# Proof of Training (PoT): Harnessing Crypto Mining Power for Distributed AI Training

**Authors:** Peihao Li  
**arXiv:** [2307.07066](https://arxiv.org/abs/2307.07066)  
**Date:** 2023-07-13  
**Summary page:** https://www.summarizepaper.com/en/paper/2307.07066v1/

## Key points

- Integration of AI with crypto mining is a prominent trend in technology
- Challenges exist in seamless collaboration between AI and blockchain technology
- Proof-of-training (PoT) protocol introduced to bridge the gap
- PoT protocol uses PBFT consensus mechanism for global state synchronization
- Aims to enhance task throughput, system robustness, and network security
- Implementation of decentralized training network (DTN) to evaluate PoT protocol performance
- PoT protocol shows potential for distributed AI training with efficiency, resilience, and data security
- Source code available under MIT License for review, utilization, and modification by researchers and developers
- Innovative approach addresses challenges in integrating AI with crypto mining and advances decentralized training networks.

## Summary

The integration of artificial intelligence (AI) with crypto mining has become a prominent trend in the rapidly evolving landscape of technology. However, this convergence faces significant challenges that hinder seamless collaboration between these two fields. To address these obstacles and bridge the gap between AI and blockchain technology, a novel approach known as the proof-of-training (PoT) protocol has been introduced. The PoT protocol utilizes the practical Byzantine fault tolerance (PBFT) consensus mechanism to effectively synchronize global states. By combining the strengths of AI and blockchain technology, this protocol aims to enhance task throughput, system robustness, and network security. To evaluate the performance and feasibility of the PoT protocol, an implementation of a decentralized training network (DTN) has been developed. Through rigorous testing and analysis, it has been demonstrated that the PoT protocol exhibits substantial potential for distributed AI training. The results indicate promising outcomes in terms of system efficiency, resilience, and data security. Furthermore, by providing access to the source code under the MIT License, researchers and developers can review, utilize, and modify the implementation of both the PoT protocol and DTN. This innovative approach not only addresses existing challenges in integrating AI with crypto mining but also paves the way for further advancements in decentralized training networks.