Modulus: Pioneering Accountable Machine Intelligence
Modulus brings AI to the blockchain. They combine the power of Zero-knowledge (ZK) proofs with machine learning to offer “accountable AI.”
“If we are increasingly going to use the assistance of or delegate decisions to AIs, we need to make sure these systems are fair in their impact on people’s lives, that they are in line with values that should not be compromised and able to act accordingly, and that suitable accountability processes can ensure this”. - (2019) European Commission A definition of AI: main capabilities and disciplines
*Enters Modulus*
AI models are all around us: from chatGPT and Midjourney to the social media algorithm that is used to show specific content on your feed. They affect our decisions and the world at large but are operated by centralized entities that avoid transparency within their system, opening the gates for manipulation. We’ve seen movements calling for decentralized social media and overall permissionless systems, but in order to create a user experience similar to that of centralized systems, we need similar tooling to help effectively bring these systems together.
Using the same example of a decentralized social network, here would be an ideal tech stack:
Decentralized storage of mutable data (IPFS, Aarweave)
Decentralized storage of immutable data (Ethereum)
Decentralized front-end
Smart Contracts
Decentralized/transparent personalization algorithms (ai) ← This is where Modulus comes in.
Most components of the stacks are already there, while other components are now being pioneered by talented teams like Modulus. This is why we are proud to be backing Modulus Labs in their $6.3M round to advance a critical component of the Web 3 tech stack; accountable artificial intelligence. The round is co-led by 1kx Network and Variant, as well as participation from Floodgate, Alliance, Inflection.xyz, the Blockchain Builder’s Fun, GCR, and angels from the Ethereum Foundation, Worldcoin, Polygon, zkSync, Celestia, Microsoft, and more.
What is Modulus?
Simply put, Modulus brings AI to the blockchain. They combine the power of Zero-knowledge (ZK) proofs with machine learning to offer “accountable AI.” This category is popularly known as ZKML (zero-knowledge machine learning).
Let’s unpack what this means.
What does it mean to “bring AI to the blockchain”?
Bringing artificial intelligence on-chain means producing an AI algorithm whereby anyone can verify that the algorithm is operating correctly/fairly. Let’s assume someone built a web3 version of ChatGPT. Since web3’s ChatGPT is on-chain, users can verify that the algorithm operates without tampering.
AI on the blockchain also means that dApps can now embed AI functionalities to offer users better experiences without having to trade off decentralization. For example: an NFT marketplace that curates NFTs you may like based on the NFTs in your wallet. Or, an AI-powered staking vault that finds the best and safest yield using on-chain data – all fully decentralized; from the back end to the AI algorithm used.
How is this even possible?
The cost of AI computation in the web2 space is already extremely high, let alone moving the compute to the on-chain world. But there is already a solution we’re pretty familiar with; ZK roll-ups.
ZK roll-ups helped scale Ethereum, increasing throughput while preserving security. Instead of putting the data directly on Ethereum, Zk roll-ups put the proof of computation on Ethereum. Layer 2 (such as StarkNet and zkSync) is the network of compute nodes that are actually processing transactions and generating these proofs. The proofs are then batched and sent to the Ethereum mainnet, where each proof is verified and the state change is accepted. Zk proofs can be verified extremely quickly, and this is what allows Ethereum to scale.
This mechanism can also be used to bring AI on-chain. If we can create a proof of our AI inference call, then we can demonstrate that we ran a specific model over a specific set of data. ("Inference call" is the process of using a trained ML model to make predictions on new, unseen data or input. When you make an inference call, you're essentially asking the AI model to use its learned knowledge to provide an output based on the input it receives).
Generating this AI inference proof means that others can now verify a model without having to trust the creators of the model, and without having to run the entire computation themselves, only needing to check the generated proof. In other words, we can have powerful AI on-chain without giving up on the decentralized and trustless nature of crypto.
Enabling Trustless Autonomy
Trustless autonomy is achieved when things work how they are supposed to, without having to trust any one person or entity.
In ZK roll-ups, complete verification is the ability to check the entirety of a sequence of computation instructions in a deterministic way. With the complete knowledge of all inputs to a particular function (which is what SNARKs/STARKs do).
A similar thinking is applied to AI on-chain. Modulus defines what complete verification means for AI models. Each model can be considered “completely verifiable” when it consists of 3 characteristics:
The model architecture: The AI architecture for our AI-powered staking vault example should be encoded within the smart contract function itself.
The inputs: This is the raw data that the model utilizes as part of its algorithm. For our staking vault, the inputs would include stuff like the impermanent loss, volume, prices, etc.
The model weights: Model weights refer to the parameters that a machine learning model uses to make predictions or classifications- in our example, you could have weights for the impermanent loss. These weights should be published separately with the ability to check them before utilizing the AI.
Although some of these characteristics may be private to users (for example: our AI staking vault may choose to hide the model weights to protect the alpha), the algorithm can still be verified completely, allowing an application to achieve trustless autonomy given that all three characteristics of a model are transparent and on-chain, the entire execution of that ML model is deterministic and thus can be verified. Most importantly, this requires that the model operator mustn’t be able to switch out weights, architecture, or inputs, and thus can’t influence the final result.
Potential use cases:
We’ve just scratched the surface of what’s possible when AI is brought to the smart contract. Some use cases include:
Matching: Recommendation and matching algorithms can be brought to Web3.
NFTs: Wallet owners can be better served based on the NFTs they currently own.
AI Oracles: A highly accurate AI model can be trained to solve any off-chain validation problem, making a single, centralized oracle trusted by all.
Tokenomics: Improving the quality of a model on-chain and being paid in tokens for the improvements, such as running a DAO based on an AI, or distributing Airdropped tokens based on the evaluation of participation by an AI.
Decentralized Health Care Ensemble: Data scientists can apply their techniques without needing to understand anything about healthcare data or outcomes.
Model Integrity: Providing assurance that the same machine learning algorithm is being run on different users’ data the same way.
Attestations: Integrating attestations from external verified parties into a model or any other kind of smart contract running on-chain.
Decentralized Inference or Training: Performing machine-learning inference or training in a decentralized way, and allowing people to submit data to a public model.
Proof of Personhood: Checking that someone has been verified, without revealing any information about that person’s identity, whether that identity is fully recognizable, or pseudonymous, like a public key.
ML-parameterized DeFi applications: Updating parameters in real-time using an ML model.
Automated trading strategies: Providing proof of the financial model inference when deploying into a specific position.
Fraud monitoring for smart contracts: Detecting possible malicious behavior and enacting a pause.
Decentralized prompt marketplaces for generative AI: Verifying the outputs of a prompt to ensure to a potential purchaser that the prompt does in fact create the desired images.
Replacing the private key with privacy-preserving biometric authentication: Abstracting the private key via facial recognition or other unique factors.
Fair airdrops and contributor rewards: Creating detailed personas of users that determine airdrop allocations or contribution rewards based on multiple factors.
Filtering for Web3 social media: A social media platform could use an open-source ML model that is agreed upon by the community and publish proofs of the model’s inference when electing to filter a post.
Advertising / Recommendations: Running a model on my tastes locally that feeds into media applications to serve me content.
New types of on-chain games: Creating cooperative human versus AI games and other innovative takes on on-chain games.
Builders behind Modulus
The Modulus Labs team consists of three talented co-founders: Daniel Shorr, Ryan Cao, and Nick Cosby. All three of them have technical engineering background. Daniel met Nick in High School and Ryan at Stanford. Modulus Labs has 3 advisors: Zhang Zhenfei (EF), Riad Wahby (CMU), Rand Hindi (Zama). The team built the first fully on-chain deployment of an AI algorithm to mainnet (RockyBot). They then went on to build the first on-chain AI game (https://www.leelavstheworld.xyz/). Learn more about the team here.
We are thrilled to be supporting this team along their journey. To learn more about Modulus Check out their extensive writings here and follow them on X for updates.
About Bankless Ventures
Bankless Ventures is a $40M early-stage Web3 Venture fund launched in 2023 to empower pioneers to explore the frontier of Web3.
If you want to invest in the Bankless Ventures fund, you can submit an interest form as a Limited Partner (LP):
If you’re a startup that wants to be reviewed by our investment team, show us what you’re building here.
Disclosure
The views expressed here are those of Bankless Ventures, LLC (“Bankless Ventures”) personnel quoted and are not the views of Bankless Ventures affiliates. Certain information contained here has been obtained from third-party sources, including from portfolio companies of funds managed by Bankless Ventures. While taken from sources believed to be reliable, Bankless Ventures has not independently verified such information and makes no representations about the enduring accuracy of the information or its appropriateness for a given situation. In addition, this content may include third-party advertisements; Bankless Ventures has not reviewed such advertisements and does not endorse any advertising content contained therein.
This content is provided for informational purposes only, and should not be relied upon as legal, business, investment, or tax advice. You should consult your own advisers as to those matters. References to any securities or digital assets are for illustrative purposes only and do not constitute an investment recommendation or offer to provide investment advisory services. Furthermore, this content is not directed at nor intended for use by any investors or prospective investors, and may not under any circumstances be relied upon when making a decision to invest in any fund managed by Bankless Ventures. (An offering to invest in a Bankless Ventures fund will be made only by the private placement memorandum, subscription agreement, and other relevant documentation of any such fund and should be read in their entirety.) Any investments or portfolio companies mentioned, referred to, or described are not representative of all investments in vehicles managed by Bankless Ventures, and there can be no assurance that the investments will be profitable or that other investments made in the future will have similar characteristics or results. Charts and graphs provided within are for informational purposes solely and should not be relied upon when making any investment decision. Past performance is not indicative of future results. The content speaks only as of the date indicated. Any projections, estimates, forecasts, targets, prospects, and/or opinions expressed in these materials are subject to change without notice and may differ or be contrary to opinions expressed by others.