AINFTize: Make Your Own AINFT

[AI Network is building a decentralized AI ecosystem. Our mission to manage artificial intelligence together with our community and launch of AINFT. Our new release, AINFTize is based on AINFT. This posting was written by AI Network developer, Jungwoo Pyo.]

In the midst of Gen AI era

• It is by far the era of Generative Artificial Intelligence. Five days after Open AI’s Chat GPT was released, the number of users exceeded 1 million and the number of MAU exceeded 100 million in over two months. Considering that TikTok took 9 weeks to achieve 100 million MAU and Instagram took 30 weeks, the impact of Gen AI is a record.

Figure 1. How Long It Took to Hit 100M Monthly Users(source)

• The prevailing idea is that the emergence of LLM (Large Language Models), which are at the center of Gen AI, will change the paradigm of the industry. Several global big tech companies are scrambling to release their LLM models. Examples include Meta’s LLaMA and Google’s Bard. They are rushing to dominate the early Gen AI ecosystem by launching the LLM model.
• In the field of Image Generation, there are numerous prompt-based image generation cases starting with Midjourney and Stable Diffusion at Stability.ai. The quality of the generated image is becoming more and more natural to the extent that it is difficult to distinguish it from the actual image. Even the detailed description pointed out as a limitation can be seen to perform well through ControlNet released by Stanford.

Prompt Composability

• As described above, Gen AI eventually became an inevitable trend. Making good use of this tech has become an important key to the success or failure of the business. It takes a considerable amount of time for a company to develop a completely new product. If you use a modularization and combination of LM models defined as Prompt, however, you can improve the product much easier than before. Prompt composability, which flexibly combines Prompt-based LM models, is an important characteristic that improves the speed of product development.
• Langchain is a key example of Prompt composability. The main components include a Prompt Template that uses prompt to define LLM characteristics and limit input format to respond in the desired direction, Chains that combine multiple prompt templates to produce refined responses from LLM and an Agent that produce a task to answer a given question; all of these characteristics are for a combination of prompt and LLM. Using this, LLM-based applications with great potential are completed.

Figure 2. LangChain Modules

Example of Prompt composability: AI NPC

• You can actually create a variety of applications based on Prompt composability. A typical one is AI NPC specialized in a specific domain. AINPC is not just an interactive chatbot that responds only to human input. Each AI NPC has its own characteristics defined by a prompt. AINPC itself can also have a proactive characteristic of thinking and acting.

Figure 3. Example of AI NPC — Modbox OpenAI-GPT3(video link)

• The above example is a screen that defines the characteristics of an NPC in an open sandbox game called modbox as prompt. The scene of communicating with the NPC can also be seen through the link.
• Let’s imagine what would happen if AINPC with the above characteristics had its own identity. AI NPCs act on their own according to defined prompts and can interact among other AI NPCs without human input. Soon, it can be abstracted into at least the same individual as humans within the ecosystem. In other words, a new ecosystem in which humans and AI NPC coexist — the Internet for AI is born. (A similar simulation was recently conducted by Google.)

AINFT in Web3

(This paragraph referred to twitter thread of Chris Dixon, partner at a16z)

• In the end, in order to implement a protocol in which abstracted objects such as AINPC interact with each other, a standardized primitive is needed. Web3’s philosophy, blockchain, and metaverse platforms are ideal infrastructures for constructing a new ecosystem and their technologies are actively developing them. The presence of primitives in a new ecosystem suitable for Web3 is essential and will be the center of expanding the ecosystem. We present the primitives of the upcoming Web3 and metaverse ecosystem as AINFTs.
• The importance of Primitive can be seen in detail by looking at the developmental process of WWW (World Wide Web) in the past.
• WWW collectively refers to a space where you can exchange information with people using computers connected to the Internet. There are standardized hypertexts written in html based on a protocol called http, and the Internet has played a role in connecting numerous hypertexts to each other. When the Internet first appeared, people thought that websites written in html were just a limited means of displaying information such as brochures and articles. But now it is a means of expressing pretty much everything we think on the Internet.
• It is NFT that can play a primary role in Web3. In the early stages of Web3 where no killer app appeared, people use NFTs only for recording offline assets such as ownership or membership of an art piece, but as the Web3 ecosystem develops and the asset in the ecosystem becomes richer, NFTs can play a primary role in defining ownership in the Web3 ecosystem.
• The protocol that connects these NFTs may exist on the blockchain. Currently, numerous blockchains are working well, and if there are primitives that are well standardized, a huge and new ecosystem can be built based on this.

AINFTize: Build AINFT on Your Own

AINFTize is an authoring tool platform that can easily produce AINFTs as desired using prompt and converting existing NFTs into AINFTs. It is currently planning to provide an image crafting service called Stable Diffusion for NFT. Using the Stable Diffusion inpainting model, users can directly select specific parts of their NFTs or parts they want to change, and enter a prompt to create a new look. The changed appearance can then be applied to the NFT or returned to the previous appearance. The process of changing the appearance of the NFT is all recorded on the blockchain, proving that the first appearance of the NFT or the changed appearances are all the same NFT. These changes can also be tracked.

Figure 4. Example of AINFTize service — Stable Diffusion for NFT

• Just as the introduction of web frameworks such as JavaScript and React has paved the way we can express on the web much more conveniently, we extend AINFTize to creative layers to make AINFT easier and more diverse for users. AINFTize is a scale-out platform. In the future, LM of various domains such as text-to-image model, LLM, and image-to-image will be introduced to provide the function of connecting NFT’s persona setting or voice synthesis model to AINFT as well as images. Through these functions, AINFT can be established as a standardized primary in the upcoming Internet for the AI and contribute to the ecosystem through AINFTize.
• When all the foundations are well established, it is the participation of the open-source ecosystem and community that is the catalyst for the explosive growth of the Web3 ecosystem. In particular, the effect of community participation in Web3 is great. The contents of the ecosystem produced by a large number of users are diverse and creative and the assets they create can increase the value of the ecosystem itself and take a virtuous cycle structure for both the users and ecosystem. We operate several communities, including AINDAO, to support the participation of interested users.
• AINFTize will become the mainstay of the Internet for AI ecosystem that will grow significantly in the future by helping to attract users to make AINFTs.

Conclusion

Figure 5. The roadmap of AINFT Ecosystem

• Common Computer continues to draw and implement roadmaps for building an AINFT ecosystem. We are currently building four components essential to the AINFT ecosystem.

1. We are developing an AI Network, a Layer-1 blockchain to run AI NPC and metaverse.

2. We are developing AINFTize, a metaverse asset authoring tool equipped with an open interface based on prompt. This makes it easy for ecosystem contributors to create and manage an asset or AI NPC, and further facilitates creative work.

3. We are building an AINFT marketplace and a creator economy system that maximizes interoperability. Within the ecosystem, creators can freely operate on various platforms, share and sell their works.

4. We are building a large-scale metaverse world led by mega influencers. Within the metaverse, influencers can interact with various participants, form and grow communities.

• In the AINFT ecosystem, which consists of well-established components, many communities are created where members interact and lead to explosive growth. The value of communities formed on top of AI and metaverse infrastructure will be enormous and the growth potential will be endless. Common Computer is running hard at the beginning of a new market.

AI Network seeks to become the “Internet for AI” in the Web3 era. A decentralized AI development ecosystem built on its own blockchain, AI Network is open to all by gathering developers, creators, and GPU resources for development.

Join as a Captain to receive $AIN rewards for providing GPU resources. Join as a developer to gain access to GPUs and earn from writing open-source AI code. Join as a Creator to utilize AI Network’s developers, open source code, and GPUs to create your own AINFTs.

We represent the future of ‘collaborative AI’ to democratize AI and make it useful and utilizable for everyone.

If you want to know more about us,

• Homepage: https://www.ainetwork.ai/
• AI Network DAO Discord: https://discord.com/invite/aindao/
• AI Network YouTube: https://www.youtube.com/@ainetwork_ai/
• AI Network Facebook: https://www.facebook.com/ainetworkofficial
• AI Network Twitter: https://twitter.com/ainetwork_ai