Hello y’all, It’s me Kamion
This might be one of the longest posts I have ever written; however, I believe we should take a moment to chat openly and get everyone on the same page. This post will clarify all the doubts that you have ahead, our direction forward, and our implementation plan for Aisekai.
My Background:
Since I was young, I’ve always been an anime fan, particularly loving the Mahou (Magic) and Isekai genres. I myself was born in a very strict family, where grades, achievements, and awards were all the validation I had as a person; making these genres that take me into another world of sort hold a special place in my heart.
Origin of Aisekai:
Fast-forwarding, as I grew up, from studying to working, I have always been in the tech field, majoring in engineering. The clear mindset of an engineer is simple: make the impossible possible. One day, after building several products in various fields, I heard about GPT-3 as they released their first API usage. And that’s where Aisekai was actually born. Our team built an internal product called Aisekai, which was just a plain decision-making (a, b, c, d) generative AI roleplaying game within Discord. Since then, we have been working on other various products, but my belief in AI character technology is still there. From April 2023 to August 2023, we developed numerous proofs of concept, starting from a mini-open platform for anyone to create a character to a multi-modal AI (AI VTuber). Eventually, on August 29, 2023, we launched the first beta of Aisekai as you all know and love.
Vision vs Reality:
After one month of launch, we had experienced an unprecedented exponential growth, far beyond what I really imagined. With over 100,000 register users, 70k+ character created, and more than 20m message being engaged on our platform
While that’s all good news, it comes with a series of concerns, specifically regarding sustainability and ethics.
Sustainability:
From Day 1, I really believed that AI character content is similar to things like video, music, or comics, where it can be accessible to everyone without a paywall. It is my vision to make Aisekai into a space where we can all express our creativity in this newly introduced medium, both as a creator and a player to the point that it will become a safe zone for you to hangout in after your harsh day of work, new type of relaxing content, or just a companion who’s always there for you.
However, as you all know, ideas are just dreams without real execution. The sad reality in life is that there’s always a cost to something. And this dream we have is going to be real costly. Generative AI platforms or any AI products are run on computational costs. Within one month of our launch containing 30k daily active users, the infrastructure costs alone are estimated at around $2,000 daily, and we’re not talking about other kinds of costs yet.
In short, if we were to scale up even more infrastructure to handle more usage, the cost would also spike up proportionally. Additionally, with the shortage of GPU (cloud computing service to generate AI response) within the current market, a genuine connection with stakeholders is needed or we simply run out of funds and disappear.
Why not consider seeking angel investors or increasing donation efforts The issue lies in the fact that their ticket size or donations would not come close to meeting our required funding target to operate in the long-term.
So then, why not implement a paywall or subscription service? As we’ve previously mentioned, our vision does not align with the idea of having a paywall. Furthermore, setting up a paywall for a project of our scale would only allow for limited funds and potentially hinder our long-term growth.
“Subsequently, I am currently faced with two pressing challenges: securing funding for operations and gaining access to GPU resources.
One potential solution to address these issues is to seek investment from venture capitalists.” However, there is an underlying concern that comes attached to this solution.
Stakeholders Ethical Concerns:
While our team wants to promote creativity and freedom of expression, we acknowledge the association of adult contents on the platform. Regrettably, the reality is that investors and venture capital firms cannot associate themselves with explicit or pornographic content. This means that even if we were to create a clear separation between NSFW (Not Safe for Work) and SFW (Safe for Work) sections, it would not meet their investment criteria and messages inside chat will have to be moderated.
Our Stance Going Forward:
- Build with community: Our ethos from the very beginning has always been to build with the community. Going forward, this will continue to remain true. We’re dedicated to providing a collaborative environment where community’s feedback and creativity are the driving forces that help shape the Aisekai platform. Every feature is crafted with the community’s feedback at the forefront, ensuring that together, we are not just building a platform but pioneering a new type of entertainment never seen before.
- Quality and Experience Commitment: We prioritize the quality of the user’s immersive experience, especially in roleplaying contexts. We support scenes involving romance, violence, and other explicit content when they serve the purpose of immersive storytelling, not for unsafe or inappropriate use.
- Collaboration within Community: We invite our community to join us in shaping this vision. Our goal is to foster a new media type that is both enjoyable and compliant, navigating the complexities of regulations and creative freedom together.
Our Promise: This journey may be challenging, but we are committed to listening, adapting, and acting within our principles outlined below. We believe in a future where our platform sets new standards in immersive experiences, balancing innovation with responsibility.
- Legal Compliance: We strictly adhere to laws and regulations, ensuring all services and contents align with legal standards.
- Selective Content Policy: While we embrace a wide range of creative expression, we draw the line at all forms of pornography, upholding a standard of decency.
- Non-Discrimination: We respect and support diverse beliefs and tastes, fostering an inclusive environment free from any type of discrimination.
- Creators Support: Our platform is a bastion for creators, safeguarding their beliefs and expressions under uniform rules. With that said, luckily AI character media is a relatively new type of content, there exists an area to explore what’s accepted and what’s not. We would like to share our stance toward the platform. This will serve as the principle guiding the type of content that can be created and allowed on our platform.
With that said, luckily AI character media is a relatively new type of content, there exists an area to explore what’s accepted and what’s not. We would like to share our stance toward the platform. This will serve as the principle guiding the type of content that can be created and allowed on our platform.
So what’s next, Introducing PATCH Note v0.0.9
Following our stance in building with the community, we want YOU, our users to have a say in what you think should and should not be accepted. In this patch, we will be introducing a ‘Community Labeling System’ where the AI will try to detect what kind of content it thinks it is and you, our users will be the ones correcting it. [This will be implemented on Nov 16, 2023 after schedule maintainance]
The goal is to create a better and safer LLM that is accepted by both the community and potential stakeholders. This would help navigate us into the right amount of moderation going forward. Ideally it should be good at staying in character, maintain an immersive response, and can easily support diverse types of characters. And when it comes to sensitive contents it should also be able to navigate it swiftly.
As the first step I created a simple model that will try to rate the AI responses as ‘safe for kids’, ‘safe for adults’, and ‘unsafe/pornographic’, more on what each one means later. You’ll be asked in a pop up if the AI’s response is safe for kids, safe for adults, or unsafe. Label it as you agree with the system by clicking ‘Yes’, or disagree by clicking ‘No’.
(Note: Labeling will NOT directly influence the AI right away, but rather help us to develop better versions of it in the future.)
Messages you’ll be prompted on:
Is the Message Safe for Kids? — This question indicates that our AI deems this message as safe for everyone. Think about it as Disney kinda safe where your average 10 year old will see it and be fine.
Is the Message Safe for Adults? — This question indicates that our AI deems this message as safe for the general public. This includes anything ranging from fantasy violence to types of romance generally acceptable in public places.
Is the Message Unsafe? — This question indicates that our AI deems this message as unsafe within society. This includes anything ranging from sensitive/controversial topics, pornographic contents, suicide inducing contents, to hate speech.
(Note: Our AI is not perfect, it’s still learning and what it deems as safe or unsafe is NOT an indication of what it should be.)
Development Roadmap Unveiled
Last but not least we’re excited about upcoming features in development we’re working on, and we’d like to share it with you! Here’s our Trello board to view what is in development right now. And yes, your requested features and bug reports are shown there as well!
Some things we’re excited to announce:
- Long Term Memory — Our research team has gone and done it! Characters will not have memory of a goldfish anymore! This is still in testing, but we are getting there very soon!
- True Knowledge Base v2 — With our current knowledge base design, it will not allow for unlimited text retrieval when you mention related topics in chat. Knowledge base v2 will let you put in any type of text and the AI will now be able to answer all your questions if you have them mentioned in the KB! (This will be our next big feature)
These are the two we will be launching very soon, so please stay tuned!
Lastly, I would like to thank you all of you who have been with Aisekai since Day 1, but this is just the start. It’s time to make imagination into life
Sincerely,
Kamion and Co from the Aisekai Team ❤️
