The convergence of AI and digital assets: A new dawn for financial infrastructure

by Forrest Colyer, John Liu, and Michael Greenwald | on

The financial landscape has been in a constant state of evolution. From stock ticker machines to algorithmic trading systems, innovation has always been at the core of finance. Yet, among these transformative changes, the confluence of artificial intelligence (AI) and digital assets like cryptocurrencies, central bank digital currencies (CBDCs), and tokenized assets has the potential to be the most disruptive change to finance in decades. This post aims to explore this intersection, focusing on how AI and digital assets are reshaping financial services and infrastructure.

The history of AI and traditional finance

AI has been used in impactful ways in the financial industry for more than two decades. Although basic computational models and statistical methods such as standard deviations and Bayesian regressions have been in use since the 1980s to generate trading signals, machine learning (ML) gained significant traction in the 2000s and early 2010s in the areas of high-frequency trading and risk modeling. This increase in traction was driven by increasing accessibility to data, dropping costs to store data, and increasing computational power. Financial institutions employed increasingly complex AI such as neural networks powered by additional datasets such as credit card data. The use of AI spread to other areas of finance, such as fraud detection, consumer credit scoring, and customer service. By mid-2010s, the impact of AI was transforming personal finance with the availability of robo-advisory platforms through startups (such as Betterment) and established financial institutions (such as Charles Schwab) alike. In addition, advancements in deep learning and reinforcement learning models continued to improve the efficiency of the financial industry. With the increasing accessibility of foundation models, such as through Amazon Bedrock , they are now able to power generative AI and revolutionize the financial industry. For example, generative AI will enable financial companies to unlock value and create new products by analyzing vast amounts of previously untapped unstructured textual data, which according to IDC’s white paper “ Untapped Value: What Every Executive Needs to Know About Unstructured Data ”, is estimated to comprise 80–90% of all existing data.

At the same time that ML utilization was increasing in finance, Bitcoin’s launch in 2008 laid the foundation for a new form of digital currency and network for exchanging value. Bitcoin, which operated on a decentralized network with a single source of truth that could be accessed by anyone, revolutionized the concept of value transfer and preservation. By 2012, Bitcoin and other altcoins such as Litecoin started gaining mainstream attention. In 2015, Ethereum went live, introducing smart contracts, programs stored on the blockchain that enables business logic to be facilitated without the need for intermediaries, opening the door for decentralized applications (dApps). Innovators created and battle-tested smart contracts with borderless digital currencies worth billions of dollars to create the decentralized finance (DeFi) industry in 2019, a non-custodial financial system that doesn’t require a central system. DeFi has enabled lending, investing, and exchanging for anyone with an internet connection. Stablecoins, digital currencies pegged to fiat currencies such as the dollar, have also enabled more efficient payment and settlement methods and have seen growing utilization from major payment processors such as Visa and, more recently, PayPal. Countries have also been moving forward with CBDC plans, with 11 CBDCs launched and 21 in pilot. Improvements to increase the transaction scalability using Layer 2 technologies and emerging zero-knowledge technologies, which share data while preserving privacy, are further enabling digital assets to support critical financial systems.

Convergence of AI and digital assets

The convergence of these two emerging technologies—AI with its ability to self-learn and extract actionable insights from vast amounts of data and digital assets, which exist on a decentralized, transparent, and automated network—is expected to further amplify the impact of one another. The following are a few ways in which combining the strengths of both technologies can develop a more efficient and robust financial system.

In order for AI to be effective, a trusted source of data needs to be accessible and on-demand. Digital assets, which use blockchain technology, allow financial transaction data to be stored in a trusted, accessible, and transparent fashion. AI can access data such as transaction history and current balances for real-time financial analysis. For example, AI trading algorithms can use Amazon Managed Blockchain (AMB) Query’s straightforward APIs to access the current and historical balances of a digital asset across all accounts on a blockchain with subsecond latency. The algorithms can then fine-tune its rules based on new transaction-patterns or market paradigms, and use AMB Access to swap digital assets when the new rules are met. AI can also combine data not stored on the blockchain (known as off-chain data) with data on the blockchain (known as on-chain data) to derive further insights. For instance, AI tools that analyze Ethereum smart contracts could track the impact of traditional off-chain market prices, such as equity markets, to on-chain lending platform activity, such as liquidations or platform withdrawals. Based on these complex relationships, the AI could generate new signals to remove liquidity from lending platforms when equity markets decrease a certain percent.

The combination of AI, programmable digital assets, and smart contracts can create a financial system that can efficiently run complex tasks and enforce financial agreements without human intervention. For example, AI algorithms can trigger smart contracts to buy and sell assets when market conditions are met or freeze digital assets from further transfers when fraudulent activities are detected. Smart contracts can automatically record each step of an AI algorithm, providing a transparent and immutable audit trail for compliance or further training of AI algorithms. As AI continues to learn from the trusted dataset, it can in turn adapt and optimize its algorithms as well as smart contracts to new market conditions. Additionally, digital assets on blockchain can also protect against fake digital assets, which are increasingly easy to create with generative AI. Users of digital assets can validate the authenticity of a digital asset by checking the issuing contract against a shared digital asset registry controlled by authorized parties.

With increased automation and digitization, cybersecurity takes on greater importance. AI can monitor transactional data in real time, identifying and flagging any unusual activities. Advanced ML algorithms can predict future attack vectors based on existing patterns, providing an additional layer of cybersecurity. As foundation models get trained on the trusted digital asset transactions and smart contract code on blockchains, generative AI can help developers write more secure code. Technologies like Amazon CodeWhisperer are speeding up coding tasks, providing inline code suggestions to improve developer productivity. In addition to code suggestions, Amazon CodeWhisperer can detect hard-to-find security vulnerabilities and flag emerging security vulnerabilities during development. Generative AI can also help generate documentation for smart contracts to make them easier to understand. This transparency will be crucial when smart contracts are potentially managing billions of dollars in digital assets and settlement. An important component enabling this innovation is the way that data is stored, accessed, and leveraged to promote learning in AI models. Since data is critical for any computer’s ability to learn, the cloud has become vital to anyone building compelling solutions in the age of AI. With Amazon Bedrock making a variety of leading generative AI models from Amazon, A121 Labs, and more through straightforward APIs, we expect more innovative applications of generative AI with digital assets.

Additionally, nascent but growing zero-knowledge (ZK) technologies which are being used to enable transaction scalability (for example, with Layer 2 chains) and protect privacy on a public blockchain, are well-positioned to bridge the gap between traditional centralized finance and DeFi, bringing even more assets into the digital world. Zero-knowledge proofs (ZKPs) enable new forms of data sharing that allow transacting parties to verify information without revealing the underlying data. The applying party submits a mathematical proof to demonstrate that they hold the needed information or meet the set criteria, and the recipient confirms this proof without ever seeing the underlying information internally. For example, if a bank customer sends proof showing that they meet the criteria for a loan, the bank can verify this proof mathematically, without ever seeing the applicant’s sensitive financial information such as income and bank accounts. ZKPs can also enable compliance with KYC/AML regulations without storing honeypots of customers’ personally identifiable information (PII). This technology has great potential in that it directly addresses the larger data privacy concerns financial services institutions have long had with decentralized finance and public blockchains. Zero-knowledge rollups (ZK rollups) are a more sophisticated application of ZK tech that roll up ZKP transactions into bundles and then run them offline (off the blockchain). Running the transactions off-chain reduces the total amount of information posted to the blockchain while still providing proof that the transactions were executed correctly, thereby mitigating security and privacy concerns while still maintaining decentralization. ZK rollups are faster and cheaper than on-chain transactions, positioning them as an especially scalable and secure alternative to fully public blockchain transactions. This increased scalability and security will be important in onboarding the even larger set of population adopting digital finance in the future.

Impact of digital infrastructure outside of US

The economic landscape is evolving rapidly with the emergence of various fiat currencies, central bank digital currencies (CBDCs), and cryptocurrencies that each represent a set of goals or priorities set forth by the issuing body. Recent developments, such as the implementation of economic sanctions have ignited isolationist sentiment across the world, leading to the fracturing of interests, values, and leadership which could create an economy that is based on a basket of currencies rather than one that hinges primarily on the performance of the US Dollar/US economy. This effect, known as “de-dollarization”, could further elevate the importance of the respective countries’ digital infrastructure. According to the 2021 edition of Facts and Figures from International Telecommunication Union (ITU) , an estimated 2.9 billion people were still not connected to the Internet and telecommunications infrastructure, with more than 95% of them living in developing countries. This statistic underscores the dramatic change that is still yet to occur across the globe as a significant portion of the world’s population gains access to the Internet. At the same time, populations that have never before been connected will leapfrog more primitive, fixed-wire Internet infrastructure and instead receive 5G capabilities that promise higher connectivity for the Internet of Things (IoT), low latency, high reliability, and efficient energy usage. Along with the increase of 5G infrastructure comes the need for resilient cloud infrastructure that serves to protect data in transit and support reliable systems functionality. As the world continues its globalizing trend and more people gain connectivity, one of the largest challenges in this transition will be financing new infrastructure projects that support telecommunications. Beyond that, governments and the private sector must work together to advance projects that provide smart, agile, cloud-based systems for their citizens. According to the study, “ The Economic Impact of the Market-Making Internet – Advertising, Content, Commerce, and Innovation: Contribution to US Employment and GDP ,” in 2020, the Internet economy contributed $2.45 trillion to US GDP—more than 10% of annual GDP—and created 17.6 million jobs. This statistic shows that connectivity to the internet provides nations with the chance to supercharge their economy. It all starts with infrastructure. Whether across modern economies or developing countries, the role of digital infrastructure will become increasingly intertwined with more and more economic activity in the twenty-first century; the companies that embrace and enable this infrastructure transformation will be the leaders of the digital economy.

These tools, along with customer support and cloud services, could greatly benefit the evolving operations of both countries and corporations as they undergo the transition from legacy infrastructure to more automated infrastructure rails. The intersection of cloud computing, AI technology, digital assets, and data storage with cybersecurity tools will define global economic competition in the twenty-first century, and the companies and technologies that create this future will be at ground zero for helping define it. The transition to automation and more efficient processes to manage these technologies starts with infrastructure investment and build.

Conclusion

The convergence of AI and digital assets could bring a paradigm shift in financial infrastructure. The mutually reinforcing advancements in AI and digital assets have the potential to democratize financial automation, enhance efficiency, and significantly ramp up security protocols. A future where AI is processing both on-chain and off-chain data will require a new infrastructure to cohesively function. Decentralized applications are frontend user interfaces built onto smart contracts, enabling transactions to be run without sharing personal information between the parties. Generative AI tools can be trained on both off-chain and on-chain data to optimize and secure the smart contract ecosystem. Today, generative AI assists smart contract parties through chatbots and virtual assistants in the writing of smart contract code and monitoring any unusual network activity that may represent a security risk. The more data these generative AI tools have access to, the more effectively they will spot errors or security risks, enhancing their ability to proactively identify and mitigate risks in real time. At the same time, the blockchain networks on which these digital assets exist provide transparency, a single source of truth, and provenance which help validate the authenticity of the information being actioned on by AI or used to update AI. Although there are still challenges that need to be solved, innovators, armed with more powerful and accessible cloud technologies, including instances purpose-built for ML such as Amazon Web Services Inferentia and Amazon Web Services Trainium , are getting closer to unlocking the combined benefits of AI and digital assets every day.


About the authors

Forrest Colyer manages the Web3/Blockchain Specialist Solutions Architecture team that supports the Amazon Managed Blockchain (AMB) service. Forrest and his team support customers at every stage of their adoption journey, from proof of concept to production, providing deep technical expertise and strategic guidance to help bring blockchain workloads to life. Through his experience with private blockchain solutions led by consortia and public blockchain use cases like NFTs and DeFi, Forrest helps enable customers to identify and implement high-impact blockchain solutions.

John Liu is the Head of Product for Web3 / Blockchain at Amazon Web Services. He has 13 years of experience as a product executive and 10 years of experience as a portfolio manager. Prior to Amazon Web Services, John spent 4 years leading product and business development at public blockchain protocols with a heavy focus on cross-chain technology, DeFi, and NFTs. Prior to that, John gained financial expertise as Chief Product Officer for fintech companies and portfolio manager at various hedge funds.

Michael B. Greenwald is a senior executive for Amazon Web Services serving as the Global Lead for Digital Assets and Financial Innovation. Michael serves as a representative member for Amazon Web Services of the U.S. Commodity Futures Trading Commission (CFTC) Technology Advisory Committee. He is an adjunct professor at Columbia University and a senior fellow at the Atlantic Council Geoeconomics Center and Center for New American Security. He was a former fellow at Harvard Kennedy School’s Belfer Center. He was the first U.S. Treasury attaché to Qatar and Kuwait, acting as the principal liaison to the banking sector in those nations for two presidential administrations from 2010-2017. He holds a Juris Doctor from Boston University, a Master’s from Boston University’s Frederick S. Pardee School of Global Studies, a Bachelor of Arts in History from George Washington University and is currently pursuing the General Management Executive Education Program at Harvard Business School.