Oppenheimer Moment of AI: How IO.NET is Revolutionizing the Decentralized AI Computing Market

Authors: Raghav Agarwal, Roy Lu, LongHash Ventures

Compiled by: Elvin, ChainCatcher

AI

Humanity is at an AI Oppenheimer moment.

Elon Musk pointed out, "As we advance our technology, it is crucial to ensure that AI serves the interests of the people, not just those in power. AI owned by the people provides a path forward."

At the intersection with cryptocurrency, AI can achieve its democratization. Starting with open-source models, then moving to people's AI, serving the people, by the people. While the goal of Web3 x AI is noble, its actual adoption depends on its availability and compatibility with the existing AI software stack. This is where IO.NET's unique approach and tech stack come into play.

IO.NET's decentralized Ray framework is a Trojan horse for launching a permissionless AI computing market to web3 and beyond.

IO.NET is leading the way in bringing GPU richness. Unlike other general-purpose computing aggregators, IO.NET bridges decentralized computing with industry-leading AI stacks by rewriting the Ray framework. This approach paves the way for broader adoption both within and outside of web3.

The Race for Computing Power in the Context of AI Nationalism

Competition for resources in the AI stack is intensifying. Over the past few years, a plethora of AI models have emerged. Within hours of the release of Llama 3, Mistral and OpenAI released new versions of their cutting-edge AI models.

The three levels of resource competition in the AI stack are: 1) training data, 2) advanced algorithms, 3) computing units. Computing power allows AI models to improve performance by scaling training data and model size. According to OpenAI's empirical research on transformer-based language models, performance improves steadily as we increase the amount of computation used for training.

Over the past 20 years, the usage of computing has surged. An analysis by Epoch.ai of 140 models shows that the training computation for milestone systems has increased by 4.2 times per year since 2010. The latest OpenAI model, GPT-4, requires 66 times the computation of GPT-3, approximately 1.2 million times that of GPT.

AI Nationalism is Evident

Massive investments from the U.S., China, and other countries total around $40 billion. Most of the funding will focus on producing GPU and AI chip factories. OpenAI CEO Sam Altman plans to raise up to $7 trillion to enhance global AI chip manufacturing, emphasizing that "computing will become the currency of the future."

Aggregating long-tail computing resources could significantly disrupt the market. Centralized cloud service providers like AWS, Azure, and GCP face challenges including long wait times, limited GPU flexibility, and cumbersome long-term contracts, particularly difficult for smaller entities and startups.

Underutilized hardware from data centers, cryptocurrency miners, and consumer-grade GPUs can meet the demand. A 2022 study by DeepMind found that training smaller models on more data is often more efficient than using the latest and most powerful GPUs, indicating a shift towards more effective AI training using accessible GPUs.

IO.NET Structurally Disrupts the AI Computing Market

IO.NET structurally disrupts the global AI computing market. IO.NET's end-to-end platform for globally distributed AI training, inference, and fine-tuning aggregates long-tail GPUs to unlock low-cost high-performance training.

GPU Market:

IO.NET aggregates GPUs from data centers, miners, and consumers worldwide. AI startups can deploy decentralized GPU clusters in minutes by simply specifying the cluster location, hardware type, and machine learning stack (TensorFlow, PyTorch, Kubernetes), and immediately pay on Solana.

Clusters:

GPUs without compatible parallel infrastructure are like reactors without power lines; they exist but are unusable. As emphasized in the OpenAI blog, the limitations of hardware and algorithm parallelism significantly affect the computational efficiency of each model, constraining model size and utility during training.

IO.NET leverages the Ray framework to transform thousands of GPU clusters into a unified whole. This innovation allows IO.NET to assemble GPU clusters without being affected by geographical dispersion, addressing a major challenge in the computing market.

Ray Framework Stands Out

As an open-source unified computing framework, Ray simplifies the scaling of AI and Python workloads. Ray has been adopted by industry leaders such as Uber, Spotify, LinkedIn, and Netflix, facilitating the integration of AI into their products and services. Microsoft offers customers the opportunity to deploy Ray on Azure, while Google Kubernetes Engine (GKE) simplifies the deployment of open-source machine learning software by supporting Kubeflow and Ray.

Ahmad showcased his work on the decentralized Ray framework at the 2023 Ray Summit

Decentralized Ray - Scaling Ray for Global Inference (video link: https://youtu.be/ie-EAlGfTHA?)

We first met him when Tory was the COO of a high-growth fintech startup, and we knew he was a senior operator with decades of experience capable of scaling startups to productive levels. After talking with Ahmad and Tory, we immediately realized this was the dream team to bring decentralized AI computing to web3 and beyond.

Ahmad's brainchild, IO.NET, was born from a moment of insight in practical application. Developing Dark Tick, an algorithm for ultra-low latency high-frequency trading, required substantial GPU resources. To address cost issues, Ahmad developed a decentralized version of the Ray framework that clusters GPUs from cryptocurrency miners, inadvertently creating a resilient infrastructure to tackle broader AI computing challenges.

Momentum:

By leveraging token incentives, IO.NET has onboarded over 100,000 GPUs and 20,000 cluster-ready GPUs by mid-2024, including a significant number of NVIDIA H100 and A100. Krea.ai is already utilizing IO.NET's decentralized cloud service, IO Cloud, to drive their AI model inference. IO.NET recently announced collaborations with several projects, including NavyAI, Synesis One, RapidNode, Ultiverse, Aethir, Flock.io, LeonardoAI, and Synthetic AI.

By relying on a globally distributed GPU network, IO.NET can:

• Reduce inference times for customers by allowing inference closer to their end users compared to centralized cloud service providers
• Improve resilience by connecting multiple data centers through a highly integrated network backbone, organizing resources into regions
• Lower the costs and access times for computing resources
• Allow companies to dynamically scale and leverage resources
• Enable GPU providers to achieve better returns on their hardware investments

IO.NET stands at the forefront of innovation through its decentralized Ray framework. Utilizing Ray Core and Ray Serve, their distributed GPU clusters efficiently orchestrate tasks on decentralized GPUs.

Conclusion

Promoting open-source AI models is a recognition of the original spirit of internet collaboration, where people can access HTTP and SMTP without permission.

The emergence of crowdsourced GPU networks is a natural evolution of the permissionless spirit. By crowdsourcing long-tail GPUs, IO.NET is opening the floodgates to valuable computing resources, creating a fair and transparent market that prevents power from concentrating in the hands of a few.

We believe IO.NET realizes the vision of AI computing as currency through decentralized Ray cluster technology. In a world increasingly composed of "rich" and "poor," IO.NET will ultimately "make the internet open again."