Inside Bittensor: Dynamic TAO & New Subnets

Key Insights

• Bittensor is a diverse ecosystem of specialized networks, or subnets, that are each focused on a specific machine learning use case or resource provision.
• The Dynamic TAO proposal aims to transition Bittensor’s emission distribution system from a subjective, manual process dominated by a few validators to an automated, market-driven mechanism for more efficient distribution.
• In partnership with the Cortex.t subnet, the Nous subnet has launched an incentivized model fine-tuning platform designed to yield an open-source model that matches the performance of GPT-4.

Bittensor is an AI model marketplace that incentivizes machine learning developers to continually optimize models tailored to a variety of targeted use cases. In doing so, Bittensor creates a permissionless platform that enables open-source AI researchers and engineers to monetize their work, thereby accelerating the growth of the open-source AI movement.

In December of last year, we released a comprehensive report on Bittensor, detailing both the internal mechanics and broader workings of the protocol. Since then, Bittensor has gone from a niche project at the intersection of crypto and AI to gaining broad recognition within the open-source AI community. This is evidenced by endorsements from industry leaders, such as the CEO of HuggingFace, and the launch of subnets by prominent open-source AI organizations like Nous Research.

Other recent developments include a proposal from the OpenTensor Foundation to transition Bittensor from a single-token ecosystem into a multi-token one, a change that would have significant downstream effects.