About OORT
OORT is based on the Olympus protocol, creating a trusted decentralized infrastructure for AI. OORT provides three major decentralized AI products for businesses and individual clients: OORT Storage, OORT Data Center (B2C, B2B), and the upcoming OORT Compute. OORT has currently secured $10 million in funding from investors including Taisu Venture, Red Beard Venture, Sanctor Capital, as well as support from Google and Microsoft.
Host bms:
Today we are pleased to invite Dr. Li, the founder and CEO of OORT, to discuss the future of decentralized AI. Hello, Dr. Li! First, could you introduce your background and experiences? How did you enter the fields of AI and blockchain?
Dr. Li:
Hello, first of all, thank you for the invitation. I am very pleased to have this opportunity to share some of my experiences, especially regarding the research and practice of decentralized AI. I am currently the founder and CEO of the OORT project, which started in the summer of 2021 and has been going on for nearly four years now. At the same time, I teach at Columbia University in New York, belonging to the Department of Electrical Engineering.
Academically, my research direction mainly focuses on AI and blockchain technology, with reinforcement learning being a subfield of AI. I have written an English textbook on reinforcement learning, which is being used by some universities. Tsinghua University’s Computer Science Department even translated it into Chinese, and the Chinese version was officially published in China about two or three years ago.
Host bms:
How did you first get in touch with blockchain, and what was the opportunity?
Dr. Li:
In fact, I officially began to delve into blockchain in 2017. At that time, the ICO craze spread globally, and students showed a surge in interest in cryptocurrencies and blockchain technology. Due to their demand, I attempted to offer blockchain-related courses at Columbia University. There were almost no teachers with such experience, so I had to gather scattered materials, study white papers and academic papers, and then organize a systematic curriculum, develop assignments and exam questions, so that students would not only understand the superficial concepts but could truly become blockchain technology experts capable of driving the industry in the future from the mathematical and engineering fundamentals.
Host bms:
In the absence of mature teaching materials and reference cases at that time, how did you address these teaching and research challenges? Did you encounter any particularly interesting or difficult parts?
Dr. Li:
Yes, at that time, the biggest difficulty was the scattered information and the lack of systematic teaching materials. I spent a lot of time searching for information online, combining white papers and limited academic papers, and delving into the underlying cryptography of blockchain. For example, why use SHA-256 instead of SHA-512 or SHA-128? The underlying theories involve mathematical concepts like the birthday paradox. These contents need to be organized into suitable teaching materials for engineering master's and doctoral students—they will become true technical experts in the future, not just beginners who know concepts.
Host bms:
You originally focused on reinforcement learning and AI, what prompted you to expand from the field of AI to the world of blockchain and further develop the idea of combining the two?
Dr. Li:
Before this, I worked in Qualcomm's R&D department on 5G chip design and began researching underlying algorithms as early as 2012. After entering the academic circle, I mainly focused on the field of AI reinforcement learning. However, due to chance, I came into contact with and studied blockchain technology and started teaching it. In exploring the intersection of blockchain and AI, I realized that combining the two has enormous potential: decentralized networks provide new infrastructure for AI's data and computing power, while AI technology can optimize the operation and efficiency of blockchain. This cross-disciplinary research pushed me to further contemplate the possibilities and practical ways of decentralized AI.
Host bms:
Recently, Vitalik talked about the drawbacks of centralized AI in an interview. What do you think are the essential differences in technical superiority and commercialization value between decentralized AI and centralized AI?
Dr. Li:
That's a great question. I believe the biggest technical advantage of decentralized AI lies in achieving transparency and traceability of the entire AI development process through blockchain technology.
Currently, in the entire AI field, openness and transparency have become hot topics. For example, OpenAI is now criticized by many for no longer being 'Open' because we cannot know what data they used for training, whether there is bias in the model, the details of the model training, and whether the backend intervenes in the output results. Such questions are difficult to answer in centralized AI.
Decentralized AI leverages blockchain technology to ensure key aspects of the training process are public and verifiable, thereby building trust. With transparency and traceability, users find it easier to trust AI's decisions and data sources. The three flagship products that OORT is about to launch are fundamentally aimed at enhancing AI's transparency and traceability through blockchain.
In terms of commercial application value, decentralized AI allows more people to participate in building and using AI together. Just like TikTok or Xiaohongshu, by aggregating the wisdom and contributions of individuals, a virtuous cycle is formed. 'Made by people, for people'—when large numbers of individuals and organizations contribute data, computing power, and resources to build AI, the final results serve everyone. This model holds immense commercial potential, not limited to the traditional centralized AI's B2B or B2C models. Decentralization allows more people to participate, making AI a truly infrastructure created by the masses and serving the masses, thereby nurturing greater commercial value.
Host bms:
Was the OORT project born in such a context? Can you introduce us to the OORT project?
Dr. Li:
Yes, the OORT project was born relatively early. Back in 2017, when I was teaching reinforcement learning courses at Columbia University, students needed to train their AI agents for their final projects, which required a lot of computing power. The cloud services commonly used by students (like AWS and Google Cloud) were very expensive, and many students complained about the heavy financial burden.
At that time, I thought: With so many idle computing resources around the world, why not use blockchain technology to integrate these distributed resources into a decentralized computing network, reducing costs while increasing flexibility?
In 2017, there was no clear concept of 'decentralized AI,' but integrating distributed computing resources was already a key step in building decentralized AI infrastructure. We published academic papers and applied for core patents in the US during our research process in the laboratory. After the patents were officially authorized in the summer of 2021, we decided to commercialize them, thus founding OORT. The company was initially named Computecoin but was later rebranded to OORT.
Host bms:
When building decentralized AI infrastructure, what do you think is the biggest challenge in terms of technology, such as computing power and data privacy storage? How is OORT responding to this?
Dr. Li:
The biggest challenge is not just privacy or data security, but how to ensure that decentralized infrastructure's performance is close to that of centralized cloud service providers.
When companies consider migrating to decentralized infrastructure, what they often care about most are performance metrics such as reliability, availability, and latency. If the performance of decentralized cloud cannot compare with traditional centralized clouds like AWS and Google Cloud, then even if decentralization has many advantages in terms of price, privacy protection, and resilience, it would be hard to make customers pay.
Taking OORT's decentralized storage as an example, we need to encrypt and distribute file shards across hundreds or even thousands of nodes. When users need to read a file, we must retrieve the fragments in parallel from multiple nodes and then decode and decrypt them.
To keep latency close to that of centralized service providers, we adopted advanced coding theory and algorithm optimization at the foundational level. In communication and storage design, we borrowed ideas from similar 5G foundational algorithms to ensure that even if some nodes go down or offline, we can still quickly reorganize the data, ultimately achieving nearly equivalent access latency to traditional cloud storage. We can control latency to around 100 milliseconds, at a cost about 50% to 60% lower than Amazon S3.
In summary, the biggest challenge is to ensure that the performance of decentralized infrastructure is comparable to that of centralized cloud services while retaining the various advantages of decentralization. This is the problem that OORT has spent the most effort to tackle.
Host bms:
Currently, the technological competition between China and the US in the AI field is receiving a lot of attention. How do you view the competition between China and the US in AI technology? Does decentralized AI have greater policy dividends and development space in this context? How should we avoid the risks that geopolitical issues may bring?
Dr. Li:
Looking globally, the significant investments in the AI field come primarily from China and the US. The US has an advantage in original technology and foundational research; from the internet to mobile internet to AI, the starting point of technological waves is often in the US. In contrast, China's advantage lies in its speed of application and implementation, with the market and application environment enabling new technologies to be commercialized at a rapid pace.
Cooperation between China and the US should have been a win-win situation. The US provides original technology, while China is responsible for rapid implementation and large-scale commercialization. However, geopolitical issues and trade wars have made this cooperation model difficult.
In this international environment, decentralized AI actually has strong resilience and vitality. On the policy level, both China and the US are directly or indirectly supporting related technologies. For example, China is vigorously promoting distributed computing pools, while the US emphasizes decentralized storage and computing from a national security perspective, thereby indirectly promoting the development of decentralized technologies.
The advantage of decentralized AI lies in its global characteristics, not relying on the policies of a single country or region. When individuals and organizations from all over the world build decentralized AI in an open and collaborative manner, this system is not easily subject to complete restrictions from geopolitical factors, just as the open-source Linux system cannot be entirely stifled by a single government.
Therefore, I believe the development space for decentralized AI is very large, not only because it aligns with the logic of technological evolution but also because it receives wide participation and support globally, making it less susceptible to the decisive influences of geopolitical fluctuations.
Host bms:
You have provided us with a new perspective, letting us realize that decentralization is not only a technical aspect but also a global mindset.
Host bms:
Recently, CZ also participated in the heated discussion about AI and blockchain. He mentioned that AI labeling or broader AI data is very suitable for being conducted on the blockchain, leveraging global cheap labor and enabling instant payments through cryptocurrency. In this context, how do you think OORT will further promote the realization of these concepts in the current decentralized AI ecosystem?
Dr. Li:
This question is very good, and it's also an excellent opportunity for us. As I mentioned earlier, OORT has three core products: OORT DataHub, OORT Storage, and OORT Compute, corresponding to data collection and labeling, data storage, and data computing. These three major links are precisely the core infrastructure for building AI models.
The OORT DataHub, which will launch in December, is the world's first decentralized AI data collection and labeling platform. It aligns closely with the philosophy proposed by CZ: through blockchain technology, we can leverage participants from around the world to collect and label data, and pay them using cryptocurrency. This not only eliminates geographic bias but also allows us to obtain high-quality data at lower costs and greater efficiency.
Additionally, OORT is a storage provider for BNB Greenfield, and our data collection processes are recorded on the blockchain to ensure data transparency and immutability. This data will be automatically stored on OORT Storage or BNB Greenfield and provided to AI companies for training or fine-tuning models. Thus, the data will not be as easily deleted or manipulated as it would on centralized servers.
On the payment side, we will use OORT's native tokens for incentives and rewards. Our target markets include Africa, Latin America, and Southeast Asia, where users have a high acceptance of crypto and the promotion costs are relatively low.
It is worth mentioning that we have already conducted practical experiments in the early stages. This summer, we decentralized the labeling of a large number of Mars surface images collected by the Curiosity rover from 2011 to 2012 through the community. This has provided us with valuable experience and proven that our solutions are not just theoretical concepts but have mature precedents.
In the future, we will continue to consolidate this first-mover advantage, allowing more users to directly use OORT's products, personally experience, and participate in building the future of decentralized AI. I believe this is not only our mission but also a crucial step for users to truly touch and integrate into this global transformation.
Host bms:
I have read several of your interviews, and you are an experienced engineer, professor, and inventor with over 200 patents and multiple academic research achievements published in IEEE journals. Which of these research experiences has been most beneficial for the current development of OORT? Or how have you successfully commercialized these academic achievements?
Dr. Li:
In fact, some of the patents we have are currently the core technologies supporting us. We have two core patents being used in OORT's products, one of which is about a consensus mechanism of honesty, namely POH (Proof of Honesty). This is a very core mechanism, and we have partially applied it in OORT's products.
However, it should be noted that there is often a significant gap between papers or patents and the actual realization of products. We need to gradually transform the theoretical ideas in patents or papers into functional product capabilities.
For example, OORT is about to launch the world's first decentralized data collection and labeling platform, OORT DataHub. In this platform, controlling data quality is crucial, and at the foundational level, we are using this POH (Proof of Honesty) mechanism. Currently, we may only apply about 30% to 40% of the patent content in our products, with the remainder needing significant development and refinement. This is the necessary path from theory to practice—few patents or papers can be directly and seamlessly transformed into products; continuous iteration and improvement are required.
However, OORT already has a clear roadmap that will gradually fully integrate the entire POH mechanism into our products. Ultimately, we hope to empower the development of decentralized AI with a high-quality technological foundation, which is precisely the direction we are currently striving for.
Host bms:
Okay, thank you very much, Dr. Li, for sharing your wonderful insights on decentralized AI with us today. Thank you for your time and sharing!
References:
Forbes: https://www.forbes.com/sites/maxli/
Past interviews:
https://cryptoslate.com/podcasts/max-li-emphasizes-blockchains-role-in-ai-trust-and-ethics-revolution/
https://www.youtube.com/watch?v=59UIxaWl46I
Recent news:
https://www.coinspeaker.com/oort-githon-technology-seal-3-year-deal-customer-satisfaction-ai-agent/
Official website: https://www.oortech.com/
Official foundation: https://www.oortfoundation.org/
