“Software is eating the world” has become one of the iconic phrases of the last decade of the software industry. Quoted in 2011 by software legend and venture capitalist extraordinaire Marc Andreessen, it synthesized the idea that companies that operated mostly in the physical world were transitioning to the digital economy in a trend that will essentially transform every company as a software company.

Jesus Rodriguez is CEO of IntoTheBlock, a blockchain and cryptocurrency market analysis firm. This article is a preview of a talk he will give this week on the Big Ideas stage at Consensus 2022 in Austin, Texas.

In recent years, the evolution of machine learning (ML) and artificial intelligence (AI) has permeated all areas of the software industry, leading many experts to claim that “machine learning is eating software.” Crypto and digital assets are rooted on the foundation of code and programmability and, consequently, are likely to be influenced by ML-AI trends. The intersection of ML-AI with digital assets is likely to usher in a new era in which intelligence becomes a native component of crypto assets.

The idea of intelligent crypto assets is conceptually trivial but full of practical challenges. Which are some of the fundamental ML trends that can rapidly impact the next generation of crypto assets? How about the main scenarios that can benefit from intelligence capabilities in crypto or some of the key technical challenges that need to be overcome for crypto to become intelligent. This essay explores some of these ideas and develops a thesis about the potential of the intersection of crypto and ML.

An important point to realize when thinking about AI-ML in the context of crypto-assets is that crypto is the only asset class in history that has the potential to become natively intelligent. AI-ML capabilities in traditional asset classes such as commodities or equities are implemented in vehicles like robo-advisors or quant strategies that live outside the asset itself. Even though there is an obvious role for those vehicles in the crypto space, crypto assets can natively embed those AI-ML capabilities in the assets. This benefit is, obviously, a side effect of the programmable and digital capabilities of crypto. Crypto assets are based on code and that code could take the form of AI-ML models.

AI-ML is likely to play an important role in the next decade of the crypto market. While the initial phases of crypto have centered around digitization and automation, the next iteration seems to be destined to be focused on intelligence. There are plenty of applications of AI-ML in crypto today, but we can’t claim that crypto-assets are inherently intelligent. In the near future, we should expect to see crypto-assets and protocols start to incorporate AI-ML as native capabilities that will allow them to learn and adapt their behavior based on their surrounding environment or markets.

The inevitability of digital assets becoming intelligent is partly dictated from the astonishing evolution of AI-ML technologies in the last few years. In the context of crypto, we shouldn’t think about AI-ML as a generic thing but rather as a group of interrelated types of methods. From that perspective, there are a small number of AI-ML schools that seem particularly well-suited for applications in the crypto space. Let’s explore some of the most popular techniques through the lens of their potential within crypto technologies.

Considered by many the most important evolution of the last decade of AI-ML, transformers are behind the revolution in natural language understanding (NLU) and are making inroads in other areas such as computer vision. Models like OpenAI’s GPT-3 or NVIDIA’s Megatron are able to generate synthetic texts indistinguishable from real, engage in highly complex question-answer interactions or even exhibit reasoning capabilities over textual forms. Models like OpenAI’s DALL-E 2 or Google’s Imagen are able to generate artistic images from textual forms bridging intelligence across multiple domains.

Understanding the impact that transformers have had in the NLU and computer vision space, it’s not difficult to imagine the influence they are likely to exert in areas like NFTs that rely on visual representations and textual interactions.

Meta (Facebook) AI Research recently referred to self-supervised learning (SSL) as the “dark matter of AI” as an analogy about the foundational role that this new type of technique can have in the next generation of AI models. Conceptually, SSL tries to enable intelligent capabilities that resemble how babies learn by observation and interaction. SSL tries to overcome some of the limitations of traditional supervised learning methods that need to be trained with large volumes of labeled data. Models like Meta’s DINO are able to classify objects in images without previous training.

The applications of learning without massive amounts of labeled data seem perfect for crypto. Decentralized finance (DeFi) could be an immediate beneficiary of these methods.

Blockchain datasets represent the biggest source of data in crypto. From a structural standpoint, blockchain datasets are natively hierarchical as they model relationships between addresses, transactions or blocks. Graph neural networks (GNNs) is the AI-ML discipline that specializes in learning over hierarchical datasets. Companies like Google’s DeepMind are using GNNs to predict traffic in Google Maps or even understand the structure of glass.

GNNs seems like a perfect AI-ML technique for crypto assets. If blockchains are ever going to become intelligent, GNNs are likely to play a key role in developing knowledge from their native datasets.

Deep reinforcement learning (DRL) became sort of pop culture after DeepMind’s AlphaGo defeated multiple time Go’s world champion Lee Sedol. AlphaGo mastered Go by playing an unfathomably large number of games against itself and correcting its own mistakes. This trial-error, learning by interaction form is the essence of DRL.

Since AlphaGo, DRL has been at the center of remarkable AI-ML achievements. DeepMind’s own AlphaFold shocked the scientific community by being able to predict the structure of proteins from a sequence of amino acids, a discovery that can unlock a new era in medicine. Another marquee DRL model from DeepMind was MuZero, which is able to master games like Go, chess or Atari without even knowing the rules.

The principles of DRL of learning by trial-and-error seems relevant to many areas of crypto such as DeFi or NFTs, in which conditions change all the time. After all, most crypto protocols are based on strong game theoretic rules and DRL have proven to excel at games.

Cyberpunk legend, science fiction writer William Gibson’s once said “‘The future is already here – it’s just not evenly distributed.” That quote could serve us as a philosophical guideline as we think about the path towards intelligent crypto assets. The creation of crypto coincided with the golden era of AI-ML research and technology developments. Today, AI-ML technologies are rapidly becoming mainstream and it’s a matter of time before they become a first-class citizen in the crypto space. The use cases seem to be everywhere. Let’s explore some of the most obvious.

There have been some applications of using AI-ML generative methods to create NFTs. However, the influence of AI-ML should expand to all areas of the NFT space. Let’s imagine NFTs that incorporate language and speech capabilities to establish a dialog with users, answer questions about its meaning or interact with a specific environment. Just like you interact with your favorite digital assistant, imagine establing a conversation with a visual NFT that can change its appearance based on the nature of the dialog. Similarly, think about using AI-ML transformer models that have been pre-trained in millions of paintings to generate unique NFTs that capture unique aspects of the style of the masters.

DeFi protocols are all about automation but they are not exactly intelligent. Incorporating AI-ML capabilities into DeFi protocols seems inevitable. We can envision a new generation of automated market maker(AMM) protocols that can adjust the balances in pools using real time predictive models based on existing market conditions. Similarly, we can think of lending protocols that adjust the size of loans based on an intelligent profile of the addresses requesting it.

AI-ML is influencing all aspects of software infrastructure such as networking, compute or storage and blockchains are unlikely to be an exception. It’s not far-fetched to think about intelligent consensus protocols that improve performance based on predictive models. Similarly, we can think of blockchains that develop intelligent economies to control the computation cost in the form of gas or other equivalents.

User experience seems to be one of the most obvious areas to introduce AI-ML capabilities. It’s a matter of time before wallets or exchanges start incorporating native intelligence capabilities that help improve investment and trading decisions that today are fully reliant on human subjectivity.

The topic of programmable stablecoins seems very prominent these days after the Terra UST collapse. What if, instead of thinking about this form of stablecoin as programmable, we could think about forms that are programmable but also intelligent? Instead of programmable stablecoins that adjust the peg based on statically defined economic gymnastics, what if they could rely onAI-ML algorithms that organically learn from market conditions. A combination of AI-ML with human supervision seems to be an interesting approach to explore in this area.

The relationship between crypto and AI-ML is more bidirectional than most people think. While the scenarios in which AI-ML can influence the next generation of crypto assets and infrastructure are fairly clear, there are some non-obvious areas in which crypto can influence AI-ML technologies.

Decentralized AI (dAI) is an emerging technology movement that looks to leverage the decentralization compute as well as tokenization mechanisms to mitigate some of the increasing centralization challenges of AI-ML technologies. A subdomain of the general dAI approach are mechanisms that leverage crypto-assets to create economies in which companies and individuals are incentivized for sharing data and AI-ML models.

Data is the electricity of AI-ML but, today, is highly controlled by a small number of incumbents and there are virtually no incentives for companies to collaborate and share data to break that monopolistic cycle. Introducing clever tokenomics and incentive mechanisms could organically help to establish channels for companies to regularly cooperate in the creation and training of AI-ML models for specific tasks and share the benefits.

Bias and fairness is another hot topic in AI-ML these days that could be hugely influenced by the use of native crypto technologies. Datasets used in the training of AI-ML models are permeated with biases, discrimination and toxic data points which can influence the knowledge of AI models.

While there have been a lot of advancements in quantifying and monitoring the fairness of AI-ML models, there are no robust accountability and benchmarking mechanisms that are trusted across the entire industry. Imagine using a blockchain layer to keep track of the bias and fairness score of specific AI-ML models and compensate for models that are improving their fairness scores. This is a low-entry point scenario for the usage of blockchain technologies in AI-ML infrastructures.

Without a doubt, AI-ML should be a foundational element of the next generation of digital asset technologies but there is also a lot of tangible value that crypto and blockchains can deliver in the world of AI-ML. Fundamentally, crypto could serve as an economic and accounting layer that helps build fairer and more democratic AI-ML solutions.

AI-ML is influencing each and every area of the software world and crypto is unlikely to be an exception. The core principles of digital asset technologies have been centered around democratizing financial services by using digitization and automation. Intelligence is one of the next frontiers for crypto and we are likely to see the impact across the entire space. From intelligent NFTs, DeFi protocols to new forms of crypto-assets, the incorporation of AI-ML is likely to unleash a new era of innovation in crypto. The technologies and use cases are already here. It’s time to start building.

The Coming InDAOstrial Revolution by Julie Fredrickson

Distributed autonomous organizations give humans a chance to build bigger, weirder things on radical timelines, just as the advent of the corporation paved the way to the Industrial Revolution.

In an era of misinformation, blockchain technology can renew our faith in evidential truth, not least during the current conflict in Ukraine, says Jonathan Dotan, the founding director of The Starling Lab.

Rhys Lindmark, a “Big Ideas” speaker at CoinDesk’s Consensus festival, on how the crypto generation could rewrite the rules of charitable giving.

More local money could lessen the incentive to “exit” the communities who need the resources, says Matt Prewitt, president of the RadicalxChange Foundation.

Quantified forecasting is an invaluable and yet underused tool, and prediction markets appear a vital tool for its adoption.

Read more about

DISCLOSURE

Please note that our privacy policy, terms of use, cookies, and do not sell my personal information has been updated.

The leader in news and information on cryptocurrency, digital assets and the future of money, CoinDesk is a media outlet that strives for the highest journalistic standards and abides by a strict set of editorial policies. CoinDesk is an independent operating subsidiary of Digital Currency Group, which invests in cryptocurrencies and blockchain startups. As part of their compensation, certain CoinDesk employees, including editorial employees, may receive exposure to DCG equity in the form of stock appreciation rights, which vest over a multi-year period. CoinDesk journalists are not allowed to purchase stock outright in DCG.

Read More