AI has become deeply ingrained in our daily lives. Neural networks and machine learning are being extensively used in various practical applications. Blockchain technology is no exception to this trend.
What is Bittensor?
Bittensor aims to be a language for creating decentralized markets, referred to as “subnets,” within a single token system. These subnets operate on the Bittensor blockchain, allowing them to interact and integrate into a unified computing infrastructure.
The primary purpose of Bittensor is to aggregate the computing power of participants for AI and machine learning tasks. Each subnet utilizes the computational resources of the required number of participants to perform specific types of work, such as translation, audio transcription, AI text recognition, stock market forecasting, and more. This project exemplifies the potential synergy between AI and blockchain technologies.
The cohesion among subnets is maintained through a shared blockchain and token called TAO.
TAO
TAO serves as the native cryptocurrency of the Bittensor ecosystem. It is essential for incentivizing subnet participants, staking, voting on protocol updates, paying commissions and fees, and most importantly, for availing services related to AI and machine learning. If a user requires a specific service offered by a subnetwork, such as text translation, they can use TAO to compensate the participants of that subnetwork (if the subnetwork supports such transactions).
What is a subnet?
A subnet is a competitive market where miners compete to efficiently complete tasks. Individuals have the opportunity to create their own subnet or participate in existing ones. To create a subnet, a registration fee (in TAO) must be paid, and a unique network identifier, called a netuid, is obtained.
Participating in an existing subnet can be done in two roles: subnet miners or validators. Participants register their computing equipment, along with their wallet, within the subnet. Subsequently, they run either a miner module or a validator module provided by the subnet owner on the same computer.
Basic roles of subnet participants
The subnet owner sets the task that miners should perform and designs an incentive mechanism for other participants to follow.
Miners execute the assigned work by employing AI or machine learning models on their computational resources. It’s worth noting that miners do not directly operate on the blockchain; the terminology used by Bittensor might be slightly misleading. Nonetheless, they receive TAO rewards for their off-chain work.
Subnet validators independently evaluate the task completed by miners off-chain. Each validator publicly expresses their assessment of the quality of the work performed. These evaluations are then fed into the Yuma consensus engine on the Bittensor blockchain.
The incentive mechanism essentially defines the architecture of a subnetwork.
-
The subnet protocol governs the communication between miners and validators.
-
The task subnets specify the work to be performed by the miners.
-
The subnet’s reward model outlines the execution of tasks and how validators should evaluate the miners’ work.
How do subnets interact with the blockchain?
The blockchain records significant activities across all subnets. The distribution of rewards for miners and validators is determined within the blockchain. This function is orchestrated by an algorithm called Yuma Consensus (YC). Validator-decided scores for subnet miners are transmitted to the YC algorithm. Once every 12 seconds (per subnet), the algorithm computes rewards based on the input data. The earned TAO rewards are then credited to the wallets of miners and validators within the subnet.
How does Yuma Consensus work?
Since Bittensor allows anyone to create subnets and devise their own incentive mechanisms, it is crucial for the blockchain to process uniform data received from different subnets. To achieve this, validators’ evaluations of miners’ work are transmitted as weights. Based on the distribution of weights, YC calculates rewards for miners.
There remains one challenge: how to fairly compensate validators. The dividends earned by validators depend on how well their scores align with the scores of other validators within the subnet. This incentivizes conscientious behavior from participants – miners’ work is evaluated by validators, and the quality of evaluations is assessed based on the consistency of validators’ judgments.
Conclusion
Essentially, Bittensor strives to create an inclusive ecosystem, enabling anyone to establish a subnet tailored towards solving specific problems. Each subnet operates with its own economy and incentive system for participants. Bittensor acts as the platform for deploying activities within these subnets. Currently, the Bittensor ecosystem encompasses 32 subnetworks, tackling diverse challenges ranging from text translation and image generation to scientific research on cellular automata.
The information provided in this article does not constitute investment advice. The opinions expressed by the authors, analytical portals, and experts might differ.