Dear Phoenix Community,
The FLC Committee and Phoenix Global Core Development would like to provide you the latest updates on our global consortium for decentralized AI, Federated Learning Consortium (FLC). As you may know, FLC has been a work in progress that has been awaiting approval as a non-profit organization by the relevant Hong Kong authorities. Due to the founding governance committee and companies consists partially of Mainland China organizations and persons, it has been difficult to secure a non-profit license, hence the delay of launch.
However, the founding members of FLC have decided to proceed with an alternate route that effectively allow the same, if not a wider operating scope. FLC would be registered as a for-profit holding company in Hong Kong, and it would effectively be the governing entity for Federated Learning Consortium (would be registered as a sub-association). The entire process is expected to take around 1 month.
Additionally, during this period of time, FLC has revamped and optimized its goals and objectives as an organization that aims to be the leading entity for the research and advancement of decentralized AI and related fields such as multi-party computation (MPC) and other privacy-preserving technologies.
Five FLC Objectives
1. Research & Development
This is one area of major change compared to the original non-profit route, which was to facilitate the resources, connections, and discovery of best practices within FL and MPC. This restructure of the organization will enable in-house researchers, engineers, and data scientists to spearhead certain projects and initiatives, or in some cases collaborate with third party enterprises to work on their proprietary projects and POCs (proof of concepts). This enables FLC to obtain a wider variety of funding for faster advancement of federated learning projects and faster discovery of use cases.
Phoenix Global, as one of the founding members, and the only founding member in the blockchain space, will work closely with FLC to accelerate use cases and quicker blockchain integration.
2. Establish an Industry-Standard for Best Practices
As a new technology niche still considered in the early phases of Gartner’s Hype Cycle, there are essentially no industry-standards, documentation, and case studies outlining technical and operational best practices for federated learning use cases. We aim to change that by working alongside enterprises, academics, and technology firms, to discover through experimentation, research, and collaboration, the best practices for FL and related technologies across multiple domains, including but not limited to finance, travel, life sciences, retail, consumer internet, and automotive sectors.
We will also perform deep-dive assessments on available federated learning and MPC related technologies such as existing FL-enabled AI frameworks and AI-supported blockchain Layer 1 networks. This includes standalone FL frameworks such as Tencent’s FedAI or FL-enabled extensions such as TensorFlow Federated.
3. Integration and Application of Blockchain
As an organization, the focus on blockchain’s application of AI and MPC is one key differentiator. In the niche FL tech space, there are variations of opinions on blockchain technology, but FLC is definitely within the camp that believes Federated Learning in its ultimate form is essentially decentralized AI with self-governance capabilities. In that sense, blockchain technology is key to helping FL achieve its ultimate form.
The typical implementation method of federated learning currently can be described as less than ideal, as to make the process operate smoothly, a centralized “facilitator” will be required to manage the process. An example; One of the leading organizations and proponents of FL in Greater China region is Tencent’s financial services arm WeBank (Tencent’s equivalent to Alibaba’s Ant Financial). WeBank utilizes FedAI (its proprietary framework) to help banks to leverage FL to achieve joint credit scoring of mutual banking customers without needing to share or expose data regarding that particular consumer. However, the entire process and platform is facilitated by WeBank, and is not self-governed and managed by the banks (the participants). This again, creates potential problems caused by centralization (black box transparency being on of them) and does not achieve truly decentralized AI.
FLC aims to work with members, organizations, and academics to explore truly decentralized implementations of FL that leverages blockchain for transparency and self-governance and self-management.
4. Building a Connected Network and Ecosystem
FLC aims to be the platform that connects an ecosystem of enterprises, professionals, academics, and technology firms that are involved, or related to FL and privacy-preserving technologies. This also includes organizations and personnel in related fields such as MPC, edge-computing, IoT, AI, and system integration.
FLC will differ from an industry or vertical-focused association in that, the goal is joint-exploration, development, and discussion of real-world use cases and already active initiatives for FL that often would require collaboration outside one particular organization. For example, a life sciences or healthcare research firm would like to develop a joint AI project to look at after effects of COVID-19 on lung structure. This would require machine learning on scan data of patients from each respective organization, but such data cannot be pooled or shared.
Perhaps first they would require consultations on the machine learning model used and how FL can be applied. Then they would need to consider how to prepare datasets and facilitate feature-selection from the data. Then a technical-implementation plan would need to be created. This first stage they would probably need to collaborate with a researcher, or an academic, that FLC can either provide them in-house, or from the FLC members network, depending on the specific case on hand, to help them complete a plan and POC (proof-of-concept). Then they would require a technology or system integration firm to help them develop or implement a solution that would be production-ready (run at scale); a core offering FLC will be able to provide.
5. Serve as a Hub for AI, Blockchain, and IoT
The interesting aspect about federated learning is that at its core, it’s able to bridge multiple disciplines and fields. Enterprise-level FL may involve technologies including AI, MPC, and blockchain. Consumer device-level may involve AI, IoT, edge computing, and blockchain as well. With the proliferation of data, AI, and now the need for privacy-preservation technologies at the same time, FL is at the intersection of multiple fast-growing disciplines. This is good because this is able to build FLC into a diverse and robust ecosystem with eclectic value propositions for its participants.
We hope you enjoyed this update on FLC and what is to come!
Have questions? Join the conversation in our Telegram channel https://t.me/APEXcommunity and Twitter https://twitter.com/Phoenix_Chain
