AI/ML, big data in finance: advantages and effect on business models/activity of financial sector participants
2.4. Integration of AI in Blockchain-based financial products
Distributed ledger technologies (DLT) such as blockchain have become increasingly popular in various industries, particularly finance. This growth can be attributed to the advantages of speed, efficiency, and transparency these innovative technologies offer. Automation and disintermediation drive these benefits. The widespread adoption of DLTs in finance can increase efficiency by eliminating intermediaries in securities markets, including issuance and post-trade clearing and settlement. It can also be helpful in payments, such as central bank digital currencies and fiat-backed stablecoins, and can facilitate the tokenization of assets more broadly. Such adoption may reshape financial operators’ roles and business models, including custodians.
The finance industry promotes the convergence of AI and DLTs in blockchain-based systems to increase efficiency. However, the actual level of AI implementation in such projects is not significant enough to justify claims of convergence between the two technologies.
In reality, instead of convergence, what’s happening is that AI applications are being implemented in certain blockchain systems for specific use cases, such as risk management. Similarly, DLT solutions are implemented in some AI mechanisms, such as data management. The latter involves using DLTs to provide information to a machine learning model, using the blockchain’s immutable and disintermediated characteristics while enabling the sharing of confidential information on a zero-knowledge basis without violating confidentiality and privacy requirements. The use of DLTs in AI mechanisms is expected to allow users of these systems to monetize their data, which machine learning models and other AI-driven systems, such as IoT, are using. Implementing such AI use cases is driven by the technology’s potential to increase further efficiency gains of automation and disintermediation in DLT-based systems and networks.
AI has the potential to significantly enhance the automation capacity of smart contracts in DLT-based finance. Specific use cases of AI in DLT networks include risk management (such as anti-fraud measures and automated restrictions on network access) and data inference and control (such as improving the function of Oracles). However, most of these applications are still in the development phase.
AI can play a significant role in enhancing the security of blockchain networks, particularly in payment applications. It can identify any suspicious activity associated with theft or scams despite the need for public and private keys to compromise user security. Additionally, AI can improve onboarding processes on a network by using biometrics for AI identification and assist in AML/CFT checks for any DLT-based financial services. Integrating AI in DLT-based systems can also aid such networks’ compliance processes and risk management. For example, AI applications can provide wallet address analysis results that can be used for regulatory compliance or an internal risk-based assessment of transaction parties. However, it is essential to note that when financial intermediaries are eliminated from financial transactions, the effectiveness of current economic regulatory approaches focusing on regulated entities may erode.
Incorporating AI-based solutions in DLT-based systems at the protocol level can aid regulators in achieving their regulatory objectives more efficiently. This can be accomplished by means such as automated sharing of regulated entities’ data with authorities in a real-time and seamless manner, as well as programming regulatory requirements into the code of the programs for automatic compliance. One of the solutions the market suggests for addressing the challenges of supervising decentralized networks without a single central authority is the participation of regulators as nodes in such networks.
Incorporating AI into DLT-based systems could enhance the quality of data inputs into the chain. The responsibility of data curation would shift from third-party nodes to independent, automated AI-powered systems, making information recording and sharing more robust and challenging to manipulate. This would be particularly beneficial for third-party off-chain nodes, like Oracles, that provide external data to the network. These nodes are vulnerable to errors and malicious data feeds, which could compromise the integrity of the network.
Incorporating AI could further increase disintermediation by bringing AI inference directly on-chain, rendering third-party information providers redundant. Moreover, AI could act as a safeguard by testing the honesty and truthfulness of the data provided by the Oracles. This would prevent cyber-attacks or manipulation of third-party data provision into the network.
Theoretically, AI applications can increase participants’ trust in the network by allowing them to verify the accuracy of information provided by Oracle and detect any system compromise. However, introducing AI does not necessarily solve the “garbage in, garbage out” problem. Poor quality or insufficient data inputs remain a challenge observed in AI-based mechanisms and applications (as discussed in Section 3.1).
2.4.1. AI expanding the powers of smart contracts
The integration of AI techniques in blockchain-based systems has the potential to impact the governance and risk management of intelligent contracts significantly. In this context, the most practical application of AI is self-regulated DLT chains that can operate autonomously. While the effects of AI on the roles and processes of DLT-based networks remain hypothetical and untested, it is clear that this technology has the potential to revolutionize the entire industry.
Smart contracts have been around for a long time and are based on simple software code. Currently, most smart contracts being used do not incorporate AI techniques. Therefore, many of the proposed benefits of using AI in DLT systems are still theoretical, and claims made by the industry regarding the convergence of AI and DLT functionalities in marketed products should be approached with caution.
Intelligent contracts are crucial to blockchain technology but are not immune to errors and vulnerabilities. This is where AI comes in handy. AI tools can enhance the security of smart contracts and manage risks by analyzing the patterns of intelligent contract execution. For instance, Natural Language Processing (NLP) can be used to detect fraudulent activities and identify flaws in the code. Additionally, AI can test the smart contract code more quickly and accurately than a human code reviewer. As smart contracts are based on code automation, it is essential to ensure that the code is flawless to ensure the robustness of the contract.
Box 2.4. Intelligent contracts in DLT-based approaches
Smart contracts are self-executing agreements written as code on Blockchain ledgers, automatically executed when predefined trigger events occur.
Smart contracts are essentially computer programs that run on the Ethereum blockchain. The code of these programs determines their operation and timing. These contracts are designed to define rules just like traditional contracts, but with the added benefit of automatic enforcement through the code once the specified conditions are triggered.
Smart contracts run on the network and execute functions when users interact by submitting transactions.
Smart contracts enable DLT-based networks to benefit from disintermediation and are a significant source of efficiency that such networks promise to offer. They allow for the automatic execution of actions such as payment or transfer of assets based on specific conditions set in the code without human intervention. However, the legal status of smart contracts is still not defined in most jurisdictions, which raises enforceability and financial protection concerns. The audibility of the code of such intelligent contracts also requires additional resources from market participants who would wish to verify the basis on which such clever agreements are executed.
The integration of AI into smart contracts holds the potential to enhance their automation capabilities. This can be achieved by increasing their autonomy and adjusting the underlying code dynamically to market and environmental conditions. NLP, a subset of AI, can potentially increase the analytical reach of smart contracts associated with traditional contracts, legislation and court decisions. NLP can further assist in analyzing the parties’ intent, as reported by Technologist in 2020. However, it is essential to note that these AI applications for smart contracts are still theoretical and have yet to be tested in practical scenarios.
Operational risks, compatibility, and interoperability challenges must be addressed between conventional infrastructure and DLT-based and AI technologies. AI techniques, such as deep learning, demand significant computational resources, which may hinder their performance on the blockchain. According to Hackernoon (2020), data storage off-chain could be a better option for real-time recommendation engines at this stage of infrastructure development, as it could prevent latency and lower costs. As the technology and applications it facilitates mature, operational risks associated with DLTs must be resolved.
2.4.2. Self-learning intelligent contracts and governance of DLTs: self-regulated chains and Decentralised Finance (DeFi)
Smart contracts powered by AI have the potential to create self-regulated blockchain networks. Researchers suggest that AI can even be used to automate and predict events in these intelligent contracts using techniques such as reinforcement learning (Almasoud et al., 2020). In other words, AI can extract and process real-time information, which can be fed into intelligent contracts to adjust their code automatically. The result is a fully autonomous, self-regulated, decentralized blockchain network without human intervention.
Decentralized autonomous organizations (DAOs) exist as independent codes on the blockchain. They have already existed but could be further facilitated by AI-based techniques. For instance, AI could provide real-time data feeds as inputs to the code. The code would calculate a desired action based on the data provided. Self-learning intelligent contracts powered by AI would be vital in adding new features to the chain’s logic. They would learn from the experience of the chain and adjust or introduce new rules, essentially defining the overall governance of the chain.
DeFi projects are typically managed by DAOs with centralized aspects, such as on-chain voting by governance token holders and off-chain consensus. Human intervention could be a management point for controllers. However, integrating AI into DAOs could facilitate different decentralization and decrease the enforceability of conventional regulatory guidelines.
Using AI to develop fully autonomous chains presents significant challenges and risks to its users and the broader ecosystem. In such environments, AI smart contracts would execute decisions and operate systems without human intervention, raising critical ethical considerations. Moreover, introducing automated mechanisms that instantly switch off the model (known as ‘kill switches’) is highly challenging in such decentralized networks, which is also a significant issue in the DeFi space. Section 2.2 provides definitions for these terms.
AI integration in blockchain technology can support decentralized applications in DeFi by increasing automation and improving efficiency in providing financial services. For instance, the application of AI models can enable the provision of personalized recommendations across various products and services, credit scoring based on users’ online data, investment advisory services and trading based on financial data, and other reinforcement learning applications on blockchain-based processes. Although deploying AI in DeFi can enhance the capabilities of the DLT use case by providing additional functionalities, it may not significantly alter any of the business models involved in DeFi applications, similar to other blockchain-based financial applications.
