San Francisco, 10.18.2023 — Hyper Oracle, a startup building cryptographically verifiable data sources for blockchain and ML, has unveiled opML (Optimistic Machine Learning), a groundbreaking framework that brings a new era of fairness and verifiability to onchain AI and machine learning. opML is not merely a technological advancement; it addresses the pressing need for transparency and validity in the AI space.
Why opML is Crucial in Today’s AI Landscape
The AI and ML industry has long been criticized for its opacity. Even when models are declared open-source, there is no concrete evidence to confirm that the specified model was actually used for a given output. This lack of transparency and verifiability in ML models and algorithms mirrors the secrecy shrouding social media platforms’ news feed algorithms for years, which have been kept closed-source under the guise of protecting “trade secrets.”
Great talk this morning at @ETHKL1 from @drCathieSo_eth about what zkML is and how we can use it to improve fairness.
Here’s the gist: pic.twitter.com/gRYmwyp7Lj
— 𓆝 𓆟 𓆞𓆝 𓆟 (@ronstedt) October 15, 2023
https://medium.com/@danieldkang/tensorplonk-a-gpu-for-zkml-delivering-1-000x-speedups-d1ab0ad27e1c
The recent move by Elon Musk to open-source Twitter’s Algorithm was a step towards transparency, but without a mechanism to prove that the open-sourced algorithm is the one in operation, skepticism prevails. This is where opML comes into play, providing a mechanism to cryptographically prove that an algorithm has been run or an ML model’s inference, ensuring that a query to a model has been made and validating the number of parameters of a model through optimistic proofs.
Zero-Knowledge Proofs: A Game-Changer for Onchain ML
ZKPs, a cryptographic mechanism that allows for the validation of a claim without revealing the underlying data, have found significant application in the blockchain sector as a scaling solution for blockchains like Ethereum. However, the cost of implementing zkML (Zero-Knowledge Machine Learning) is currently exorbitant, making it impractical for widespread use.
opML emerges as a viable alternative, enabling AI model inference onchain using an optimistic approach, which is significantly more cost-effective, faster and efficient compared to zkML. It adopts a verification game to guarantee the decentralization and verifiability of the ML computations, ensuring that ML can be performed onchain with flexibility and performance.
opML: Towards ethical AI
Hyper Oracle’s opML is not just a breakthrough; it’s a tool for ushering in a new era of establishing more ethical AI practices in the space. The introduction of opML means that we can now challenge the previously unassailable stance of companies that claimed it was impossible to open-source their ML models and algorithms. With opML, we can utilize optimistic proofs to bring unparalleled transparency into the AI space, ensuring that the AI industry becomes more transparent and just.
The opML framework, which is open-sourced and available for contribution here, is a testament to Hyper Oracle’s commitment to making AI onchain a reality, ensuring that AI not only makes it onchain but also brings about a reason as to why AI should be onchain. It is a first step on a journey towards making zkML a reality.
About Hyper Oracle
Hyper Oracle is building a programmable zkOracle protocol that decentralizes and secures dApps. zkOracles power smart contracts with arbitrary compute, trustless automation, and rich data sources. Hyper Oracle abstracts away the complexity of zero knowledge technology and offers the simplest, developer-friendly way to create next-gen dApps.
Follow us on our social networks and keep up to date with everything that happens in the Metaverse!
Twitter Linkedin Facebook Telegram Instagram Google News Amazon Store
