Project Overview
FLock is a decentralized AI model training and verification network that aims to break the monopoly of technology giants on AI models. It enables decentralized development of AI models by providing modular computing power, data, and training methods. It democratizes model creation and adjustment by incentivizing the community to provide training data and model feedback. Protocols and developers can use FLock's platform to train models specific to their own use cases, such as AI companions, function call models, sensor analysis, etc.
The core concept of FLock is to activate the potential of private domain data and push artificial intelligence to a new breakthrough point. It innovates the data calling and model training methods of AI, and extends the traditional training model that relies on public data to the private data field.
At the technical and product level, it focuses on decentralized training and adopts federated learning technology to ensure that the training data is decentralized and stored while the model is learning, without leaking data privacy. At the same time, it guides training needs to decentralized computing platforms such as Akash and IO.net, further realizing the decentralization of model training.
At the same time, the FLock team has been deeply involved in the field of artificial intelligence for many years, with top academic backgrounds such as Oxford University and BCG and rich industry experience. The team has contributed core technologies to many well-known open source AI projects and published many papers in top scientific journals such as the Royal Society and NeurIPS, with solid technical strength.
This month, the FLock project also received great news in terms of financing, and has won the favor and support of many well-known investment institutions. According to the latest official news, FLock has successfully completed a round of strategic financing, led by Grayscale's parent company Digital Currency Group (DCG), and followed by a number of strategic partners including Lightspeed Faction, Animoca Brands, Fenbushi Capital, GnosisVC, GSR Ventures, OKCoinJapan, Bas1s Ventures and A41, to jointly promote the vigorous development of FLock. A total of US$11 million in funds were raised, which will be used specifically to launch FLock's upgraded test network and federated learning client, laying a solid foundation for the future development of FLock.
FLock has always been committed to the lofty mission of democratizing the creation, governance, and ownership of AI agents, and is committed to taking power back from the monopoly of large global companies, returning more decision-making power and rewards to the community, and achieving true sharing and co-governance. As one of the most active and visionary investment institutions in the last bull market, DCG has a wide and in-depth investment layout in the field of AI, including TAO, WLD and other leading AI projects.
This investment is undoubtedly another important layout of DCG in the AI field. The new funds will provide impetus for FLock's innovation and development in the field of AI infrastructure, help it move to a new stage, and jointly explore new possibilities for the future of deAI.
FLock architecture and product innovation
As the only AI infrastructure project funded by the Ethereum Foundation, FLock allows AI models to be trained while data is stored locally. To date, the FLock training platform has brought together 1,500 verified active Github AI engineers to jointly promote the advancement of decentralized AI.
FLock's product suite covers the entire life cycle of decentralized AI development, including AI Arena, FL Alliance, and AI Marketplace. AI Arena is a competitive platform for basic model training, where developers can use decentralized computing power to optimize models; FL Alliance is a privacy-protected collaborative platform where data contributors and developers can jointly optimize basic models while ensuring data privacy; AI Marketplace is a model hosting and trading platform where users earn revenue through data, computing resources, and algorithm contributions. 
FLock Edge Compute is an important part of the FLock project, which emphasizes the use of high-quality private data, lower training data latency, and distributed computing power. Through this innovative technology, AI engineers can request encrypted data from users' edge devices in real time to create more accurate and richer models. At the same time, ordinary users can not only protect their privacy in the process of contributing data, but also get rewards by participating in model optimization.
FLock Edge Compute: Activating the Potential of Private Domain Data
FLock firmly believes that private domain data holds great potential for the development of artificial intelligence. In order to fully unleash this potential, FLock has made disruptive innovations in AI data calls and model training methods, extending training tentacles to the field of private domain data.
By integrating edge computing and federated learning technologies, FLock has opened up a new model training and data processing path, deploying training tasks directly to users' edge devices, such as mobile phones, tablets, and laptops. At the same time, with the support of distributed computing power, the model can obtain encrypted model parameters from each edge device and integrate them, thereby avoiding direct leakage of original data.
Compared with the traditional centralized model training method, edge computing can make full use of high-quality private domain data, improve the accuracy and richness of the model, significantly reduce the delay of training data, and make the model closer to real time and better meet user needs. At the same time, the flexible use of distributed computing power also makes model training more efficient and flexible, providing AI engineers with unlimited creative space and possibilities.
Suppose a retail company wants to improve the accuracy of its personalized recommendation system. In the past, the company may need to upload the user's shopping data to the central server for model training. However, through FLock's edge computing and federated learning technology, the company can directly train the model on the user's mobile phone, use the mobile phone as an edge device, encrypt and process the user's shopping data, and upload the encrypted model parameters to the server, while protecting the user's privacy, improving the efficiency and accuracy of model training.
JIT: A new way to innovate data calls
FLock also revolutionizes the data calling process of AI, from advance calling (AOT) to point-to-point just-in-time calling (JIT). JIT compilation is a way of compiling code while the program is running, which is dynamic, efficient and has strong optimization capabilities. Through blockchain-native JIT data calling, AI engineers can access original domain data in real time to improve training, while ordinary users can get rewards by contributing data while still retaining ownership.
Application Scenario
FLock relies on its unique technical advantages in model training. It not only supports efficient training while retaining data locally, but has also been successfully applied to multiple cutting-edge fields.
In the healthcare sector, FLock has worked with University College London Hospital to deploy a blood sugar prediction algorithm to provide more accurate health management solutions for diabetic patients. At the same time, it has worked with Moorfields Eye Hospital to optimize the ophthalmology detection algorithm and use decentralized computing power to improve the early diagnosis rate of ophthalmology diseases. In addition, applications such as Request Finance's on-chain credit scoring system and Morpheus and Ritual's trading robots have also achieved rapid model iteration and performance optimization through the FLock platform.
To simplify the model training process, FLock provides a one-click deployment template, which allows developers to easily train and verify models on distributed computing platforms such as Akash. This innovation not only lowers the training threshold, but also attracts a large number of Akash community members to participate in FLock mining and contribute computing resources together.
In addition to the healthcare field, FLock's technology is also widely used in other scenarios. For example, IO.net uses PoAI to achieve node consensus and trains anime image generation models based on this, bringing users more vivid and realistic visual effects. At the same time, FLock has also trained a Move language programming assistant specifically for developers for Aptos, further promoting the popularization and application of blockchain technology. In addition, Animoca Brands and its products also use FLock to train internal dedicated models, covering due diligence, investment analysis, market making, and operational support. Many fields, give full play to the advantages of privacy-preserving AI, and help business innovation and efficiency improvement.
Ecosystem Construction
FLock is committed to building an open and inclusive AI ecosystem to attract all kinds of users to participate. Through a decentralized training platform and incentive mechanism, FLock encourages developers, data contributors, and ordinary users to actively participate in the training and optimization of AI models.
For developers, FLock provides convenient and efficient platforms and tools to support model development and training. Developers can use FLock's edge computing power and federated learning technology for training and optimization, while also obtaining rich training data and model feedback resources. In addition, FLock also provides community resources and support services to help developers solve problems and challenges during the development process.
For Holders, they can support project development by holding FLock tokens or participating in community governance and receive corresponding returns. FLock's token economic model has been carefully designed to motivate participants while ensuring system security and decentralized governance. Holders can participate in project decision-making through voting and governance, while enjoying benefits such as token dividends.
Ordinary users can also participate and earn income through FLock AI Arena. They can choose to become a Delegator and entrust their FML to outstanding Training Nodes and Validators. This will not only help excellent models stand out faster, but also enjoy a rate of return close to that of managed nodes. If you accurately select potential nodes, the income will also be considerable, and there is no need to train the model yourself or bear the hardware costs.
Latest developments and future prospects
According to official news, the Web3 AI project FLock.io announced that it will launch the token $FLock on Bybit on December 31 and go online for a week-long launchpool. $FLock is the native token of the FLock ecosystem and the core of the entire AI training platform, providing support for its AI training platform AI Arena, collaborative node network FL Alliance, and AI Marketplace.
Currently, the token economics of FLock has been announced, and the total supply cap is set at 1 billion $FLock.
According to the token economic model, 47% of the tokens will be allocated to the community to incentivize all users who contribute to FLock, including those who participate in AI Arena, FL Alliance, AI Marketplace, and future Flock ecosystems. In order to recognize active participants in the testnet phase, $FLock will be airdropped when the mainnet is launched, and future incentives will be systematically minted within 60 months and will decrease by 1% per month.
Development prospects
FLock continues to optimize community building and incentive mechanisms to attract more users to actively participate in ecological development. With the continuous growth of user scale and the gradual improvement of the ecological system, FLock is expected to occupy a place in the field of decentralized AI and promote further prosperity of the ecology.
With its unique advantages and forward-looking ecological layout, FLock has already made its mark in the AI infrastructure track. It combines cutting-edge technologies such as edge computing and federated learning with the concept of decentralization, providing innovative solutions for AI model training and processing. At the same time, FLock has created diversified application scenarios and a complete incentive mechanism, which is bound to attract a large number of users to participate in ecological construction and development.
In the future, FLock will continue to focus on developing more practical and easy-to-use products around retail and consumer user groups. These products will fully consider user needs and usage habits, lower the threshold for participating in AI model training and data contribution, attract more users and developers to join the construction of the AI ecosystem, and promote the innovation and application of AI models and technologies.
