Here some words of Bill Li (Chief Network Architect) when he visited the Matrix main telegram channel (https://t.me/matrixainetwork)
Since English is not my Mother tongue, maybe I will use Chinese and translate it with Google translation.
A community member: I got a question about the technical progress of the last weeks – Will there be more technical writeups/articles regarding the development progress of the chain itself? Would be very interesting to get a more detailed dive into the technical problems/solutions of the transition of the chain
我们目前倾向于采用QEMU或者类似的代码，模拟一个标准的ARM A9 CPU或者RISC-V CPU。然后用这个运行计算程序。当系统稳定后，增加智能合约的支持，此时智能合约不再是solidity，而是大家自由开发的程序，只要能够对账户系统访问即可。我们实际上是不同意采用WASM（webAssebler)方式的。
1, the current main development is divided into two categories:
THE MAINTENANCE OF THE MATRIX MAIN CHAIN, INCLUDING BUG FIXES, FAIRNESS REVISIONS, ETC.
In addition, the development of suitable for everyone to facilitate the development of contracts and deployment of things;
2, for MATRIx ecological development
Mainly includes: 1. Real computing system. This includes the deployment of an open computing system for the calculation of computing power by ordinary nodes.
We now prefer to use QEMU or similar code to simulate a standard ARM A9 CPU or RISC-V CPU. Then use this to run the calculation program. When the system stabilizes, increase the support of smart contracts, at which point smart contracts are no longer solidity, but freely developed programs, as long as they can be accessed to the account system. We actually don’t agree with the WASM approach.
2, to provide some APP and RPC interface, easy to develop AI APP applications.
Maybe I can add some explanation.
First task, we need maintain the main chain and provide some function for developer.
Secondly, we want our blockchain become a CPU cluster and provide cloud computing for everyone.
So we need some plan to do it.
A community member: is there any plan for the natural language programming on matrix or did you remove that?
We developed a voice-interactive robot, but the effect was so poor that the APP was not released. We are now ready to use a template-like approach, performed through a voice menu. This development process will have to wait until this release is completed and the distribute computing system is complete.
We you contact our pr, he can give you some demo about this process. The apk demo is not nice, therefore we decide that it should not allowed to release.
A community member: Thanks for the reply – The maintenance and fixing of miscellaneous bugs is still all in an private repository I guess?
Will there maybe be a little more transparency regarding the progress/developments so we can see those finished Issues/Tasks?
The public github repository gets updated rarely just with major releases – If you track the Issues/Tickets internally with Jira or any other issue tracking software, would it be possible to get a short summary of done Issues? Would be nice to see, maybe once a month, what different Issues where found and fixed within this timespan.
yes, we use gitlab to do it.
we construct a bugzilla system.
A community member: Nice to hear, will it be publicly viewable or is it currently only for internal use?
I think we should debug via github issue system.
In our gitlab system, we use CI/CD to test and auto release it.
After this version release, I will try github system to do it.
And maybe, if some bug exist, you can use issue and we will solve it.
It’s internal use.
I can show our bugzilla email snapshot to you.
wait a moment.
(later we got the image released because the vpn went down):
A community member: so is this correct, you want to build a CPU cluster, and this is the one performing the actual AI calculations? And the introduction of GPU compute/the apocalypse miner that is based on GPU, too, which functionality will those have on the final mainnet?
Because in the near future, we don’t want to design and produce CPU IC, we must something to enhance our mainnet function. So we choice the norm CPU to do our first step. If first step is ok, then we need to do some AI aided CPU, then it can do actual AI calculations.
And we must identify these two methods. Using normal CPU with AI coo-processor or booster, we can Make an protocol in mainnet. It’s a consensus.
But we use GPU, that is the second method, we don’t need consensus, just a application. Maybe one day, somebody can control it or change it. we need add some protocol to realize it.
However, Gpu solution is the most convenient to do actual AI calculations.
A community member: thats interesting. most of the more serious masternode miners I am communicating with are running quite powerful CPU setups inluding myself (AMD Epyc of the latest generation) which would be favourable in our case. we were just preparing for the release of GPU mining algorithm. so both options, CPU with AI co-processor and GPU will be running in parallel at some point in time, or will it be test phases for both options? And will the MCMC based mining algorithm still be integrated in the future?
We just make some RPC solution via some framework, and the add a Sales settlement system. It works.
In fact, we are ready to deploy to Xinjiang in this package.
A community member: when will the sales settlement system be published? and will payments be mad in $MAN currency?
Bill Li: we must use MAN currency when somebody use this system
Why do we have to deploy this system? Because other blockchains are just virtual currencies, we’re working hard to make a difference to the world’s computer programs and to be able to really make a difference with other blockchain system.
This idea has some feasibility, but the actual operation is more difficult. First, we must have a solid machine that is running steadily, not a cloud computing machine. In addition, we need to build a system that can perform various task distributions. There are a lot of simple implementations, such as docker, but this threshold is a bit high for the application, so we also have a headache. Then there’s how to prevent hackers from clearing systems.
The implementation step is roughly divided into three pieces: (1) the power output system, which can provide docker implementation on the regular CPU, but as mentioned earlier, this is more demanding for the user; Users just need to write some simple apps themselves. (2) Settlement system, this has more ready-made program, this Chinese New Year, I am trying to learn this piece. I will provide this part of the learning code as soon as possible; (3) task distribution scheme, which can first adopt a centralized automatic distribution system, using the bidding method, but need to support anyone to build this system, and also to send the task function.
A community member (in reply to “we must use MAN currency when somebody use this system”):
would you be willing to share some documentation on the more specific token economics that will be part of the network in the future? I am aware that
1.) data providers get paid by data users
2.) algorithm providers get paid by algorithm users
3.) dapp/application providers get paid by application users.
so by implementation of smart contracts, lets assume one iteration of the use of a dataset or data point will be paid for by the user, is that correct? Are there more specific mechanism documents on that marketplace ecosystem available?
Today’s smart contracts are hardly supported to such a fine level, so they are based on the design of an internal program, or based on a real CPU code (virtual code can also be used). At present, only the equity configuration can be implemented. Try to achieve the return each user can get. This is the focus of the settlement system. The settlement system, we think, is at its core to take advantage of the fairness of blockchain implementation, if pure lying contracts, too easy to attack, must be some degree of oralce, and then external import implementation.
In addition, we set up this system, I would like to be together to discuss how to deal with, how to achieve. I will write this implementation plan after the release of the new version, I hope you will participate together, implement together.
Does anybody have some question? if no answer in 3 min, I will quit.
A community member: maybe one last question, when will the new mainnet version be published?
Because the system has a lot of creativity, so I do my best to plan the path and choose a reasonable solution. If there is anything wrong, I hope you can point it out, and then we can finish it together. I wish we can create a beautiful situation.
Actually, on January 1 this year, the version was fixed, but you know, I’m still working from home because of the new corona pneumonia, basically only a few employees are working properly, the rest are in the idle state, just email can be normal.
This is it for today. If possible Bill Li and Owen Tao will visit the channel again…