In recent years, the blockchain field has been booming, and the field of artificial intelligence (AI) has been booming as well. However, these two revolutionary technologies seem to have little intersection. But conceptually, there are many complementary aspects between blockchain and AI. For example, the inherent decentralization of blockchain technology may help solve the centralization problem of AI, and the transparent and verifiable nature of blockchain may help solve the opacity problem of AI models.
Recently, the concept of "Blockchain X AI" has caused a lot of hype, leading to a significant increase in the market value of related cryptocurrencies, as shown in Figure 1. This indicates that the market is very optimistic about this combination, and investors are also confident.
Figure 1: Market value of cryptocurrencies in different subfields, data as of April 18, 2024
However, the integration of blockchain and AI has also exposed some conflicts between the two. For example, AI requires intensive computing and large amounts of storage, while the distributed ledger architecture of blockchain emphasizes redundancy - each node stores and computes the same information.
Recently, a research team from Tsinghua University and Fraunhofer HHI and other institutions published a paper titled "Blockchain and Artificial Intelligence: Synergies and Conflicts", analyzing the technical synergy and conflicts between blockchain and AI. It is worth noting that the team did not focus on theoretical analysis, but instead focused on the cryptocurrency market, analyzing "Blockchain X AI" projects and some specific use cases with a market value exceeding 10 million US dollars.
Figure 2: Complementarity and opposition between blockchain and AI
Now let's take a look at what this paper talks about and what interesting or useful insights it provides.
Blockchain X AI: Synergy and Conflict#
Figure 2 shows the complementarity and opposition between blockchain and AI.
Synergy between Blockchain and AI#
Decentralization vs. Centralization: The training and maintenance of large-scale language models (LLMs) such as GPT require a lot of computing power, electricity, and data resources. For example, the training cost of GPT-3 released in 2020 was approximately 4.6 million US dollars. Such high costs make AI large models the playing field of only a few large technology companies - they have basically become monopolists in the AI market. This monopoly may hinder competition, which is a concern frequently expressed by policymakers in regions such as the United States and Europe, where they are very active in enforcing anti-monopoly laws to maintain market balance and prevent market dominance by a single entity. In contrast, blockchain technology is decentralized; this feature may be used to solve the centralization problem of AI systems. If deployed properly, the decentralized nature of blockchain can prevent any party from controlling the entire network. This feature can help achieve some form of regulatory mechanism within AI systems, achieve a more balanced distribution of power, and promote collaboration among all parties. Therefore, integrating blockchain technology is expected to address the debates on regulation and monopoly in the AI field, making AI governance more inclusive and fair.
Transparency vs. Black Box Nature: Another major feature of blockchain technology is transparency, and the transactions and records on it are verifiable and tamper-proof. On the other hand, AI is like a black box - it is difficult for people to understand the reasoning process behind its decisions. Perhaps we can use blockchain ledgers to record the decision-making process of AI, achieve transparent audit trails, and enhance the credibility of AI applications. In addition, blockchain can also integrate advanced encryption technologies (such as zk-SNARKs) or use secure hardware (such as Trusted Execution Environments/TEE). These technologies can help verify whether specific computational steps are faithfully and accurately executed.
Data Management and Dependence: Blockchain can regulate data and data access permissions through smart contracts and protocols such as the InterPlanetary File System (IPFS).
Open Source vs. Closed Source: Blockchain can achieve shared ownership through encryption protocols, thereby enabling fine-grained privacy configurations and addressing the limitations of proprietary AI models. If shared AI systems (trained and controlled by participants) can achieve performance comparable to commercial models, the transparency of AI development will be greatly improved. This can also promote the creation of more fair and comprehensive AI solutions.
Conflict between Blockchain and AI#
Although there are synergies between blockchain and AI as mentioned above, there are significant conflicts in their operational requirements, which hinder the integration of the two.
Computational Cost and Load: For large language models (LLMs) such as GPT-4 and Llama 3, both training and inference require a large amount of computing resources. The consensus mechanism, cryptographic operations, and unfavorable data structures of blockchain will increase the computational burden, thereby affecting scalability.
Storage Limitations and Data Intensity: While the decentralized nature of blockchain ensures security and redundancy, it also leads to significant storage requirements, which are undoubtedly costly and inefficient for data-driven AI systems. In general-purpose blockchain systems like Ethereum, each node must store all information because redundancy ensures the security and resilience of the blockchain network, but it is not conducive to scalability. Since new data on the Ethereum Virtual Machine (EVM) is stored in transaction format, common data structures on the EVM may hinder retrieval speed. On the other hand, AI applications generate and process a large amount of data, requiring efficient and scalable storage solutions.
Pseudonymity and Security Challenges: Blockchain allows permissionless and pseudonymous access through asymmetric encryption, and protection against potential Sybil attacks can be achieved by setting computational or financial barriers. In addition, some use cases use blockchain as a platform to enhance privacy protection and distributed AI training, using technologies such as federated learning. However, if these use cases support pseudonymous participation in the training process, there may be risks. These methods are vulnerable to adversarial federated learning attacks, and it is very difficult to determine the identity of malicious attackers because, by design, contributions submitted to the overall AI model are private and difficult to measure.
Operational Mismatch: Most blockchain virtual machines use fixed ledger operations to ensure deterministic results - this is important because financial transactions involve money. Floating-point operations may introduce precision loss in calculations, especially when multiple values with vastly different magnitudes are involved. However, a common practice in AI training is to normalize floating-point parameters to a range of 0 to 1, as this helps achieve stable and effective gradient flow and provides implicit regularization, thereby improving overall training performance.
Blockchain X AI: Use Case Studies#
Based on the synergy and conflicts between blockchain and AI mentioned above, let's take a look at some use cases. The research team surveyed some of the best projects in integrating blockchain and AI. They focused on existing products that have issued tokens and have a market value exceeding 10 million US dollars. There are also some projects with novel use cases but a market value below 10 million US dollars. They classified these projects based on three research questions:
- How well does the project integrate blockchain and AI technologies?
- What is the role of blockchain in the project?
- What is the role of AI in the project?
The clustering analysis results are shown in Figure 3, which includes four main clusters: AI as a peripheral technology to blockchain, AI participating in blockchain, blockchain managing AI processes, and blockchain as the core infrastructure for AI.
Figure 3: Clustering analysis of projects integrating blockchain and AI
AI as a Peripheral Technology to Blockchain#
AI can help improve user experience in interacting with blockchain, enable intelligent analysis, and simplify the development process of blockchain applications.
AI Participating in Blockchain#
AI can actively participate in the blockchain ecosystem and governance structure. The team presents two exploration directions in the paper: one is to let AI agents join the distributed network as participants or stakeholders, such as letting AI trade on decentralized exchanges (Dex) by itself; the other is to involve AI in governing decentralized autonomous organizations (DAOs), although this is currently challenging.
Blockchain Managing AI Processes#
More and more people are using blockchain technology to manage AI processes, creating a decentralized framework for resource sharing, data management, and application deployment.
Blockchain as the Core Infrastructure for AI#
General-purpose blockchain systems like Ethereum face the triangle trade-off of scalability, security, and decentralization.
Taking Ethereum as an example, its security is ensured by thousands of nodes distributed worldwide, with over 1 million validators. To achieve consensus and finality, every new information block must reach every node in this global network and be verified by each node. Each block is of kilobyte size and created every 12 seconds, resulting in high storage and computational costs.
Therefore, it is not practical to directly execute or store computationally intensive AI operations on the chain. However, Layer2 rollup is becoming a popular paradigm. Simply put, Layer2 rollup refers to processing transactions off-chain and then aggregating the results before recording them on-chain. This solution can improve throughput and reduce costs, making it cost-effective.
Similarly, blockchain specifically developed for AI use cases must (1) overcome challenges related to high computational and storage costs, public access, and underlying virtual machine limitations, and (2) use blockchain purely as a management, governance, and security layer. Table 1 shows new systems that use blockchain as the core infrastructure for AI.
Table 1: Blockchain as the infrastructure for AI, where DAI = Distributed Artificial Intelligence, BC = Blockchain, DT/FL = Distributed Training/Federated Learning, C-Layer = Computing Layer, TA = Technical Analysis, DM = Distributed Management, PoS = Proof of Stake, DPoS = Delegated Proof of Stake, dBFT = Delegated Byzantine Fault Tolerance, FL = Federated Learning, DID = Decentralized Identity, ZK = Zero Knowledge, DePIN = Decentralized Physical Infrastructure Network, DC = Distributed Computing, DD = Distributed Data, ASBS = Application-Specific Blockchain Systems, IPFS = InterPlanetary File System, * = Low Maturity/No Public Code, ? = Information Not Available