The demand for affordable computing power exponentially increases in burgeoning fields such as AIGC, Web3, and the Metaverse. Computing power, often called the “oil” of future human society, is essential. Yet, numerous independent data centers around the world are underutilized.
For example, the US’s average utilization rate hovers around 12% to 18%. This underutilization leads to bottlenecks, which has a cascading effect of escalating prices for GPU computing.
Consequently, several projects are underway, aiming to construct a decentralized cloud network to maximize these untapped compute resources, with a focus on AI business applications. This article highlights some of the most reputable of these projects.
Understanding AI Model Training and Inference
Compute power is predominantly utilized in AI model training and inference. AI model training is the process of employing a machine learning algorithm to build a model using a training dataset. Meanwhile, AI inference involves applying the trained model to a new dataset, generating outputs or “predictions” based on the intelligence acquired during the training phase.
This allows for understanding new data and making real-time decisions. We reviewed each project in this article, specifying whether it focuses on model training, inference, or both.
Akash Network (Founded in 2017 – Training, Inference)
Akash Network, an innovative open-source supercloud network, provides a decentralized marketplace for cloud computing. It enables users to buy and sell computing resources securely and efficiently, offering a “reverse auction” system where customers can submit their desired price, with providers vying for their business. This results in competitive pricing. The network, leveraging technologies such as Kubernetes and Cosmos, employs the Akash Token (AKT) for transactions and governance. It boasts features like limitless storage, dedicated IP addresses, and high levels of privacy and security.
Cortex Labs (Founded in 2017 – Inference)
Cortex is an open-source, peer-to-peer, decentralized blockchain platform that seamlessly integrates AI models. AI developers can upload their models onto the blockchain and incorporate them into smart contracts. Cortex has accomplished significant milestones, such as launching the world’s first AI on the blockchain and enabling AI inference on-chain. It is recognized as an innovative blockchain for its AI integration and machine-learning capabilities. It empowers decentralized applications to utilize on-chain inference and fulfill smart contracts’ objectives.
Bittensor (Founded in 2019 – Inference)
Bittensor, an open-source protocol, powers a decentralized blockchain-based machine-learning network. Reminiscent of Bitcoin’s mining network, it provides censorship-resistant access to a decentralized network of machine learning models. Bittensor’s vision is to establish a pure market for AI models, creating an incentivized environment for the transparent exchange between consumers and producers of these valuable models. Its focus on encouraging AI model participation and fostering markets that create value sets it apart within the industry.
Gensyn (Founded in 2020 – Training)
Gensyn is a machine learning computing network aiming to unite the world’s computational resources into a global supercluster accessible to everyone. It enables developers to train deep learning models over a network of interconnected devices, utilizing idle computing power in a verifiable way. The network employs blockchain technology to validate the completion of deep learning tasks, initiating token-based payments. Gensyn aims to democratize compute power with decentralized technology, providing AI developers with a cost-effective, scalable, and environmentally conscious solution.
OORT (Founded in 2021 – Training, Inference)
OORT is a pioneering decentralized data cloud designed to optimize privacy and cost-efficiency by integrating global computing and storage resources. It offers enterprise-grade decentralized cloud-based solutions tailored for generative AI and data-driven businesses. With a substantial network of over 29,000 nodes across more than 100 countries, it boasts prestigious clients and partnerships, including DELL, Lenovo Image, BNB Chain, Tencent Cloud, and Cardano. OORT’s ‘Talk to Data’ (TDS) service has dramatically enhanced Lenovo Image’s customer support, boosting customer satisfaction rates from 30% to over 90%. OORT is at the forefront of transitioning traditional non-crypto users to blockchain-based decentralized cloud solutions.
Io.net (Founded in 2022 – Training, Inference)
Io.net has established a decentralized physical infrastructure network to source GPU computing power for AI and machine learning applications. It streamlines infrastructure deployment by leveraging an open-source library used by OpenAI to distribute ChatGPT training across over 300,000 CPUs and GPUs. Furthermore, it utilizes Solana’s blockchain for transactions, rewarding machine learning engineers and miners for contributing computing power. Unlike centralized services like AWS, Io.net offers a more adaptable approach, analogous to Kayak in the airline industry, by facilitating the booking of compute resources.
Article by Max (Chong) Li – Follow on Twitter or LinkedIn. Faculty at Columbia University; CEO of Web3 Data Cloud “Oort”
