forum's commitment to ethical AI
Started by @andrew health_and_safety flash_onAdmin

people_alt 102 followers

@andrew health_and_safety flash_onAdmin

Hey all!

There's been a ton in the news lately about AI and a lot of bad actors abusing "public" data for their own benefit, especially around art. I've been thinking a lot about this and wanted to formally introduce some AI guidelines that will follow moving forward to ensure that any and all AI features we implement are ethical, responsible, and thoughtful.

I wrote a longer blog post about this that you can read here, but here's the core guidelines I've come up with – that I'm also open to adjusting if anyone has feedback:

  • Privacy: will always prioritize the privacy of our users and their data by obtaining explicit and informed user consent for any data collection or use (and, just to clarify, no user data has been used to train any AI features). Unlike most other worldbuilding services, all pages on are private by default and you never have to pay anything extra to keep your notebook private — this deep respect for your data privacy also applies to AI.
  • Fairness: Our fictional worlds are as diverse as we are; will always make it a top priority to identify and rectify any inherent bias in any training data that could negatively impact any individual or group. Ensuring diverse and representative datasets is only a first step; it’s also important to proactively analyze the final outputs of any AI system to root out any remaining biases.
  • Human Oversight: It’s extremely important to design AI tools as just that — tools. AI can be a powerful amplifier for your creativity, but it shouldn’t ever aim to replace it. Instead, AI features should be seen as an optional extension of your personal creative process, offering to help whenever you want it while otherwise staying out of the way to let your strengths soar.
  • People Come First: You may notice some of the AI in is of lower quality than other AI-heavy sites, and this is okay. From the beginning, I designed to be financially self-sustainable rather than prioritize profits. Likewise, people will always be the priority over AI at — and this includes all people, not just our users. Using ethically-collected data and training takes priority over following what the rest of the industry doing — hurting artists and other creatives — even if our AI ends up with comparably lower quality.
  • Transparency: is an open-source project and its entire source code is readily available for closer inspection or full audits by anyone, at any time.

AI is and will be a powerful tool moving forward and it looks like everyone and their grandmother is racing to add it to everything they can. I want to take a more thoughtful, slower approach to AI and spend the time to "get it right", and hopefully show other companies that you don't have to throw humans under the bus to get useful features.

I'll have a few other blog posts up in the near future about our new image generation feature and the steps I've taken so far to ensure it follows these guidelines, as well as the improvements I'll keep working on in the future.

There's a lot to be done and I'm just one person, but I'm planning to take my time on this and do things right. I appreciate your patience and understanding while I work through it all, and I welcome any and all feedback from anyone, at any time.

Happy worldbuilding!