When ZKP Meets AI: Is zkML the Hidden Next Hype Narrative?
TechFlow Selected TechFlow Selected
When ZKP Meets AI: Is zkML the Hidden Next Hype Narrative?
zkML = ZKP + ML, which refers to AI machine learning models based on zero-knowledge proofs.
Navigating Web3 tides with focused insightsAuthor: hitesh.eth
Translation: Frank, Foresight News
zkML might be the next big narrative after artificial intelligence.
However, zkML can be a bit complex for many people. In this article, I'll explain it in the simplest way possible.
What is zkML?
In short, zkML = ZKP + ML
Where: ZKP = Zero-Knowledge Proof, ML = Machine Learning.
Therefore: zkML = Zero-Knowledge Proof Machine Learning
In one sentence: using ZKP technology on machine learning models to generate outputs without revealing sensitive data used during training, while ensuring computational correctness.
What is a machine learning model? A machine learning model is a computer program trained to make predictions based on large amounts of data.
For example, large language models like ChatGPT are built upon machine learning models.
What is inference? Inference is the process of analyzing user prompts, attempting to understand context, and using trained data models to provide results.
Let's take ChatGPT as an example:
The first step in the inference process is writing input—for instance, entering the prompt "Write a crypto rap song in Drake's style."
Second, ChatGPT analyzes the context—"a crypto rap song in Drake's style"—then activates its trained model according to the user's request, identifies patterns within the training data, and generates a Drake-style crypto rap song as output.
What Can zkML Do?
Throughout the entire inference process, two types of privacy issues may lead to sensitive data exposure:
-
Membership Inference Attacks: attackers can analyze model outputs to infer whether specific data points were part of the training process;
-
Model Inversion Attacks: by crafting specific prompts, attackers may attempt to reconstruct fragments of training data from outputs;
How does zkML help here? zkML enables inference on sensitive data without exposing the training data itself.
This is achieved using ZK proof systems such as Plonky and Halo2. Currently, Plonky2 is the fastest ZK proof system.
With zkML, attackers will never have direct access to training data.
Current State of zkML Development
As of now, zkML remains in its early stages, with several startups actively building zkML infrastructure.
Risc Zero is collaborating with Spice AI to build a complete zkML solution suite for developers.
Ingonyama is developing hardware specifically designed for ZK technologies, which could lower the barrier to entry into the ZK field, and zkML may also be applied during the model training process.
Modulus is applying zkML to on-chain inference processes. They currently have six partners building various zkML use cases:
For example, Upshot has built a price prediction model; Worldcoin is using Modulus for private authentication; and AI ARENA is leveraging zkML in gaming economic models.
Privacy-focused blockchain projects such as Oasis Protocol, Secret Network, and Aleo are also exploring zkML-based use cases within their ecosystems. Additionally, NOYA.ai is using zkML to build fully on-chain DeFi strategies.
OraProtocol is building a trustless, ZK-based machine learning inference protocol. Developers will be able to use zkML inference to build any decentralized application powered by machine learning and secured by Ethereum.
The entire zkML narrative is still in its infancy, but I expect that in the coming months, during this bull market, there will be a hype cycle around this concept. Therefore, now is an excellent time to closely track this space and prepare accordingly.
Join TechFlow official community to stay tuned
Telegram:https://t.me/TechFlowDaily
X (Twitter):https://x.com/TechFlowPost
X (Twitter) EN:https://x.com/BlockFlow_News
hitesh.eth
