Matrix AI CEO Answered questions from the community! — October 2023 AMA

Matrix AI Network
7 min readOct 31, 2023


Q1: Can you elaborate on how the significant increase in neural network parameters enhances Morpheus’s performance and capabilities compared to the previous version?

For deep neural networks, generally speaking, the number of parameters directly determines the network’s capabilities. The larger the number of parameters, the more powerful the neural network becomes. This is actually similar to the human brain’s capabilities — the more active neurons there are in the human brain, the stronger its abilities. Typically, in the process of simulating the workings of the human brain, the number of parameters is used as a reference to the number of neurons in the human brain. So, in this new version of Morpheus, the number of parameters has increased fifteenfold. Morpheus will have better thinking and response capabilities, but the response speed will decrease due to the increased computational load caused by the increase in parameters.

Q2: With the training corpus now 35 times larger, could you share some practical examples of how Morpheus benefits from this extensive knowledge base in real-world applications?

Generally, an increase in the size of the training dataset has several positive effects on Morpheus, including the following:

  1. 1.Improved Generalization: Increasing the amount of data helps Morpheus better understand the underlying patterns and principles of the problems it’s learning, thus enhancing its generalization ability. By discovering more patterns and regularities within more data, Morpheus can make more accurate predictions and classifications on new data. Reduced
  2. 2.Overfitting Risk: Overfitting occurs when a model performs well on the training data but poorly on new data. A larger dataset can lower the risk of Morpheus overfitting because it becomes more challenging for Morpheus to memorize all the details of every training sample, encouraging it to learn more general features.
  3. 3.Suppression of Randomness: In cases with small datasets, randomness can lead to unstable training results for Morpheus. As the dataset size increases, Morpheus’ performance becomes more stable across different training sets, making Morpheus more reliable.

However, it’s not always the case that a larger corpus or dataset has a purely positive impact on the model. For example, having a large amount of data doesn’t necessarily guarantee positive results if the data quality is low. In such cases, a large dataset with low-quality data can introduce noise or even mislead the model’s learning, resulting in incorrect outcomes. We’ve added 35 times more data to our dataset, but we’ve also performed quality screening and content filtering to improve data quality accordingly

Q3: AI is still in its infancy and a large number of companies have fallen victim to a shortage of computing power. How do you plan to democratize crypto while utilizing AI and break barriers and boundaries in web3 space?

The high cost and scarcity of computational power were among the initial reasons for designing MATRIX. We aimed to create a decentralized computational power distribution platform for two primary purposes.

Firstly, by using distributed privacy computing, we sought to enhance the security of AI computations.

Secondly, we aimed to reduce the cost of computational power through mining incentives, thereby providing more affordable computational power to users interested in AI development and services in the blockchain and Web3 space.

Furthermore, we believe that artificial intelligence is a crucial tool for achieving the democratization of cryptocurrencies. Currently, cryptocurrency democratization is mainly evident in two aspects: blockchain consensus mechanisms and the rise of DAOs (Decentralized Autonomous Organizations). However, in the face of thousands of concurrent transactions per second in the blockchain field, human decision-making efficiency appears to be quite low. The future virtual world is bound to be information-driven, and true democracy requires participants to have sufficient information acquisition and processing capabilities. Therefore, we need AI as our assistant. The goal of MATRIX’s Avatar Intelligence is to become a user’s virtual presence in the virtual world, enabling real-time data and information acquisition and processing for users.

Finally, we designed Intelligent Contracts to effectively lower the entry barriers for Web3, allowing users without programming skills to access the core contract functionality of Web3. This also enables Web2 developers to easily port their applications into the Web3 world without extensive learning and research

Q4: What practical and tangible outcomes do you foresee in the near future as a result of Morpheus’s advancements in the context of achieving Avatar Intelligence, and how will these advancements impact the daily lives of people worldwide?

The design goal of Avatar Intelligence is to empower users to thrive and excel in Web3 or the metaverse. Firstly, we believe that Web3 and the metaverse are the inevitable trends of the future. The future virtual world will undoubtedly be a digitized realm dominated by vast amounts of information. Participants in this world will need to have the capacity to acquire and process information effectively to be competitive. It’s clear that humans alone do not possess this capability, so we need AI as our assistant. The objective of MATRIX’s Avatar Intelligence is precisely this — to become a user’s virtual representation, enabling real-time data acquisition and processing, thereby enhancing users’ quality of life and enabling them to generate more value in the virtual world.

Q5: Could you provide insights into the goals and objectives of the community test, and what valuable feedback or contributions are you looking to gather from participants?

The purposes of the testing can be divided into several parts:

  1. User Experience Testing: We hope that community users can provide us with more feedback through testing to make Morpheus’s user interface more user-friendly and intuitive.
  2. Bug Collection: Involving more users in testing helps us discover issues that we may not have found otherwise.
  3. Stress Testing: During the testing process, we aim to evaluate Morpheus’s performance under usage pressure. This will help us find the optimal balance point between the number of parameters and response speed.

Q6: Can you explain the significance of adding multiround optimization for Intelligent Contract contract generation and how it enhances the overall efficiency and performance of the platform? AND What specific challenges or complexities does multiround optimization address in the context of generating Intelligent Contracts, and how does this benefit users and developers?

If you have used Morpheus or ChatGPT, you probably understand the principle of multi-turn optimization. When interacting with large models, if the output content doesn’t fully meet our needs, we can provide feedback to the model to have it correct its output. This is the working form and principle of multi-turn optimization. When we are not satisfied with the contract content generated by the model, we can request it to make modifications and optimizations, gradually getting closer to our desired goal.

Q7: The Hybrid PoS + PoW consensus mechanism makes use of value-added computation through the use of the Markov Chain Monte Carlo (MCMC) computations. Can you give a detailed explanation of this?

We have not fully achieved this goal yet. Our original intention was to use the MCMC (Markov Chain Monte Carlo) algorithm to replace the hash algorithm. The primary reason for this choice is that MCMC is a more suitable algorithm for mining. It has the advantages of convergent and predictable results. If we use MCMC instead of the hash algorithm, mining computations can become useful rather than purely wasting energy and computational resources.

Currently, we have implemented a hybrid algorithm. The mining algorithm on the MATRIX mainnet is a hybrid algorithm combining AI and hash. This means that the computations involved in mining on the MATRIX mainnet include a portion of useful computations. Currently, this useful computation involves tasks related to graphic and image recognition.

Q8: How does Matrix AI Network view the convergence of virtual and physical worlds through technologies like the metaverse, and how might it play a role in this evolving space?What unique approaches is Matrix AI Network taking to mitigate the environmental footprint associated with blockchain technology and energy consumption?

The primary goal of MATRIX 3.0 is to provide users with a better experience in virtual worlds like the metaverse. We plan to use brainwave technology and large model technology to create an Avatar Intelligence for users in the virtual world. This Avatar Intelligence aims to enhance users’ efficiency and effectiveness in the virtual world and become a crucial link between the physical and virtual realms.

Regarding the second question, we address it primarily through MATRIX’s consensus mechanism and mining algorithm. We are dedicated to transforming the mining hash algorithm into an artificial intelligence algorithm, so that mining computations can become useful rather than pure energy and computational resource waste. Additionally, MATRIX’s mining utilizes an election mechanism based on a random clustering algorithm, which does not require every node in the network to participate in mining. The computational power of nodes not selected can be leased to users in need. Currently, we use MANTA to provide AI computational power, further preventing the wastefulness of computational resources

How will Morpheus benefit from $Man coin volume? So, as a result, if every work or product done in the project is done with the aim of increasing the price of $Man. Will you add features that make Morpheus different and allow payment with $Man? @SatoshiMANato A: Because Morpheus’ ultimate goal is not to create a chatbot similar to ChatGPT, we aim for Morpheus to be a part of the infrastructure for Avatar Intelligence. This means that if a user wants to create a virtual alter ego based on AI and brainwave technology, they would need to interact with Morpheus to teach it their language habits, thought processes, values, and other aspects. In the future, using Avatar Intelligence and migrating across different platforms will require payments in MAN (MATRIX’s native cryptocurrency). Part of these fees will be considered the cost associated with Morpheus

The Matrix AI Network was founded in 2017. In 2023, we enter Matrix 3.0 blending neuroscience with our previous work to realize the vision of the Matrix films.


Website | GitHub | Twitter | YouTube

Telegram (Official) | Telegram (Masternodes)

Owen Tao (CEO) | Steve Deng (Chief AI Scientist) | Eric Choy (CMTO)



Matrix AI Network

The Matrix AI Network was founded in 2017. In 2023, we enter Matrix 3.0 blending neuroscience with our previous work to realize the vision of the Matrix films.