Will AI + blockchain be the next ascension path?

Original Title: AI x Blockchain The Next Level

Original Author: Steve Vassallo, General Partner at Foundation Capital

Translated by: Qianwen, ChainCatcher

Photo taken on March 31, 2021, workers are transferring mining machines

Blockchain and artificial intelligence are two of the most defining technologies of our time. Each technology is a powerful, disruptive force, akin to the monsters Godzilla and King Kong, nearly unstoppable. However, when King Kong and Godzilla face enemies that neither can defeat, they occasionally join forces. (Next year they will collaborate again in "Godzilla vs. Kong: The New Empire.")

The combination of these two great technological breakthroughs, artificial intelligence and blockchain, has the potential to unlock infinite possibilities and tackle the tough challenges we face together.

I recently had the opportunity to explore the immense potential of combining these two directions. Not long ago, I hosted a panel discussion at Coinbase's inaugural "Machine Learning (ML) and Blockchain" conference. The panel brought together four experts from academia and industry to interpret how these rapidly evolving technologies can merge and unleash tremendous potential. Our conversation covered many topics, including how blockchain can accelerate the development of artificial intelligence, the complexities of collaborating with blockchain data, and the prospects of large language models (LLMs).

One significant benefit of combining artificial intelligence and blockchain is that when it comes to issues of data falsification and content falsification (as artificial intelligence evolves, fake data and fake content have increasingly become pressing problems), blockchain can serve as a solution, using cryptographic digital signatures and timestamps to combat misinformation, helping people discern what is real and what is false.

At the same time, artificial intelligence can enhance the efficiency of blockchain networks, improve their security, and unlock new functionalities, such as allowing protocols to make decisions based on real-time, on-chain data.

To convey the points more clearly, the following content is excerpted from my colleague's remarks, presented after editing.

Bhaskar Krishnamachari, Professor of Electrical and Computer Engineering and Computer Science at USC

In my view, there are two main areas of intersection between blockchain and artificial intelligence. The first is applying ML models to address challenges within blockchain, and the second is leveraging blockchain to solve pressing issues in artificial intelligence.

In the first case, ML models can sift through complex patterns in blockchain data, helping to improve the performance of decentralized applications on the chain. By analyzing transaction data, ML models can reveal potential misconduct, such as wash trading and illegal fund transfers, and detect emerging security threats. In addition to helping ensure the security of blockchain networks, ML models can also enhance network performance. For example, they can dynamically adjust transaction fees based on transaction volume and optimize system resources during peak usage.

What is less discussed is how blockchain can aid the development of artificial intelligence. As a borderless, internet-native payment system, blockchain can create economic incentives for people to contribute data and computational resources to train ML models. We have been conducting research on decentralized data markets at USC to achieve this goal.

In recent years, we have seen a handful of tech companies capturing an increasing share of global data and artificial intelligence. This has raised concerns about privacy, bias, and security: blockchain, as a decentralized, transparent, and publicly auditable system, can address all these issues. For example, blockchain can track the sources of data used to train artificial intelligence models and cryptographically verify their authenticity. By confirming that these inputs are unaltered and fair, blockchain can help enhance trust in the recommendations provided by artificial intelligence systems.

Leo Liang, Head of Data Platform and Services at Coinbase

At Coinbase, most of the challenges my team faces are related to data. Specifically, we need to extract data from the blockchain and convert it into a format usable by ML models. I like to think of the blockchain as an onion, with countless intricate layers. Its decentralized nature means that data is distributed across many nodes, each independently verifying and adding new blocks. When multiple blockchains are in play, the networks become even more complex—now you are dealing with an interconnected onion network. Synchronizing and ensuring data consistency in this chaotic, decentralized ecosystem is no easy task.

Moreover, the blockchain is a relatively isolated system, unable to access data beyond its boundaries. To enable ML models to make predictions about the real world, we need to combine on-chain data (data stored on the blockchain) with off-chain data (data outside the blockchain, such as stock prices, exchange rates, weather patterns, etc.). This is akin to connecting the blockchain to the internet, which is undoubtedly an exciting yet daunting engineering challenge.

Sam Green, Co-founder and Research Director at Semiotic Labs

At Semiotic Labs, I am responsible for the AI development work for The Graph, a decentralized protocol for interacting with and utilizing blockchain data. Simply put, The Graph reads data from the blockchain, processes it, and creates an index, which is like an alphabetically arranged list of encyclopedia entries. This organizational structure simplifies data retrieval on the blockchain. By indexing blockchain data, The Graph transforms it into a format that is easy to query, analyze, and apply to downstream applications.

Transactions on The Graph involve two main participants: one is the data seller, known as the indexer, and the other is the data buyer, known as the consumer. These entities interact through what we call a "gateway." When a consumer sends a query to the gateway, the gateway allocates the query among indexers based on factors such as bid price, service quality, and latency. Indexers earn revenue by servicing these queries and providing blockchain data to consumers. With the help of artificial intelligence, we have established algorithmic pricing agents to help indexers maximize their revenue while ensuring consumers receive reliable, high-quality service.

In many ways, blockchain is an ideal environment for training artificial intelligence agents. The rules defined by smart contracts, as well as the actions of users recorded in transactions, are publicly visible on the chain. Since these rules and actions are known, we can create simulations of this blockchain environment and use them to train AI agents, which can then be deployed. The secret to success lies in a rapid feedback loop: the faster the learning through trial and error, the quicker the agents can improve their performance.

In the future, we believe that integrating LLMs into The Graph will unleash tremendous potential. Currently, users must query The Graph using a specialized language called GraphQL. In contrast, LLMs allow users to express their requests in natural language (Note: as opposed to programming languages, this refers to the languages people use in everyday life). LLMs enable anyone to interact with The Graph using simple English, further democratizing access to blockchain data.

Paul Bohm, Founder of Teleport

Teleport is developing an open marketplace for shared cars. Currently, the shared car market is a closed system, making it difficult for users to switch between different services. If email systems were as closed as shared cars, users of Microsoft's Outlook and Apple's iCloud would be unable to send emails to each other. Similarly, if web systems were closed, Apple's Safari browser would not be able to communicate with Microsoft.com.

An open shared car system means bringing this market back under the oversight of internet norms. In an open system, participants can choose from a variety of applications from many different providers and communicate with each other. In a closed market, fair pricing is often absent, with prices set by providers to maximize the value they can earn. Open shared mobility and removing intermediaries mean that drivers can earn more profit, passengers pay less for each ride, and ultimately more funds can circulate in the local economy.

For an open market to succeed, it must rely on user trust. Engineers often focus first on various technical aspects, such as speed or novel features. However, when building markets for the real world, we must start from users' needs for safety, security, and privacy. Only then can we determine the best technologies to meet these needs without going down the wrong path.

Of course, these assumptions are just possibilities and merely the beginning of this series of conversations, which focus on what aspects will be released, enhanced, solidified, and elevated to new heights when blockchain and machine learning converge. Digital consensus technologies like blockchain can achieve fair, trustworthy, secure, and verifiable systems.

While artificial intelligence may further undermine trust, blockchain will strengthen trust networks, providing a powerful mechanism to ensure the integrity of sensitive data. At the same time, artificial intelligence allows people to explore the deep sea of distributed data, which makes blockchain cumbersome and complex during large-scale adoption. By deploying artificial intelligence to tackle highly scalable problems, we can bring blockchain to a billion users.

For entrepreneurs in blockchain or artificial intelligence, these are exciting prospects: we are discussing not just a single technology, but the combination and promotion of two technologies. Artificial intelligence and blockchain, like Godzilla and King Kong, the combination of the strengths of two giant monsters, will undoubtedly be unstoppable.