Blog:
HUMAN awards grant to DataUnion to build Reputation & Recording Oracles
HUMAN Protocol is delighted to announce the awarding of a grant to DataUnion! DataUnion is a Machine Learning project that will be using the grant to build free and open source Recording and Reputation Oracles for data labeling work completed on HUMAN Protocol. While this is the scope of the initial grant, the interests and capabilities of DataUnion could promote further collaborations with our project.
DataUnion is designed to get more value out of data. Founded in November 2020, DataUnion’s first commercial move was to operate in the data marketplace on Ocean Protocol, on which they would capture, annotate, and verify data. They realized that only single entities could offer their data on this marketplace, which created silos, and blockades to the usage and curation of insights from datasets.
They have since built a product for people to build and sell datasets together. This is achieved by ascribing a ‘value share’ to all those who contribute to a dataset. If an individual annotates a single image, and that image is sold, the individual can earn for their role. When data is sold via DataUnion, the data is not simply removed from the data pool, and sent off to be siloed elsewhere. Instead, DataUnion takes a collaborative approach. Sold data is shared data; and DataUnion, along with the buyer, can work on the data to curate more insights.
This sharing of data can benefit everyone. It is one of the core principles of ML that quantity of data is a quality of its own; DataUnion’s method unifies, rather than fragments, data pools.
They recently closed a seed round, backed by Outlier Ventures.
A Recording Oracle records answers and reserves the funds in the smart bounty for the Worker.
A Reputation Oracle checks the work of the Recording Oracle, which checks the work of the Exchange Oracle, which itself may perform checks on the data submitted by a Worker. The Reputation Oracle determines if the Worker will get paid; and assigns a reputation score to the Worker.
In the case of this grant, DataUnion will build a Reputation Oracle specifically to check work that comes from the hCaptcha Exchange Oracle.
The Oracle will perform a human-in-the-loop service. Data that comes through hCaptcha is often worked on by an algorithm; we do not want an algorithm checking an algorithm’s work – for they are likely to fall into the same mistakes. Instead, the hCaptcha algorithm passes on data labels which it cannot determine are correct or not. The DataUnion Reputation Oracle will leverage its networks of humans to check the work.
In time, however, this potentiates many new possibilities. DataUnion will be using the datasets to train algorithms which could themselves either label or check different kinds of data in various scenarios.
In Machine Learning, sharing, connecting, and collaborating can benefit every party. This is a core principle shared by HUMAN Protocol and DataUnion. It’s a principle that can open up scenarios for further collaboration between the two projects.
While the grant is for DataUnion to check the quality of work with a Reputation Oracle, the roles could be reversed, for example. DataUnion could carry out data-labeling work, and hCaptcha could provide the HUMAN-in-the-loop service as a Reputation Oracle. The purpose of this is to utilize each network’s capacity and capabilities to the maximum, and thereby create more potential ways to derive value from datasets.
Looking further ahead, DataUnion’s image-annotation checking oracle could potentiate further oracle work, both within the image annotation vertical, such as that provided by Intel CVAT, and new verticals of data labeling work altogether, such as video and text. Data Union is working on building algorithms for these verticals.
This is only one reason we are happy to begin collaborating with DataUnion; for as they build algorithms to verify more kinds of data, so they can support more kinds of work for Workers to complete via HUMAN Protocol.
For the latest updates on HUMAN Protocol, follow us on Twitter or join our Discord. Alternatively, to enquire about integrations, usage, or to learn more about HUMAN Protocol, get in contact with the HUMAN team.
Legal Disclaimer
The HUMAN Protocol Foundation makes no representation, warranty, or undertaking, express or implied, as to the accuracy, reliability, completeness, or reasonableness of the information contained here. Any assumptions, opinions, and estimations expressed constitute the HUMAN Protocol Foundation’s judgment as of the time of publishing and are subject to change without notice. Any projection contained within the information presented here is based on a number of assumptions, and there can be no guarantee that any projected outcomes will be achieved.
