game is really hard to implement (and a shitload of work) so i'm building my own MTG inspired game to try and balance it using genetic deck building. [0] https://ntnuopen.ntnu.no/ntnu-xmlui/bitstream/handle/11250/2...
edit: I mistakenly thought you authored the paper, but if you do make progress on your game definitely update us
I bet WOTC would find this very useful!
I am extremely confident WoTC is aware of this and is working on and/or using similar
Earlier quoted context omitted.
This seems like a question of linguistic semantics, but I'd argue that groups of humans are only good at this at the scale where its not so much a skill of the individual humans but a property of the larger system itself. Millions of humans individually trying to figure out how to exploit some new cards aren't going to be anywhere close to as good as millions of people playing games against each other to try things o…
While this is true to some extent, there are absolutely known "hackers" within the community that are particularly good at finding these sorts of interactions. The millions will shake out the rest though for sure.
Earlier quoted context omitted.
Midjourney is a limited set of models and (to me at least) much easier to spot than Stable Diffusion output, with it's giant array of custom models, LoRA's, textual embeddings and tricks like HiRes fix and upscale script.
I used SD1.5 when it came out and left it alone until a few days ago. The improvements are incredible and hard to articulate. If anyone put SD down for a good period of time, they should really give it another go. Use the full suite of options and make sure you are using a model fitted for your purpose.
https://images.squarespace-cdn.com/content/v1/6213c340453c3f...
Earlier quoted context omitted.
There has been a paper on using genetic algorithms for deckbuilding in magic[0] I'm working on at least partially reproducing and improving upon the paper but the reality is that MTG as a game is really hard to implement (and a shitload of work) so i'm building my own MTG inspired game to try and balance it using genetic deck building. [0] https://ntnuopen.ntnu.no/ntnu-xmlui/bitstream/handle/11250/2...
Interesting! Is there a way I can keep up with this? edit: I mistakenly thought you authored the paper, but if you do make progress on your game definitely update us
Can a MCTS with enough compute time beat a human consistently enough? Can MCTS reliably exploit super efficient combos? How can I optimize MCTS Strategies for stronger and stronger players? Can you use Machine learning to optimize further? How do you encode a TCG for a neural net? How do I deal with a potentially infinite play field (MTG can generate token cards). Can genetic deck building find the best decks? If not, are there other options? ...
My ultimate goal here is to create a better Magic. Fixing all the Gameplay issues from having grown over decades, creating a much more streamlined ruleset, a much broader stable meta, etc. etc. But yeah, right now i'm in the "thinking a lot about this" phase.