Where AI & Blockchain Intersect

As an article about AI, I thought it fitting for this to be my first article cowritten with ChatGPT.

AI and blockchain have a bad track record. On-chain computation is expensive and AI is computationally heavy, and most projects at the intersection have either failed or rugged. 

However, there’s not much that isn’t starting to be affected by the recent advancements in AI, and the crypto market is not an exception.

A few of the catalysts include (or will soon include):

• LLMs as middleware
• A creativity surplus (devs can focus more attention on the core or high-level tech)
• OSS + LLMs + Crypto Bounties = a worldwide community of potential contributing developers (with no dev knowledge required!)

Below are a few blockchain niches that go well with, and will be positively impacted by AI.

Analytics & Monitoring

One of the significant areas of interest at the intersection of artificial intelligence and blockchain technology is analytics and monitoring. The recent advancements in AI, specifically LLMs, open new possibilities for analyzing and processing on-chain data.

An example of this application is Dune's LLM Roadmap, which includes SQL query explanations and natural language search capabilities. By leveraging LLMs, Dune Analytics could process vast amounts of blockchain data more effectively, allowing developers and researchers to derive valuable insights that were previously buried deep in a sea of data.

Search Engines

Another promising area for AI integration into blockchain technology is generalized cross-chain blockchain search. As the number of blockchain networks and decentralized applications continues to grow, there is a pressing need for a comprehensive search tool capable of accessing and retrieving information across multiple chains.

By employing AI algorithms, a "Google" of blockchain data could be developed, enabling users to search and discover relevant information across disparate networks with ease. Some projects have already been funded around this concept.

Imagine, for example, searching ‘show me who’s on the Curve multisig’ and getting a list of addresses matched with their Twitter and Lens profiles where possible.

Fun

We created Pixelwhimsy, an on-chain NFT collection, without any help from developers, and it’s an excellent showcase of what’s possible with ChatGPT without in-depth specialty knowledge.

The combination of LLMs and diffusion models allows developers to integrate art with their code, and for creatives to self-publish their art, games, and immersive experiences.

In my article, How to Change History with NFT Art, I made the case that simply by keeping crypto fun, adoption will follow. Never did I imagine that only three months later we’d have these new tools at our fingertips to build more fun.

Taxes and Accounting

One of the barriers to mainstream adoption of DeFi is the complexity of managing taxes and accounting for crypto transactions.

The integration of LLMs as middleware to categorize on-chain events for tax purposes could significantly simplify this process. By parsing calldata and categorizing trades, income, and expenses, LLMs can potentially offer a more efficient and accurate solution than existing tax software.

This improvement would lower the barriers to entry for DeFi usage, leading to increased adoption and a more inclusive open finance ecosystem.

Risk Detection & Monitoring

Preliminary Auditing

LLMs can contribute to enhancing the security and reliability of blockchain networks by assisting in the preliminary auditing of smart contracts.

Testmachine AI, for example, is supposedly developing an AI-driven platform focused on finding and fixing smart contract bugs. Although LLMs will not eliminate the need for human audits, early detection by developers will allow auditors to dedicate their time to looking deeper for bugs.

Real-Time Risk Monitoring

Real-time risk monitoring is another critical aspect of blockchain security that can benefit from AI integration. Forta, an ML-based threat detection toolset, is already monitoring dApps with an aggregate value of $44 billion. As competition among auditing services intensifies, incorporating on-chain monitoring services could become essential for maintaining a competitive edge in the market.

Smarter Simulations

Introducing AI-driven agents in agent-based simulations can enhance the effectiveness of risk assessment and analysis. Companies like Gauntlet can leverage human-like agents to conduct more accurate and reliable simulations, providing valuable insights into potential risks and vulnerabilities in blockchain networks and dApps.

Distributed Social

The integration of LLMs into social media platforms opens the door for experiments in distributed AI content curation. By allowing users to customize their content preferences and employing AI algorithms to curate feeds accordingly, a new era of personalized and decentralized content consumption can be realized. This development has far-reaching implications for the future of social media, as detailed in my blog Pseudonymity and Social Media After LLMs.

Code Translation

An often-overlooked aspect of LLMs is their potential for code translation. By facilitating the translation of programming languages, LLMs can enable us to port our favorite dApps to specialized Layer 2 (L2) solutions, independent of language constraints. Of course, we’ll still need devs and audits, but the process should allow much faster proliferation of dapps without linear scaling requirements.

Conclusion

While challenges remain, the convergence of AI and blockchain technology holds great promise for a future that is more efficient, data-driven, secure, and engaging. Further research and exploration of these synergies will undoubtedly unlock new opportunities and drive innovation in the blockchain space.