#Question *from* OAI - Scaling this community

26 messages · Page 1 of 1 (latest)

frozen holly
#

various schemes all launched by the government of Gujarat under the leadership of Chief Minister Narendra Modi in February 2002.

frozen holly
#

Narendra Modi

untold jay
#

o

low flower
#

Can i start asking?

tribal turret
#

hello

lavish compass
flint bluff
#

do i need to log in openai when running openai in discord?

mossy bramble
#

what is housing market in Yilan ludong area

pulsar tusk
#

definition of external functional analysis

jovial meadow
dense yew
#

Hi everyone! Lately we’ve been adding a lot more users at a faster pace.

This means that we expect that a lot more users will join the Discord as well; I have yet to send out the link to the latest large batches of folks we added.

Some questions for you all that would help us figure out how we scale this community:

  1. What do you like about being in this Discord?

  2. Any tips for our team to ensure that that doesn’t change as we scale the Discord community as well?

  3. What are some great large-scale Discord servers we can learn from, and how do they manage them?

  4. More tactical question: we’ve been using a hacky, manually intensive way to verify that the user is in DALL-E. What’s the easiest way to auto-verify users (have you been in other servers where they take your email and a bot verifies you)?

(Thanks for your patience as most of us started from not using Discord at all! 💕)

languid dust
#

  1. the people are really friendly and collaborative

  2. just making sure people that are "unsavory" are dealt with

3.a good moderation team is great and a great way is to incorporate bots into the sever for general management

  1. as rundy said a verification bot should be quite is to find if not make as long as there's someone able to add a list of all the emails then it should all be fine.
eager arrow
#

About question 4, it should be definitely possible to create a discord bot which verifies you automatically, as Discord's bot API is extremely handy and there's a lot of modules which help with using it, many different servers use bot-based verification methods for many different purposes and they work very smoothly.

minor shell
#

  1. i love being surrounded by those who are creating like me, learning about people and what they do in #introductions, and being connected to the OAI team who are able to help with questions pertaining to the site if needed (natalie helped me before with something actually shy)

  2. right now it is quite calm, but eventually chats might require moderating on a level that employees will not be able to zero in on as much. moderators in the server would ensure people can have basic questions answered (credibly) if no one is available to answer them, and just the general server serenity which i appreciate a lot

  3. as abtrax noted, bots are a great way to manage a scale of users that humans would not easily be able to. there's a lot of options out there for bots, my favourite being carlbot turtlemelon

  4. to backtrack off of rundy, having them input an email which a bot would approve/give the verified-user role. a button asking for human help afterwards if not accepted would be a good idea too!

kindred urchin
#
  1. Cozy. Folks are kind, happy to help and open to collaboration. No egos on display. 🙂
#
  1. The server will change with more people - its inevitable. Things that will help set the tone for the community:
  • a charter on what goals OpenAI has for DALLE2 as it relates to users in the preview/artist access program
  • guidelines on what is acceptable discourse (community rules/behaviors)
  • guidelines on DALLE2 generations - rules on content, sharing, and eventually license/usage
    o To that end, having a FAQ with info would help
#

To help offset that change, you'll need engagement by individuals that are trustworthy and that are empowered to help facilitate. In other words, mods. 😄

#
  1. FULL TRANSPARENCY STATEMENT - I am a mod on the NightCafe discord; we have currently 15k members. I am happy to lean in and offer advice on discord moderation. In general, it's about moderating content, not people. In the event of a situation use timeouts on folks who are not acting in accordance to the rules to clean up and discuss with involved parties. I'm big on restorative action - and I think it could work here, as this is about building and maintaining a community for folks who have an incentive to continue having access to the service.

You will want bots to assist mods as other have recommended. You'll ideally have a moderation team that has near 24x7 availability through informal schedules/cycles.

#

Some immediate advice - if interest grows in DALLE2 and the number of invitations going out remains small then it is inevitable fan groups will spring up. There is already another discord like that for DALLE2. Midjourney has a gigantic unofficial Facebook group. These will in time develop their own groupthink, not act in your behalf, speak freely amongst themselves and may become sources of misinformation and confusion.

#

And when those members do gain access they may bring that attitude into the discord.

#

And for those in the other discord - I'm not saying that it is happening there, but it can. The unofficial Midjourney Facebook group is quite toxic at times.

#

Also, the way DALLE2 implements any plans for paid subscriptions/access will also change the tone of the discord. Be up front about changes. Move your updates to the top of the list. Post status updates about downtime (planned or otherwise).

swift warren
# dense yew Hi everyone! Lately we’ve been adding a lot more users at a faster pace. This m...

I can't speak as to 1 specifically since I just joined, but I moderate the official Minecraft speedrunning server with maybe 25K people or so.

I'll echo the other suggestions of using bots like Dyno, Carl-Bot, or Zeppelin, and by making sure that moderators are around 24/7. If I had to guess right now, this Discord's members are clustered in America and Europe, but that's likely to change as access becomes wider. What we do in the Minecraft server is to have clear delineations of channels between general, casual chatter and specific topics, and enforce heavily (1-day mutes) people who end up in the wrong channels. We also have a lot more measures for things like anti-spam than here would need.

I think having a good onboarding process for new members unfamiliar with DALLE (e.g. the doc that Lucas and Rundy have made) would be great, especially with some Q and A (e.g. is it acceptable to inpaint one's face, and are there consequences if this is flagged by the filter?). This will also make new members more willing to "play nice," because of the social connection formed, and less willing to suddenly come in stirring up trouble. I also agree with WereSloth's points about monitoring fan groups (e.g. I'm active in a DALLE 2 fan Discord) and both using them to keep a pulse on how things are overall, but also making sure that people aren't going to come in here and start trying to generate Elmo at D-Day or whatever.

People are also very paranoid as a whole about access either becoming paid or more limited for existing users. Transparency is important, as is giving some indication (even just a 👍 reaction to suggestions) that a real OpenAI human is paying attention to discussion and we aren't just shouting into the void.

kindred urchin
#

yeah, Q&A/office hour would be a great thing to have. doesn't need to be anything specific with an agenda. one hour a week, could even have folks submit questions ahead of time.

hasty blaze
#

  1. The members are intelligent, creative, enthusiastic and highly interested in generative art and technology more broadly.

  2. Curate membership. Maintain a good verification stage. Establish an effective channel layout, in order to keep discussion relevant. Ideally, moderators work for openai and have a strong connection to the community - otherwise enlist active and mature moderators who can assist with keeping things civil (but without being heavy handed and making the place feel over-regulated).

  3. Huxley, DRP. ? They do most of the above.

  4. CollabLand, Chatsight