Crypto’s Next Frontier: AI & DeFi
Authors: 0xbeachball and AmberShi5
AI has captured the attention of millions of people around the world. Rapid improvements in publicly available technology has driven fascination with models including DALL-E 2 and GPT-3. The potential applications of artificial intelligence are practically limitless and will eventually touch every aspect of our societies. These machine models are already surpassing the capabilities of humans in the realm of data analysis and prediction and will only improve exponentially as time goes on.
Classification, pattern recognition, and predictive capabilities make AI models an ideal fit for financial applications that depend on large, real-time data sets. This requirement makes AI a natural fit with decentralized applications and blockchains. Sufficiently trained AI models could create unmatched insights when provided with the ordered, immutable, and fully transparent data created by blockchain ecosystems.
At New Order, we’re focused on building next generation decentralized projects. So, what’s possible when it comes to the intersection of artificial intelligence and decentralized applications?
Predictive AIs Will Dominate Market Making & Liquidity Control
Concentrated liquidity within AMMs (like that deployed in Uniswap v3) are an effective way to combat impermanent loss and slippage within trading pools. That said, concentrated liquidity still requires active management from liquidity providers and is often far from optimized. AI models today are generally built in a predictive format, meaning that they respond to data by providing the next most likely output in a given sequence. This architecture makes AI models an ideal fit for predicting the concentrated liquidity needs within an AMM at a given moment in time. The use of reinforcement learning in market making models is already taking place. We expect that AI systems will dominate market making and liquidity routing across DeFi as models evolve.
AI Auditors Will Augment Their Human Counterparts
Smart contract hacks and protocol exploits continue to be the number one risk vector in decentralized finance markets. Detecting vulnerabilities in smart contracts has emerged as a massive industry that has a major influence on the release of innovative protocols. Deep learning and progressively more sophisticated neural networks will ultimately lead to models significantly augment the capabilities of their human counterparts. These models could eventually surpass their users capabilities, however we view this technology as a powerful tool to make existing audit processes more exhaustive and effective. These methodologies are already being explored with some researchers achieving sub-second vulnerability detection with 97% precision.
Black Swan Detection Will Dramatically Improve Sector-Wide Security
As a result of their highly pattern-focused nature, AI models are uniquely good at recognizing anomalies in a constantly evolving dataset. In the context of DeFi, AI-powered anomaly detection models could be used to anticipate large hacks and pause contracts before they happen. For example, large bridge hacks (like those that occurred with Wormhole and Nomad) could be predicted and mitigated in real-time. Some pre-transaction anti-fraud systems already exist today (ex. LayerZero’s Pre-Crime) and we expect these systems to become significantly more effective with AI capabilities built in.
Arbitration-Based Models Will Become Scalable With AI
The decentralization of governance has led to a shift in decision-making, where instead of decisions being made by a single individual, decisions are made by a distributed group of people. For instance, in the case of Nexus Mutual, a decentralized insurance protocol, claims are evaluated by a panel of arbitrators that comes to a consensus on claim legitimacy. This process could be improved by the integration of a sufficiently advanced AI. AI models can analyze claim-related data and provide recommendations to arbitrators, or even automate the arbitration process by identifying relevant data and issuing rulings.
The potential applications of artificial intelligence in the crypto sector are significant. The ability for models to be both generalized or niche prediction machines along with the wealth of readily available, on-chain data is a powerful combination. We do not expect AI to immediately take over every field (ex. auditing), however we do foresee rapid adoption of these models in liquidity provision and attack detection in the near future. These technologies will ultimately improve the efficiency and security of crypto-native infrastructure and financial markets.
If you’re a founder or builder looking to launch a project at the intersection of AI and DeFi, get in touch.
