Decentralized finance (DeFi) has been capturing the minds and hearts of the crypto community in the last few months. DeFi offers a new and transparent playground to automate financial primitives such as lending or market making without the need of intermediaries.
The on-chain centric nature of DeFi protocols generates a data footprint unprecedented in capital market vehicles. This data can be used by machine intelligence models to better understand and benefit from the behavior of DeFi protocols.
Currently, this level of DeFi intelligence is used mostly as part of third-party applications such as arbitrage bots or yield generating strategies. As DeFi evolves, intelligence should move from a third-party application to a core capability of DeFi protocols.
In simpler terms, while the current generation of DeFi protocols has brought us unprecedented levels of programmability and automation, the next wave will focus on intelligence as a core capability. We are on the verge of the intelligent DeFi era.
Software is eating the world and artificial intelligence (AI) is eating software. Most of the world’s software is being rewritten, with AI as a first-class citizen. Just like databases or APIs, AI is becoming a fundamental building block of modern software applications. Business and consumer applications across diverse domains are becoming intelligent in nature and so is the infrastructure powering them. DeFi is likely to undergo a similar transformation.
The cornerstone of the first generation of successful DeFi protocols has been the automation of financial primitives such as lending or market-making in the form of programmable interfaces. That level of programmability and automation has started to push the boundaries of innovation creating vehicles such as flash loans, automated market-makers (AMM) or on-chain insurance that are impossible to even imagine in other capital markets.
The focus of this generation of DeFi protocols has been on programmability, but the next generation should look towards intelligence. There is something beautiful, almost seductive, about the simplicity of the x*y=k balance in UniSwap or the liquidation logic in Compound. But the evolution of the market is likely to push us towards more sophisticated DeFi protocols that leverage machine intelligence as a first-class citizen. These types of protocols won’t only enable robust levels of automation, but they will be able to learn and evolve their behavior based on market activity.
What does DeFi intelligence really mean? In plain terms, imagine a group of DeFi primitives that intelligently evolve their behavior based on data about the historical performance or conditions of the crypto market. These intelligent DeFi primitives would not only bring new levels of efficiency to that market but they can also be combined to enable a new generation of smart, decentralized asset management vehicles.
Let’s explore a few ambitious ideas in this area.
As its name indicates, the current generation of AMMs such as UniSwap, Balancer or PancakeSwap are focused on automation. On top of AMMs, traders have built all sorts of smart strategies ranging from questionable front-running mechanisms to incredibly clever arbitrages. What if that intelligence could be incorporated as part of the underlying AMMs?
Let’s imagine a new generation of AMMs that adjust the balances based on oracles that use predictive models to forecast the relationship between different assets. This type of intelligent AMM will learn from factors such as historical market performances of the behavior in a given set of liquidity pools to determine an asset distribution function that correctly reflects the current market conditions. Among its many benefits, intelligent AMMs can help reduce the excessive levels of arbitrage that we are experiencing in the current generation of AMMs.
Intelligent DeFi lending
Lending in DeFi protocols is another area that is likely to incorporate intelligence capabilities as a first-class building block. Products like flash loans have already proven the possibilities of DeFi to enable lending models that are not possible in traditional market vehicles. The current generation of lending protocols such as Maker, Aave or Compound are, for the most part, based on collateralization models and still ignore many other factors that can enable robust borrower-lender dynamics.
As the DeFi lending evolves, we should expect the intelligence of lending protocols to evolve with it. We could easily envision DeFi lending protocols that would factor not only collateralization levels but factors such as liquidation history, interactions with other lending protocols and transactional patterns to develop an intelligent and quantitative profile of borrowers and lenders. Think about this trend as an intelligent credit rating system for DeFi protocols.
Insurance is becoming an established component of the DeFi ecosystem. Platforms such as Nexus Mutual are introducing a level of stability in DeFi by insuring smart contracts across different platforms. But current insurance models are based on the characteristics of specific DeFi platforms but, at the moment, leverage very little intelligence.
In a world in which the interactions with DeFi protocols are completely transparent, that data footprint can be used to create smarter insurance models. Let’s imagine a model in which the insurance of specific smart contracts or addresses is based on intelligent models that use quantitative DeFi factors such as the number of liquidations in lending protocols the smart contract has been exposed to or the types of pools specific addresses are supplying liquidity. These factors could provide a very intelligent profile about specific addresses that could be incorporated into decentralized insurance models.
DeFi intelligence is not optional
Willian Gibson’s famous quote “The future is already here – it’s just not very evenly distributed“ certainly applies to the role of intelligence in the next generation of DeFi protocols. At first, the idea of intelligent DeFi protocols might seem like one of those “AI for X” ideas (add your favorite term: recruiting, accounting, CRM, etc.) that attempts to insert AI into any meaningful software trend in the market. However, the reality of the fintech space in general and DeFi in particular paint a very different picture.
Intelligence is not a fancy requirement of a key component of the next phase of DeFi protocols. Functionally, DeFi is attempting to automate and dis-intermediate the core building blocks of financial services. At the same time, the rapid evolution of machine learning technologies is making these traditional services in capital markets become more intelligent.
Think about the increasing levels of machine intelligence displayed by robo-advisors like Betterment, active managers like Two Sigma or even brokerage services like Interactive Brokers. Competing with that level of intelligence requires more than automation and programmability. It requires intelligence.
The frantic evolution of DeFi together with the mainstream adoption of machine learning technologies are forging the path towards intelligent DeFi protocols. This new generation of DeFi protocols won’t automate only financial primitives in a decentralized manner but will intelligently evolve their behavior based on market conditions and utilization.
DeFi is one of the most disruptive technology trends in modern fintech. Machine learning and AI technologies are becoming a key building block of any modern software system. The combination of these two movements seems inevitable and can unlock a new level of innovation in DeFi that powers a brand new generation of decentralized, trust-less, intelligent financial services.
The leader in news and information on cryptocurrency, digital assets and the future of money, CoinDesk is a media outlet that strives for the highest journalistic standards and abides by a strict set of editorial policies. CoinDesk is an independent operating subsidiary of Digital Currency Group, which invests in cryptocurrencies and blockchain startups.