The Federated States of AI: A Proposal
The role decentralisation has to play in future AI development
We are not opposed to the development of larger and better models โ indeed, it is because we want to see AI technology continue to flourish that we are fighting to prevent the sector from being dominated by two or three huge corporations in collaboration with government.
For reasons mentioned here, we believe a diverse ecosystem of participants, encompassing a wide range of players in all domains, including AI model designers, application developers, and knowledge/data providers, serves humanity's interests more effectively.
There are already several projects working to decentralise AI using Web3 technology and principles.
Bittensor: Decentralized machine learning Akash Network: Decentralized marketplace for cloud resources SingularityNET: Decentralized AI services marketplace fetch.ai: Open network for AI agents Ocean Protocol: Privacy-preserving data sharing protocol for AI Render Network: Decentralized GPU cloud rendering
We are taking a different - but complementary - approach:
KIP Protocol: Decentralised base layer for AI apps, models and data owners to build on, and transact and monetise safely in Web3. This is done through the creation of Web3 Knowledge Assets, granting true digital ownership rights to all AI value-creators.
We believe that a system of fair and transparent revenue sharing between all AI value-creators is the key to encouraging a vibrant ecosystem that incorporates AI players of all sizes.
And we believe that the essential first step towards accomplishing that is the creation of digital ownership rights over knowledge & data through the blockchain.
Our logic is straightforward:
Each component of the AI value chain is dependent upon the others, and currently whoever has greatest market power will also tend to control the channels via which value is redistributed, thus obliging the others to accept its conditions;
By creating an open framework to recognise the ownership of knowledge - whether in the form of data, models or apps - a system of fair-dealing will make it technically and commercially easy for everyone to participate in the economic benefits of model improvement;
The sharing of economic benefits will encourage more people to contribute their knowledge, thus furthering the improvement of AI.
This will result in smaller AI model developers and AI app makers being able to participate in the new ecosystem, further incentivising innovation in the AI space for all and offering users a wider variety of products.
Equitable sharing of economic benefits is the first crucial step to creating awareness and action against AI monopolies.
By implementing a Web3 Framework with blockchain technology for digital ownership rights, we aim to achieve several key objectives:
Ensure Data Provenance and Integrity: On-chain accounting systems will be used to securely and transparently track the origin and usage of data, ensuring its authenticity, reliability, and traceability.
Enhance Privacy and Control: Data owners will have greater control over their information, including by setting usage permissions and conditions, thus ensuring respect for privacy preferences - especially for sensitive data - in AI training. Similarly, small model developers will have the option to sell their work in a closed-source format, rather than simply sharing it freely or keeping it entirely private, as now tends to be the case.
Facilitate Fair Compensation and Attribution: If provenance and control are assured, it becomes possible to ensure that creators and owners of data, models and apps will automatically receive fair compensation for their contributions via an ERC-3525 standard token redistribution framework. This will create a more equitable data sharing ecosystem for AI development, and encourage more creators to contribute their data for AI development.
Boost Transparency and Trust: The transparency of the decentralised ledger allows for the monitoring and auditing of data usage, models and apps, building trust among data providers and enabling suppliers to check up on developers and users. This is crucial for the correct functioning of all such systems, but particularly in sensitive fields such as healthcare and finance.
Facilitate Collaborative AI Development: A fair and secure decentralised data sharing platform will foster a collaborative-but-competitive environment for AI development, producing more innovative and diverse AI solutions: the online ideal of communities made up of small pieces loosely joined.
Last updated