MANTA — The Brain of Tomorrow’s Metaverse (4/4)
Matrix provides computing power for the Metaverse through the separation of computation and consensus.
Even after years of development, there are currently only a small number of projects and businesses engaged in the practical implementation of the Metaverse. And of these, most are only offering watered-down versions that by no means reflect the true potential of the Metaverse and are plagued by numerous technical issues.
For instance, Decentraland, one of the earliest virtual games on Web 3, held a Metaverse meetup last December. During the meetup, the server suffered an overload with merely a few thousand people online so that many players experienced a temporary loss of video and audio signals. In the end, to make the meetup worth it for the attendees, Decentraland had to give in and resort to the centralized Tencent Meeting app instead, which deprived the event of its specialness. This is but one of many such fails.
In fact, gaming and social media are naturally where the Metaverse should gain its first foothold. To develop a game, first, we need to set up the framework of a virtual world so that people can interact and have fun there. With the framework in place, we can introduce an economic system into the game based on the real world. The only challenge here is that, compared to traditional games, network and data exchange latency will be the biggest problem the Metaverse has to solve.
Traditional centralized games support high-frequency interactions between players and the server. Each action players take is synced to the server within tens of milliseconds, which demands much of the upload and download speed of the server.
To decentralize without sacrificing performance sounds like an impossible task, but this is where a technological revolution must take place. Since a decentralized network can theoretically have wide coverage of edge nodes. These nodes can be closer to players than traditional servers, and this way latency can be kept as low as possible. But if so, why are latency still restricting the development and implementation of decentralized blockchains?
To explain this, we need to look at the consensus mechanism of blockchains. Consensus mechanisms are what guarantee the security of blockchains. The programs in the server are calculated and cross-checked by the tens of thousands of nodes in the network. In contrast to the once-and-for-all approach of traditional servers, consensus mechanisms will cause many times more data and computational tasks to flood the network, which makes millisecond-level response speed nearly impossible. Even though players may get assigned a node located closer to them, the benefit of this will be easily outweighed by the extra burden of running consensus algorithms.
With this in mind, if we can find a consensus mechanism that guarantees the security and stability of programs without centralized management, then the mechanism can be deployed onto distributed networks to utilize their computing power resources to the maximum.
This technology is called the separation of computation and consensus. Under this arrangement, consensus algorithms only need to do the thing they are best at doing, while computation will be assigned separately to each node in the decentralized network.
Matrix has designed a wholly new consensus mechanism (Hybrid-PoW) to make this idea come true.
Hybrid-PoW is essentially a marriage of DPoS and PoW. From a mathematical perspective, the long-term expected yields of elective PoW and global full-time PoW are mostly the same, which means the introduction of an election-based mechanism would affect miners much. One challenge, though, is to guarantee the fairness of elections. Most DPoS election mechanisms currently adopted by projects can’t achieve true randomness and fairness. To solve this problem, Matrix has introduced random clustering algorithms (RCA), which categorize nodes by their characteristics and select representatives from them through true random elections.
With the help of RCA, Matrix’s Hybrid-PoW will be more secure, better at guaranteeing fairness and easier to implement. All miners in Matrix AI Network don’t participate in PoW mining in every mining cycle. Only elected miners do PoW computing so that Matrix can export the extra computing power on its platform to clients who need it, just like a cloud computing platform would do.
In a network built on an HPoW consensus structure, users don’t have to go through blockchain to interact with nodes. This avoids latency caused by consensus mechanisms and brings gaming ping down to a millisecond level. Meanwhile, Matrix’s global network of nodes also works as a powerful server to run games on.
In the future, Matrix will open up its entire computing power network so that users can contribute their extra computing power and become a part of the Matrix ecosystem. With good hope, this will become the world’s largest distributed computing power network, and thanks to its support, the full potential of the Metaverse may finally be realized.
MANTA — The Brain of the Metaverse
The Metaverse is essentially a gigantic virtual world where people interact with each other online. Computing power is the foundation of this virtual world, as the rendering of virtual environments and users’ interactions are made possible by computing power. To make the Metaverse truly worthwhile, user privacy must be protected, and the way to do this is through distributed privacy-preserving computation. If the Metaverse were a human, then the computational network would be the brain that commands and provides energy for the whole body.
MANTA of the Matrix ecosystem offers the perfect solution for building such a brain for the Metaverse.
MANTA (MATRIX AI Network Training Assistant) is a distributed auto-machine learning platform built on Matrix’s high-performance blockchain. The platform is essentially an AutoML application together with its deployment system, which uses distributed network technologies for acceleration. The AutoML network searches for a high-precision and low-latency deep model, which is then accelerated through distributed computing. MANTA has two core functions: auto-machine learning (Auto-ML) and distributed machine learning. The framework for the latter is a perfect distributed privacy-preserving computation network.
The speed of network searching and training will be further accelerated through distributed GPU parallel computing on MANTA. MANTA’s distributed machine learning is essentially a set of distributed parallel algorithms, which support parallel data and parallel models for acceleration. The idea behind all this is to distribute the data generated in each iteration to different GPUs for forward and backward computing. In each iteration, each GPU will be sampling the same sub-model. Distributed parallel model algorithms are different from parallel data in that, for each iteration, MANTA allows different nodes to sample different sub-networks and conduct a global exchange of gradient information after each node has completed its gradient integration.
Due to the characteristics of MANTA and Matrix’s unique consensus mechanism, this network will be capable of connecting the idle computing power of the world. Your iPad or iPhone could contribute computing power to Matrix while you sleep at night. At the same time, MANTA has a principle of using your own resources to serve yourself first. When your own devices are idle, the system will prioritize using their computing power to support your Metaverse activities. This allows a lower-level privacy strategy to be used to improve efficiency. When using computing power from unknown devices, a higher-level privacy strategy and distributed computing involving more nodes will kick in. We call this smart edge computing, which may eventually prove to be the solution for the huge amounts of privacy-preserving computation necessary for the Metaverse.
Matrix + the Metaverse: A Restructuring of Data and Computing Power
Without a doubt, the Metaverse we are talking about today has far exceeded what the 1992 novel Snow Crash could imagine. In the past decades, we experienced:
· The technological revolution brought about by advances in modern digital information technologies as well as the social and economic changes that followed;
· The social and political environment since the industrial revolution challenged and reforged by the Internet, ushering us into the era of Web 3;
· Advances in VR, AR and MR technologies with ever-growing power to create a virtual world comparable in detail and complexity to our physical world;
· The invention of blockchain, and the new economic possibilities it brought in combination with DeFi, NFT and other financial inventions;
Looking back, the landmark events in the real history of the Metaverse may be different from what people originally had in mind, whether it’s the level of decentralization, the closeness to reality, customizability, or the maximum number of online users the Metaverse can take. In order to have the Metaverse span all platforms and industries, we must keep looking for creative solutions.
The birth of every new technology in the history of mankind is made inevitable by the technological breakthroughs that preceded it. In essence, the Metaverse is a restructuring of data and computing power. Current limitations are caused by a lack of computing power and effective measures to collect and manage data. The demand for limitless amounts of computing power has given rise to the concept of decentralized privacy-preserving computation.
As far as we can see, the scale of data and computing power will be the single biggest limiting factor for the development of the Metaverse. This challenge will be present at every step in building the Metaverse, and we must find out how to restructure and reassemble data and computing power in order to guarantee user privacy in the midst of the lightning-fast exchange of information. We must also know how to integrate idle computing power from nodes around the globe so they can work seamlessly. These are the focuses of the Matrix team.
We cannot put our finger on by what exact time we’ll gain sufficient and stable computing power to bring the Metaverse to its full potential. But we know that MANTA’s approach to distributed large-scale computing power will play a key role in this, which won’t just benefit the Metaverse, but also be critical for the success of Web 3. The Metaverse is still in its infancy, and Matrix still has a long way to go. But now, as the brain of tomorrow’s Metaverse, MANTA offers us a glimpse of what the Metaverse could be and a future of endless possibilities.