Crypto and AI: Shaping the Future of the Internet

Scenario 1: Incremental advancements in AI and crypto.

AI and crypto technologies evolve but result in only modest efficiency gains for enterprises. The growth of internet-of-things (IoT) applications is limited while Decentralized Physical Infrastructure Networks (DePINs) leverage blockchain tokenomics (currency that can be used automatically within blockchain code) to enable autonomous real-time interactions within physical or online networks. Even with limited AI expansion, these networks could play an important role in optimizing usage of already-scarce computational resources and become more efficient as they expand with improved training data. DePINs, in 2024, are seeing some application of these technologies.

Scenario 2: A rapid expansion of AI that exacerbates centralization risks.

AI is centralized within a few, large technology incumbents, leading to a concentration of power. AI models play a dominant role in how businesses and consumers communicate and transact over the internet, intensifying the importance of data ownership and identity verification. Blockchain technology is used to decentralize information and facilitate privacy maintenance but with limited effect. AI applications extend beyond crypto into traditional financial markets, with associated and significant risks. AI’s concentration within a few corporations with potentially opaque governance structures engender reduced transparency and diminished accountability and give rise to risks that threaten to erode public trust (see figure 2).

A decentralized internet emerges that uses blockchain architecture to distribute data and decision-making across multiple nodes, reducing the risk of bias, censorship, and privacy that are associated with centralization (see figure 3). Blockchain’s transparency underpins the integrity, immutability, and traceability of data and AI decisions, with blockchain providing the means to record and later audit both. This enhances accountability and reduces the likelihood of personal data exploitation. Blockchain could also be used to verify the authenticity of AI-generated content and distinguish between humans and bots, potentially mitigating the risks of deepfakes and misinformation. Blockchain’s ability to enhance data privacy while ensuring regulatory compliance allows AI models access to larger volumes of training data and enhances its security.

The potential of these cryptographic technologies to enhance privacy and security, while also enabling AI-driven insights and collaboration, will have wide-ranging uses. For example, in the handling of medical data: homomorphic encryption could allow sensitive information to be processed and analyzed without being decrypted, meaning it can be used by AI models to predict disease outcomes and personalize treatments while protecting patient confidentiality; multi-party computation could enable medical networks to combine research data for the training of AI models and joint analysis for medical research; and zero-knowledge proofs could ensure an AI-generated medical images do not contain any embedded personally-identifiable information.

The intersection of AI and crypto offers potential for improvements across a range of industries, notably those that face complex (and sometimes currently intractable) challenges, or where there exists significant (including currently unrecognized) potential to unlock new efficiencies. The extent of the potential benefits from the new technologies will however vary across different uses and will be dictated by advancements in the two technologies, as detailed (earlier) in our three scenarios.

A summary of the relevance of AI and crypto technologies across five key areas, and under our three scenarios, is provided in the table below (see figure 4), while a more detailed review follows.

The problem

The increasing sophistication of cyber threats (coupled with greater use of AI by malevolent actors) means organizations face an increased cadence of attacks, which threaten financial and reputational damage, as well as rising cyber insurance costs.

Opportunities

AI threat detection and automated responses can significantly enhance cyber security in blockchain networks, and especially in DePINs and decentralized finance (DeFi) applications. AI can also help ensure the security of code, which is particularly important for DeFi applications, including by assisting in smart contract validation and through code testing and verification. For example: deep-learning models can analyze large datasets for unidentified vulnerabilities, known as zero-day vulnerabilities; and large language models can be trained on libraries of malware to detect patterns of attack and pinpoint vulnerabilities in existing code.

Risks and challenges

Data bias in training sets can affect AI performance, potentially leading to false negatives. AI models could be targeted and manipulated to compromise their effectiveness. To mitigate these risks and maintain AI-driven security, continuous updates and validation are essential.

The problem

Traditional financial markets' exposure to inefficiencies, the risk of fraud, and manual oversight increases costs, can hinder or delay real-time transactions due to reliance on intermediaries, and raises the potential for errors and criminality. Addressing inefficiencies through automation can lead to additional risks.

Opportunities

Smart contracts (coded sets of rules and conditional actions stored on a blockchain) could improve financial markets' transparency and efficiency. The contracts automatically execute based on pre-set conditions and can be integrated with verified real-world financial data through information bridges, called oracles. AI's ability to process and analyze large datasets provided by oracles can be used to efficiently generate pertinent inputs for smart contracts. The combination of smart contracts and AI could streamline markets by automating routine tasks, such as financial settlement, contract execution, and compliance checks, reducing the need for manual oversight and minimizing human error. Multi-party computation protocols can be used in the creation of decentralized oracles that ensure the security and accuracy of data across blockchains operating in a trustless system. AI compliance tools can play an important role in enhancing security in automated financial markets by identifying anomalies and potential fraud. They can monitor transactions in real-time and automatically trigger smart contracts to take preventive actions, such as halting suspicious transactions​.

Crypto wallets can allow AI agents to transact with each other through on-chain payments. This could, for example, enable an AI trading bot to acquire inferences from another AI model that is trained on a data set that is not generally available. Obviously AI agents do not have access to bank accounts in the traditional payment system, but they can be set up with crypto wallets and smart contracts, allowing them to exchange with each other, for example using tokens to pay for data. In September 2024, Coinbase’s CEO announced the first such AI-to-AI transaction. Potential applications of such interactions could also be much broader than in financial markets.

Risks and challenges

Vulnerabilities in oracles, such as susceptibility to data manipulation, can compromise the integrity of oracle-connected financial systems. Determining liability in such scenarios can be complex, due in part to the uncertainty of the legal framework and regulatory environment for crypto and AI-driven financial systems, particularly across international borders. AI models' complexity can render decision making opaque, posing audit and accounting challenges. Smart contract enforceability may be legally uncertain, potentially limiting their application in traditional financial systems.

Compute power (the capability of a computing system to perform computations and process data) and data storage are often siloed, leading to inefficient resource utilization. The expanding use of AI is creating significant additional demand for compute power to run data centers, and the trend is set to continue (see figure 7). This additional demand could strain already scarce energy resources and further challenge efforts to reduce emissions.

Opportunities

DePINs can facilitate peer-to-peer resources exchanges, including of storage and processing power, and incentivize use by rewarding participants with crypto tokens. This enables the monetization of excess capacity/ the purchase of resources, leading to optimized hardware use and reduced idle time. Tokenization of machine learning data and compute power may enhance distributed and collaborative AI systems that utilize high-speed and low-cost blockchain architecture. AI-driven pricing can effectively match infrastructure with demand, reducing computing costs and improving efficiency. Crypto is often portrayed as a drain on energy resources due to its use of the blockchain that underpins bitcoin, the largest cryptocurrency by market capitalization. It is important to note that energy consumption is specifically a feature of Bitcoin’s proof-of-work consensus mechanism and not inherent to all crypto. That means not all crypto is designed to consume as much energy. That said, bitcoin miners’ large computing infrastructures are often located in areas where energy costs are low, and particularly in locations (such as Texas) where they can use excess energy generated by solar and wind sources. Some bitcoin mining companies are using their access to cheap energy to offer AI services to diversify their revenue streams. AI data center firms are also looking to acquire or collocate with miners’ infrastructure to access cheap energy.

Risks and challenges

Token price volatility can affect the reliability and attractiveness of DePINs, potentially deterring participation and hindering data collection. Aggregating responses from multiple AI models may reduce errors and increase reliability, but can be complex, especially for sophisticated tasks such as integrating with blockchain platforms. Bitcoin mining and AI computations have different hardware requirements. Miners access cheap and clean energy because they have flexibility and economic incentives to adjust their demand according to the needs of the grid, increasing demand at times of low usage (when energy prices are cheaper) and decreasing at times of peak usage (when energy is expensive). Data centers for AI use do not have as much flexibility to switch units on and off in response to externalities. This limits the potential synergies between bitcoin mining and AI services.

The problem

Smart devices have so far mainly generated convenience benefits for individual consumers and are yet to fulfil their potential to build smart networks. At a municipal level, inefficient use of collective data and underutilized resources can lead to higher operational costs and suboptimal services. Effective networks of smart devices could help (see The Rise of AI-Powered Smart Cities, May 18, 2024).

Opportunities

By leveraging crowd-sourced sensors (dashcams, energy meters, toll road monitors, and water pipe flow monitors) municipalities can gather decentralized data, while rewarding contributors with tokens as payments. AI-driven data analysis can be used to optimize infrastructure management software, including mapping of energy grids, traffic flow, and water/sewer systems, resulting in cost savings for government agencies. AI can also dynamically adjust resource allocation to match real-time demand, reduce waste, and predict maintenance needs. For wireless infrastructure, AI can improve efficiency and security by reducing latency and mitigating cyber risks through intelligent traffic rerouting.

Risks and challenges

Tamper-proofing is critical to ensure data integrity and operational reliability when securing physical assets and sensors within DePINs. Privacy concerns can arise due to the processing of large volumes of data, especially where information is sensitive, necessitating compliance with data protection regulations. As the number of IoT devices grows, managing and scaling network infrastructure becomes increasingly complex. Maintaining high data quality and accuracy is crucial to avoiding inaccurate AI predictions and operational inefficiencies.

The problem

Manual processes and a lack of real-time data integration mean supply chains face costly inefficiencies and operational risks due to delays and excess inventory.

Opportunities

AI can be used to predict delays and dynamically (including in real-time) adjust operations to reduce excess inventory and optimize routes. That would be underpinned by smart contracts, which can automate instant payments, facilitate compliance, and offer real-time tracking with the aid of DePINs, resulting in a seamless and transparent logistics system with immutable records.

Risks and challenges

Evolving legal frameworks across different jurisdictions introduces regulatory uncertainty and creates compliance difficulties. The complexity and opacity of AI models can pose audit and accounting challenges, while the enforceability of smart contracts remains legally ambiguous. Data privacy and security concerns must be addressed to protect sensitive information and the integration of new technologies with existing legal systems can be costly and complex. Scalability issues may arise as supply chains expand and ethical considerations regarding job displacement and decision-making biases need to be managed.