Please make sure to check the Glossary and workflow before going through these questions.
Q: How do people create their own Mine?
A: We plan to provide a variety of deployment options for the generic Mine software. The general theme is a downloadable that you can easily run on one of the available devices we plan to support. General candidates for platforms we plan to support are your own cloud node, local machine, smartphone, or possible even a raspberry pi. However, we have yet to curate a final list of places to store your data and are quite open minded about the possibilities.
Q: How do people load data into their Mine?
A: We plan to provide an set of adapters that allow easy access to existing data sinks like Twitter Accounts, Smartwatches and others. However a large thriving ecosystem of awesome developers is required to provide all adapters required to create mines of huge value. In addition to machine generated data there might be web interfaces where Miners can further enrich their data by providing personal details.
Q: How do I know which data is already available in the Mines?
A: All data schemas will be publicly available and the before mentioned adapters will create de-facto standards. OpenMined might show a counter of available data sets and data points per schema.
Q: What is the current development state?
A: Currently the demo shows the overall workflow of OpenMined. In addition all the functional steps are implemented using available open source libraries or open code from our side. The next step for OpenMined will be a large(r) scale proof of concept. While running several mines on the platform we will identify several issues regarding performance, usability and overall platform design. Those will be tackled in a second development phase.
Q: Why do you need a Blockchain?
A: The blockchain technology is currently used to mediate between Farmers and Miners. It takes care of pointing Miners to the correct network to train and also acts as a trust for the Farmers Bounty to be fairly (which is not necessarily evenly) distributed across all participating Miners.
Q: How can a data scientist design his model?
A: Early on the syft library will come with several basic models. Most available libraries like scikitlearn or tensorflow do not have homomorphic encryption capabilities. Therefore they do not work with OpenMined out of the box. We will focus on bringing more ways to build your models later on during the development.
Q: Isn't homorphic encryption slow?
A: If we tightly integrate homomorphic encryption with what you’re trying to with machine learning and stay focused on the end use case, things look pretty good performance-wise. If you take default Tensorflow models and slap homomorphic encryption on it... things won’t look so good 🙂
If you have a new question feel free to open an issue in this repository asking us to clarify or come visit us in slack.