Forum Replies Created
-
# 5 years, 10 months ago
Question from Ivan, Ivan on Tech interview
Q Ivan: What kind of capacity will the network have? How will the scalability issues be compared with ethereum today, or how many transactions per second (TPS) can we get? How can we get any feel, for what kind of applications this system will be able to handle.
A Billal: The TPS question is something we get a lot, it’s the golden question. We’ve actually made several tests with our mvp her in paris. We actually reached 1 million tps, but we have to explain what that means. For example, you can find on wikipedia that visa transactions are 65k tps and that’s enough for all customers worldwide. A lot of blockchain projects are claiming they can reach 1m or 2 m tps, but what do they actually mean with tps? From the first step of the transaction, I omit a transaction, it will go to the network, it gets validated by the old board, after that goes to the blog, and after I write in the blockchain etc. We have to be more precise, because in our protocol we solve the problem of selection and election. The problem with the consensus of bitcoin and ethereum, is that they are using proof of work. Proof of work is not a consensus algorithm for selection. I will select or elect the node that will validate the block, but this will take a lot of time. You need to use hash function the first time, then you need to inverse the hash. This will take a lot of time, because it’s a brute force approach. For neurochain it’s based on thermodynamics and integrity. It’s only calculations and it will take less than a second to calculate all the parameters of all the bots in the network. To conclude, yes we solved the problem of selection and election. This means that we don’t have a problem of performance. The number of transactions, that I can include in the network depends on the size of the block.
Bruno: The energy issue that the blockchain is experimenting with right now, is something that we have in our address merely. One by speed and second by the different type of consensus, it will take a lot less energy. Your phone could even be holding a bot and create the whole consensus, without using huge machines.
in reply to: Transaction speed -
# 5 years, 10 months ago
Question from Ivan, Ivan on Tech interview
Q Ivan: Will the quality of those decisions be better with time, because that is one of the feature of machine learning. The more data they receive, how better they get.
A Billal: Yes, it will be a lot better over time. The application will gain more data over time, and my historical data will have more reach, also my algorithm will be more precise. I will also use the experience of the other bots, such as data. Like I said, it’s such a virtuous circle.
Q Ivan: With smart contracts you can define conditions to give exact orders what to do. You don’t have the flexibility with smart contracts when something unexpected happens, because it’s all set in stone. With artificial intelligence you have that flexibility, but how can we trust that the AI doesn’t do something completetly irrational.
A Billal: We are using trading algorithms. All the banks around the world are using trading algorithms. The idea is that I will deblock an algorithm, which buys/sells stocks on the market without the control of any human. What we at neurochain do, is that we define the framework/limits of these algorithms. The possibilities of making mistakes would increase with the algorithm running for a longer time. When the bot would meet those limit, it will stop working. The bot would ask the human for advice to continue, so in this case it would ask for the human decision.
Bruno: When you’re talking about artificial intelligence there is always a human factor that comes in play. At some point you need to define what the limits of the questions are and the algorithm always at some point will reach the limits. You will have to basically start to define what will be the behavior after that. We at neurochain strongly believe that this behavior need to be defined with values that we believe in. That’s why our messages and how we are creating our different business applications are always done in a certain way, where artificial intelligence is always based at the end of the different values that we own and caring. There’s a big competition in the world, but not to define what would be the blockchain that will be the most used. In regard to neurochain, more specifically about the one that will be employing artificial intelligence concepts on which basically human values it will really reside. We believe it’s a huge fight and something that should require a lot of attention by our politicians, political leaders, economic leaders etc. That’s why we’re very involved politically here with the governments and other european governments, because they see that this is not just a technology play. This is also a moral play and goes way beyond simply technology.
Q Ivan: Will these machine learning models, mathematical models train together, so all the bots are updated together with new data and experience, or will they independently evolve with their own experience?
A Billal: There will be a specific development for each bot. If your bot is somehow involved in an application, it will have more algorithms than the other bots. The core of the bot will always be the same for every bot and the intelligent part of the bot will depend on their actions.
in reply to: Difference between Smart Apps and Smart Contracts -
# 5 years, 10 months ago
Hi Alchemy,
Thank you for this good question.
You are right the security of our futur dapps is a major point to the adoption of NeuroChain.
Some coutermeasures are designed. The first is that the dapps are not running in the core of the bot, it is an independant layer dedicated to dapps. This means dapps will not interfer directy with the ledger. They can read the ledger, use alogrithms based on data from it and then create transactions. It is not fully integrated.
Dapps can have some bad stuff inside their design, but this is not a specific problem of blockchains, it could be the same in a basic website.
With ML and AI and the consensus based on the trust of nodes, dapps like this could be easily put in quarantine.
Thank you -
# 5 years, 10 months ago
Hi Jlat,
We had some discussions on TPS. It is a tricky subject because every blockhain has its own criteria. What is TPS ? transactions added to the ledger? transactions created? transactions validated? The global average speed of internet is 6Mb per second, less than 1MB. If you want ot create a worldwide blockchain with a real distributed system, this technical limit is the first you will reach. Computing transaction is not difficult. If you want to validate a deep calculation on a mathemaical subject, it could be more tricky. And then, the next question is : is there an interest in the real time validation of a transaction ? This technical limit is the same for all the blockchains. So the difference is in the time to process informations. To give you a number, we have tested our consensus (calculation of entropy, entalpy and integrity) and the result is that a medium power laptop can process 60,000,000 transactions within a minute. The performance measurements have to be, always, made in the context of the whole consensus operation. Just to give you some benchmarks, for Visa system, they actually expect 65. 000 tx/s in theory. .
https://usa.visa.com/dam/VCOM/global/about-visa/documents/visa-facts-figures-jan-2017.pdf
Another metric we have to consider is the volume of transaction in a block. In NeuroChain, this parameter is adjustable depending on the sytem evolution. the Block-size is an important parameter for the network operation and for its security Machine Learning will continously improve velocity of NeuroChain.in reply to: Transaction speed
