The journey from web1 to web2 was like moving from a “read-only” state to interactivity. Web2 was characterized by the rise of social networking and e-commerce, the dominance of a few big tech companies, and concerns about data ownership and user privacy. The next internet evolution, web3, promises a more dynamic internet, offering widespread asset ownership, decentralized infrastructure––meaning without any intermediaries––and more privacy and transparency than ever before.

As web3 matures at the infrastructure level and permeates more verticals, developers need more sophisticated technology to make it easier to build at scale. New blockchain protocols, decentralized applications, smart contracts, and even non-fungible tokens (NFTs) require intelligent ways to detect fraud, consume large amounts of data, and analyze user behavior. To do this, web3 developers need to leverage the power of computing to analyze large datasets, understand patterns, and accurately predict future steps. This is where artificial intelligence (AI) can play an important role. Through pattern recognition and learning from large amounts of data, AI can help web3 builders with predictions and automation, reducing repetitive tasks and human errors. 

In this article, we explain how AI can be applied to web3 at the infrastructure level and highlight the top verticals poised to benefit.

Are we at an inflection point?

In a McKinsey report on recent technology trends––such as clean tech, mobility, bioengineering, applied AI, and web3, among others––applied AI ranked highest for sector innovation, based on patent filings and research publications, and relatively high for interest, based on news mentions. Web3, on the other hand, scored relatively high for interest, but much lower for innovation, suggesting there’s been far more patents and published research in AI than in web3. So, since both AI and web3 have attracted considerable investment, how can web3 become more innovative? The answer may lie in the intersection of web3 and AI. 

Although developers are continuing to build on the blockchain, wider adoption of web3 continues to be hindered by challenges in its security, scalability, interoperability, and user experience. That’s where innovation in AI can be utilized—to make web3 more efficient. As AI continues to improve different industries, it’s only a matter of time before it touches web3. One key driver: both missions are compatible. The promise of AI is to help individuals and companies make sense of complex data, while the promise of web3 is to enhance trust and privacy without intermediaries—AI can help deliver on that promise. We believe that growth in AI and web3 will build on each other and only compound over time.

Infrastructure level use cases 

In web3, the primary levels of infrastructure are blockchains, decentralized applications, and smart contracts:

• Blockchains are shared databases––they are ledgers in which a distributed network of nodes validates transactions or data through consensus mechanisms, such as proof-of-work. Data recorded in the blockchain is immutable, fostering trust among participants.
• Decentralized applications (DApps) are consumer-facing applications that interact with blockchains.
• Smart contracts are computer programs stored on the blockchain that automatically run when certain conditions are met.

Given the decentralized nature of web3 and its infrastructure, developers must consider how to leverage data to improve their operations. We anticipate three infrastructure level use cases where AI can enhance web3:

1. Intelligent Smart Contracts : A smart contract is a self-executing code on the blockchain, triggered when predefined conditions are satisfied without any third party intervention. Currently, smart contracts are static and rule-based decision systems that lack dynamic decision-making capabilities. That’s where AI can help: AI models can learn from past actions or large amounts of data to make smart contracts more adaptive in nature. Blockchains that run on AI models, such as Cortex, can be used to analyze actions of network participants and dynamically define new sets of rules based on those actions to attain new business goals. AI can also be used to recognize flaws or potential legal issues in contracts, and it can strengthen security efforts, which we’ll discuss more below. 

2. AI for web3 security: Reports estimate that web3 projects lost as much as $3.9 billion of funds to hacks in 2022. Smart contract exploits and bridge hacks accounted for about half of total hacks and total lost funds, respectively. Security is a significant hurdle to adoption, and these numbers indicate an urgent need for security solutions in the web3 space.

AI can be used to detect and prevent cyber attacks and improve the overall security of web3 projects. By analyzing past attacks, AI security providers like AnChain.AI can identify patterns and anomalies, adapt to new threats, and provide insights making it possible to identify and mitigate potential threats before they become a problem. AI-based cybersecurity tools such as CyVers can help in the detection of fraudulent activity by analyzing large volumes of user data created by blockchain and web3 apps. In smart contracts, AI products such as Quantstamp can be used to review code, identify weak spots and vulnerabilities, and prevent potential exploits and hacks.

3. Intelligent consensus protocols: Because they lack a central authority, blockchains rely on a network of validators––often individual people incentivized with cryptocurrency as a reward for checking the validity of transaction data––to achieve consensus at a protocol level (e.g., Bitcoin). Consensus mechanisms—like proof-of-work—determine the efficiency, security, and scalability of a blockchain. 

Researchers and startups are developing new approaches to consensus mechanisms. For example, Inery is working on an AI-based consensus mechanism that will organize block validation in the most efficient order based on the uptime (how often the blockchain is live and operational, rather than systemically down), computational power, stake, and other performance metrics of the blockchain's nodes. Velas is working on an Artificial Intelligent Delegated Proof-of-Stake (AIDPoS) algorithm that automatically adjusts the configuration of the blockchain to current circumstances within the network by embedding trained models at every node based on data collected at the previous epoch. 

Others have proposed models that, while still theoretical, suggest the potential to enhance consensus mechanisms. Coin.AI proposes to train deep learning models where a block is only mined when the performance of such a model exceeds a performance threshold. WekaCoin, relies on a new approach called proof-of-learning. That relies on machine learning competitions to validate blocks of transactions. Both alternatives seek to be more efficient with energy consumption than traditional proof-of-work protocols.

Vertical use cases

Different verticals are already using web3 technology to build decentralized applications, or DApps, which are web or enterprise applications that allow users to execute smart contracts. Because DApps don't depend on a centralized server, their data is stored across the blockchain network ensuring users with greater privacy, security, and control over their data. By using AI and machine learning (AI/ML) techniques, DApps offer a personalized experience by analyzing user behavior and preferences. Additionally, AI can be used to analyze blockchain data to identify patterns, predict future trends, and automate tasks. Below we will explore some top verticals where AI and web3 can add value.

1. Decentralized Finance (DeFi) enables financial transactions between two parties without the need for an intermediary. The global DeFi market was about $12 billion in 2021 and is set to grow to approximately $232 billion by 2030, reflecting a CAGR of almost 43%. The sheer volume of the DeFi market alone makes it an attractive opportunity for AI penetration. As DeFi systems become more mainstream, the massive amount of data they generate can be used to train and develop AI models. 

Through its predictive analytics capabilities, AI can analyze market conditions in real time and better assist DeFi systems to make more informed decisions. AI can be applied in DeFi lending to create credit score ratings and manage risk, as well as automate compliance checks with various regulatory requirements, such as anti-money laundering (AML) and know your customer (KYC). We think DeFi is one of the most promising areas where AI can add value, from personalized financial advice to risk management, automation, and security. However, government regulations, privacy, and AI explainability issues need to be tackled to realize the full potential of AI in DeFi systems.

2. Gaming: Unlike traditional gaming, where players don’t actually own anything, web3 gaming enables players to own their in-game assets through non-fungible or fungible tokens and transfer them without third-party intervention. The added benefits of asset ownership in web3 gaming include transparency and security. Web3 games can also be made more responsive, adaptive, and challenging by using AI in the game development process. AI can be used to generate large open-world environments, reducing both design and development time and effort. In games where users actions influence the storyline, AI can help generate stories and scenarios based on past story lines and user interactions. Recent research indicates that NPCs (non-playable characters) will dramatically improve as they can learn, adapt, and interact. We can envision a future in which we generate our own virtual and immersive worlds with interactive agents.

3. Others (Retail, Social Media, Healthcare, and Supply Chain): 

Retail: In retail, AI-generated visuals can be used in the metaverse to give consumers the real-world fitting room experience with the help of 3-D assets. 

Social media: With AI, web3 social networks can analyze large amounts of on-chain and off-chain data to provide more personalized content to their users. From a marketing perspective, AI can analyze social network user data and recommend more relevant products and services from advertisers. Ultimately, the combination of AI and web3 will strengthen the premise that users will be able to move across different social networks, carrying with them not only their data, but also their connections and social capital. 

Healthcare: In healthcare, while AI can be used to analyze medical records and provide insights into the patient’s health and lifestyle, blockchain or distributed ledger technology (DLT) can be used to securely store sensitive data and help with medical staff identification and verification (credentialing). 

Supply chain: While blockchain is used in supply chains to track products right up to their original source, AI can use historical sales data to generate demand forecasts and plan distribution routes. Combining AI with web3 technologies such as Blockchain/DLT can result in a safe, immutable, and decentralized system for crucial data that AI models must collect, store, and use. 

Decentralized AI

So far, we have explained how AI can add value to web3. But, there is one area where web3 technologies can be of use to AI: decentralized AI. Traditional AI can be decentralized by using crypto to distribute storage, compute power, and governance. For instance, through decentralized storage, blockchain technology can be used to record data changes in AI models as they learn new information (e.g., Filecoin, Arweave). Through distributed and decentralized computing power, we can create on-demand platforms to optimize machine learning. Those offering idle GPU for computing power can be rewarded for training models, and such rewards can then be used to use these models. Given the openness of decentralized networks, miners can exchange information to improve their models. Governance processes, too, can be based on consensus among multiple nodes, providing a more transparent and secure alternative to traditional AI systems. Networks like Gensyn and SingularityNET are using blockchain to decentralize AI—in particular, the latter recently announced HyperCycle, a proposal for a ledgerless blockchain to vastly improve how validators achieve consensus and therefore efficiency.  

Creating value using web3 and AI 

As AI helps all sectors reduce costs, increase revenue, and build moats, here’s how we think it can improve existing web3 business models:

1. Reduce costs: Web3 business models that act as shared databases to improve interoperability among their participants can use AI to improve their efficiency and reduce costs. For example, in supply chain, blockchain can provide a single source of truth for different enterprise resource planning (ERP) systems that track product shipments. AI can add value by automating existing workflows or repetitive tasks, and capture and act on real time information. 

2. Increase revenue: Web3 business models that remove intermediaries (e.g., cryptocurrency transfers) and facilitate peer-to-peer transactions (e.g. gaming) can use AI to increase revenue. For example, in web3 gaming businesses that earn revenue by selling digital assets or allowing players to transfer digital assets, gaming companies can use AI to predict the assets players will purchase, resulting in additional revenue from transaction and transfer fees.

3. Build moats: As businesses use AI to increase user interactions and retention, they’ll also collect more data, which can help build better models and more personalized services. 

It’s possible that web3 companies can use AI to continuously improve their products, leading to product stickiness and defensibility within the web3 ecosystem. One thing to remember here is that your core value proposition as a web3 system matters most––one that will attract the users in the first place––and AI is just a tool on top of that to retain users.

These are just suggestions and we leave it to the creativity of individual entrepreneurs as to how to use AI to improve their web3 business models.

Challenges for web3 and AI builders

Emerging technologies come with emerging challenges––here are three important considerations for web3 builders working with AI:

• Data quality: AI models depend on reliable data to deliver intelligent information and services. In the web3 ecosystem, “oracles” connect smart contracts stored on blockchains with third-party data sources to provide necessary inputs. So, it matters how trustworthy these oracles are if the same data is used to train the AI models.

• Bias: AI bias refers to the tendency of AI models to deliver biased results based on data that may include biased human decisions or reflect historical or social inequities. This can result in discrimination and encourage mistrust toward AI systems. Techniques such as counterfactual fairness and humans-in-the-loop can be used to reduce biased results in AI models.

• Privacy: In AI, the more data, the better––and large amounts of data are susceptible to privacy breaches. As we explained above, despite web3’s emphasis on privacy, through solutions such as secure multi-party computing and zero-knowledge proofs, its ecosystem is still prone to hacks, too. As researchers explore privacy in AI––think homomorphic encryption, federated learning, and differential privacy––we expect some of these advances to reach AI’s intersection with web3, too. 

Building an intelligent web3 future

AI can greatly benefit web3: it can automate tasks to reduce error, improve usability by providing more personalized results, add an extra layer of security, improve scalability, and enhance decision-making capabilities for web3 applications. As AI technology matures, adoption challenges such as AI bias, privacy, and data quality will improve. Today, even though there’s a lot of academic research on using AI for web3 processes such as consensus mechanisms, few companies are actually implementing AI into their web3 protocols and projects.

We think there’s potential for AI to enhance web3 across the board, and in particular, we see smart contracts as a likely beneficiary in the near-term, as we expect them to become more ubiquitous in the future. AI-based security providers also stand to benefit from web3. At the application level, verticals ranging from DeFi and gaming to retail can benefit from implementing AI and web3 into their offerings. AI is becoming an integral layer of technology, and we believe that AI will eventually be woven into all aspects of web3 infrastructure and applications.