#Hack-AI-ton

1 messages · Page 1 of 1 (latest)

mighty elk
#

(working title, name is not finalized)

Overview: Make Neuro-sama community attempt to create their own AI’s and let Vedal laugh at their futile attempts.

Benefits:

  • More people will appreciate the work vedal does and how difficult it is (or isn’t neurOMEGALUL).
  • General knowledge of AI in community will increase. Stonks 📈
  • Vedal can steal implement some interesting models/architectures for himself
  • Potential for healthy competition in the future (Vedal has a big headstart, so no worries :Clueless:)
  • Slight chance of Vedal hiring an interesting contestant
  • Great content

Directions:

  1. Generic theme. Just a generic, broad, open-ended theme. Make them go wild and create anything that implements AI in some shape or form as long as it fits the theme
  2. Architecture. Traditional hackathon focused on one or a group of AI architectures. Might be based on vedal’s needs for inspiration and/or research. Can result in datasets that he can later feed to Neuro
  3. AI-Centered Game Jam. Just a traditional game jam, but the core mechanic of the game must involve AI, from LLM to computer vision
  4. Enabling Neuro. Modding games so Neuro can play them (or play alongside Vedal). Would require specifications from Vedal, or we can postpone the thing until he drops that “Neuro gaming” system that he was talking about

Part 1. Feel free to add new ideas or give feedback on existing ones. Part 2 about possible issues and things to decide is lower.

#

Potential problems:

  1. Judging. Running models or inputting keys for testing may cause issues for judges. We’ll need to decide on a testing medium before determining the criteria.
  2. Copyright. We need to make sure contestants don’t violate copyright. A tutorial on finding properly licensed datasets and following scraping etiquette (e.g., respecting robots.txt) might be necessary.
  3. Resources. Not everyone has good hardware or access to AI services APIs in their region. Sponsorships with cloud services or such could help us out with that (Mythic might be able to help here).
  4. Controversy. Vedal doesn’t like them and “uncontrolled” (by vedal) AI use in certain fields might force him into unwanted discussions. Clear, thought out rules and guidelines can help manage that bs.

Something to decide later on:

  1. Judges
  2. Judging criteria
  3. Submission platform
  4. Awards
  5. Maximum team size
  6. Deadlines and timeline
stoic temple
#

eh you probably wouldn't get that many submissions, plus they might be too similar

mighty elk
#

Eh, Swarm Big, Problem No. But I see your point

Submission quantity. With different marketing strategies we can reach different results. Like if we just announce it once on Discord there will be some amount of people. If we get Vedal to mention it on stream, there will be more, if he will actively encourage it, mods ping it multiple times on Discord, clippers unite and Neuro asks (gasp) there will be much more.

Quality and diversity (? tell me if I got you wrong). This is a big problem yeah. But I think this community can still surprise us with it's creativity and imagination. Also, diversity heavily relies on the direction that we will take. For example, if we choose to go purely with "Architecture" and Vedal decides that he needs more facial recognition data, then there might be a problem with too much similar submissions and too narrow of a theme for people to be interested in. I think if we carefully choose more of a broad subjects or pick out multiple categories then it can be more diverse. Besides, the goal of this event is gaining knowledge, not the resulting products.

Will I get :essaying: again i wonder

stoic temple
#

tbf the essaying threshold is quite low imo

old rune
steel leaf
#

you may wanna check out the melba toast thread to see how the neuro community's AI endeavours went lmao

quasi river
#

we've already done this. and no one gives a shit about other AIs, don't even try.

steel leaf
#

I wouldnt say nobody gives a shit just that right now theres no possible way to compete

#

in order for someone to compete with Neuro theyd have to find huge collab partners and do something even bigger than what neuro is currently capable of

mighty elk
#

This is not about competing with Neuro, this is a community event for doing something with AI, be it a little machine learning or LLM fine-tunes