Post

It might be a losing battle, but I'm pondering what #LDJam #NewLD could do to sustain a "Compo-like" vibe. Game submission metadata: [ ] NO AI WAS USED to write code [ ] NO AI WAS USED to create visual content [ ] NO AI WAS USED to create audio content [ ] NO AI WAS USED to generate other data 🤔

0 reposts26 likes

I like this! The compo jam I did in the most recent LD was 100% AI free but i know people struggle with the "what about code assist automating the shit I hate doing" convo when it comes to jams and stuff. For me the challenge is the point, but I get it, I guess. All to say: not a losing battle!

It saddens me, but I don't think we can assume people didn't use AI by default. The proliferation of AI in everything means you have to try to NOT use AI, meaning we should assume AI was used UNLESS they specifically say it wasn't. Also I just realized "create" vs "generate" is a debatable word. 🤦

"Generate" as a term would be inclusive of your various content generators (Suno, Stable Diffusion, etc), but *could* be interpreted as excluding AI touch-up brushes, or AI's that insert or filter content (i.e. prompt: make the character glow). "Create" I want to say is all encompassing. 🤔

Something we easily overlook is AI translation. As an example, someone might only be able to participate because they used AI translation. AI translation is nothing new, but today's LLMs are dramatically better than previous generation ML translation models. 🤷

Another easily overlooked issue is base-code/libraries. We allow 3rd party libraries and middleware, but there is an impossible number of dependencies to have to check whether AI was used on them or not. Hypothetically you could bypass an "AI was not used to write this" restriction with this. 🤷

The Brackeys Game Jam made everyone respond to “What code and assets did you not make from scratch during the jam (if any)?” I think it might work to ask people to disclose “How did you use AI while making this game?” and let that effect their ratings however other participants see fit.