Some players may not know why NEAR is often mentioned when discussing AI concepts. Here’s a little-known fact: NEAR started as an AI company. Co-founder Illia Polosukhin has nearly a decade of experience in AI and was one of the eight authors of the groundbreaking paper “Attention is All You Need.”
In 2017, Illia and another co-founder, Alex Skidanov, founded NEAR.ai with the goal of creating the first “AI programmer.” They aimed for natural language communication with computers, which would then automatically program. However, due to the limited capabilities of AI models at that time, this attempt failed.
During this process, they encountered smart contracts and found them to be an interesting subset of programming, but there were many other challenges with blockchain technology. Thus, in 2018, NEAR strategically pivoted to first build a genuinely useful decentralized development platform, NEAR Protocol. The pivot, initially estimated to take six months before returning to AI research, ended up taking six years. While other blockchain projects are now “strategically pivoting” to AI, NEAR is finally returning to its roots.
Recently, NEAR’s official website unveiled its AI tech stack, divided into three main layers: Application Layer, Infrastructure and Model Layer, and Data Layer.
Under these three layers, NEAR has gathered 11 of the latest ecosystem AI projects. BlockBeats will now briefly outline these 11 projects to provide an overview of NEAR’s AI ecosystem.
Application Layer: Bitte, Cosmose, Jutsu
Bitte: AI Agent Wallet
AI Agent is a hot direction in AI application development. The consensus is that in the future, various AI Agents will replace humans in executing various transactions on the blockchain.
Bitte tries to take a small step forward using existing technology by integrating OpenAI’s model API, allowing users to command Agents to perform various on-chain operations through a ChatGPT-like interface.
For example, if the user types “Mint an NFT with AI of a rocket going to the moon”, Bitte will call the DALL-E 3 API to generate an image of a rocket going to the moon and mint it on the NEAR blockchain. In addition, users can have the Agent Swap and transfer tokens for them, create contracts and NFT sets, and more.
Developed by the Mintbase team, a project born during the 2022 NFT boom.
Cosmose AI: AI Shopping Guide Platform
Cosmose AI, an e-commerce company that uses AI to predict and influence offline shopping behavior, received investment from the NEAR Foundation last April.
Partnering with NEAR Foundation, Cosmose launched Kai-Ching (KAIC), a native stablecoin for payments, cashback, and rewards on its e-commerce platform KaiKai, operating on the NEAR network.
Jutsu: AI Agent Marketplace
Still in the whitepaper stage, Jutsu aims to be a marketplace similar to ChatGPT’s GPT Store, where developers can publish AI Agents, and users pay with the platform token JUT to use these Agents.
Originated from the 2021 Eth Denver hackathon as a developer tool called “NEARpad.”
Infrastructure and Model Layer: Exabits, Hyperbolic, Nevermined, Pond
Exabits/Hyperbolic: GPU Computing Power Rental Platform
Both aim to be the io.net on NEAR and are part of the first batch of NEAR Horizon AI incubation programs.
Essential for any chain aspiring to build its AI ecosystem is having its own GPU computing power rental platform.
Nevermined: AI Payment Protocol
A payment platform enabling AI developers to monetize various products, including AI models, agents, and datasets, through smart subscriptions that specify access parameters.
For example, price and time limits, basically the functions of the Web2 subscription platform are moved to the chain in the form of NFT.
Currently, the Nevermined application is deployed on the Polygon, Gnosis and Arbitrum networks, and in the future it should be expanded to NEAR as a payment infrastructure to support the development of the platform’s AI ecosystem.
Pond: Decentralized GNN Model
Focused on Graph Neural Networks (GNN), suitable for analyzing graph-structured data like social networks and chemical molecular properties.
Aims to build a decentralized GNN model to learn from blockchain data and predict user behavior based on learned interaction patterns.
Data Layer: Masa, MIZU, Nillion, Ringfence
Masa Network: Decentralized Data Marketplace
Masa is a subnet on Avalanche that allows users to contribute data and computing resources by running work nodes to earn token rewards. These work nodes scrape, structure, transform, annotate, and vectorize data from sources like Twitter, Discord, and podcasts. Developers (Oracle nodes) can access this data and LLM services to build AI applications.
However, this is not the project’s biggest highlight. Besides the “Node to Earn” model, Masa has been promoting “Surf to Earn.” Before pivoting to AI, Masa was initially developing a decentralized credit scoring protocol based on SBT (Soulbound Tokens). They later innovatively introduced the concept of zkSBT (Zero-Knowledge Soulbound Tokens).
Unlike traditional website cookies, zkSBT allows users to share their browsing data completely anonymously for data analysis and model training, in exchange for token rewards. To facilitate this, Masa plans to launch a Chrome extension, although this extension appears to be more challenging to develop than initially expected.
MIZU: Decentralized Synthetic Data Generation Network
I don’t know why, but every project loves to emphasize that it is the first in various aspects. Mizu claims to be the first and largest decentralized open data network, which is essentially a decentralized synthetic data generation network.
Based on user-contributed datasets, the network incentivizes the community to create prompts to generate large amounts of synthetic data, which, after verification, are submitted to a data repository. This compensates for the lack of real-world data and provides more targeted training data.
According to their roadmap, the testnet is expected to go live in August. Those interested should keep an eye on it. Data is likely to become a new key infrastructure track for decentralized AI, following computing power.
Nillion: Decentralized Secure Computing Network
Nillion is a decentralized public network designed for secure computing and data storage without relying on blockchain technology. It introduces a new cryptographic primitive called Nil Message Compute (NMC), which allows nodes in the network to process data securely and privately without needing to communicate with each other or maintain an immutable ledger like traditional blockchains.
NMC is the core technology behind Nillion, enabling the network to partition and distribute data across nodes, perform secure computations without decryption, and achieve near-centralized server processing speeds while protecting privacy.
Overall, Nillion’s introduction of a new cryptographic primitive has significant application potential in private AI model inference and training.
Ringfence: Data Asset Monetization Platform
AI companies scraping user data for model training has always been a controversial issue, as it is difficult to protect the rights of creators in this model. Ringfence proposes a creative solution to this issue—rNFT. Data uploaded by users to the platform becomes NFT assets, authorized for use as NFTs.
Traditional NFTs usually represent ownership of a single item, while rNFTs are like a folder or collection containing multiple sub-NFTs (called cNFTs). Through smart contracts, rNFT owners can easily commercially license the entire rNFT or specific components within it.
The Ringfence platform allows users to contribute rNFTs for neural network training and receive rewards. Its long-term goal is to build the first neural network trained entirely with authorized work.
Summary
NEAR’s AI projects are still in their early stages, many at the proof-of-concept level, requiring substantial effort before full deployment. Unlike Arweave’s decisive AI transition with project AO, NEAR’s AI return is more understated.
NEAR is recognized as a high-performance blockchain, and its connection to AI mainly revolves around its co-founder Illia. However, NEAR has continuously built AI infrastructure, such as Chain Abstraction, crucial for future Agent-based on-chain interactions.
NEAR does not lack the brand, technology, or funds to make strides in AI. The challenge lies in establishing a vibrant AI ecosystem beyond infrastructure investments.
-
-
-
-
-
-
-
-
