Author: Danny Sursock, Head of Archetype Fund; Translation: Golden Finance xiaozou
1. Platform conversion
The world is shaped by periods of dramatic change in technology or infrastructure, and when they do, unleashing generational functional innovation, a new world is shaped. Think of telegraph and railways, fiber optic cables and the Internet, or mobile phones and 3G networks.
We believe that the intersection of two groundbreaking fields, artificial intelligence (AI) and blockchain, is a similarly transformative moment.
The three theoretical pillars of this article are as follows:
① The rise of AI will increase the demand for blockchain technology
② AI will accelerate the maturity and adoption of decentralized applications
③ Open source innovation in decentralized infrastructure will shape the future of AI
2. Blockchain can provide better design space
There are many areas where AI has a high impact, but they can be roughly summarized into three main categories:
① User-oriented intelligent systems, products or applications
② Improve the operational and/or capital efficiency of the enterprise
③ Eliminates marginal costs of content creation (and idea generation)
In particular, generative AI presents unique challenges and opportunities where we believe blockchain technology can play to its strengths.
To understand why, it’s important to consider the core inputs that drive the evolution of intelligent systems. Machine learning (ML) is fundamentally driven by data (lots of it, but of increasing quality), feedback mechanisms, and computational power.
Major players in the AI/ML space, such as OpenAI (backed by Microsoft) and Anthropic (backed by Google and Amazon), are already pooling resources and building moats around their models and data. But despite early advantages in compute, data, and distribution, this approach risks undermining the collaborative development cycle that gave rise to the industry in the first place, stifling momentum.
Blockchains like Ethereum offer a viable solution and have emerged as reliable, neutral data and computation systems that drive open source innovation. Blockchains already underpin a range of digitally native primitives that are well-positioned to play a key role in a world increasingly shaped by generative AI.
We believe that blockchain has a great opportunity to become a major force in open source research and development in the field of artificial intelligence.
3. Current market conditions
This year, a lot of money has been invested in core infrastructure, model layers, and even user-facing applications like chatbots, customer support, and coding assistants. Still, it’s not obvious where value will accrue (and to whom) in traditional sectors in the long run.
In the current paradigm, AI has the potential to become a centralizing force that continues to expand the dominant web2 market. Especially at the infrastructure and model layers, the name of the game is expansion — expansion in hardware and capital resources, data access, distribution channels, and unique partnerships.
From cloud service providers like AWS to hardware manufacturers like Nvidia to established giants like Microsoft, many players are moving towards a full-stack model, either through mergers and acquisitions or through patent collaborations.
The top players are in a race for scale and profits, but the market for ultra-expensive, high-precision enterprise API models may be limited by economics, open source performance convergence, or even trends toward low-latency workload requirements.
Meanwhile, a large portion of the mid-market has seen a commoditization trend towards “OpenAI API wrapper” products that are functional but difficult to differentiate.
4. Open source builds momentum
Open source datasets for pre-training, training, and fine-tuning, as well as free access to foundational models and tools, encourage businesses of all sizes to unleash their creativity directly using open source systems and tools.
A leaked Google article outlines how the gap between the closed-source and open-source code worlds is rapidly narrowing. Notably, 96% of code bases today use open-source software, a trend that is particularly evident in the fields of big data, artificial intelligence, and machine learning.
At the same time, the time may be ripe to disrupt the cloud services oligopoly.
The big three, AWS, Google Cloud, and Azure, have historically entrenched themselves in enterprise competition by layering tools and services to capture the market. This dominance has created many challenges for enterprises, from restrictive operational dependencies to excessive costs associated with cloud infrastructure, especially given the premiums charged by the major providers.
Existing companies facing expense pressures from restructuring their operations, coupled with attempts to experiment and integrate the growing amount of open source AI, will create a window to restructure their businesses using decentralized alternatives.
As such, the emerging intersection of open source AI and blockchain technologies offers an extraordinary area for experimentation and investment.
5. Encryption and AI: A two-way value relationship
We are extremely excited about the potential symbiotic relationship between AI and blockchain.
Cryptographic middleware can greatly improve the information input of the AI supply side by establishing efficient computation and data markets (for supply, labeling, or fine-tuning), as well as proof or privacy tools.
In turn, decentralized applications and protocols will reach new heights by absorbing the fruits of this labor.
It is undeniable that crypto has come a long way, but protocols and applications are still hampered by still-unintuitive tools and user interfaces for mainstream users. Likewise, smart contracts themselves can be limited, both in terms of manual workload demands on developers and in terms of overall functionality fluidity.
Web3 developers are a very productive bunch. At its peak, just 75,000 full-time developers created a trillion-dollar industry. Coding assistants and ML-augmented DevOps promise to power existing jobs, while no-code tools are quickly enabling a new class of builders.
As machine learning capabilities are incorporated into smart contracts and brought on-chain, developers will be able to design more fluid and expressive user experiences and ultimately design entirely new killer apps. This step-change improvement in on-chain experience will attract new — and potentially larger — audiences, catalyzing a significant adoption feedback flywheel.
Generative AI could be the missing link for cryptocurrency, changing UI/UX and spawning a new wave of technology development. In turn, blockchain technology will harness, generalize, and accelerate the potential of artificial intelligence.
6. Use blockchain to build a better data market
(1) Data is the basic information input for machine learning
Yes, huge improvements in computing infrastructure have been helpful, but it’s massive databases like Common Crawl and The Pile that make it possible for the underlying models to gain worldwide attention today.
In addition, companies will use this data to refine the underlying models of their product offerings or build future competitive moats. Ultimately, data will become a bridge between users and personal models that run locally and continuously adapt to individual needs.
Competition for data is therefore an essential frontier where blockchain can take advantage – especially as quality becomes an important attribute in shaping the data market.
(2) Quality over quantity
Early studies suggest that up to 90% of online content could be synthetic in the next few years. While synthetic training data has certain advantages, it also introduces significant risks associated with deteriorating model quality and reinforcing bias.
There is a real risk that machine learning models may run out of non-synthetic data sources in the coming years. Cryptocurrencies’ coordination mechanisms and proof primitives are inherently optimized to support decentralized markets where users can share, own, or monetize the data they use to train or fine-tune domain-specific models.
Therefore, web3 may be a better and more efficient source of artificially generated training and fine-tuning data.
(3) Progress
Blockchain-enabled decentralized training, fine-tuning, and inference processes can also better preserve and leverage open source intelligence.
Smaller open-source models are being improved using an efficient fine-tuning process and are now comparable to larger models in output accuracy. As a result, the trend has begun to shift from quantity to quality in terms of source and fine-tuning data.
The ability to track and verify the lifecycle of raw and derived data promotes reproducibility and transparency, which in turn drives higher quality models and inputs.
Blockchain can build a lasting moat as a leading domain with diverse, verifiable, and tailored datasets. This is especially valuable in situations where traditional solutions over-index algorithmic progress to cope with insufficient data.
(4) Content Tsunami
The coming wave of AI-generated content is another area where cryptocurrency’s first-mover advantage will come into play.
This new technological paradigm will empower digital content creators at an unprecedented scale, and Web3's plug-and-play infrastructure makes it all simple and straightforward. Cryptocurrency has home field advantage, thanks to years of development around primitives that establish ownership and immutable provenance of digital assets and content in the form of NFTs.
NFTs can capture the entire content creation lifecycle, but can also represent digital native identities, virtual assets, and even cash flow.
As a result, NFTs enable new user experiences like digital asset marketplaces (OpenSea, Blur), while also rethinking business models like written content (Mirror), social media (Farcaster, Lens), games (Dapper Labs, Immutable), and even financial infrastructure (Upshot, NFTFi).
The technology could even combat deepfakes and computational manipulation more reliably than the alternative: using algorithms. A clear example of this is when OpenAI’s detection tool was shut down due to accuracy failures.
One final note: Advances in concise and verifiable computation will also upgrade the dynamic landscape of NFTs as they incorporate ML outputs to drive smarter, evolving metadata. We believe that AI tools and interfaces based on blockchain technology will unlock comprehensive value and reshape the digital content landscape.
7. Use zero-knowledge proofs to harness the infinite knowledge of machine learning
Zero-knowledge (ZK) proofs have made great progress as the blockchain industry seeks technical solutions that enable resource-efficient computation while maintaining trustless dynamics.
While originally designed to address resource bottlenecks inherent in systems like the Ethereum Virtual Machine (EVM), ZK proofs offer a range of valuable use cases related to artificial intelligence.
An obvious example is a simple extension of an existing use case: efficiently and concisely verifying computationally intensive processes such as running ML models off-chain, so that the final product (such as model inference) can be integrated on-chain via smart contracts in the form of ZK proofs.
Proof of storage combined with collaborative processing can go a step further and make on-chain applications more flexible and agile without introducing new trust assumptions, greatly enhancing their functionality.
Of course, this also allows completely new functionality to be implemented.
When called through an API, a ZK proof can be used to verify that a specific model or data pool was actually used to generate inferences. It can also hide the specific weights or data used by the model in customer-sensitive industries such as healthcare or insurance.
Companies can even collaborate more effectively by exchanging data or IP, benefiting from shared learning while still maintaining ownership of their resources.
Finally, ZKPs have real applicability in the increasingly relevant (and challenging) area of distinguishing between artificial data and the synthetic data discussed earlier.
Some of these use cases depend on further development around technical implementation and the need to find sustainable economies of scale, but zkML has the potential to have a unique impact on the trajectory of AI.
8. Long-tail assets and potential value
Cryptocurrency has already proven its role as a preeminent architect of value flows in traditional markets such as music and art. Over the past few years, on-chain liquid markets have also emerged that represent off-chain tangible assets such as wine and sneakers.
The successor will naturally involve advanced ML capabilities as artificial intelligence is brought on-chain and made accessible to smart contracts.
ML models combined with blockchain rails will redesign the collateralization process behind illiquid assets that were previously inaccessible due to lack of data or buyer depth.
One approach is for machine learning algorithms to query a large number of variables to assess hidden relationships and minimize the attack surface for manipulators. Web3 is already experimenting with creating markets around new concepts like social media relationships and wallet usernames.
Similar to the impact that AMMs had on unlocking long-tail token liquidity, ML will revolutionize price discovery by acquiring large amounts of quantitative and qualitative data to uncover hidden patterns. These new insights can form the basis of smart contract-based markets.
The analytical capabilities of artificial intelligence will be embedded in decentralized financial infrastructure to discover potential value in long-tail assets.
9. Decentralized Infrastructure Layer
Cryptocurrencies’ strengths in attracting and monetizing high-quality data solve one problem. The other side — the infrastructure behind AI — holds similar promise.
Decentralized Physical Infrastructure Networks (DePINs) like Filecoin or Arweave have built systems for storage that themselves incorporate blockchain technology.
Others like Gensyn and Together are addressing the challenges of distributed network model training, while Akash has launched an impressive P2P marketplace that connects supply and demand for excess computing resources.
Beyond that, Ritual is building the foundation for open source AI infrastructure in the form of incentive networks and model kits, connecting distributed computing devices for users to perform inference and fine-tune.
Crucially, DePins like Ritual, Filecoin, or Akash can also create a larger, more efficient market. They do this by opening up the supply side to a wider range of people, including passive suppliers who can unlock latent economic value, or by aggregating less powerful hardware into pools that compete with their higher-performing peers.
Each part of the technology stack involves different constraints and value preferences, and there is still a lot of work to be done in field-testing these layers at scale (especially in the emerging area of decentralized model training and computation).
However, the foundations exist for blockchain-based computing, storage, and even model training solutions that could eventually compete with traditional markets.
10. Conclusion
The union of crypto and AI is quickly becoming one of the most inspiring frontiers for design, already impacting everything from content creation and cultural expression to enterprise workflows and financial infrastructure.
In summary, we believe these technologies will reshape the world in the coming decades. The best teams will natively combine permissionless infrastructure, cryptoeconomics, and AI to improve product/service performance, enable entirely new behaviors, or achieve competitive cost structures.
Cryptography introduces unprecedented scale, depth, and granularity of standardized data to collaborative networks, and there is often no obvious way to derive utility from this data.
At the same time, AI transforms the pool of information into a vector of relevant context or relationships.
When these two fields come together, a uniquely mutually beneficial relationship can be formed, laying the foundation for builders of the decentralized future.