The Whitepaper of MintAI Protocol is Published!

The Whitepaper of MintAI Protocol is Published!

This whitepaper is a combination of the Proof-of-Training protocol (POT) and DeAI protocol research documents. The network structure of POT is largely used as the layer-1 structure of MintAI. Other designs can be found in our DeAI research. The idea behind the MintAI protocol is awesome. We will develop it to the fullest extent.

Until now, integrating web3 into any traditional web2 scenario applications has always been a tedious task, and in many cases, we may wonder, "Yeh, this new app is different, but is it better?"

By default, most web3 applications need wallet integration to support their business model. This intrinsically ends up being more complicated and counterintuitive for most users compared to simpler solutions such as easier login and payment options.

So, we would expect a lot of questions if we build platforms in DeAI, with some expected questions like:

Why are you building protocols for decentralized AI services? Is it in any way better than traditional centralized AI services like Anyword, Midjourney, etc.? Web3 interaction requires chain interaction, which almost always requires a wallet. How can web2 users access this platform? No lunch is free. Even if services are built upon web3 infrastructure, how is the design better than existing ones?

It's important to cover these questions in the whitepaper because the ability to answer them can determine how far the MintAI protocol can go.


Questions that highlight the advantages of MintAI

  • Why are you building protocols for decentralized AI services?

    AI and Web3 integration is always a hot topic given their huge market caps in each. Since gaining popularity in 2021, Web3 has attracted a large audience and community members. Apart from that, the Web3 world has some great resources that can help AI naturally, such as millions of GPU cards ready for use. Previously, people loved to use those cards for crypto mining, but since Ethereum transformed to PoS, the usage in mining has massively shrunk. Meanwhile, the emerging trend of LLMs in 2022 has also brought AI into the spotlight. The Web3 infrastructure and resources are inherently useful for spawning AI services and support in a peer-to-peer manner. That's what the most successful current projects like io.net, Render, and Bittensor are trying to do. So, we also want to be part of the game by contributing useful projects to the community :)

  • You are building decentralized AI services (DeAI). Is it in any way better than traditional centralized AI services like Anyword, Midjourney, etc.? Most AI companies base their revenue model on providing services, either by subscriptions, unit prices per use, selling service credits, etc. However, this causes tremendous transaction friction for customers when they are charged every time they use the service (like image generation) or switch to a different type of AI service, as most of the time a new subscription is required. Moreover, for people in less developed regions, payment itself might even be an issue, as discussed in our whitepaper.

    We designed our network infrastructure to be L1-L2, so that all network operations (excluding asset transfer) can happen without the value store chain such as Solana or Ethereum, thus free! These operations include but is not limited to:

    1. Client/miner registration/unregistration into the network.
    2. New service requests and job handling and routing for clients/miners.
    3. Client ratings for services they request, which are to be logged permanently.
    4. Most network node interaction operations.

    Moreover, the service is free for clients in a much broader sense. For users to access AI services, they must stake their MintAI (MAI) tokens in the smart contract designated for AI services. Think of this staking process as making a fully refundable security deposit. Users can retrieve their tokens by releasing the service providers from the obligation to provide further AI services to them.

    Of course, no free lunch. We have revealed how we made this mechanism work in the whitepaper :)

    In summary, unlike most existing AI products (either centralized or decentralized), our new products do not attempt to collect micropayments for every user request. Instead, users can access services and bandwidth without spending cryptocurrency. This ensures that the services are readily available for the general public to access for free via their browsers. Additionally, with the rating system deployed, the quality of services is consistently high, since all participating miners (given large-scale enterprises, reliable home computer and GPU providers) will offer resources with fast internet connections trying to get higher rating from the clients. Users are well-informed about the quality of the services they are using, and they are not charged. These features eliminate transaction frictions and promote large-scale commercialization and adoption of decentralized AI apps.

    Perhaps one more benefit is that cryptocurrency is more accessible around the world compared to credit cards, and MAI tokens can make use of technologies such as cross-chain bridges to be accessible by all cryptocurrency mainnets. Web3 payments will be made so simple!

Image
  • Web3 interaction requires chain interaction, which almost always requires a wallet. How can web2 users access this platform?

Yes, it turns out the most commonly used OAuth frameworks, such as Google, LinkedIn, and Facebook, can actually be combined with a user-set password to create a secure web3 wallet—all handled with just a few clicks.

What to expect from here on out

"Talk is cheap. Show me the code."

— Linus Torvalds

In the following six months, the MintAI network team will create L2 smart contracts for MAI token utilities, as well as L1 infrastructure implementations for the DeAI service to function. This will include a p2p module, wallet module, consensus engine, and blockchain module. Once everything is ready, one or two demos running on the MintAI protocol will be developed for demonstration purposes.