AI + Blockchain? Accenture Got it All Wrong!
Blockchain technology often confuses people, and this confusion mainly stems from the several misconceptions that I have already mentioned in the two previous articles (8 most common misunderstanding of blockchain part 1 and part 2).
The misconceptions are hindering laymen and sometimes experts to understand the technology and build real solution on it.
Knowing the kind of mistakes we are committing is just as important as knowing how to fix them. Today, I attempt to explore how to build a real application on blockchain by taking the Accenture report (issued on 11th Oct, 2018), which contains numerous mistakes, as an example.
This article is written in two parts; the first is about the mistakes in the paper, and the second provides suggestions on how different industries can tap into the value of blockchain.
1. Blockchain Does not help Data Sharing
“Now imagine a world where, through a blockchain-based system, every party involved in the move could see the data and information pertaining to the end-to-end relocation process with appropriate permissions granted by the stakeholders- in this instance, the family.
Organizations could access just the data they need, and because the information reflects blockchain’s enhanced trust levels, parties no longer need to engage in messaging and reconciliation. As organizations widely implement these kinds of models, the possibilities for AI systems will grow significantly” — page 4
In layman’s terms, the above sentence suggests that companies surrender the data they own (or steal), encrypt it by users’ private key and upload it on blockchain. After that, these companies can access this data only with users’ permission. In the event that a user loses his/her private key, his/her data will be lost permanently because nobody controls the system. How this system even makes sense is unknown, but it sounds legitimate with the word blockchain.
People often think blockchain helps data sharing because of two keywords: security and trust, which are also misconceptions that I am going to explain in points 4 and 5.
2. A smart contract is not a contract for businesses in digital form
“In the future, wider access to data across an ecosystem and the advances in automated business logic via smart contracts could enable new and greater access for AI machines to traverse business ecosystems and deliver more comprehensive solutions to customers.” — page 5
A smart contract is merely a program on blockchain. A smart contract that is running on a public blockchain is an immutable, censorship-resistant program. By contrast, a smart contract running on a permissioned chain is merely another regular program.
Smart contracts running on a public blockchain are sometimes called decentralised applications (DApp). Decentralising applications can be helpful. For example, we can look at decentralised gambling applications (which account for 40% of total blockchain transactions [Standard Kepler Research]).
These applications do not have licenses; users do not know the operators, and no protection is in place for gamblers. Yet, users can trust them because decentralisation guarantees the immutability of the code. That is, the agreed program code cannot be altered.
Decentralised applications place the core logic on blockchain and establish fully automated execution. Consequently, these applications behave like ‘contracts’, and users can trust the program to deliver what is claimed. However, the mere placement of a program code on blockchain does not decentralise the application.
Moreover, we need to understand that decentralisation is not necessary to achieve automation unless the goal is creating an application or service that cannot be controlled or wiped out by anyone (e.g. a gambling software without a license or an application for use by North Korean generals in buying and selling properties in New York).
I do not see much business use cases here. Thus, I will not spend considerable time on it, given that this article is written for business owners.
3. Decentralised artificial intelligence (AI) versus full automation
“The processes of a decentralized management, evolution, adaptation and cooperation in MAS, consisting of a large number of reactive agents, are investigated and stimulated. “ — page 7
The beauty of decentralisation is censorship resistance. If a decentralised AI is to be deployed on public blockchain (given sufficient public blockchain throughput), then an immutable AI that is controlled by nobody must be expected. However, an uncontrollable AI does not mean anything to a business organisation.
Meanwhile, an AI program deployed on a permissioned blockchain will not differ from any other privately owned AI program.
4. Does blockchain redefine trust?
“Blockchain is a new type of database system that maintains and records data so that multiple stakeholders can confidently and securely share access to the same data and information. As such, it is changing the nature of boundaries between organizations. Since the invention of modern databases in the 1950s, the governing business model concerning them has centered on trust. For example, Party A needs to have confidence that Party B (or anyone else) hasn’t unilaterally changed any data. Consequently, companies traditionally build data systems they can fully control and operate using a “messaging” based business model. In this case, Party A sends its view of the world in a “message” to Party B, and vice versa. Only when both parties can reconcile those views will they complete the business transaction. Blockchain is changing that concept of trust in data. Through blockchain and other types of Distributed Ledger Technologies (DLT), companies can now access a common shared data set that they and other stakeholders know they can trust.” — page 12
This is one of the most common misconceptions that I have mentioned in my previous article — ‘By its nature, every piece of transaction data stored on blockchain should be self-evident. But not every piece of data stored on blockchain is transaction data. Moreover, blockchain technology itself cannot make general data self-evident (e.g. the coffee is from Ethiopia).’
The above misconception leads us to think about which data should be and not be recorded on blockchain.
Only data that can be verified by previously stored data should be stored on blockchain. I call this kind of data recursive data, which means that if the previous data is correct, then the current data must be correct.
Only data that can be verified by previously stored data should be stored on blockchain. E.g. Transactions, ownerships and account balances.
The chain characteristic allows us to easily trace the history of the data down to its origin. Hence, data that can be verified by previously stored data can be proven right or wrong without others’ help. Transactions, ownerships and account balances are data that can be verified by previous data. If all previous transaction records are correct, then the account balance must be correct.
However, this self-evident ability does not apply to other type of data, such as medical data, electricity consumption and farm records. One cannot determine whether maize is organic merely by checking the previous record, even with the immutability of the data.
5. Blockchain improves data security. However, you may not want to pay the price
“Security: Protection and control can be implemented at the data element level instead of the database or data table levels making it much more difficult to penetrate.” — page 12
This statement is true. It means that every data is encrypted, and they can be decrypted by users only. Blockchain is secure, but a company must ensure that it indeed needs this level of security and is willing to make the relevant sacrifices in terms of data usability.
The logic of the paper lies in AI’s need for training data and blockchain’s ability to facilitate the data sharing across different parties.
Most companies can already share data with third parties as long as they are willing to do so. We do not need to move data to blockchain simply because we want to share it across different parties.
Therefore, the question is as follows: Can storing data on blockchain create additional benefits regarding security, data authenticity, data availability and privacy protection? We often ask the wrong question of what blockchain can do, when the question should be ‘What can blockchain do better?’
Instead of “What blockchain can do”, we should ask “What can blockchain do better?”
The Accenture report lists six use cases of blockchain in different industries, namely, smart energy and smart buildings, public science, supply chain, smart devices, identity and healthcare. In the next article, I will explain how these industries can tap into the power of blockchain technology and transform their core businesses.
Originally published at https://keplerlab.io on May 15, 2019.
AI plus Blockchain? Accenture is doing it wrong! — Kepler Blockchain Lab was originally published in Hacker Noon on Medium, where people are continuing the conversation by highlighting and responding to this story.