Artificial intelligence is advancing at an unprecedented rate, achieving new milestones almost daily. From generative AI to self-supervised learning, AI models are becoming larger, more powerful, and achieving superhuman levels of performance on increasingly complex tasks. However, fueling this AI revolution requires massive and growing amounts of computing power and data storage that are beginning to outstrip available infrastructure.
Research models today require exponentially more resources than just a few years ago. DeepMind’s protein folding model released late last year reportedly took over 1.6 million GPU hours to train. As models continue scaling in size and complexity to tackle more tasks, compute and storage needs will swell further still.
The consequences of this “AI compute crisis” are huge. Without a rapid expansion of supply, progress could grind to a halt, further centralizing the sector under a few mega-corporations that control infrastructure. Smaller players would be frozen out due to resource constraints.
Fortunately, a new class of decentralized protocols collectively termed DeVIN (Decentralized Virtual Infrastructure) aims to alleviate resource scarcity and democratize access. By leveraging globally distributed unused computing and data resources, projects like io.net, Gensyn and Grass are building decentralized resource markets optimized for AI.
Individuals and organizations can contribute spare GPUs, CPUs, bandwidth and storage to these networks in exchange for tokens or fiat. This dramatically multiplies available capacity by tapping into billions of dollars worth of idle hardware. Researchers pay a fraction of centralized clouds’ fees to access these resources.
Aside from magnitude-scaling supply, DeVIN brings additional benefits for distributed AI. Immune to downtime or censorship, datasets and computations can be replicated redundantly across thousands of nodes worldwide, improving uptime and security versus centralized providers.
The market-based incentive models of projects like Gensyn help ensure work is completed accurately too, building trust in outsourced computations. Transparency from integrating with blockchains also builds confidence in accounting and finances throughout these new ecosystems.
While compute solutions like io.net and Gensyn are laying foundations, data is rapidly becoming AI’s biggest bottleneck, necessitating storage-focused alternatives. Leaders in decentralized storage include Filecoin and Arweave. Filecoin has amassed over $200 million in capital against its vision to empower global data markets through an IPFS-integrated network.
Meanwhile Arweave, currently in beta, pursues a permanent web archive securing information across centuries through its native AR token. With estimations the world will store over 175 zettabytes of digital assets by 2025, innovative storage protocols are critical to keep pace with informational growth fueling AI.
That said, DeVIN platforms launching solutions remain early-stage works-in-progress facing major hurdles around speed, usability, and mainstream adoption before comparably rivaling major clouds. And centralized infrastructure providers like Google are actively pursuing hybrid approaches combining on-premise hardware with proprietary software and data ownership models.
Overall though, the demand and enabling technology for blockchain-powered decentralized infrastructure seems poised to transform how organizations access compute and storage resources globally at massive scales. If protocols can gain meaningful user traction in the coming years, they may help shift dynamics of an industry otherwise forecast to centralize further under tech giants. Only time will tell which solutions gain product-market fit, but crypto and AI’s merging again shows potential for deep sectoral disruption.
ZMSEND.com is a technology consultancy firm for design and custom code projects, with fixed monthly plans and 24/7 worldwide support.
