Path to the InferChain: Spectral’s 2024 Roadmap

Tl;dr - In November 2023, Spectral’s Machine Intelligence Network went live on mainnet, bridging AI/ML inferences onchain and founding the Inference Economy. In 2024, we’ll be scaling our platform capabilities and diversifying ML inferences, ultimately progressing towards the InferChain, optimized to serve the rapidly growing demand for verified ML outputs.

The vision behind Spectral’s Machine Intelligence Network

The basic premise behind Spectral is that as artificial intelligence (AI) and machine learning (ML) become more important for society, so too does a need for a transparent, private, secure, verifiable source of ML inferences (i.e. verifiable machine intelligence). Spectral is bridging this gap between ML, AI, and blockchain by creating a two-sided marketplace offering high-quality inference feeds for smart contracts. Spectral uses zero-knowledge machine learning (zkML) to allow for privacy preservation and a unique validation process to protect the intellectual property (IP) of Modelers contributing to the platform and enable reward-sharing based on consumption of these models. To learn more please read our Vision Paper.

We launched the Spectral Network on mainnet in November 2023 with our first challenge: Web3 Credit Scoring. Using a dataset and training model based on our flagship MACRO Score that has been used by 100,000+ wallets since 2021, we invited the ML modeling community to create a better web3 credit score. 

Our launch was met with great interest from the broader web3 and data science communities; more than 200+ Modelers have signed up on our platform to build models for our first challenge. Forty Modelers have also signed up for our Early Modeler Program, a dedicated engagement for providing us with feedback. Modelers on our platform are actively building models for our first challenge. Consumption for inference feeds is slated to begin in February 2024.

Building on this traction, we’re announcing our 2024 roadmap, featuring our vision for a thriving Inference Economy, catalyzed by The InferChain — built for web3 to integrate with AI in a trustless, verified way. To that end, we’ve set specific goals for our phased product development across 2024.

Phase 1: Inception (Q1 2024)

Goal: Launch Machine Intelligence Network (Alpha) on Mainnet and establish a strong base of early adopters.

We begin Q1 by launching the Decentralized Architecture for Machine Intelligence, a platform capable of conducting competitive ML challenges, delivering inferences for consumption, providing verifications using zkML, and distributing bounties and rewards to Modelers and Validators based on the usage of their model.

By February, as the model evaluation window for the first challenge comes to an end, we will commence Consumption of the First AI/ML feed on-chain. Consumers will have the chance to play with these models in browser for free, or consume them within smart contracts for a fee.

To incentivize validation of ML feeds on our network, and improve the robustness of our governance, we’re also launching SPEC governance features. These will give consumers the ability to submit proposals and vote for improving Spectral’s product roadmap.

Once we’re live with the consumption for Challenge #1, we’ll be launching Challenge 2 — Solidity Code Generator — a generative AI model trained on thousands of smart contracts, capable of producing functional smart contract code on demand.

Phase 2: Scaling (Q2 2024)

Goal: Scale and grow network usage by building advanced platform features for Modelers and Consumers on the network

To capitalize on the momentum generated by early adopters (Modelers and Consumers) on the Spectral Network, Phase 2 of our product development will be dedicated to building advanced capabilities that catalyze the usage of inference feeds within smart contracts.

To begin with, we’ll launch Permissionless Creator and Validator Onboarding, a feature that will allow anyone to design, initialize, and launch ML challenges on our platform OR to join as a Validator responsible for verifying the integrity of inference feeds.

To incentivize competitive behavior amongst Modelers, we’ll launch Modeler Reputation Rankings, which will rank all Modelers on the platform as a function of their model scores, model consumption statistics, and other parameters.

By this time, we expect a significant number of consumers using our network to consume inference feeds. To further protect the integrity of our network, we’ll be Securing the Network using SPEC, where each Modeler must stake SPEC to be eligible for additional earnings on the network.

We also anticipate launching a Web3 AI Agent for Solidity, where a user’s wallet connects to an AI Agent that can deploy and execute smart contract logic. For example, anyone can use their agent to write a bot to trade BTC/ETH, find the most optimal route for sourcing decentralized liquidity on Uniswap X, or deploy a new coin onchain. 

Phase 3: Diversification (Q3 2024)

Goal: Enrich the platform by curating versatile inference feeds available for a diverse set of use cases.

During this phase, we expect to have substantial traction on both sides of the marketplace. The product roadmap will focus on improving retention on our platform by providing consumers with a versatile set of inferences that can be customized on demand.

Customizable Inference Feeds are the instruments of the Inference Economy. A Consumer can denote the type of feed they want to consume, the SLA they’re comfortable with, the fees per inference they’d like to pay, and other parameters directly through smart contracts.

We are also in discussions with multiple firms to solicit new ML challenges, such as NFT Recommendations Engine, Uniswap X Collaborative Fillers, Multimodal solidity LLMs, Smart Contract Anomaly Detection, Validator Classification, MEV Searcher/Solver optimization, etc. that will help the broader web3 community realize the value of AI within their smart contract layer.

Phase 4: The InferChain (Q4 2024)

Goal: Actualize the Inference Economy by improving the speed, cost, and efficiency of onchain feed consumption

This year we aim to radically optimize the cost and speed of bringing inferences onchain via an InferChain. Our goal is for smart contracts to be able to consume trusted inferences, verify their provenance, and have certainty in their quality. 

As we continually bridge the gap in bringing useful inferences on-chain, look out for an early release of our Inferchain vision in 2024.

We plan to launch the InferChain in 2024 on testnet, followed by a Mainnet launch in early 2025. Feel free to join our community on Discord or Slack to stay in touch with the latest developments on our network.