Can Trustworthy Artificial Intelligence be Developed with Blockchain Technology?
Introduction
Although it has witnessed hundreds of important events over the centuries, it is always remembered with the development that triggered those events. For instance, the 20th century witnessed two major world wars, the collapse of states, walls, and ideologies, as well as astronomical discoveries, but the trigger for all these developments is the industrial revolution that took place at the beginning of this century. Again, the century we are in has witnessed a pandemic that killed and will go on killing millions at the very beginning, I am sure that it will witness colonization on other planets, new world wars, maybe star wars, the collapse of thoughts, states, and beliefs, notwithstanding this will not be mentioned with anything I have mentioned, the thing that triggered them. It will be remembered for its web revolutions. The 21st century started very quickly, and suddenly, our lives began to change with the inclusion of computers and the internet, and these technologies brought forth artificial intelligence and blockchain technology. We are discussing how to complement the cons of these two technologies, which we will now call revolution, by using the pros of each other. In this article, I will explain how we can solve the trustworthiness and explainability problems of artificial intelligence technology by using the reliability, transparency, and explainability features of blockchain technology, namely decentralized artificial intelligence.
Blockchain Technology
With the widespread use of computer technology, the concept of data began to become popular in our lives. This popularity turned into a problem with the spread of internet technology and the concepts of data storage and data security entered our lives. Something else happened at that time, the economic crisis in 2008 destroyed trust in central banks and financial institutions, and a white paper was published. This white paper was published by Satoshi Nakamoto and introduced a digital currency encrypted using blockchain technology. Nakamoto attributed this 2008 crisis to the callousness of central banks and the greed of financial institutions. According to its, current technology already gives us the opportunity to manage our money without these institutions. Satoshi Nakamoto claimed in the white paper it published that an encrypted digital currency would solve the problems of financial institutions and central banks until now, but how? All of the fiat currencies we use now were actually digital, what did Nakamoto offer different from this digitality?
In that white paper, Nakamoto put forward the central banks and financial institutions as the problem and highlighted the cryptology, that is, the blockchain technology as a solution, not digital money technology, he was right. Because if we look at recent history, the occasion for most economic crises is the central banks that print too much money and the financial institutions that fill their pockets with this money. Howbeit, the results, namely the spiral of high inflation and debt interest cause big problems for people. Bitcoin, and rather the technology behind it, offers limited supply, distributed management, mining, transparency, as well as immutability. Blockchain technology allows the transfer of value, experts do not want to qualify the transfer of a value as information transfer in this regard, because blockchain technology provides the opportunity to transfer value in Bitcoin. Information transfer in the digital environment can be done very easily and simply, but the decentralization and reliability feature that highlights the blockchain technology prevents this transfer process from being done as quickly as information transfer in the digital environment. These include some components in the system; Block, ledger, hash, nodes, mining, consensus, PoW (Proof of Work), PoS (Proof of Stake), and decentralized data storage with smart contracts.
The system, that is, the chain, is created by miners. Bitcoin miners create a series of numbers every 10 minutes. As the miners perform the mine operation, blocks, and as the information to be created on these blocks are processed, ledgers are formed. We call this ledger creation process hash. We can compare the blockchain to a grocery store's ledger, miners mine a page every 10 minutes for this ledger, and blocks are formed, then the grocer fills these pages and hashing takes place, and these filled pages create ledgers. These records are now kept and controlled by the authority that manages the system in the central system. Currently, a bank keeps the transactions of its customers in its own database and manages these transactions through these databases. We can say that the situation is the same in Bitcoin, there are blocks that we can call a database, the ledgers formed by recording data on it, and the blockchain formed by these ledgers, the data recorded in this database and which we describe as value, namely cryptocurrencies, that make up this database, managing and it’s the people, not the person holding it, that’s the solution Satoshi Nakamoto offered. Currently, the monetary system is in the hands of central banks, which monopolize the management of those ledgers, and financial institutions that process data to ledgers, that is, in the hands of a specific, centralized, single authority. The situation is different with cryptocurrencies. Because the system does not consist of a single huge database, it is a chain consisting of thousands of scattered databases, that is, blocks with hash operations, connected by encryption one after another. The chain has miners scattered around the world that enamel this system. Miners create the chain with the mining process, enter the values into the ledgers and perform hashing, in other words, they manage this management process with the approval of more than one miner, and they hold the chain while doing these operations. In other words, instead of a database, a miner that processes values into that database and a management mechanism that manages these values, as in the central system, multiple blocks that are created and managed in a dispersed manner, and a chain with multiple copies in which these blocks are linked to never be changed, are decentralized, unchangeable and It offers a reliable system thanks to Bitcoin blockchain technology.
This technology has not remained in its unchanged form as published in the Bitcoin white paper. Satoshi Nakamoto sparked a spark with that white paper that day, and every other entrepreneur after its shed wood on that spark, and the blockchain as well as web 3 trend, which we can perhaps call a revolution, emerged today. Vitalik Buterin introduced the Ethereum cryptocurrency after 2010. What makes this crypto money stand out is smart contracts that will integrate this technology into our lives rather than blockchain technology. Although Satoshi Nakamoto boasts of the encrypted and decentralized side of Bitcoin, the encrypted digital currency, this money also has a digital side and needs to be integrated into our lives. Buterin solved this problem in his Ethereum white paper by decentralizing Nick Szabo’s smart contract method to blockchain technology. Nick Szabo defines smart contracts as neither smart nor contracts. It is a kind of software, the simplest form of which is coffee vending machines, on the basis of which the demands of the contracting parties are carried out by the software code. Vitalik Buterin has evolved this technology to integrate decentralized applications into the chain. I would like to give an example for you to sit in your mind; You have a land and you want to sell it for Ethereum cryptocurrency, first, you convert this land into a value as a none fungible token and you mint the Ethereum chain, your buyer should be the owner of Ethereum and wants o buy a land with the Ethereum in his hand. Thanks to a decentralized application and its smart contract, you own the buyer’s Ethereum at the time of sale, and the buyer owns the tokenized land, that is, the valuated land, and the ownership is done by hashing the chain at the time of sale, and the smart contract performs the entire hashing process. Smart contracts became bridges between cryptocurrencies, Bitcoin or Ethereum, and real-life, bridges that manage transactions between the chain and real life. This technology will be the actor that manages the transactions between artificial intelligence, decentralized application, and blockchain for decentralized artificial intelligence, which is the subject of this article.
That spark was set that day, once by Satoshi Nakamoto, entrepreneurs like him, smart people had no intention of stopping. After the Web 2 revolution took place, data became the center of our lives, and storing and managing it became a big problem. The storage problem seems to have ceased to be the problem of individuals with the spread of cloud technologies, but this time security has become a big problem for corporate and individuals. Data, access to data, data security, and management is also big problem for a group. This group of artificial intelligence developers, I will talk about this in detail while explaining the artificial intelligence technology below. So what is data? Is it just sets of numbers that take seconds to transfer today, or the values that makeup cryptocurrencies? If preserving, storing, and managing it is so important to us today, what is called data is more than sets of 1 and 0 numbers. So why don’t we resort to blockchain technology when we manage, store or protect it, which we do. Although there are new startups, crazy entrepreneurs who threw wood into the web 3 revolution fire found a solution to this issue by using blockchain technology. It provides decentralized data storage service using distributed databases, smart contract protocols, and a transparent, distributed, reliable and encrypted management model thanks to blockchain technology. Distributed data storage is another indispensable component of distributed AI, like smart contracts.
Artificial Intelligence Technology
The features that people have, such as decision making, perception, problem-solving, which belong to humans and prove the existence of intelligence are brought to the machine with artificial intelligence, machine learning, and deep learning methods. In fact, we can say that the machine should be intelligentized or evolved. By teaching all the words, rules, and patterns in two languages, you can make a machine translate between two languages, or you can teach all lung cancer forms that have been experienced so far, and you can have software diagnose lung cancer from radiological images. In fact, the essence of the matter is that we can transform a machine into an intelligent being like us by using the data we have known and archived so far and using two methods, namely machine learning and deep learning methods together or separately. It should not be overlooked that they may be intelligent beings like us, not a better or different kind of intelligent being because we teach them to be intelligent by showing the data we have experienced so far. In fact, this is the main reason for artificial intelligence, which is my purpose for writing this article, and the problems that pose a danger to us. The problem is not artificial intelligence or their working methods, the problem is us, we teach them the mistakes we have experienced so far and will experience in the future, and we cannot foresee and prevent mistakes due to the complexity of learning methods.
I would like to elaborate on the problem mentioned above by me, let’s say we are coding an artificial intelligence-supported algorithm of smart facial recognition systems biometric data taken from perhaps millions of people will be used to perform machine and deep learning of this algorithm. However, the extraction of these data is a process that has been going on for maybe 70 years after the 1950s. We extracted this data, to be more precise, human beings who have discriminated against blacks and women for centuries. Artificial intelligence that will be taught with all of that data will identify a white-skinned male person with 99 percent success, and a white-skinned female person with a rate of maybe 95 percent, but this rate will decrease as the skin color gets darker. It may not even describe a Kenyan woman, and there has been a case of discrimination. Because we have always discriminated against black phylon and women for years, we once deprived them of even the basic rights that a person should have, and the data created in this period or created under the influence of this period will necessarily contain that discrimination, and that discrimination is as if we coded it as our reflection in the mirror. will appear in artificial intelligence algorithms. Artificial intelligence technology first entered our lives in movies and later in simple studies, when we saw it in movies, we were excited by the skills of that technology but the fact that it entered our lives and did not recognize us as a human or defined us as a monkey caused disappointment, and this is just the beginning. People had a problem of trust in artificial intelligence at the first meeting, but how can this be solved? Can artificial intelligence be reliable, explainable, and transparent as asked in the title of this article? With the onset of the pandemic, we’ve all been away from our jobs and schools for almost a quarter of a year. In England, students had to continue their half-term education and take exams in order to be able to settle in the university, but this was not possible due to the pandemic and the British government followed a different path. According to them, an AI-powered algorithm could calculate the fairest grade using students’ old exam grades and other parameters, and the day came when the grades were announced and the students were shocked. The algorithm reduced the grades of students who lived in slums and went to schools in those areas, because the data on which the algorithm would do deep learning and study included discrimination that students living in those areas would fail, resulting in discriminatory learning and study.
We have more or less understood the learning method, working method, and reason for discrimination of artificial intelligence. The algorithm learns bias and works with discrimination if that data contains discrimination. Let’s focus a little more on the cause before the solution, the reasons why learning and working involve discrimination. I also wrote about the learning methods at the beginning of the topic; machine learning and deep learning. Machine learning is a less complex method, discrimination can be controlled in this learning method, but this is not the case with deep learning. It is a form of learning that works and learns the deep learning method with deep neural networks. Post-learning discrimination occurs in these networks. We define deep neural networks and the algorithm that takes place here as a black box because, in this process, the input and output cannot be reconciled, and the results of the output do not match the input. In the academic articles published, black box and transparency are defined as two opposite concepts in learning that take place with deep learning. I will share a deep neural network model below for your better understanding.
As you can see in the model, it is very difficult to predict the problems that occur in the data obtained during the deep neural networks, that is, the learning process with deep learning, in the learning part or after the output.
Both the data collected so far that make artificial intelligence intelligent and learning models cause many problems in the outputs of artificial intelligence, and explainability, interpretation, and transparency are very difficult in current models. Well, how we will ensure transparency, I will explain in detail while explaining the decentralized artificial intelligence issue below.
Decentralized Artificial Intelligence
At the beginning of the article, I claimed that we can use the pros of one technology to cover the cons of another technology. The learning method of artificial intelligence has many problems due to the data used in this method and while working with artificial intelligence, but due to its current popularity, it has gone through discrimination in this article. So, may artificial intelligence, where learning, working, data storage, collection, and management of all these works in a decentralized and dispersed model, cause as many problems as it is now? So can we solve these problems by completely decentralizing artificial intelligence?
The diagram above shows the decentralized artificial intelligence learning and working scheme with blockchain technology. Let's go over the components of the schematic first. The first is Fronted DApps. Thanks to Fronted DApps, different users, interested persons, and parties will be able to support the process of configuring the training and working processes of artificial intelligence, various parameterization, data accessibility, decision-making of outputs, development, and recording processes and interpretation. The second is the Access layer. Thanks to the link layer, more than one data transfer protocol can be used. For example, thanks to the Web3 interface, a connection can be established between DApps and the blockchain. Or, with the JSON-RPC API, a connection can be established between web-connected platforms and the Ethereum blockchain network. Thanks to Rest HTTP, it will be possible to establish a connection between the cloud data centers and the blockchain. The third is the AI layer. This layer is the layer that carries out the artificial intelligence process, this layer contains two types of artificial intelligence models or rather predictors. One of them is the normal AI predictor, the other is XAI, which works on the black box, this is the AI predictor that will work in more predictable, explainable ways. Two types of predictive nodes work in conjunction with server-fronted DApps. One of the AI servers works on the raw data, while the other manages less interesting operations such as data cleaning, noise removal, outlier extraction or feature extraction, or dilemma mitigation. Or servers work with learning models directly on processed data and draw conclusions. The AI predictor processes the data through the conventional black-box algorithm, while the XAI, that is, the explainable artificial intelligence predictor, works with a different mechanism to explain the results of the study and also in communication with the normal AI in this process. Another section is Support services. This section consists of two components, one is the registration service, and the other is the reputation service, that is, the scoring service. Registry records control the actors and members in the ecosystem. The scoring service, on the other hand, evaluates the scores from the artificial intelligence predictors and gives the rewards and penalties accordingly through cryptocurrencies. Of course, all of these management processes take place through smart contracts. The last part is Blockchain platforms, smart contracts, and decentralized data storage systems. This unit exists to store data and manage the whole ecosystem, smart contracts manage all services and units separately, all within themselves. The model includes 4 types of smart contracts, reputation and registration contracts that manage the support unit, the AI-task contract that manages artificial intelligence studies, and finally, the aggregator contract that acts as a bridge between layers.
In the first diagram, we got to know the layers and the actors of the system. In this diagram, we will examine how these actors work with each other, and as a result of this study, we will examine whether a transparent, explainable, and reliable artificial intelligence can be created that will answer the question I asked in the title of the article. Decentralized applications depend on support services and decentralized platforms running in the background, i.e. smart contracts, and decentralized data backup system thanks to the connectivity layer. Thanks to these applications, interested parties and parties have the ability to configure, configure various parameters, select the predictor type and number, access data, perform scoring and recording operations and interpret them. Decentralized applications have the ability to select registered predictors and define the SLA (service-level agreement) that will govern the system. In addition, decentralized applications upload AI-Task and aggregator smart contracts to the system, which will perform system management operations such as managing the SLA agreement, providing data flow, selecting predictors. And in all these processes, the communication takes place on the blockchain network, so the system becomes reliable thanks to the explainable, immutable, and transparent features of the blockchain. If the predictors accept the SLA agreement or the AI prediction agreement from decentralized applications, the aggregator performs the transfer of the agreement with the smart contract. “This transfer might be a kind of crypto money transfer” (Salah and Hasan 2018.) It is actually a deposit that will disappear if the process is left unfinished, but if it continues successfully, the reward is like a gain that will continue to increase with cryptocurrencies. Because when the positive and negative results and predictions from the predictors are sent to the smart contract, a process in which crypto money will be sent in return for the transfer will continue.
The layer with predictors is actually a layer where there are more than one traditional artificial intelligence and targeted explainable artificial intelligence servers and will work with reliable databases. Thanks to these predictors, different types of artificial intelligence systems, and different types of learning on their basis, different types of results will emerge for the same data, and these results will reach smart contracts and decentralized applications and those concerned and can be audited, compared and explained. Finally, according to the feedback from decentralized applications, artificial intelligence systems will continue to work according to the best results. Since the whole process, I have described will take place on the blockchain, these transactions will always be transparently auditable without any hashing, data, or results being destroyed at all stages.
Sunghyuck Hong believes that the black box problem can be solved with a blockchain-based decentralized data storage method. According to Hong, the data stored in the blockchain can be checked before entering the artificial intelligence algorithm, and if there is a problem after exiting the algorithm, it can be checked again thanks to the transparency and immutability features of the blockchain technology. According to the first method, Hong planned a simpler decentralized artificial intelligence. With the help of IOTs, the internet of things, raw data will be collected from outside, then there will be an intermediate server that will manage this raw data so that this server can process the data and thanks to the connection layer. can hash it to the blockchain. Afterwards, the process of pulling the data from the blockchain and entering the artificial intelligence algorithm, and hashing the data from the algorithm to the blockchain will be auditable at any time, since the whole process will be on the chain. Experts working in this field believe that reliable artificial intelligence can be created thanks to blockchain technology. Below, I will evaluate the areas of use that can be damaged due to the disadvantages of artificial intelligence technology, but where these losses will be eliminated thanks to blockchain technology.
Conclusion
Every new technology offers pros and cons from the very first day. If we use these pluses to make up for the minuses of others, we create a huge interconnected chain, as in blockchain technology. No innovation can be thrown aside just because it has exploratory damages. The important thing is to adapt those discoveries to our lives so that we can develop in those discoveries. The WEB 3 giant takes us from a single center and places us in a big chain. Now my data has been stolen, central banks have created inflation, artificial intelligence has excluded me, more than worries, my data is safe on the chain, miners give me the opportunity to control our money against inflation, artificial intelligence has completed the evolution of the blockchain to be smarter than us.
References
- Thang N. Dinh, My T. Thai/ AI and Blockchain: A Disruptive Integration/ COVER FEATURE FUTURE OF AI/ 2018
- Dr. Sakthi Kumaresh, Dr. K B Priya Iyer/ Decentralised Artificial Intelligence Enabled Blockchain Network Model/ Turkish Journal of Computer and Mathematics Education/ 2021
- Alejandro Barredo Arrietaa, Natalia Díaz-Rodríguez b, Javier Del Ser , Adrien Bennetot, Siham Tabikg, Alberto Barbado h, Salvador Garcia g, Sergio Gil-Lopez, Daniel Molina, Richard Benjamins, Raja Chatilaf, Francisco Herrera/ Explainable Artificial Intelligence (XAI): Concepts, taxonomies, opportunities and challenges toward responsible AI/ Information Fusion/ 2020
- Mark G. Core, H. Chad Lane, Michael van Lent, Dave Gomboc, Steve Solomon and Milton Rosenberg/ Building Explainable Artificial Intelligence Systems/ Conference: Proceedings, The Twenty-First National Conference on Artificial Intelligence and the Eighteenth Innovative Applications of Artificial Intelligence Conference/ 2006
- Vasco Lopes, Lu´ıs A. Alexandre/ An Overview of Blockchain Integration with Robotics and Artificial Intelligence/ 2018
- Plamen P. Angelov, Eduardo A. Soares, Richard Jiang, Nicholas I. Arnold, Peter M. Atkinson/ Explainable artificial intelligence: an analytical review/ WILEY/ 2021
- Mohamed Nassar, Khaled Salah, Muhammad Habib ur Rehman, Davor Svetinovic/ Blockchain for explainable and trustworthy artificial intelligence/ Wiley interdisciplinary reviews: data mining and knowledge discovery 10/ 2020
- Sunghyuck Hong/ Explainable Artificial Intelligence Study based on Blockchain Using Point Cloud/ Journal of Convergence for Information Technology/ 2021
- Erik Karger/ Combining Blockchain and Artificial Intelligence — Literature Review and State of the Art/ Conference: 2020 International Conference on Information Systems/ 2020
- Energy consumption of bitcoin mining/ Cambridge working papers in economics: 1948/ Sinan Küfeoğlu, Mahmut Özkuran/ 24 Mayıs 2019
- Recent developments in blockchain technology and their impact on energy consumption/ Johannes Sedlmeir, Hans Ulrich Buhl, Gilbert Fridgen, Robert Keller/ 24 şubat 2021
- A comparative analysis of current cryptocurrencies/ ICISSP 2018/ Lara Mauri, Stelvio Cimato, Ernesto Damiani
- Theories of cryptocurrency, blockchain and distributed systems and environmental implications/ Cryptocurrencies and blockchain technology applications/ P. Mary Jeyanthi/ 22 mayıs 2020
Join Coinmonks Telegram Channel and Youtube Channel learn about crypto trading and investing
Also, Read
- Bookmap Review | 5 Best Crypto Exchanges in the USA
- The Best Crypto Hardware wallet | Bitbns Review
- 10 Best Crypto Exchange in Singapore | Buy AXS
- Red Dog Casino Review | Swyftx Review | CoinGate Review
- Best Crypto to Invest in India | WazirX P2P | Hi Dollar Review
- Best Crypto Trading bots in Canada | KuCoin Review
- Crypto Trading Signals for Huobi | HitBTC Review