AI is becoming increasingly integrated into modern life. Neural networks and machine learning solutions have found widespread applications in various areas of human activity. Blockchain is no exception.
Bittensor is a project that positions itself as a language for creating decentralized markets, or “subnetworks within a single token system.” These subnetworks operate using the Bittensor blockchain, allowing for interaction and integration into a unified computing infrastructure.
The primary purpose is to combine the computing power of participants to work with AI and machine learning. Each subnet can utilize the computing power of the required number of participants to perform specific tasks such as translation, audio transcription, AI-generated text recognition, stock market forecasting, and more. This project exemplifies the potential synergy between AI and blockchain technologies.
The unity of the various subnetworks is achieved through a shared blockchain and token called TAO.
TAO serves as the native cryptocurrency within the Bittensor ecosystem. The coin is used to reward subnet participants for their work, stake tokens, vote on protocol updates, pay commissions, and facilitate payments related to the use of AI and machine learning services. If a user requires the services of a subnetwork (e.g., text translation), they can use TAO to compensate the participants of that subnetwork (if the subnetwork supports such transactions).
A subnet refers to a competitive market where miners compete to efficiently complete tasks. Anyone can create or participate in an existing subnet. To create a subnet, one must pay a registration fee in TAO and obtain a unique identifier called a netuid.
Participation in an existing subnet can occur in two roles: as a subnet miner or as a validator. Participants register their computing equipment and wallet with the subnet owner. Subsequently, they launch either a miner module or a validator module on the same computer, provided by the subnet owner.
There are distinct roles for participants within a subnet:
– The owner of the subnet defines the tasks for miners and develops an incentive mechanism for other participants.
– Miners execute the assigned tasks using their computing power and AI or machine learning models. It is important to note that miners do not operate on the blockchain itself but are compensated in TAO for their off-chain work.
– Subnetwork validators evaluate the work performed by miners independently, also off-chain. Validators publicly express their assessment of the quality of work, which is used as input for the Yuma consensus engine on the Bittensor blockchain.
The incentive mechanism represents the architecture of the subnetwork.
- The subnet protocol governs how miners and validators communicate with each other.
- The task subnets determine the specific work assigned to miners.
- The reward model of the subnet defines how tasks should be performed and how validators should assess miners’ work.
Subnets interact with the blockchain by recording essential activities. The distribution of rewards for miners and validators is determined on the blockchain. The Yuma Consensus (YC) algorithm facilitates this process. Validator-assigned ratings of subnet miners are sent to the YC algorithm, which calculates rewards based on this input every 12 seconds. TAO rewards are then allocated to miners’ and validators’ wallets.
The YC algorithm requires data from different subnets to be of the same type since Bittensor allows individuals to create their own incentive mechanisms and subnets. Therefore, validators’ assessments of miners’ work are transmitted in the form of weights. The YC algorithm calculates miner rewards based on the distribution of weights.
However, a challenge remains regarding how to fairly reward validators. Dividends, or rewards for validators, are determined by comparing their scores with those of other validators within the subnet. This ensures conscientious behavior, where validators assess the quality of miners’ work, and the quality of these assessments is examined based on how uniformly validators judge.
In conclusion, Bittensor aims to create a unified ecosystem where anyone can create a subnet tailored to solving a specific problem. Each subnet has its own economy and incentive system for participants. Bittensor serves as the platform for deploying these subnets. Currently, the Bittensor ecosystem includes 32 subnetworks addressing various challenges, from text translation to cellular automata research.
Please note that this material and the information provided do not constitute individual or investment advice. The opinions expressed may differ from those of the author, analytical portals, and experts.