Although AI is the most disruptive technology of our time, just like web2, most innovations now happen in centralized companies. @NEARProtocol wants to become the premier ecosystem for next-generation AI research and applications to address this phenomenon. Another important component of Near's growing AI ecosystem: the foundational model and AI payment infrastructure.
@PondGNN is an advanced AI model designed for Web3 that uses graph neural networks (GNNs) to analyze complex graph-structured blockchain data. By integrating the concept of GPT, Pond enhances its capabilities to provide accurate insights and predictions while solving the unique challenges of Web3. Potential applications include behavioral analysis, DeFi security, personalized recommendations, anti-money laundering, market prediction, etc. Pond is an industrial-scale model that can support a fully automated AGI world and serve as the on-chain brain for advanced AI-driven services and applications in the Web3 ecosystem.
Pond has built a comprehensive model that can enhance the processing of text and graphic data. To meet the challenges brought by Web3 data, Pond is using graph neural networks as the basic model.
GNNs are a class of neural networks designed to process and extract meaningful information from graph-structured data. They exploit the topology of graphs to capture relationships and dependencies between nodes, which makes them particularly effective for tasks involving interconnected data points, such as Web3 transactions and addresses.
Pond innovatively combines GNN with generative pre-trained GPT to maximize their capabilities. Instead of abandoning existing technologies such as GPT, Pond takes a collaborative approach. It integrates the attention mechanism and time/position encoding in the language model into GNN and uses GPT to directly process text data. The GNN model serves as the source of truth for retrieval-enhanced generation, alleviating the hallucination problem in GPT and enhancing its knowledge. By combining the natural language understanding of GPT with the graph analysis capabilities of GNN, Pond unlocks hidden insights in Web3 data.
Pond will develop the base model and its associated toolkit to enable third-party models to be integrated into the larger model, a technique pioneered by one of his team members, Liangxi Liu, in his paper, “A Bayesian Federated Learning Framework with Online Laplace Approximation.”
The "super model" will support third-party use cases and allow them to customize or fine-tune their own models based on it. In addition to developing their own ecosystem, they are also working with third-party distributors such as Ritual, Allora, and Phala Network to further expand their developer and application ecosystem.
Another important source of revenue for Pond is the use of APIs to call its underlying models. Key stakeholders include third-party distribution channels such as Ritual and Allora, as well as various models, AI agents, and applications. These APIs can seamlessly integrate Pond's advanced analytical capabilities into various services, providing necessary insights and functionality. This approach not only expands Pond's coverage in the Web3 ecosystem, but also ensures stable income through API usage fees, thereby increasing the overall value and applicability of its platform.
Nevermined envisions a world where AI agents become primary consumers. They feel that the advancement of AI depends on the ability of agents to access and process the services and data they need. However, AI agents lack traditional bank accounts, and currently the payment user experience between agents and humans is poor.
Nevermined is solving these problems, enabling AI agents to seamlessly transact, create value, and interact with their communities. AI agents will not have bank accounts, but will use blockchains, digital assets, and crypto wallets to transact. This means that services, models, datasets, workflows, etc. from third parties or even other agents also need to be able to transact with digital assets. Nevermined's solution is to create a platform that allows AI developers, data publishers, and model providers to wrap their APIs in access tokens, thereby controlling usage. This service, called Smart Subscriptions, allows AI builders and agents to monetize their AI services.
AI builders and creators can monetize their APIs and also set time and price access rules. Once the builder/creator sets the subscription terms for API access, a decentralized ID is issued that is associated with the creator's wallet and can be tied to a specific subscription.
As a bonus, the Nevermined App automatically generates a widget with subscription details for discovery by marketplace users, but creators can also add this widget to their own sales channels, such as a website. Purchasing subscription access means the asset’s smart contract gives the buyer a unique subscription token that can be copied into their app.
Smart Contracts + Subscription Logic = Smart Subscriptions! Smart Subscriptions enhance blockchain smart contracts and NFTs, providing greater utility than traditional NFTs. While NFTs typically represent unique assets in a 1-to-1 relationship, Smart Subscriptions allow for a one-to-many relationship, creating buckets of assets under one token.
This innovation is particularly useful for AI work that requires access to multiple assets such as datasets, models, and analytical services. A single Smart Subscription can represent and coordinate the entire AI service pipeline without the need for separate NFTs for each component.
Nevermined has also added time-based access parameters to Smart Subscriptions. This feature supports token-gated access for different durations, ranging from a few hours to a few years, providing flexible and customizable access control for digital assets and services.
Near intends to work with Pond to incubate a set of NEAR native basic models trained based on historical transaction data. Other developers in the ecosystem can then use these models to create various use cases around it.
Near plans to work with Nevermined to create a NEAR native AI agent framework that can be integrated into ecosystem applications. This will bring Nevermined’s payment and coordination platform into the Near ecosystem, empowering and compensating AI developers and agents.
Near is working closely with various AI projects to sow the seeds of an AI ecosystem. As we mentioned, the availability of input data is critical, but so are the underlying models and payment and coordination mechanisms (like what Pond and Nevermined are building). With these in place, you can start to see the ecosystem take shape.
