SPACE - Distributed AI Network
Problem: exponential increase in the cost of training and maintaining LLMs. Solution: creating a decentralized AI network, based on blockchain, using resources of user GPUs, Mining Machines, Phones - to train and maintain LLMs of IT and AI giants
Artificial Intelligence And Digital Services
Artificial intelligence
The practice will be founded in the fourth quarter of 2024 with the establishment of an RnD team consisting of scientists, cryptographers, mathematicians, software engineers and blockchain developers. In the second and third quarters of 2025, the practice will test the blockchain on the GAUSS test network. In 2026, the practice will complete beta testing of the system and client architecture (Windows and Linux). In 2027, distributed learning of the selected LLM will be completed, support for PyTorch framework will be added, 2 master nodes and more than 500 test compute nodes will be installed.
An international project that will be implemented worldwide, attracting the maximum number of users' devices
1. Endless scaling possibilities: Creating a cluster with computing power capable of supporting the training of models with a huge number of parameters. This will overcome the limitations of centralized scaling and use millions of GPUs. 2. Privacy and Security: Data storage, transmission and processing are carried out using asymmetric encryption. In addition, the system provides monitoring and certification of the qualification of computing nodes. 3. High availability: The decentralized architecture avoids the problems associated with single-point failures. 4. Cost savings and speed improvements: Using a distributed cluster with 4070Ti graphics cards instead of a centralized cluster with A100 allows you to reduce costs by more than 40% and increase the learning rate by more than 30%. 5. Convenient monetization: Thanks to the distributed architecture and the pay-as-you-go monetization system, you will be able to effectively use your resources and earn income from their use. 6. Proof-of-Activity: The hybrid Proof-of-activity algorithm, which also uses Proof-of-Work and Proof-of-Stake, will allow you to use the maximum number of user devices: from phones to servers of stake holders.
Sorokin David. 3 years of executive management, 3 years in creating startups. Own projects: “SPACE” - decentralized AI network, Antibacterial door handle. Winner of the all-Russian contest “Tvoi Khod” ($10,000 grant), entered the top 30 technology startups in Russia according to RosAtom State Corporation, organized cooperation on his startup with the government organization “National Technology Initiative”, finalist of the national award “Business Success”, participant of the show. Grant winner and winner of the federal program “I'm in Business”, won a $5,000 grant to create a school co-working space. Studied at the British Academy of Business on the MINI-MBA in London.
N.A.
