NFTs rely on the cryptographic security of public blockchain networks for properties like traceable provenance and ownership. This is what gives us the ability to "own" digital data, and by extension, gives them value.

When it comes to AI generated art, however, the full extent of the provenance paradigm is currently unfulfilled. Until zkMon, AI art had to be created fully off-chain, then brought on-chain manually — despite the art generation process being digitally native. This means that, until now, outputs generated from an AI artist could never have been cryptographically linked to that original art model.

The architecture of the generator is an upsampling convolutional network, complete with residual blocks and classic 3x3 convolutional filters all throughout.

The overall training pipeline follows that of SNGAN, using the spectral norm as a regularization parameter on the discriminator–an alternative way of approximating the Lipshitz constraint introduced in the foundational WGAN work.

For more details, including training hyperparameters and samples, see the zkMon code repo.

The Modulus Explorer (link to come!) is an AI verification dashboard. Built as a marriage between traditional indexers and analytics dashboards, the Explorer surfaces both on-chain verification details specific to a particular AI output–like the zero-knowledge proofs and records of an AI artist–as well as rich information about the model which powered the result. It’s like a "Nutrition Facts" label, but for verified AI outputs.

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.

NFTs rely on the cryptographic security of public blockchain networks for properties like traceable provenance and ownership. This is what gives us the ability to "own" digital data, and by extension, gives them value.

When it comes to AI generated art, however, the full extent of the provenance paradigm is currently unfulfilled. Until zkMon, AI art had to be created fully off-chain, then brought on-chain manually — despite the art generation process being digitally native. This means that, until now, outputs generated from an AI artist could never have been cryptographically linked to that original art model.

The architecture of the generator is an upsampling convolutional network, complete with residual blocks and classic 3x3 convolutional filters all throughout.

The overall training pipeline follows that of SNGAN, using the spectral norm as a regularization parameter on the discriminator–an alternative way of approximating the Lipshitz constraint introduced in the foundational WGAN work.

For more details, including training hyperparameters and samples, see the zkMon code repo.

The Modulus Explorer (link to come!) is an AI verification dashboard. Built as a marriage between traditional indexers and analytics dashboards, the Explorer surfaces both on-chain verification details specific to a particular AI output–like the zero-knowledge proofs and records of an AI artist–as well as rich information about the model which powered the result. It’s like a "Nutrition Facts" label, but for verified AI outputs.

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.

NFTs rely on the cryptographic security of public blockchain networks for properties like traceable provenance and ownership. This is what gives us the ability to "own" digital data, and by extension, gives them value.

When it comes to AI generated art, however, the full extent of the provenance paradigm is currently unfulfilled. Until zkMon, AI art had to be created fully off-chain, then brought on-chain manually — despite the art generation process being digitally native. This means that, until now, outputs generated from an AI artist could never have been cryptographically linked to that original art model.

The architecture of the generator is an upsampling convolutional network, complete with residual blocks and classic 3x3 convolutional filters all throughout.

The overall training pipeline follows that of SNGAN, using the spectral norm as a regularization parameter on the discriminator–an alternative way of approximating the Lipshitz constraint introduced in the foundational WGAN work.

For more details, including training hyperparameters and samples, see the zkMon code repo.

The Modulus Explorer (link to come!) is an AI verification dashboard. Built as a marriage between traditional indexers and analytics dashboards, the Explorer surfaces both on-chain verification details specific to a particular AI output–like the zero-knowledge proofs and records of an AI artist–as well as rich information about the model which powered the result. It’s like a "Nutrition Facts" label, but for verified AI outputs.

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.