Bittensor: Decentralization AI Network Pioneer Leading the Web3+AI New Wave

The new wave of the AI revolution

The rapid development of artificial intelligence technology is leading us into a data-driven new era. Breakthrough advancements in fields like deep learning and natural language processing have made AI applications ubiquitous. The emergence of ChatGPT in 2022 sparked a wave of AI enthusiasm, followed by a series of innovative AI tools, such as text-to-video and intelligent office assistants. The concept of "AI+" is also widely discussed and applied. The market value of the AI industry is experiencing explosive growth, with expectations to reach $185 billion by 2030.

However, the current AI industry is mainly dominated by a few tech giants, leading to a series of challenges such as data centralization and uneven distribution of computing resources. At the same time, the decentralized concept of Web3 offers new possibilities for addressing these issues. Under the distributed network architecture of Web3, the development landscape of AI is expected to be reshaped.

As the AI industry flourishes, a number of high-quality Web3+AI projects have emerged. For example, Fetch.ai utilizes blockchain technology to build decentralized economies, supporting autonomous agents and smart contracts to optimize AI model training and application. Numerai leverages blockchain and a community of data scientists to predict market trends, incentivizing model development through reward mechanisms. Velas focuses on creating a high-performance smart contract platform that combines AI and blockchain, offering faster transaction speeds and higher security.

AI projects typically consist of three main elements: data, algorithms, and computing power. Currently, areas like Web3+ data and Web3+ computing power are developing rapidly, but the field of Web3+ algorithms is relatively lagging behind, with various projects often fighting their own battles and struggling to form a synergy. Bittensor has keenly identified this gap and has built an AI algorithm platform with a built-in competitive filtering mechanism through blockchain competition and incentive mechanisms, which is expected to promote industry development while retaining high-quality AI projects.

Bittensor: Pioneer of the Decentralized AI Network

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

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

2. Fair Incentive Mechanism: The token rewards provided by the network to the subnet are proportional to contributions, and the reward distribution within the subnet also follows the same principle.

3. Open machine learning resources: The network provides services for individuals who need machine learning computing resources.

4. Diversified digital goods trading: Initially focused on trading machine learning models and related data, it has now expanded into a platform where any form of data can be traded.

The development history of Bittensor is unique; unlike many highly valued venture capital projects, it resembles a fair, interesting, and meaningful geek project. Its development history can be summarized as follows:

• 2021: The project was launched by a group of technology enthusiasts and experts dedicated to promoting a decentralized AI network.
• 2022: Launched the Alpha version of the network, validating the feasibility of decentralized AI. Introduced Yuma consensus, emphasizing the principle of data unawareness.
• 2023: Release of the Beta version, introducing the TAO token economic model to incentivize network maintenance.
• 2024: Utilize DHT technology to enhance data storage and retrieval efficiency, strengthen promotion and expansion of subnet and digital goods market.

The token of the Bittensor network is TAO, with a total supply of 21 million coins, halving every four years. TAO is distributed through a fair launch mechanism, without pre-mining or team reserves. Currently, a block is generated approximately every 12 seconds, with each block rewarding 1 TAO. These rewards are distributed to various subnetworks based on contribution, which are then allocated to owners, validators, and miners.

TAO can be used to purchase computing resources, data, and AI models on the network, and it also serves as a credential for participating in community governance. Currently, the total number of Bittensor network accounts exceeds 100,000, of which nearly 80,000 are non-zero accounts. In the past year, the price of TAO has increased by several dozen times at its peak, and its current market value is approximately $2.278 billion, with a unit price of $321.

Bittensor's core: subnet architecture

The Bittensor protocol is a decentralized machine learning protocol that allows network participants to exchange machine learning capabilities and predictions, facilitating the sharing and collaboration of models and services. The protocol includes several components such as network architecture, subtensors, and subnet architecture. The network consists of multiple nodes, which are managed by subnets and operate on a survival of the fittest mechanism.

Subnets are the most critical component of the Bittensor network. They can be viewed as segments of independently running code that establish specific user incentives and functionalities while maintaining the same consensus interface as the mainnet. Currently, there are a total of 45 subnets, excluding the root subnet. It is expected that the number of subnets will increase from 32 to 64 between May and July 2024, with 4 new subnets added each week.

The subnet includes three roles: subnet owner, miners, and staking validators.

• The subnet owner is responsible for providing the underlying code and setting up the incentive mechanism.
• Miners are responsible for iterating on the server and mining code to maintain competitiveness.
• Validators are responsible for assessing the contributions of the subnet and ensuring their accuracy.

Subnet emission is the reward distribution mechanism in the Bittensor network, where typically 18% is allocated to owners, 41% to validators, and 41% to miners. Each subnet has 256 UDI slots, with 64 allocated to validators and 192 to miners.

After a subnet is registered, there is a 7-day immunity period, and the first registration fee is 100 TAO. When all subnet positions are filled, new subnet registrations will eliminate the subnet with the lowest emissions that is not in the immunity period. Therefore, subnets need to continuously increase the staking amount of validators and the efficiency of miners to ensure survival.

Bittensor's Innovation: Consensus and Proof Mechanism

The Bittensor network employs various consensus and proof mechanisms, among which the most distinctive are the Proof of Intelligence ( PoI ) mechanism and Yuma consensus.

The PoI mechanism is a unique verification and incentive mechanism created by Bittensor, which proves the contributions of participants through smart computing tasks. Miners complete tasks assigned by validators, who score based on the quality of completion. This mechanism ensures network security, data quality, and efficient use of computing resources.

Yuma consensus is the core consensus mechanism of Bittensor. The scoring of validators is processed by this algorithm, where validators with a higher amount of staked TAO have greater scoring weight. The algorithm filters out results that deviate from the majority of validators, ultimately distributing rewards based on the comprehensive score. The characteristics of Yuma consensus include:

• Principle of Data Unawareness: Protects privacy and security during the data processing process.
• Performance-based rewards: Allocate rewards based on node performance to encourage efficient and high-quality computing and data processing.

In addition, Bittensor introduces the MOE( mixture of experts ) mechanism, integrating multiple expert-level sub-models within a single model architecture. This approach allows different sub-models to work collaboratively, yielding better results than a single model. With the cooperation of the Yuma consensus, validators can score and rank the expert models and allocate rewards, thereby promoting continuous optimization and improvement of the models.

Bittensor subnet ecosystem

Currently, Bittensor has a total of 45 registered subnets, of which 40 have been named. As more subnet slots are opened, registration competition has eased, but the subnet elimination mechanism ensures the survival of high-quality projects in the long term.

Apart from the root subnet, subnets 19, 18, and 1 are the most关注, with emission shares of 8.72%, 6.47%, and 4.16%, respectively.

Subnetwork 19, Vision, focuses on decentralized image generation and inference, providing access to top-tier open-source LLMs and image generation models. Currently, the average daily earnings per node are approximately $866.

Subnetwork 18 Cortex.t is dedicated to building a cutting-edge AI platform, providing high-quality text and image responses through APIs. Currently, the daily average earnings of nodes are about $553.64.

Subnetwork 1 is Bittensor's first subnetwork project, specifically designed for text generation. Despite facing skepticism, it still maintains a high ranking.

From the perspective of model categories, the top-ranking subnets are mostly generative models. Additionally, there are large models for data processing, trading AI models, etc. For example, subnet 22 Meta Search analyzes Twitter data to provide market sentiment, and subnet 2 Omron optimizes staking strategies through deep neural networks.

From the perspective of return and risk, successful operating nodes yield considerable profits, but new nodes require high-performance equipment and optimized algorithms to survive in the competition.

Future Outlook

1. The combination of AI and Web3 will remain a focal point of market attention for a long time, attracting significant investment.

2. Bittensor, as a non-traditional VC project, with both technical strength and market recognition, is expected to maintain its growth momentum.

3. Its innovative subnet architecture provides AI teams with a convenient pathway to access decentralized networks, facilitating quick profit acquisition. The competitive elimination mechanism will also promote subnet projects to continuously optimize models and increase staking amounts.

4. As the number of subnets increases, the registration threshold may lower, increasing the likelihood of low-quality projects entering. At the same time, the TAO rewards obtained from existing subnets may decrease, which could affect expected returns if the TAO price does not rise accordingly.