Machine learning models trained on data taken from blockchain-based marketplaces have the potential to create the world’s most powerful artificial intelligence. They combine two potent primitives:

  1. Private machine learning: private machine learning is he one which allows for training to be done on sensitive private data without revealing it.
  2. Blockchain-based Incentives: This allows the systems to attract the best data and models to make them smarter.

The result is open marketplaces where anybody can sell their data and keep them private. The developers can use incentives to attract the best data for their algorithms to them. Constructing these type of systems is challenging. It is believed that these marketplaces will transit us out from current web era to new web era for data and algorithm, where both are directly monetized.

Origin

The origin leads to 2015 from talking with Richard of numerai. Numerai is a fund that sends encrypted market data to any data scientist who wants to compete to model the stock market. Numerai combines the best model submissions into a model called “Metamodel”. It trades that metamodel and pays data scientists whose models perform well.

Having data scientist compete is a powerful idea. So, you can create a fully decentralized version of this system that could be generalized to any problem.

Construction

For example, consider a fully decentralized system for trading cryptocurrencies on decentralized exchanges.

 

It is one of many potential constructions:

  • Data: Data provides stack data and makes it available for the modelers.
  • Model building: According to the data chosen by the modelers, the models are created. A secure computation method. This method allows the models to be trained without revealing the underlying data. Even the models are staked.
  • Metamodel building: An algorithm that takes into account the staking of each model is used to create a metamodel. The creation of the metamodel is optional. There can be models which can be used without being combined into a metamodel.
  • Using the metamodel: A smart contract takes the metamodel and trades programmatically. It is done through decentralized exchange mechanisms.
  • Distributing gains/losses: Trading produces a profit or loss after certain period of time pass. This profit or loss is divided amongst contributors to the metamodel. It is based on how much smarter they made it. When the models are negatively contributed then some or all their stacked funds are taken. Similar distribution or stake slashing is performed to their data providers by the turned models.
  • Verifiable Computation: Computation for each step is either performed centralized but verifiable and challengeable using a verification game like Truebit or decentralized using secure multiparty computation.
  • Hosting: Data and models are either hosted on IPFS or with nodes in a secure multiparty computation network, as on-chain storage would be too expensive.

There are many potential constructions for a fully decentralized system for trading cryptocurrencies on decentralized exchanges. This is the description of one of them.

 

 

Anusha
WhatsApp chat