Bittensor: Decentralization AI network leading the wave of Web3 innovation

The Historical Background of the AI Revolution

In recent years, artificial intelligence technology has developed rapidly, and we are entering a data-driven new era. Major breakthroughs in fields such as deep learning and natural language processing have made AI applications ubiquitous. The emergence of ChatGPT in 2022 sparked explosive growth in the AI industry, followed by a series of innovative AI tools, such as text-to-video and automated office solutions. The widespread application of AI technology has also been put on the agenda. With the booming development of the industry, the market value of AI continues to rise and is expected to reach a scale of 185 billion dollars by 2030.

However, the current AI industry is mainly dominated by a few tech giants. While technological advancements have occurred, they have also brought a series of challenges such as data centralization and unequal distribution of computing resources. Against this backdrop, the decentralized concept of Web3 offers new possibilities for solving these issues. The distributed network of Web3 is expected to reshape the current landscape of AI development.

As the AI industry thrives, a number of high-quality Web3 + AI projects have emerged. These projects are dedicated to combining blockchain technology with AI to optimize the training and application of AI models within decentralized networks. However, most projects are still limited to a single application direction. Bittensor has keenly seized this opportunity by building an AI algorithm platform with endogenous filtering capabilities through blockchain competition and incentive mechanisms, to retain the most high-quality AI projects.

Bittensor Project Overview

Bittensor is a decentralized incentive machine learning network and digital goods marketplace. It has the following notable features:

• Decentralized: Bittensor operates on a network composed of thousands of distributed computers, effectively addressing issues such as data centralization.

• Fair incentive mechanism: The $TAO tokens provided by the network to the subnet are proportional to their contributions, and the reward distribution within the subnet follows the same principle.

• Machine Learning Resource Sharing: The decentralized network provides services for individuals who need machine learning computing resources.

• Diversified digital goods market: Initially focused on trading machine learning models and related data, it has now evolved into a marketplace for trading any form of data.

The development of Bittensor reflects the characteristics of a fair, interesting, and meaningful technology project:

• In 2021, created by a group of technology enthusiasts and experts, built a blockchain using the Substrate framework.

• In 2022, the Alpha version network was launched to validate the feasibility of decentralized AI. The Yuma consensus was introduced, emphasizing data privacy protection.

• In 2023, the Beta version was released, introducing the TAO token economic model to incentivize network maintenance.

• In 2024, utilize DHT technology to improve data storage and retrieval efficiency. Strengthen the promotion of subnet and digital goods market.

Essentially, Bittensor is an AI computing and service project powered by GPU miners.

Bittensor's tokenomics

The native token of Bittensor is TAO, which is similar to Bitcoin in several aspects:

• Total supply is 21 million coins
• Halved every four years
• Fair launch distribution, no pre-mining

Currently, a block is generated every 12 seconds, with a reward of 1 TAO for each block. About 7200 TAO are produced daily and distributed to various subnets and their internal validators and miners based on contribution ratios.

TAO tokens can be used for:

• Purchase computing resources, data, and AI models within the network
• Participate in community governance

As of now, the total number of accounts on the Bittensor network has exceeded 100,000, with non-zero accounts reaching 80,000. Over the past year, the price of TAO has risen significantly, with a current market value of approximately $2.278 billion and a unit price of $321.

Bittensor's subnet architecture

The Bittensor protocol is a decentralized machine learning protocol that allows network participants to exchange machine learning capabilities and predictions in a peer-to-peer manner.

The core of the protocol is a network architecture composed of multiple subnets. Each subnet is responsible for managing a group of nodes, adopting a survival of the fittest mechanism. Underperforming subnets and nodes will be eliminated.

A subnet is a piece of independently running code that establishes specific user incentives and functional rules. Currently, there are 45 subnets, and it is expected to increase to 64 between May and July 2024.

The main roles in the subnet include:

• Subnet Owner: responsible for providing the underlying code and setting up incentive mechanisms
• Miner: Runs servers and mining code, can operate nodes across multiple subnets.
• Validators: Assess subnet contributions and receive rewards, can stake TAO for additional benefits.

Subnet emissions are the reward distribution mechanism for miners and validators. Typically, 18% is allocated to owners, 41% to validators, and 41% to miners.

There is a 7-day immunity period after subnet registration. The initial registration fee is 100 TAO, and the fee will double afterward but will decrease over time. When the number of subnets reaches the limit, the subnets with the lowest emissions and that are not in the immunity period will be eliminated.

Bittensor's consensus mechanism

Bittensor adopts a variety of innovative consensus and proof mechanisms:

Proof of Intelligence ( PoI ) mechanism:

• Miners prove their contribution by completing intelligent computing tasks.
• Validators assign tasks and evaluate miner performance

Yuma Consensus:

• Consider the validator rating and staking amount comprehensively
• Exclude abnormal results and allocate rewards based on the comprehensive score.
• Follow the principle of data unknowability to protect privacy
• Rewards are distributed based on node performance and contributions.

Hybrid Expert ( MOE ) mechanism:

• Integrate multiple expert-level sub-models
• Different sub-models work together to enhance overall performance
• Validators can score and rank expert models.

These mechanisms collectively ensure the security, efficiency, and incentive rationality of the network.

Main Subnet Project

Currently, Bittensor has a total of 45 registered subnets, 40 of which have been named. As slots gradually open up, competition between subnets will become more intense. The top three subnets are:

Subnet 19 Vision:

• Focus on decentralized image generation and inference
• 24-hour earnings approximately 627.84 TAO
• The average daily return of the new node is approximately 866 US dollars.

Subnetwork 18 Cortex.t:

• Build an AI platform, providing text and image APIs
• 24-hour earnings approximately 457.2 TAO
• The average daily income of new nodes is approximately 553.64 USD

Subnetwork 1:

• The earliest text generation subnet
• Previously questioned, considered their method inefficient.

In addition, there are various types of subnets such as data processing and trading AI. Overall, the successful operation of subnets can yield considerable profits, but new nodes face intense competition.

development prospects

• The combination of AI and Web3 will attract market attention for a long time.
• The Bittensor project architecture combines technical and market support.
• The subnet mechanism lowers the entry barrier for AI teams.
• Competitive elimination mechanism promotes continuous optimization of projects
• The increase in the number of subnets may bring the risk of reduced profits.
• Balance the quantity and quality of projects to maintain long-term development.