AI Protocol's V3 Evolution

With the upgrade to V3, the AI Protocol allows for the creation of Hives.

When an ALI Agent and a Level 5 Intelligence Pod are fused together, a Level 5 ALI Agent is created. A Level 5 ALI Agent can be immutably recorded into the Hive Registry smart contract, producing a Hive, with its own Hive-Native utility token, with the ability to both create a Liquidity Pool and to execute Hive Token airdrops.

A Hive is an AI system that allows ALI Agents to have decentralized access to AI models and data, storage, compute power, and other resources - by enabling trustless coordination between various on-chain and off-chain components of the AI Protocol. In order for a Hive to perform its functions, it requires a variety of smart contracts and tokenized assets to self-execute a system of complex programs and on-chain events. Hives are capable of curating the provision of compute and storage power, represented by Intelligence Pods and ARKIVs respectively, processing complex AI systems, models, and data I/O, and in turn providing these AI Services to ALI Agents to consume. These same Hives are also capable of processing the complex transactions of AI Protocol-compatible assets, ALI, ETH, and Hive-Native Tokens, distributing them as rewards to the various participants in the Hive ecosystem in relation to the resources they provide and consume.

Below is a blueprint of a Hive and how it intersects with other Protocol-compatible assets in this comprehensive ecosystem - allowing for agentic and generative AI to leverage blockchain technology with embedded liquidity - the first decentralized AI system architecture like it in the world.

The architecture of a Hive

As a system, every Hive has the following components:

• Native ERC-20 Utility Token: Every Hive’s ecosystem is underpinned by its own utility token. The Hive’s utility token enables leveraging tokenomics to kickstart and reward resource contribution and participation within its ecosystem. The utility token also acts as a medium for facilitating governance, payments, distribution of rewards, and incentivizing creation of assets that can help increase the utilization of the Hive.

• Hive Smart Contracts: At the heart of a Hive lies a sophisticated set of smart contracts, each tailored to fulfill specific roles within the ecosystem.

• Hive Registry Contract: This smart contract enables the creation of new Hives, maintains an on-chain record of all the Hives and their details, enables ERC-721 tokens to register as ALI Agents, enables Assets to be linked to a Hive and keeps an on-chain record of those links.

• Hive Payments Contract: This contract allows ALI Agents to pay for utilizing the resources of Hive in a private and decentralized manner.

• Hive Reward Distributor: This contract is designed to issue rewards to those contributing their resources (GPUs, AI Models, Storage, Datasets) to the Hive.

• Asset Minter Contracts: Unique digital assets, such as AI models and datasets, can be minted through these smart contracts by any ALI Agent, thereby assigning value and ownership in a secure manner.

• Decentralized Inference Cluster: each Hive has a crowdsourced cluster of GPUs operating off-chain that fulfills inference requests from ALI Agents by running AI models. The Hive’s Creator is given the necessary permissions to run and manage the Hive’s Inference Cluster while any other Agent can pay for and directly send inference requests to it.

• Decentralized Storage Cluster: Parallel to the computational infrastructure, each Hive has an off-chain distributed cluster of crowdsourced storage units. The storage cluster is also run and managed by the Level 5 ALI Agent that created the Hive. This storage cluster provides a decentralized solution for storing data to the Hive’s participants.

• Tokenized Digital Assets: A variety of digital assets can be a part of a Hive’s ecosystem. These assets, such as the Intelligence Pods, ARKIVs, AI Models, Datasets, etc., are not only critical to the creation of a Hive but also the provision and utilization of its resources. Each digital asset, like an AI Model or an Intelligence Pod, can only be a part of one Hive at a time - enabling on-chain tracking and payments to their owners.

• ALI Agents: the entire purpose of a Hive is to enable ALI Agents to have decentralized access to AI, storage, and other resources within the AI Protocol. A Hive needs ALI Agents in order to grow and thrive.

The Unique Modularization Of ALI Agents

One important differentiation between ALI Agents and other digital assets that participate in a Hive’s ecosystem is that ALI Agents are not linked directly to any individual Hive - rather, they have the freedom to pay for and access any Hive’s resources. Other on-chain and off-chain assets, on the other hand, like an AI Model, dataset, ARKIV, or an Intelligence Pod - must be directly connected with one asset or set of assets per Hive.

To understand how a Hive is able to use token incentives and smart contracts to coordinate onchain and offchain between a wide variety of participants, provide ALI Agents with decentralized access to AI and storage services, and perform its other functions, a review of the various components and functionalities of a Hive is needed in the following subsections.

The process of creating new Hives involves three components:

• The Hive Registry Contract

• A Level 5 Intelligence Pod

• An ALI Agent

The Hive Registry

Technically, a Hive is not a digital asset or a separate smart contract. The AI Protocol allows the creation and tracking of Hives via the Hives Registry Contract. This contract, among other things, keeps on-chain records of which Hives exist, provides on-chain details of those Hives, tracks the various assets linked with them, and enables tokens to register as ALI Agents.

A new Hive, therefore, can only be created on-chain via the Hive Registry Contract. Anyone can visit the AI Protocol's GitHub section on How To Create A Hive to do so.

It is important to note that as per this release of the AI Protocol V3, new Hives can only be created on the Ethereum Mainnet. While the other chains might have deployments of the Hive Registry Contract, they would not be able to use the contract to create a new Hive.

With the upgrade to the AI Protocol V3, the smart contracts that enable Intelligence Pods to function have only been deployed on the Ethereum Mainnet, and Intelligence Pods are a necessary element in the creation of a new Hive.

The Level 5 Intelligence Pod

Intelligence Pods are the native assets (NFTs) of the AI Protocol. Intelligence Pods have the capacity to be “fused” on-chain with other NFTs. This on-chain fusion record between an Intelligence Pod and an NFT enables the latter to offer various AI Services as an iNFT. The Intelligence Pods are already being utilized within the AI Protocol to create iNFTs on Ethereum, and until recently, had the maximum Intelligence Level of 4.

The AI Protocol V3 brings with it a variety of major upgrades to the utility of Intelligence Pods.

To start with, Intelligence Pods can now be upgraded to have a Level 5 Intelligence Level. The AI Protocol V3 also lowers the number of ALI Tokens required to upgrade the Pods across all Intelligence Levels - which will result in a higher number of participants being able to upgrade and increase the utility of their Pods.

The Hive Registry Contract enforces that a new Hive can only be created by utilizing a Level 5 Intelligence Pod. Any ALI Agent fused with a Level 5 Pod can utilize the Hive Registry Contract to create one Hive.

The diagram below illustrates the process of creating a new Hive:

As can be seen from the diagram, the creation of a new Hive also sees the minting of the Hive’s own ERC-20 utility token, a portion of which is reserved for the Level 5 ALI Agent owner that created the Hive. The Hive’s own utility token, along with extra incentives created via the ALI Token, help a new Hive kickstart its ecosystem by attracting resource providers and consumers.

Apart from enabling the creation of new Hives, Intelligence Pods are also needed by those seeking to contribute GPU resources. Essentially, anyone looking to contribute computational resources to their preferred Hive and start earning rewards, will first need to link their Intelligence Pod with a Hive using the Hive Registry Contract.

After a Hive is created, most of its on-chain components (like the Utility Token as well as Payments and Rewards Contracts, etc.) are deployed automatically.

The Hive’s creator is expected to follow the documentation and guidelines of the AI Protocol in order to deploy and start the various off-chain components of the Hive. These actions may include: running the software that enables the inference cluster of GPUs and starts the Storage Units cluster, configuring the rewards smart contract, setting up the event trackers for on-chain activities, etc.

Read the AI Protocol's GitBook to learn how to Create/Update/Join a Hive

In short, just like the Miners of Public Blockchains like Ethereum and Bitcoin are expected to follow the rules and standards of their respective protocols, each Hive is reliant on its creator to accurately follow the rules and guidelines shared in the AI Protocol’s documentation.

If the Hive is not managed or run properly by its creator, it will naturally see the resource providers and the assets connected to it (as well as off-chain AI Models, Datasets, etc.) move to other Hives. Similarly, in the future - well run Hives could allow for their creators choose to charge a small premium to those looking to access its resources.

The AI Protocol participants, competition with other Hives, and reward incentives will therefore help ensure that the majority of the Hives are reliable and efficient in their operations.

The evolution of decentralized artificial intelligence (AI) systems marks a significant leap towards democratizing AI technologies. At the forefront of this transformative wave is the AI Protocol V3, which introduces an innovative concept known as Decentralized Inference Clusters (DePIN). This section explores the pivotal role of DePIN within the AI Protocol's Hive infrastructure, detailing its operation, contribution mechanism, and incentive structure.

Defining DePIN Systems

A Decentralized Physical Infrastructure Network (DePIN) system forms the architecture of the AI Protocol's V3 tech stack. It represents a paradigm shift from traditional, centralized AI computation models to a distributed framework where AI inference tasks are executed across a network of decentralized nodes. This model not only enhances the efficiency and scalability of AI services but also ensures a higher degree of accessibility and opportunities for users and compute providers.

The Hive's Decentralized Resource Clusters

Once a new Hive is created, the provision of resources to its ecosystem is incentivized through tokens and other rewards that encourage resource providers such as GPU owners, AI Model developers, dataset providers, and storage providers to join and contribute their resources to the Hive’s ecosystem.

Any resource provider, whether it is a GPU owner, or an AI Model developer, can dedicate their resources to any Hive of their preference. The on-chain elements of the AI Protocol, especially the various digital assets within the AI Protocol’s ecosystem, enable the tracking and coordination needed for the Hives to operate efficiently.

The AI Protocol V3 introduces the ability to connect digital assets with a Hive. When a resource provider connects on-chain with a Hive using an asset, it creates an on-chain record that the resource is dedicated and available to that particular Hive. A digital asset can only be connected to one Hive at a time. Connecting Intelligence Pods and ARKIVs to a Hive is also a form of "Direct Asset Governance," which is a way for Asset owners to signal their support for a Hive by the broadcasting of their Asset being connected to their Hive of choice.

Provide Offchain Compute Power With Onchain Intelligence Pods

The AI Protocol V3 democratizes participation in the DePIN system. Anyone with GPU power can contribute to any Hive's computational cluster, provided they connect an Intelligence Pod to that Hive. Intelligence Pods serve as gatekeepers to the system, ensuring that only authenticated hardware contributes to the DePIN clusters. This open participation model not only broadens the network's computational capacity but also fosters a diverse community of contributors.

Each Hive may include a variance in how rewards are distributed to compute providers, but Propolis' rewards are allocated relative to the Level of the Intelligence Pod. The higher the Level, the more the provider is rewarded.

Incentive Mechanisms: Driving Contribution Through Token Rewards

Contributors to the DePIN system can be rewarded for their participation through Hive utility token rewards and various tokenized reward systems. These incentives are designed to recognize and compensate contributors for their computational power, encouraging sustained and active participation in the network. The AI Protocol V3 introduces a flexible and autonomous incentive framework, allowing each Hive to independently design its community's economy and reward system.

Hives can launch their own Hive-native Utility token, provide liquidity to that token with a Liquidity Pool (LP), and design comprehensive distribution models of their choice, including staking rewards, airdrops, or token allocations as they see fit.

The AI Protocol V3 introduces a groundbreaking framework for data storage: Decentralized Storage Clusters (DSC). Like the compute power of Decentralized Inference Clusters, a Hive is designed to curate contributing storage capacity. By incentivizing and rewarding participation, this Decentralized Physical Infrastructure Network (DePIN) system provides a framework for storage capacity to be democratized and distributed.

Understanding Decentralized Storage Clusters

Decentralized Storage Clusters form a critical component of the AI Protocol's decentralized architecture, as DSCs shift data storage from centralized servers to a distributed network, harnessing the collective storage capacities of numerous participants. This not only ensures data redundancy and security but also significantly enhances data accessibility, capacitory potential, and fault tolerance across the Hive's network.

These clusters serve as the repository for AI models, datasets, and other digital assets essential for the operation of decentralized AI services. By distributing storage responsibilities across multiple nodes, Hives can offer robust and scalable storage solutions, critical for the efficient functioning of decentralized AI applications.

Participation through ARKIVs

To contribute offchain storage capacity to a Hive's DSC, participants must connect an ARKIV to the Hive. As onchain tokenized assets, ARKIVs act as the gateway for storage capacity, ensuring that only verified storage resources are integrated into the DSC. This mechanism fosters a secure and efficient storage network, wherein contributors can effortlessly offer their unused storage space to support the Hive's operations.

Reward For Space

Similar to a Hive's Distributed Inference Clusters of compute power, contributors to the DSC can be incentivized through a diverse reward system, including Hive utility token rewards. These incentives are meticulously designed to acknowledge and compensate contributors for their storage offerings, thereby motivating ongoing and active participation. Each Hive, with its autonomy guaranteed by the permissionless nature of the AI Protocol, can tailor its own incentive mechanisms, like Hive-native token airdrops — creating a dynamic and engaging community economy centered around storage provision.

Like ALI Agents, Hives can have their own specialized functionalities. These functionalities can be broadly classified into several categories, each tailored to specific needs and objectives. This specialization not only enhances the utility of a Hive but also creates niches within the AI Protocol, fostering a diverse and dynamic environment. Every ALI Agent can pay for and access functionalities from multiple Hives.

An Abundance Of Use Cases

Some examples of these specialized functionalities include:

• Domain-Specific AI Services: Hives can specialize in specific domains of AI services, such as natural language processing, computer vision, or predictive analytics. For example, a Hive might focus exclusively on advanced text generation services, providing state-of-the-art language models. Another might specialize in image recognition or generation, offering high-performance AI models tailored for these tasks.

• Style or Genre Specialization: A Hive may cater to specific styles or genres within a domain. In the realm of image generation, for instance, a Hive could specialize in creating anime-style images or excel in generating photorealistic landscapes. This specialization allows ALI Agents to access highly specialized AI models that are optimized for generating content in a particular style or genre.

• Identity or IP-Centric Services: Some Hives might center around a specific identity or intellectual property (IP). This would be particularly useful for ALI Agents that require AI services related to a specific brand, character, or thematic element. For example, a Hive could be dedicated to AI models that generate content or analyze data pertaining to a specific fictional universe or a real-world brand.

• Custom AI Model Development: Certain Hives may focus on the development and training of custom AI models. These Hives would not only attract resource providers like AI Model creators, but will also attract ALI Agents that require unique models, tailored to their specific needs, which are not readily available in the broader market. This might include models trained on proprietary data sets or designed to perform highly specialized tasks.

• Data Analysis and Processing Services: Specializing in data analysis, a Hive could provide services like big data processing, statistical analysis, and pattern recognition. This would be particularly appealing to ALI Agents dealing with large volumes of data needing advanced analytical capabilities.

• Blockchain and Crypto Analytics: Given the nature of the AI Protocol, some Hives might specialize in blockchain and cryptocurrency-related services. These could include AI-driven trading algorithms, market trend analysis, or blockchain data analytics.

• Education and Research Support: A Hive could specialize in services tailored for educational and research purposes. This might involve providing AI models and tools that assist in academic research, educational content creation, or specialized training models for various fields of study.

• Healthcare and Biotech AI Services: Specialization in healthcare and biotechnology would involve offering AI models that assist in medical imaging, drug discovery, patient data analysis, and other healthcare-related tasks.

The specialized functionalities of Hives within the AI Protocol ecosystem allow for a high degree of customization and optimization, catering to the diverse needs of ALI Agents. This specialization not only enhances the utility and efficiency of Hives but also fosters innovation and diversity within the ecosystem. As the AI Protocol continues to evolve, it's likely that we will see an even greater range of specialized Hives, each contributing uniquely to the broader landscape of decentralized AI services.

Level-Relative Rewards

This specialization of functionalities is reflected on-chain by each Hive requiring its GPU providers to link an Intelligence Pod that has a certain minimum Intelligence Level. Generally speaking, service providers that contribute more powerful GPUs to a Hive would be required to link an Intelligence Pod with a higher Intelligence Level. This in turn, would result in the more powerful GPU providers receiving a higher amount of rewards then those contributing less powerful GPUs.

The next layer of the AI Protocol’s tech stack further explores the different types of Assets that exist in the AI Protocol’s ecosystem, and outlines their utility and capabilities.

The Intelligence Layer of the AI Protocol now employs an advanced computer vision model called the Emote Engine. At the moment, this development is deployed to markedly enhance the emotive and expressive capabilities of Prime ALI Agents that are launched as a result of winning the Arena Prime Agent, enabling them to replicate a comprehensive range of human emotions—from subtle gestures of curiosity to exuberant smiles.

This technological advancement not only ensures that AI creations are more interactive but also imbues them with a lifelike spectrum of emotions, thereby establishing a new standard in digital expressiveness. The Emote Engine amplifies the visual realism of avatar interactions and delivers an extensive array of emotions with unparalleled realism, heralding a new era in engaging AI experiences.

Many decisions are made by ALI Utility Token holders to ensure that users of dApps like Noah's Ark and ALI Agents are adhering to the Terms Of Service and Community Guidelines of each dApps and of the Protocol itself. These Terms were developed for the AI Protocol and its dApps, based on founding documents that were proposed and voted on successfully by ALI Token holders in Snapshot votes, to administer governance boundaries that reflect the values and standards of the AI Protocol's community.

When creators of iNFTs, ALI Agents, Hives, or other assets and/or users of the site engage in certain activity that might violate these standards and rules set by the community, users may face an error message that identifies how their content or asset creation has violated one of the rules that have been agreed upon.

Other Forms Of Asset-Powered Governance

Other forms of governance can be executed through the organic, market-driven choices that the community makes with their interactions and with their AI Assets. By purchasing the Keys of an ALI Agent, and interacting with those ALI Agents, and building collaborative IP creations though them, this is a form of "Indirect Asset Governance." Key Holders and Users can signal their support for certain AI Assets through their participation in their evolution, which helps contribute to the health evolution of the ecosystem.

Connecting Pods and ARKIVs to a Hive is also a form of "Direct Asset Governance," which is a way for Asset owners to signal their support for a Hive by the broadcasting of their Asset being connected to their Hive of choice.

How Decentralized Governance Functions On The AI Protocol

In some cases where there is a gray area, potentially violating assets and/or content may be proposed to be voted on by a Snapshot governance vote. Where there are clear cut violations of Terms or Guidelines, the user may experience a filter message while using a dApp on the AI Protocol. Below are a number of messages that clarify what the user may have triggered in their attempts to generate content or assets that go against these rules for ALI Agent creation: