If we trace the origins of the world's most successful technology companies, we will find an interesting pattern: almost all of the top technology companies were born in an academic environment. Google was born at Stanford University, and Larry Page and Sergey Brin developed a search engine while pursuing their doctoral degrees. Facebook was founded by Mark Zuckerberg in a dormitory at Harvard University. It was originally intended to connect classmates, but later developed into a giant that changed social media. Many other innovations, such as VMware and today's "first AI chip stock" Nvidia, can also be traced back to university laboratories, lecture halls, or student dormitories.
The business that OORT is focusing on today is no exception. The decentralized AI platform that the team is building actually has a deep academic foundation. Its starting point is a course I taught at Columbia University in 2018 - "AI Reinforcement Learning". This course gave birth to a forward-looking idea, which is now gaining more and more attention.
Classroom Challenges
The final project for this course requires students to train an AI agent. For non-technical readers, this process involves teaching an AI model to learn and make decisions through data. This is similar to providing structured training to a digital system, allowing it to improve its performance through input, adjustment of responses, and continuous iteration.
However, training AI agents is extremely resource-intensive. It requires powerful computing power and large amounts of storage space to process data. Traditional cloud services, such as those provided by Amazon and Google, are expensive, making them unaffordable for most students. Although students demonstrated creativity and technical skills, the required infrastructure was beyond their capabilities. This raised a key question: Is there a way to bypass expensive centralized cloud services and develop a more affordable and accessible solution to give students the chance to complete their final assignments?
Prototype of decentralized solution
So, we began to explore how to use blockchain as an incentive layer to create a decentralized cloud solution for AI development services to help students complete their final projects more realistically.
We have taken the following measures:
Utilize idle resources around the world: The platform utilizes underutilized resources around the world, such as idle hard drives in offices and excess bandwidth on personal computers;
Based on blockchain: Blockchain technology provides a transparent and secure network for the integration of these distributed resources;
Cryptocurrency Payments: The platform uses cryptocurrencies for micropayments because the current financial system does not support instant micropayments on a global scale.
In simple terms, this platform is like the “Airbnb” of infrastructure. Just as Airbnb allows homeowners to rent out their spare rooms, this platform allows individuals to contribute their spare storage or computing power, significantly reducing costs.
This decentralized experiment developed for Columbia University students in 2018 became the prototype for today’s decentralized AI (DeAI) and decentralized physical infrastructure network (DePIN) concepts. In essence, DePIN is the infrastructure that supports the effective operation of DeAI systems, while DeAI is the application layer that utilizes decentralized infrastructure.
DePIN focuses on the physical layer - the infrastructure of the decentralized ecosystem. It includes a globally distributed network of storage, computing, and bandwidth, connected by blockchain technology. Think of DePIN as the "foundation" and "plumbing" of the decentralized ecosystem, making this ecosystem possible.
DeAI is built on top of this infrastructure. It leverages these decentralized resources to enable AI development and deployment in a distributed manner, rather than relying on a single tech giant to train and run AI models. DeAI provides a more affordable, scalable, and fairer way to access AI resources through DePIN.
Here are some common advantages of decentralized solutions:
Reduce AI training and deployment costs
Enhance data transparency, privacy and security
Building more global, diverse, and unbiased datasets
Improve disaster recovery and business continuity
Uncertainty and possibility
As we discussed in previous articles, the recent rise of decentralized AI has addressed many of the concerns people have about centralized AI. It is widely believed that with the help of blockchain technology, AI can be truly open source and transparent.
Of course, this process is not easy. Building decentralized infrastructure brings many unprecedented technical challenges. From optimizing network reliability to ensuring data security in distributed systems, we are solving the problems step by step. In addition, as speculative capital pours into this field and the AI competition between the United States and China intensifies, it is expected that by 2025, AI projects will mushroom and failure cases will surge.
Still, the potential of DeAI is promising. It paints a picture of a world where access to AI tools is not limited by geography or economics, and is both realistic and technically feasible. This means that a student in New York, a teacher in Buenos Aires, or a small business in Nairobi could train an AI model or store data with the same affordability and ease as a large company.
The journey from a Columbia University classroom to decentralized AI has been filled with unexpected opportunities, challenges, and rewards. What started as a solution to help students complete their projects has evolved into an effort to rethink infrastructure, accessibility, and innovation.
As more and more professionals from different fields realize the potential of decentralized AI (DeAI) to transform various industries, the interest and momentum in this field will continue to grow. 2025 is expected to be a key year for the development of DeAI. The combination of blockchain and AI will gradually go mainstream, laying the foundation for future technological advancement and bringing far-reaching impacts.
Author: Dr. Chong Li, Founder of OORT and Professor of Columbia University
Originally published in Forbes: https://www.forbes.com/sites/digital-assets/2024/12/27/how-a-columbia-class-project-became-decentralized-ai-solutions/
