“# Problem
The past years have seen a steep rise in generative AI systems thatclaim to be open. **But how open are AI Models really** ([ref](https://pure.mpg.de/rest/items/item_3588217_2/component/file_3588218/content))? AI model training will be regularized as much as the traditional finance domain and we would like to provide a solution to overcome it.
# Solution
**Unstoppable code** represents a paradigm shift in how we think about software and applications: some ideas in our world will be implemented, and launched, and we will not have a single point to shut the code down. There are already unstoppable things like crypto-mixers with all the pros and cons.
We expect that there are a lot of people around the world who would like to share for free(or not) part of their **computing resources for some time for the public good** ([ref article](https://arxiv.org/pdf/2106.10207#page20), [ref post](https://habr.com/ru/companies/yandex/articles/574466/)). We allow them to fund our protocol with computing power (stake with their computing capacity).
To make AI-model be built by the community in an unstoppable manner we need to ensure an infinite progress of model learning and prevent it from degradation. To achieve this we design our on-chain protocol smart contracts.
## Features
– Continuous Improvement: The model can only be improved, as updates are verified for correctness before acceptance.- True Open Source: The initiative ensures that the model and its updates remain open and accessible to everyone.- Censorship Resistance: Once an AI model is published freely, it cannot be regulated or stopped by any central authority (unlike GPT-3).
# Product Solution## Description
We present the design for the Unstoppable Models and example implementation for all parties involved in the protocol:
**Worker**Â – attempts to perform a weight update step (or k steps), put a stake on the correctness of the update. The Worker then publishes the updated weights on IPFS.
**Validator(DAO)**Â – Token holders who resolve disputes. Validators receive rewards regardless of the outcome but are incentivized to act honestly to promote project growth. Validators may share part of profits with Workers.
**Fraud-Proofer**Â – Independently validates selected steps within a specific timeframe and can escalate disputes to the Validator if suspicious activity is detected. If the Validator confirms a correctly identified error, the Fraud-Proofer receives the Worker’s stake.”
**As a result,** we will present a workable MVP with a frontend, smart contract, indexer for data aggregation, python SDK for anyone who wants to learn public model, and:
– there would be enough information for any **worker** to start improve the model according to the protocol, and the **delegator** to stake on that worker- there would be enough information for anyone to try to find that the learning process is incorrect and worker should be stopped, collateral removed from the worker- The contracts will be deployed on zkEVM testnet (Cardona) and on **Polygon zkEVM**- Frontend code will be open-source and anyone could deploy indexer for the data required for that frontend (we are going to use **StreamingFast** and create scheme for our custom contract to be indexed and deployed on The Graph’s Decentralized Network, thus we will support key features of the project [Censorship Resistant and True Open Source] on reprentation layer too). Our Frontend replica will be hosted on the domain we will choose.
