Proof of Useful Intelligence (Poui): Blockchain Consensus Beyond Energy Waste
Zan‐Kai Chong, Hiroyuki Ohsaki, Bryan Ng · 2025
Blockchain technology anchors decentralized systems by enabling secure, transparent data management across distributed networks, powering a wide range of applications from foundational cryptocurrencies like Bitcoin to the recently emerging tokenization of real-world assets (RWAs), such as property and commodities. However, its scalability and environmental sustainability depend on consensus mechanisms that maintain network integrity without imposing excessive computational or energy burdens. Proof of Work (PoW), a prevalent mechanism seen in Bitcoin, relies on miners performing energy-intensive cryptographic computations to ensure robust security, yet driving significant resource demands. In contrast, Proof of Stake (PoS) selects validators based on the amount of cryptocurrency they stake, as exemplified by Ethereum post-Merge, providing a markedly more energy-efficient option than PoW. While PoW excels in delivering decentralized security through computational effort, it does so at the cost of high energy consumption; PoS, meanwhile, enhances participation accessibility and reduces resource use but introduces potential centralization risks due to wealth concentration among larger stakers. The rapid rise of artificial intelligence (AI) models, with their substantial energy consumption, underscores a growing strain on computational resources. Hence, it inspires us to propose a new consensus mechanism, namely, Proof of Useful Intelligence (PoUI). PoUI is a hybrid consensus mechanism where workers execute AI-based tasks, such as natural language processing or image analysis, to earn coins, which are then staked to secure the network, seamlessly integrating security with real-world utility. This system leverages decentralized functional nodes, i.e., job posters who submit tasks, market coordinators who oversee jobs distribution, workers who perform computations, and validators who ensure accuracy, all orchestrated by smart contracts for task execution and reward allocation. Our energy analysis benchmarks PoW at$\mathbf{3. 5 1 ~ k W h} / \mathbf{m i n e r}$, PoS at$\mathbf{0. 1 ~ k W h} /$validator, and PoUI at$\mathbf{0. 6} \text{kWh} /$worker - yielding a 97 % energy reduction from PoW while adding value. Simulations further demonstrate that PoUI's dynamic reward adjustment regulates worker participation in the job market, which subsequently encourages a sufficient number of validators in the network.