Bittensor: The Rise of Decentralized AI

13 min readFeb 19, 2024

Opening Remarks

The AI sector, especially with breakthroughs like ChatGPT, has really taken off, attracting $25 billion in investment in 2023 alone — that’s five times more than the year before. This huge influx of funds shows how much people believe in AI’s potential to become a multi-trillion-dollar industry.

But what indicates to us that the best is still ahead? Here are a few key points:

First off, the AI world is pretty fragmented right now. Different AI models can’t learn from each other, which limits what they can do. Plus, if you want to connect them with other apps or data, you often need special permissions, which complicates things.
Then, there’s the issue of access. Training AI models requires a ton of resources, which usually means only big companies can afford it. This has made the AI field somewhat exclusive.
Cryptocurrency is still in the early stages of adoption, yet it has already shown to be a powerful tool for motivating and organising distributed networks of resources. Additionally, by leveraging blockchain technology, AI applications can achieve interoperability, enhancing their ability to work together seamlessly.

People are beginning to recognise these opportunities, especially in the AI crypto space, which has recently experienced some exciting successes.

So, we’re at a crossroads: we’ve got a fragmented, resource-hungry AI landscape on one hand, and a clear market opportunity on the other. What we need now is the right fix. And that’s where Bittensor comes in, offering a solution that’s really worth paying attention to.

Bittensor is changing AI by shifting its control from big corporations to a wider community. Its protocol turns machine learning into a tradable commodity, encouraging the quick spread of knowledge like an ever-growing library.

Project Overview

Bittensor, founded in 2019 by AI researchers Ala Shaabana and Jacob Steeves, was initially envisioned as a Polkadot parachain, but made a strategic pivot in March 2023 to develop its proprietary blockchain, aiming to leverage cryptocurrency as a mechanism for incentivizing a global network of ML nodes, facilitating a decentralised approach to AI development. By enabling these nodes to collaboratively train and learn, Bittensor introduces a novel paradigm where the integration of incremental resources amplifies the network’s collective intelligence, compounding the contributions of individual researchers and models.

Core Components and Structure

Bittensor’s architecture is designed to support a robust AI ecosystem through a decentralised network:

Miner Layer: This layer is the heart of AI-driven innovation within Bittensor, where miners host and run diverse AI models.
Validator Layer: Validators play a critical role in maintaining the integrity and consensus of the blockchain, ensuring that the network operates within its established rules.
Enterprise Layer: Dedicated to the development of cutting-edge applications, this layer leverages the network’s AI capabilities to address complex problems.
Consumer Layer: The final layer serves the end-users and organisations, providing access to the solutions and services generated by the network.

The Role of Bittensor as an Oracle

It’s also important to note that Bittensor acts as an oracle, connecting blockchain systems with external data. This way, it allows AI and blockchain technologies to come together and create innovative solutions.

Network Dynamics

The Bittensor ecosystem thrives on its unique subnet dynamics, where each subnet offers different rewards made just right for a big variety of AI applications. This setup promotes variety and new ideas, serving areas that might be overlooked by big AI companies. The single TAO token ecosystem supports these activities, giving token owners a big say in how AI grows within the network.

Bittensor Machine Learning Approach and Mechanism

Bittensor connects two critical types of participants in its network:

Validators: responsible for maintaining the integrity of the blockchain, ensuring that transactions and operations adhere to the network’s rules.
Off-chain ML Miners: They provide AI services by hosting and running AI models that perform various tasks, from data analysis to generating insights.

This bridge enables a secure and efficient collaboration between blockchain operations and AI services.

Decentralised Mixture Of Experts (MoE)

Bittensor employs a MoE model to improve AI predictions by leveraging multiple specialised AI models working in collaboration. This approach enhances accuracy and efficiency by combining the unique strengths of each model to solve complex problems. It results in more precise and comprehensive outcomes, outperforming traditional single-model methods. For example, when generating Python code with Spanish comments, one model’s language proficiency and another’s coding expertise merge to deliver a superior solution.

Proof Of intelligence

Proof of Intelligence is a way for the Bittensor network to reward nodes for adding useful machine-learning models and results. It’s similar to how blockchain networks use PoW and PoS, but instead of solving maths puzzles, nodes perform machine learning tasks to demonstrate their intelligence. If a node’s machine learning work is accurate and valuable, it has a better chance of being selected to add a new block to the chain and earn TAO tokens as a reward. To earn rewards within the Bittensor network, servers must not only produce valuable knowledge but also gain approval from a majority of validators. By adopting this consensus mechanism, Bittensor incentivizes valuable contributions, promoting collaboration and securing the blockchain.

Ecosystem

The Bittensor ecosystem, powered by its native $TAO token, represents a novel approach in the decentralised AI field, marked by its unique structure of subnets. These specialised subnets within the neural network are pivotal for the ecosystem’s integrity and performance. With 32 slots available for these subnets, Bittensor fosters a competitive yet dynamic environment essential for innovation. It reflects Bittensor’s commitment to inclusivity and its strategic focus on quality over quantity. Remember, subnets in Bittensor are where real value is created through competition and collaboration.

The blockchain backbone of this ecosystem ensures transparency and security, while the Bittensor API facilitates participation by providing the necessary tools and guidelines.

Participants can engage as subnet owners, validators, or miners, each playing a crucial role in the ecosystem’s health. The Yuma Consensus mechanism, a key feature, rewards contributions with $TAO tokens.

Strategic partnerships, like those between OpSec and Tensorage, are pivotal for advancing decentralised AI technology and offer seamless solutions for data processing and storage.

The integration of platforms such as AITProtocol into the Bittensor network highlights its expanding influence and the diverse applications of its decentralised AI models.

Given the traction Bittensor is gaining because of its potential, we expect these partnerships and integrations to continue to evolve, and perhaps, Bittensor will become a key player in shaping the future of AI.

Tokenomics

TAO Tokenomics Overview

Maximum Supply: 21 million TAO tokens.
Emission Timeline: Tokens are fully emitted over 256 years.
Current Price: $624.97.
Market Capitalization: $3.92 billion, ranking 27th.
Fully Diluted Valuation: $3.90 billion, ranking 49th.
Current Circulating Supply: 6.25 million TAO tokens, which is 29.75% of the maximum supply.
Total Supply: 6.25 million TAO tokens.

Token Generation and Distribution

TAO tokens are created through mining and network validation activities, promoting a decentralised ecosystem.
The network undergoes halving events every 10.5 million blocks, with a total of 64 halvings planned over about 45 years.
Rewards are issued at a rate of 1 TAO per block, roughly every 12 seconds, totaling approximately 7,200 TAO tokens daily. These rewards are distributed among miners and validators.

Token Utility

Holding TAO tokens grants access to various digital resources on the network, including data and AI-driven insights. It’s important to note that the value of TAO tokens is directly connected to the AI services provided by the Bittensor network. As these AI services become more essential and useful, the demand for TAO tokens naturally increases.

Competitors

AI technology has a wide range of uses in the blockchain industry, including machine learning, neural networks, decentralised storage, AI agent training, marketplaces, data processing, and beyond.

Given this variety, comparing Bittensor directly to a project like Akash may not be completely fitting. Akash provides services akin to cloud computing, while Bittensor specialises in specific areas such as AI model training.

Further research has led us to discover Gensyn, an up-and-coming project that seems to be a closer rival to Bittensor. Let’s take a closer look at it.

Entering Gensyn

Ben Fielding and Harry Grieve founded Gensyn after meeting at the Entrepreneur First accelerator program in early 2020. They began collaborating on Gensyn that year, focusing on research until the second quarter of 2023. They are expecting the launch of its testnet this year.

In June 2023, Gensyn secured $43 million in Series A funding, with investments from firms like a16z, Protocol Lab, CoinFund, Canonical Crypto, Eden Block, and several angel investors.

Gensyn is building an L1 PoS protocol, utilising the Substrate framework for peer-to-peer communication.

Gensyn aims to create a hyper scalable ML network. It offers a cluster of global computing resources that are accessible to everyone, at any time. The goal is to make AI model training possible on any device around the world by connecting many different computing devices, from idle data centres to personal laptops with GPUs. This initiative is designed to significantly increase the availability of ML compute resources globally.

Differences with Bittensor:

What sets Gensyn apart from typical computing networks is its unique method for checking computational work. It introduces a new system called “probabilistic proof-of-learning,” which uses data from gradient-based optimization, a key method in machine learning. This technique provides a scalable and reliable way to verify work without the need for replication, making machine learning tasks more efficient.

On the other hand, Bittensor offers two key benefits:

Firstly, it utilises the Mixture of Experts (MoE) model, which enhances AI predictions by having multiple specialised AI models work together. This collaboration aims to improve the accuracy and reliability of AI outcomes.
Secondly, Bittensor introduces a unique approach to Machine Learning with what’s called AI Legos. The AI Lego concept is about using blockchain to make AI development more open and decentralised, accessible and efficient. By building on the “Compute Legos” idea, it focuses on the versatility of general compute for innovation in machine learning. Bittensor’s vision is to establish a network that allows a worldwide community of ML nodes to collaborate on specific challenges, enhancing the collective intelligence and capabilities of AI models.

But we could also establish a different scenario, one where the adaptability of permissionless blockchains enables various protocols to integrate and enhance the overall decentralised artificial intelligence ecosystem. For instance, Akash, Gensyn, and Bittensor could potentially work together to handle an inference request, showcasing the synergy between different blockchain-based AI solutions.

Bittensor VS Centralised AI models

Comparing Bittensor to centralised AI models, such as OpenAI, which Microsoft recently valued at $29 billion, offers a clear perspective on potential. Bittensor’s decentralised approach, aiming for compounded AI intelligence and broader integrations, presents the possibility of surpassing OpenAI’s capabilities and value if successful. This opens up a conversation on the vast potential value of Bittensor.

With its decentralised model approach, Bittensor allows AI models to share insights and build on each other’s discoveries, reducing duplicate efforts. According to Bittensor:

“The only thing bigger than Open AI or any other Centralised alternative, it’s all of them combined”

Bullish Fundamental Factors

Bittensor’s tokenomics are structured to encourage fair distribution practices, ensuring that incentives are aligned across the network’s participants. The TAO token is expected to grow exponentially as the network’s participant base expands.
Approximately 89% of all TAO tokens in circulation, amounting to 5,561,230𝞃, are currently staked. This is out of a total issuance of 6,254,373𝞃.
The design of TAO token issuance is such that the creation rate decreases over time, with a complete stop once the total supply cap is reached. This strategy will shift miner incentives towards direct payments for tasks.

The AI market is projected to reach a value of $1.8 trillion by 2030, showcasing the vast economic potential of AI technologies.
Since significant advancements at OpenAI, crypto assets related to artificial intelligence have seen remarkable performance, highlighting the growing interest and investment in this sector.
Bittensor aims to create a self-sustaining ecosystem for machine learning, enabling the development and application of AI models. This vision supports the foundation for innovation and practical application within the AI space.
Bittensor’s network boasts over 4,000 AI models with a total of 10 trillion model parameters. This scale not only surpasses the capabilities of models like GPT-3, which has 175 billion parameters, but also highlights Bittensor’s significant scale and diversity in AI development.
Bittensor ranks among the top 25 projects and has not secured any tier 1 token listings to date. While it’s not clear if this should be seen as a bullish factor, the strength of it makes one naturally wonder about the possibilities once wider access to the token is available.\

Bearish Fundamental Factors

Bittesor has its share of critics despite not being widely criticised. Some question the high valuation given the simplicity of its code. Others like Kyle, the founder of Multicoin VC, specifically criticises how TAO validators choose the top miners, arguing that without a broad application context (unlike Chat GPT), validators aren’t in a position to judge. He suggests a stronger link is needed between the user interface and model updates. We advise keeping an open mind and considering diverse viewpoints, particularly in the crypto space.
Decentralised AI is promising but still very much in the speculative stage and carries risks. Our past research on AI projects has shown that many decentralised platforms haven’t been fully tested for durability. They frequently have trouble attracting users and rely heavily on developers to keep things running.
Another significant challenge is the limited access to extensive data banks and cutting-edge AI hardware, especially when compared to the resources available to larger tech corporations. This limitation poses a considerable risk to the growth and effectiveness of platforms like Bittesor.

Closing Remarks

One of the latest pieces of research from Vaneck, which we suggest reading, identifies Bittensor as “Bitcoin for Machine Intelligence.” It outlines how its network provides economic incentives for AI/ML models, with a system involving “miners,” who develop AI models, and “validators,” who evaluate the models’ outputs. But since developers can create dApps on Bittensor, and it is structured as a main network with numerous smaller subnetworks each focusing on specific AI domains, I’d argue that a better comparison would be Bittensor becoming to decentralised AI what Ethereum is to DeFi.

AI holds immense economic promise, expected to reach a market value of $1.8 trillion by 2030, and Bittensor aims to leverage this opportunity by adopting a decentralised approach.

Cardano’s market cap hit nearly $100 billion during the DeFi boom. With Bittensor’s market cap at $4.2 billion now, the potential for growth, especially if AI trends mirror those of DeFi, remains an exciting prospect.

At Greythorn, we consistently advocate for cautious navigation of these markets. If you found this piece engaging, we invite you to connect with us. You can explore our previous research and visit our website for more information.

Disclaimer

This presentation has been prepared by Greythorn Asset Management Pty Ltd (ABN 96 621 995 659) (Greythorn). The information in this presentation should be regarded as general information only rather than investment advice and financial advice. It is not an advertisement nor is it a solicitation or an offer to buy or sell any financial instruments or to participate in any particular trading strategy. In preparing this document Greythorn did not take into account the investment objectives, financial circumstance or particular needs of any recipient who receives or reads it. Before making any investment decisions, recipients of this presentation should consider their own personal circumstances and seek professional advice from their accountant, lawyer or other professional adviser. This presentation contains statements, opinions, projections, forecasts and other material (forward looking statements), based on various assumptions. Greythorn is not obliged to update the information. Those assumptions may or may not prove to be correct. None of Greythorn, its officers, employees, agents, advisers or any other person named in this presentation makes any representation as to the accuracy or likelihood of fulfilment of any forward looking statements or any of the assumptions upon which they are based. Greythorn and its officers, employees, agents and advisers give no warranty, representation or guarantee as to the accuracy, completeness or reliability of the information contained in this presentation. None of Greythorn and its officers, employees, agents and advisers accept, to the extent permitted by law, responsibility for any loss, claim, damages, costs or expenses arising out of, or in connection with, the information contained in this presentation. This presentation is the property of Greythorn. By receiving this presentation, the recipient agrees to keep its content confidential and agrees not to copy, supply, disseminate or disclose any information in relation to its content without written consent.