zkML is an exciting confluence of zero-knowledge proof cryptography (ZK) and artificial intelligence (AI). Zero-knowledge proofs facilitate the confirmation of a statement’s validity without revealing its content or the method used to unearth the truth. Within web3, zero-knowledge proofs play a crucial role in scaling blockchains like Ethereum, which are typically slow and expensive at the base layer.
Large computations executed on-chain, on Ethereum for instance, are gas-intensive and might even surpass block gas limits. This is where zkML can be an absolute game-changer. Complex computations can be performed off-chain, with only a zk proof of the computation brought on-chain. Ethereum validators check the validity of the proof, effectively ensuring that the off-chain computation benefits from Ethereum’s decentralization.
Translating this to AI, it’s currently impossible to run an AI model directly on Ethereum. But what if you could simply prove that you’ve executed a specific AI model? The AI then inherits the security of Ethereum validation, or in simpler terms, becomes “on-chain.”
NFTs: The Digital Collectibles
Humans have a rich history of collecting, ranging from historic artifacts to luxury items like watches, cars, art, trading cards, and toys. Despite their tangibility and use in the real world, physical collectibles also have their shortcomings. One significant drawback is their depreciation due to wear and tear, and verifying their authenticity can also be challenging.
Contrastingly, Non-Fungible Tokens (NFTs) offer an array of benefits. Their authenticity can be verified and traced using smart contracts, like the Ethereum blockchain-based CryptoPunks. NFTs can be held, bought, and sold using an Ethereum wallet, eliminating the need for any centralized party.
The Rise of AI Created Collectibles
The term “AI Artist” is becoming more familiar in the NFT realm. An AI Artist uses AI to create unique art, and this “artist” can either be a combination between person and model or a software service alone which creates art based on external inputs. This AI-generated art also provides an opportunity to capture traits and subjective artistic quality.
In our work at Polychain Monsters, we’re experimenting with AI booster packs. These packs generate unique traits for NFTs based on on-chain randomness. Although some NFTs may share traits, the AI-rendered artwork will always be distinct and subjective, making the NFT ranking dependent on AI output rather than physical quality.
Going Autonomous with zkML and On-Chain AI
Currently, the creation of AI-generated collectibles still depends on centralized servers, human supervision, and consistent funding. While this works for many use cases, it compromises the true promise of NFTs — autonomous art with provable provenance — and that’s why we’re exploring the possibility of eliminating these limitations entirely.
With zkML, an AI model can be executed off-chain and proven on-chain according to certain rules, like the version of the AI model, the configuration, training data, etc. Imagine a monster collectible whose visual representation not only evolves but whose character develops through AI, influenced by on-chain inputs. With zkML, your collectible could effectively “live,” mutate, and develop a unique personality without any centralized control, secured by the decentralization of Ethereum and cryptography.
In this new paradigm, your collectible could transform into an autonomous creature that nobody can turn off. It might exist in the metaverse for centuries, acting according to the initially programmed rules.
If you’re interested in delving deeper into zkML and our experimental uses of it at Polychain Monsters, we recommend reading the blogpost by our partner, Modulus Labs.
