Matrix AI Network 2.0 — A Chat with AI Architect Steve Deng (2021)
With the crypto market heating up this year, it seemed like a good time to check back in with the Matrix AI Network team. Since I wrote a 2019 Year in Review article about Matrix, Professor Steve Deng, Chief AI Architect for the project, was open to answering some questions from me.
Some will remember that Matrix AI Network first appeared on the global crypto stage in 2017 at the end of the last great bull run with an ICO. Their design was to reengineer Ethereum into a hybrid PoW/PoS network that would dedicate most of the network’s computing resources into AI processing. The vision was to democratize AI by offering cheaper and more secure AI services as compared to Google, Amazon, and Tencent.
Within a couple years, Matrix had successfully built a stable blockchain and was preparing to roll out an AI platform and offer services to companies in China when Covid-19 hit. The global pandemic triggered a massive shutdown in China and exacerbated a crypto bear market that drastically depressed prices for all crypto tokens. Many crypto projects struggled during this period and Matrix was no different. Work slowed to a crawl and some team members left the project.
Fast forward to early 2021 and the crypto market is humming once again. I recently checked back in with some members of the team and with the English Telegram group and it appears that Matrix AI Network has emerged more dedicated than ever to realizing their goals. The team has been reorganized, a new roadmap issued, a new website launched, and communication stepped up. In addition, they are working on new projects including NFTs for AI created art and data / algorithm ownership, and AutoML that allows novices to train AI algorithms.
Most importantly, professor Deng is still actively involved in the project. Deng is commonly viewed as one of the foremost experts in AI in China and he has an impressive resume to show for it. He has authored over 50 papers at Tsinghua University, which is considered the MIT of China. His textbook is used at the university and his team of students have won numerous international AI challenges against teams from Google and Facebook.
Deng has received significant grant money from the Chinese government for his work and he has participated in multiple national level research projects. Since 2016 he has also served as a principal architect for one of China’s largest rail companies. For Matrix AI Network, his vision for a worldwide, decentralized AI network drives the project.
Now on to some questions.
QUESTION: Before we get started, I wanted to say thank you professor Deng for taking my questions. My first question is how would you explain the benefits of Matrix AI Network?
ANSWER: It’s my pleasure. In answer to your question, Matrix is designed to serve as a platform for AI computing. First, it offers cheap data storage and also guarantees the privacy and accountability of data usage. Second, it allows the trustworthy transaction of data and AI models on the Matrix blockchain. Third, it enables cost-efficient computing for AI model building. Google & Co. have the financial power, they can generate enormous computing power, they already have a lot of data possession, but they are centralized.
QUESTION: Do you believe that Matrix AI Network is going to advance the field of AI?
ANSWER: Yes, Matrix is designed to take AI to a new level. The basic idea is to protect the ownership of data and AI models by leveraging the power of blockchain and at the same time offer cost-efficient computing power for AI model building. Without blockchain, people do not want to share data and AI models because the usage cannot be traced and the returned value cannot be fairly assigned. Meanwhile, blockchain networks can potentially integrate significant computing power. Such power would be essential for AI applications (over one half of AI companies claim that they need more computing power but the cloud services are too expensive). From this point of of view, we want Matrix “to be the ultimate platform for AI usage in the future”.
QUESTION: Could you please describe in detail why Matrix will be able to challenge the big players? Is the key to Matrix’ success its decentralization?
ANSWER: Yes, the key is decentralization. Big players are not perceived as trustworthy service providers as they can potentially abuse user data. In addition, as all big players in data storage are also key players in data service, it’s easy for them to cross ethical borders. For instance, we already know that Facebook had such cases.
QUESTION: Why should Matrix be the first choice for AI scientists, data holders and AI clients?
Let’s say I am an AI scientist. I create AI algorithms for a certain issue. Now it’s up to me. Shall I use Google Cloud? They may steal my algorithms or they may steal my data. If I use Matrix AI, all algos (algorithms) and data will be safe. And I will earn money if others use my algos and my data. So yes, Matrix should be my choice over Google and others.
Another example: If all airports worldwide put their data into the Matrix Network each airport will profit from the mass of data (e.g. weather forecast). They will all use algos from the Matrix Network and of course the computing power This will be like a big orchestra where every participants profit from each other. And of course the algorithm developers will earn money if the airports use their algos.
QUESTION: What is the approach to bring data to the Matrix AI Network?
ANSWER: At the present time we are developing a mechanism to attract people to contribute their data and AI models to the platform by building a platform that maintains and guarantees the “digital right” of these data and models on the blockchain. Given a sufficient amount of data, an ecology with value can be built.
QUESTION:
Which field of technology will be the first for Matrix to gain real-world mass adoption?
ANSWER:It’s really hard to say. I would say AutoML and NFT(non-fungible tokens) may be such applications.
QUESTION: How will Matrix AI compare to Google and Tencent in terms of performance?
ANSWER: Let me be specific. The cloud performance can be measured in various metrics. The most common metrics are latency (the time between the submission of data and the completion of computation), throughput (the overall jobs finished in a given period of time), and efficiency (number of jobs finished on a given node within a given period of time). The clouds of Google and other big players tend to have a lower latency and efficiency as the hardware are centralized, i.e. highly optimized and standardized. On the other hand,the throughput can be higher if we do have a sufficiently larger number of computing nodes in the network.
QUESTION: How fast is the Matrix AI network?
ANSWER: The stable Transactions Per Second (TPS) measured from the main net is 12499 (12.5K). It is achieved by a special hierarchical mining mechanism in which miners can collaborate to attain a high TPS. It is the fastest blockchain.
QUESTION: Does the speed of AI Calculating, and thus data supply for the AI algorithms, also depend on the TPS? Or is it just the internet bandwidth?
ANSWER: The speed of AI computation does not relate to TPS. The data supply depends on Internet bandwidth, not on TPS.
QUESTION: In what additional ways can the community contribute to Matrix AI Network?
ANSWER: That’s a very good question. We need community members to help us identify important applications and use cases in different countries. Also let’s see if we can get involved in various projects.
Thanks again to Professor Deng for answering a few of my questions. I’m looking forward to watching the next phase of this project evolve as Matrix AI Network works on bringing real world AI usage to their blockchain.
For more information visit: