Join GitHub today
GitHub is home to over 28 million developers working together to host and review code, manage projects, and build software together.Sign up
Welcome to the Computable Protocol Wiki! We will create living documents here early and often to keep our community updated on research & development, proposals, tutorials, FAQs and more. You will get an early look at our work in progress, sometimes in a raw form, because we think openness is crucial for bringing collaborators together around our shared goals. In addition, we will publish formal technical papers on an ongoing basis as our research solidifies. We invite feedback, questions, and contributions both here and in our other repos -- just open an issue and we will review!
Decentralized Data Marketplace for Artificial Intelligence Applications
Making the World Computable
Artificial intelligence (AI) has unlocked a new computing paradigm based on probabilistic inference, where machines can deduce insights directly from data without hardcoded rules and filters. As a result, computational science has extended to real-world systems that have been too multivariate, complex, and stochastic to model with math theorems and simple rules-based computing. In effect, AI has made the world more computable.
However, AI relies on data, which remains an incredibly fragmented resource even in the age of the Internet. As long as data primarily exists in silos, inaccessibility will rate-limit the progress that AI can make. Fortunately, blockchain technologies offer new Internet primitives that can break these walled gardens down. Public blockchains enable transparent data provenance, making fair attribution on the web a far easier task. Permissionless and immutable, they can garner trust and participation that cuts across traditional corporate, national, and cultural boundaries. Finally, protocol tokens working in concert with blockchains offer economic incentives for sharing data that thus far have been missing. Taken together, these features will redefine what it means to own data on the Internet and unlock new global markets based on transacting data.
Computable Labs is leveraging these tailwinds to develop a new kind of decentralized network that crowdsources data and makes it available for learning agents, whether human or autonomous. Our protocol token design economically incentivizes one group of participants to supply data to the network and another group to curate data for relevance and quality. Network data will be organized so that everything can be queried, transacted, and transmitted within a single coordination layer. In particular, all data on the network will be algorithmically accessible, enabling a future in which computational agents can autonomously request and procure data required to meet their objectives. Without such universal access, data would remain disparate and hard to use even if openly available on the web. The network becomes increasingly valuable as data accrues on it over time, which also magnifies the importance of the network’s ability to query across all that data in a unified way. If the network is successful in better organizing the world’s data, we can unleash public data resources that will be more discoverable, usable, and ultimately more valuable on-network than off-network.
In the near term, network use cases include data markets, data on demand and as a service, and data cooperatives for scientific disciplines e.g. genomics. The long-term ambition of this project is to encode the physical world into structured data so intelligent agents can learn and simulate any kind of real-world complex system from large cities to human biology.
[Data Commons and Cooperatives]
Develop and launch the Data Markets Protocol. Incentivize communities to form around crowdsourcing and curating useful and valuable data sets. Formal technical publication coming soon. (Alpha in 2018)
Create a Data Terminal protocol so that the data collected into Data Registries can be securely shipped to users and buyers. This enables active data markets in the near term, but these data markets will include data leakage since raw data is transferred to buyers. Early implementation will focus on encrypted end-to-end data transfer from local clients. Local clients as defined here can mean local computers but also personal storage services like Google Drive, Dropbox, AWS and more. The intention is to also expand storage nodes to decentralized providers like Filecoin and others when those networks are technologically ready. Data Registry interoperability relies on the Data Terminal protocol to create a single query layer and enable a universal data market that cuts across all registries.
Build on the ability to securely ship data from Phase II to realize privacy properties through secure blockchain-coordinated distributed computing. Achieve privacy-preserving computation so that use cases are unlocked for highly sensitive data where data leakage is intolerable. [Note: We have some very early, but very exciting plans in the works for this...]