Bittensor: The Algorithm Pioneer of the Web3+AI Track Decentralization AI Ecosystem Analysis

The historical context of the AI revolution

The rapid development of artificial intelligence technology is leading us into a data-driven new era. Breakthrough advancements in fields such as deep learning and natural language processing have made AI applications ubiquitous. The emergence of ChatGPT in 2022 ignited the AI industry, followed by a series of innovative AI tools, such as text-to-video and smart office solutions. The widespread application of AI is also on the agenda. The market value of the AI industry is experiencing explosive growth, and it is expected to reach $185 billion by 2030.

The current AI industry is mainly dominated by a few tech giants, such as Nvidia, Microsoft, and Google. While technological advances have brought about challenges such as data centralization and uneven distribution of computing power, the decentralized concept of Web3 offers new possibilities for addressing these issues and is expected to reshape the landscape of AI development.

As the AI industry flourishes, a number of high-quality Web3+AI projects have emerged. For example, Fetch.ai creates a decentralized economy through blockchain technology, supporting autonomous agents and optimizing AI model training and applications with smart contracts. Numerai uses blockchain and a community of data scientists to predict market trends and incentivize model development. Velas builds a high-performance smart contract platform that combines AI and blockchain, offering faster transaction speeds and higher security.

The three main elements of AI projects are data, algorithms, and computing power. The development of Web3+ data and Web3+ computing power tracks is in full swing, but the Web3+ algorithm direction is relatively lagging behind. Bittensor seizes this opportunity by building an AI algorithm platform with a built-in filtering function through the competitive incentive mechanism of blockchain, aiming to gather the highest quality AI projects.

The Development History of Bittensor

Bittensor is a decentralized machine learning incentive network and digital goods marketplace. Its main features include:

• Decentralization: Operates based on a distributed computer network, addressing issues such as data centralization.
• Fair incentive mechanism: Network rewards are proportional to contributions.
• Open machine learning resources: providing individuals with the necessary machine learning computing power.
• Diversified digital goods trading: not limited to machine learning models, various types of data can be traded.

Unlike many high-valuation VC projects, Bittensor is more like a geek-driven fair and open project. Its development history is as follows:

• 2021: Project launched, building a blockchain based on Substrate.
• 2022: Released Alpha version, validating the feasibility of decentralized AI. Introduced Yuma consensus to protect user privacy.
• 2023: Launched the Beta version, introducing the TAO token incentive model.
• 2024: Apply DHT technology to enhance data processing efficiency, expand subnetworks and the digital goods market.

During the development of Bittensor, there has been little participation from traditional VC, which helps avoid centralization risks. The project incentivizes nodes and miners through tokens to maintain network vitality. Essentially, Bittensor is an AI computing power and service project driven by GPU miners.

The native token TAO of Bittensor is similar to Bitcoin in many aspects. The total supply is 21 million, and it halves every four years. TAO is fairly distributed, with no pre-mining or team retention. Currently, a block is generated every 12 seconds, with a reward of 1 TAO per block. About 7,200 TAO are produced daily, distributed among various subnets and their participants based on contribution.

TAO can be used to purchase computing resources, data, and AI models within the network, and it also serves as a certificate for participating in community governance. Currently, the total number of Bittensor network accounts exceeds 100,000, with nearly 80,000 non-zero accounts. In the past year, TAO has increased by several tens of times, with a current market value of $2.278 billion and a coin price of $321.

The subnet architecture is gradually being implemented.

The Bittensor protocol is a decentralized machine learning protocol that supports network participants in exchanging machine learning capabilities and predictions, facilitating collaborative sharing of models and services. The protocol includes elements such as network architecture, sub-tensors, and subnet architecture.

Subnets are the most critical component of the Bittensor network. They can be viewed as a segment of independently running code that establishes specific user incentives and functional rules. Currently, there are a total of 45 subnets, excluding the root subnet, and this number is expected to increase to 64 by May to July 2024.

The main roles in the subnet include:

• Subnet Owner: Provides the underlying code and sets up incentive mechanisms.
• Miners: Run servers and mining code, competing with each other to stay ahead.
• Validator: Assess subnet contributions and earn rewards, can stake TAO for additional benefits.

Subnet emission is the distribution mechanism of the TAO token, typically allocating 18% to owners, 41% to validators, and 41% to miners. The subnet contains 256 slots, with 64 allocated to validators and 192 to miners. The performance of validators and miners determines their status and rewards.

There is a 7-day immunity period after subnet registration. The initial registration fee is 100 TAO, and the price doubles for repeated registrations. When a subnet reaches full capacity, a new subnet will replace the one with the lowest non-immunity period emissions. Therefore, subnets must continuously increase their staking amount and efficiency to avoid being eliminated.

Thanks to this architecture, the decentralized AI data network Masa has been launched on Bittensor, becoming the first dual-token reward system and securing $18 million in funding.

Consensus and Proof Mechanism

The Bittensor network employs multiple consensus and proof mechanisms:

Proof of Intelligence ( PoI ) mechanism:

• Miners prove their contribution by completing smart computing tasks.
• Validators assign tasks and assess miner performance.

Yuma Consensus:

• Comprehensive validator rating, considering the weight of staking amount.
• Exclude abnormal ratings and allocate rewards based on the overall score.
• Follow the principle of data unawareness to protect privacy.
• Reward allocation based on performance to encourage efficient computing.

MOE mechanism:

• Integrate multiple expert-level sub-models.
• Sub-models work collaboratively to enhance overall performance.
• Validators can evaluate expert models and allocate rewards.

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

Subnet Project Overview

As of now, Bittensor has 45 registered subnets, 40 of which have been named. In the past, the number of subnet spots was limited, and the registration fee was as high as one million dollars. With the increase in spots, the stability and effectiveness of new subnets may not be as good as that of older subnets, but in the long run, high-quality subnets will emerge victorious through a competitive mechanism.

Aside from the root subnet, subnets 19, 18, and 1 attract the most attention, with emission ratios of 8.72%, 6.47%, and 4.16%, respectively.

Subnet 19 Vision focuses on decentralized image generation and inference, providing access to open-source LLM and image models. Currently, daily earnings can reach 866 USD.

The 18th subnet Cortex.t is committed to building an AI platform that provides text and image services through an API. Currently, the daily income is approximately $553.

Subnet 1 is the first text generation subnet, which has faced skepticism but still maintains a high ranking.

Other subnets include various types such as data processing, trading AI, and more. Overall, successful operating subnets can yield considerable profits, but they require high-performance equipment and continuous optimization to survive in the competition.

Future Outlook

• The integration of AI and Web3 will continue to be a market focus.
• Bittensor combines technological innovation with market recognition, and has a broad development prospect.
• The subnet architecture lowers the barrier for AI teams to enter the decentralized network.
• The competitive mechanism encourages subnets to continuously optimize to avoid being eliminated.
• An increase in the number of subnets may bring risks such as lower registration thresholds and decreased profits.

Overall, Bittensor provides a new paradigm for the integration of AI and blockchain, which is expected to play an important role in future development. However, one must also be wary of potential risks and participate with caution.