The Merger of Crypto and AI: A Comprehensive Overview

Artificial Intelligence is the hottest tech trend of the moment. What if we combine it with blockchains and cryptocurrencies? It turns out that the convergence of the two digital technologies actually holds significant potential in myriad ways. While it’s still early for this merger, we can already identify various potent combinations and numerous projects aiming to turn these into valuable products.

According to various industry heavyweights, including a16z, Pantera Capital, Messari, Blockworks and us at Flagship the merger of AI & Crypto is one of the top investment trends for 2024. According to Messari, “The two most critical inputs for any AI are data and compute power, and as such, it seems likely that “AI will transact in a currency that preserves its energy purchasing power over time.” That’s bitcoin in a nutshell.``

There is a lot packed in that statement, just like we have seen various innovations exploring areas of mutual reinforcement between AI and crypto over the past few years. In this article, we provide a detailed overview of what has been taking place, highlighting the potential in this emerging conglomeration as well as projects that are leading the way in innovation.

The Basics of AI

The current excitement surrounding AI stems from its potential to revolutionize various aspects of our lives. Its ability to automate tasks, improve efficiency, and solve complex problems has captured the imagination of individuals and organizations alike. This rapid adoption is fueled by the increasing availability of data, advancements in computing power, and ongoing research breakthroughs in AI algorithms.

AI aims to replicate human-like abilities such as learning from experience, recognizing patterns, and making decisions. This is achieved by developing algorithms that can process and analyze vast amounts of data to identify patterns and relationships.

Modern AI systems leverage advancements in hardware, particularly the use of Graphics Processing Units (GPUs), which are well-suited for the parallel processing tasks involved in such algorithms.

Can AI Think?

While AI doesn't achieve true "thinking" in the same way humans do, it can exhibit behaviours that appear intelligent by learning and adapting to different situations. Two major techniques that underpin AI include:

• Machine Learning: a branch of AI where algorithms learn from data without being explicitly programmed. This learning process involves identifying patterns and relationships within the data, allowing the algorithm to make predictions or decisions on new, unseen data.
• Deep Learning: a powerful subset of ML inspired by the structure and function of the human brain. It utilizes artificial neural networks, which are interconnected layers of processing units that can learn complex patterns from data. It excels at tasks like image recognition, natural language processing, and speech recognition, where large amounts of data are available.

What AI Can Do

When people think of AI today, they think of ChatGPT and DALL-E, but those are a very small part of what AI can do. Those are called Generative AI, referring to the capacity of AI to creatively generate content in different formats, including text, images, videos, and even code.

However, AI can solve even bigger problems. AI plays a crucial role in self-driving cars, facial recognition technology, virtual assistants, fraud detection, e-commerce personalisation, even medical diagnoses, and a lot more.

How Blockchains, Smart Contracts, and Tokens Can Be Valuable for AI

AI and blockchain technology both seem quite promising in terms of revolutionizing how we learn, work, and connect. There's an interesting discussion happening around finding ways to combine them – to fuse the innovative potential of AI with the transparency and security of blockchains. Of course, challenges remain in implementation, but the enthusiasm is palpable in exploring what these technologies can build together.

Centralised vs Decentralised Ownership, Control, and Understanding

Currently, AI is dominated by a few big tech corporations, such as Google, Amazon, Microsoft, OpenAI, Apple, etc., with the attendant risks of centralization which can manifest in various ways, not the least in the form of algorithmic bias. For instance, Google was forced to suspend the just-released image generation feature on its Gemini (formerly Bard) app following mounting reports that the tool was generating inaccurate images in a backfiring attempt to subvert racial and gender stereotypes.

Don’t forget that AI is just a tool, and ultimately, it will reflect the values and perspectives of the creators. If these creators are limited to the dominant players, then there is a problem. For one, these big tech companies already have the advantage of network effects; the more users and data they acquire, the greater their power. Often, such power leads them to influence regulatory frameworks in their favour, in the process, stifling innovation from others and starving the competition. The fact that they have amassed extensive data repositories does not help either, since it means small players have to deal with a higher-than-normal barrier of entry.

However, blockchain can solve these problems by introducing decentralisation to democratise AI development. Distributing ownership and control across a network of participants prevents any single entity from wielding undue influence. This will pave the way for the development of diverse and unbiased AI solutions, fostering greater inclusivity and fairness in the field. Another concern that blockchain addresses is that the currently prevailing approach of AI development where vast amounts of personal data are in the hands of a few big corporations is a huge security risk; a single data breach can have catastrophic consequences. Blockchain technology can introduce transparency into the AI development process and eliminate the existence of a single point of failure.

Open-source, Distributed Collaboration with Incentives

Decentralized collaboration allows for faster adaptation and innovation. There is a need to have smaller players on the field since their agility and flexibility in experimenting with new ideas come in-built. Ways in which smaller entities and even individuals can be encouraged to contribute to enhancing the quality of datasets for AI models include incentivising them to check for biases and anomalies to ensure a more diverse and representative dataset as well as verify data accuracy.

In addition, decentralized AI ecosystems are not restricted by geographical boundaries. By leveraging blockchain technology, individuals from all over the world can participate and contribute to the development process. By incorporating diverse viewpoints and data sources, the potential for biased AI models can be significantly reduced.

In addition, as we have already seen with deepfakes, as AI capabilities advance, the ability to manipulate content becomes increasingly sophisticated. This raises concerns about misinformation, intellectual property theft, and trust in digital media. Blockchain technology offers a solution through its ability to create a tamper-proof and transparent record of content creation and ownership. By anchoring content authenticity on the blockchain, the origin and integrity of digital content can be readily verified by consumers, creators, and institutions.

While communities like Hugging Face democratise access to AI models, blockchain-powered marketplaces go a step further. They incorporate token-based incentives, rewarding various participants in the AI ecosystem.

The AI Token Tech Stack

The AI tech stack is a pyramid that starts with the base protocols and extends to resources, infrastructure, applications, and models, each representing various layers of the stack. Let’s dive into each of them.

Base Protocols

These are the foundational layers upon which the entire AI token ecosystem is built. Base protocols provide the core infrastructure and technical specifications for building and deploying AI applications, models, and services on the blockchain. They are distinguished by their capacity to support multiple projects and tokens to create a network effect by facilitating collaboration, interoperability, and innovation among AI developers and users.

Examples of such protocols include SingularityNET, which aims to create a decentralised Artificial General Intelligence (AGI) as well as Ocean Protocol (OCEAN), an open-source protocol that connects data assets on the blockchain and enables providers to monetise access to their services. They aim to create a network effect by facilitating collaboration, interoperability, and innovation among AI developers and users.

However, perhaps the best way to think of base protocols is Ethereum for AI, a term that refers to the use of the Ethereum blockchain as a base protocol for AI projects and tokens. According to Ryan Adams, the founder of Bankless, AI agents will be the primary consumers of Ethereum blockspace by the 2030s, with Ethereum providing services to AIs that nations provide to humans.

Resources

Resources represent the essential elements that fuel the development and operation of AI applications within the token ecosystem and are analogous to raw materials in the construction industry. One major class of resources is the GPU (Graphics Processing Unit). More powerful than CPUs, GPUs are specialized processors crucial for efficient training and execution of complex AI models, especially those involving deep learning techniques. Blockchain-based solutions are exploring ways to optimize and incentivize the utilisation of GPUs for distributed AI processing.

Beyond hardware resources, high-quality data is vital for training and improving the performance of AI models. Typically, pre-trained models can serve as starting points or building blocks for developing new AI applications. With blockchain-based solutions, the aim is to incentivise data sharing through a decentralised model while upholding data privacy and security concerns. In short, ensuring access to sufficient and efficient computational resources, high-quality data, and relevant models is crucial for fostering a thriving AI ecosystem within the blockchain space.

Infrastructure

At the technical infrastructure layer, we refer specifically to the components that enable seamless interaction and exchange of data, models, and services within the AI token ecosystem. Thus, interoperability is crucial at this level. Applications and services need to be able to integrate with various blockchains, tokens, smart contracts, and dApps, fostering a more interconnected ecosystem. A well-developed infrastructure layer will promote collaboration, innovation, and adoption within the AI token ecosystem. Interoperability allows different projects and platforms to leverage each other's strengths, fostering a more robust and versatile ecosystem.

Applications

This layer represents the practical applications and services built on top of the underlying infrastructure and powered by AI models and resources. At this level, applications utilise the capabilities of AI to solve real-world problems and provide value to users. Such applications include the following:

• AI agents that help automate everyday tasks
• Autonomous services like on-chain governance systems powered by AI to facilitate decentralized decision-making
• Decentralized Finance (DeFi) applications can leverage AI to improve aspects like risk assessment, fraud detection, and more.

A concern here is that while applications might be the first to reach adoption due to their tangible use cases, their long-term growth potential may be limited due to the lack of strong network effects. Additionally, the utility of the associated token might not scale effectively with broader adoption.

Models

Maintaining and improving the performance of AI models requires ongoing effort and resources. We have spoken about efforts to decentralise AI model-building. Ultimately, what’s important is ensuring transparency and explainability in their decision-making processes. This helps build trust and address concerns about potential biases or unintended consequences of AI applications.

Models include algorithms for various tasks, such as deep learning for image recognition or natural language processing for sentiment analysis as well as open-source networks that promote collaboration and innovation within the ecosystem. For instance, Numeraire leverages the "wisdom of the crowd" to develop and improve AI models for financial forecasting. Participants submit their predictions and receive rewards based on the accuracy of their forecasts, contributing to the collective intelligence of the platform's models.

Sub-sectors and Popular Projects

In this section, we discuss subsectors emerging out of the merger as well as the current major projects driving innovation in the various sub-sectors.

Protocols

While protocols hold immense potential, several challenges need to be addressed for widespread adoption. These include establishing robust security measures, ensuring data privacy, and fostering clear governance models for decentralized decision-making within the network.

Within protocol ecosystems, applications are designed to be interoperable, allowing for seamless integration and interaction between different AI modules and services. They provide shared resources, such as computing power, datasets, and algorithms, accessible to participants within the network. This shared infrastructure enhances efficiency, reduces redundancy, accelerates innovation, and creates network effects in AI-Crypto protocols.

• Bittensor: A decentralized machine learning protocol that allows users to monetize their intelligence and data. It uses a native token called TAO to reward participants and enable transactions.
• enqAI: the enqAI project aims to provide unrestricted AI services with image/audio generation and large language models, powered by a decentralized GPU network.
• Commune AI: a protocol that aims to facilitate interoperability in software development by enabling developers to easily share their tools and connect with others.
• HyperCycle: this allows AI machines to interact with other AI machines and for their creators to easily monetise them, whether remote or local. The idea is to create the ‘Internet of AI’.

Compute

In the past few years, and more so in the past few months, Nvidia’s revenues and stock price have soared massively and the GPU manufacturer is now the third most valuable company on Wall Street by market capitalisation, only behind Microsoft and Apple. This is not a coincidence as the demand for GPUs has increased with the recent rise of AI.

The traditional approach to accessing computational resources for AI model development involved establishing and managing dedicated GPU farms, which could be prohibitively expensive and complex. However, blockchain-based platforms can democratise access to compute resources by offering a decentralised network of GPUs. This enables individuals and smaller organisations to engage in high-level AI model creation without the need for significant infrastructure investments. Plus, owners of GPUs get to monetise their hardware by participating in a mutually beneficial model.

Ensuring the scalability of these decentralised GPU-sharing networks to accommodate growing demand and evolving AI requirements is essential for their long-term viability.

• Render: the Render Network decentralises access to GPU cloud rendering on the blockchain for 3D content creation. It also enables granular digital rights management.
• Akash: another decentralised compute marketplace. The Akash Supercloud is built on Kubernetes and offers persistent storage as well as censorship-resistant app deployment.
• Nosana: organisations with compute hardware can monetise it by becoming a Nosana Node. It promises costs up to 85% lower than traditional public clouds.
• Netmind Power: is a machine learning platform powered by a global GPU network, allowing users to rent virtual machines, train their own models, and earn with idle GPUs.

Data

Blockchains, by design, record every transaction immutably, creating a historical record of immense value. This data encompasses details like sender, receiver, amount, timestamp, and potentially even additional relevant information depending on the specific blockchain. This comprehensive record provides a unique window into user behaviour, network activity, and market trends when analysed using AI approaches.

AI models can be trained on historical data to predict future trends within the blockchain ecosystem. This can be particularly valuable for investors seeking to make informed decisions in volatile cryptocurrency markets. For instance, AI models might analyse historical price movements, transaction volumes, and network activity to predict future price fluctuations or identify promising investment opportunities.

Models and Algorithms

AI models and algorithms can help crypto platforms to improve their performance, scalability, and usability. For example, AI can help optimize blockchain consensus protocols, enhance smart contract functionality, and enable decentralized oracles. In turn, crypto can help protect AI data and models from unauthorized access, tampering, or leakage, and enable verifiable and transparent AI outcomes. Some projects at this intersection include:

• Numeraire: An AI-based hedge fund that uses the collective intelligence of data scientists to predict the stock market via machine learning. It has so far paid over $70 million to data scientists.
• Federal AI: the aim of this project is to decentralise accessibility to hardware and training of models via its Federated Learning approach, which reinforces AI with blockchain.

Storage

Training AI models requires vast amounts of data, often including sensitive information. Cryptography and blockchain technology can provide secure storage solutions for this data, ensuring immutability. Data stored on a blockchain cannot be tampered with, guaranteeing its authenticity and integrity for training reliable AI models. In this regard, balancing the need for secure storage with data privacy regulations is crucial, particularly when dealing with sensitive AI training data.

AIgents

AIgents (or AI agents) are soaring on the promise of automating repetitive tasks currently handled by human users, not just by following pre-set instructions, but by being able to make innovative decisions in real time. Unlike humans, AI agents operate 24/7, constantly monitoring markets and reacting to opportunities, potentially leading to a more efficient and responsive crypto ecosystem. They can leverage their access to vast amounts of data and analytical capabilities to make data-driven decisions.

However, regulatory frameworks surrounding both AI and crypto are still nascent, and security concerns regarding the potential for AI agents to be exploited for malicious purposes need to be addressed. For one, the potential societal and ethical implications of AI agents controlling vast financial resources warrant careful consideration and mitigation strategies.

Why can’t these AI agents be deployed in the mainstream financial system instead? The financial system is currently fragmented and inefficient, compared with the promises of crypto; various institutions and regulations hinder seamless data flow and automation. The blockchain, on the other hand, offers a unified and transparent platform where AI agents can access and manipulate data efficiently. And the immutable nature of transactions provides a secure and reliable foundation for AI agents to operate on.

Ethereum founder Vitalik Buterin considers AI as a player in a game as the most viable level of the deployment of AI agents, compared to AI as the interface, rules, or objective of the game. As players, AI agents themselves can act as traders, miners, validators, or arbitrageurs in decentralised markets, platforms, or networks, where they can earn rewards or fees for providing services or contributing resources.

• Fetch AI: Fetch.ai describes itself as the first open network for AI agents, enabling businesses to build and monetise AI apps and services through its AI Engine, Agent Services, and other features.
• Autonolas: Also known as the Olas network, or simply Olas, Autonolas is trying to be a single network for all the building blocks of crypto and it is built on autonomous agent technology.
• SingularityNET: SingularityNET is a decentralised, blockchain-based AI marketplace, allowing users to integrate AI services from the community of providers on its platform.
• Balance AI: a marketplace for driving innovation through AI models by helping creators share and monetise their efforts. Its current focus is on enhancing the quality and credibility of AI models admitted to its network.
• AIgent X: AIgentX offers various AI-based solutions for trading analysis, business support, community management, and so on, intending to help businesses streamline their processes via AI dApps on its platform.
• Paal AI: a Google Cloud-supported platform that allows users to create and integrate advanced crypto and AI bots for online communities.

DeFi

AI excels at processing vast amounts of data efficiently, making sense of the complex DeFi landscape. Intelligent algorithms can identify patterns and trends within the data, enabling them to predict market movements and optimise strategies. For instance, even basic bots have historically proven more effective than humans in performing arbitrage due to their speed and accuracy. With the latest advancements in AI, the potential for sophisticated arbitrage strategies becomes immense.

Beyond the basics of arbitrage, though, AI holds immense potential in various DeFi applications in the following ways:

• Price Discovery: analysing market data to determine the "fair" price of an asset.
• Yield Optimization: compare yields across different DeFi protocols and automatically shift investments to maximize returns.
• Risk Management: identify and assess risks associated with investments, allowing users to make informed decisions.
• Portfolio management: AI can automatically manage tokens on behalf of traders. However, reputable crypto portfolio management platforms such as Flagship Vaults still use humans in order to avoid the current instability of AI for such delicate tasks involving private data.
• Fraud Detection: detect and prevent fraudulent activities within DeFi ecosystems, enhancing security and user protection.

Some of the leading projects in this area include:

• Kyberswap: by connecting one’s preferred wallet to KyberSwap, a user can trade efficiently, personalise their trading interface, and even earn trading fees and additional rewards.
• Hera: the Aggregator feature of Hera Finance uses AI to enable users access to swap transactions efficiently to earn maximum returns.

Mozaic: Mozaic simplifies yield farming by leveraging an intelligent AI named Archimedes to automate farming activities across multiple blockchains.

Generative AI

Generative AI is being used to create unique digital assets like NFTs, fostering new avenues for artistic expression and potentially influencing the NFT market. Despite its promise, generative AI raises ethical concerns regarding copyright infringement, misinformation, and the potential for misuse, particularly in the creation of deepfakes and deceptive content.

The rapid advancement of generative AI technology also poses challenges in regulating its use and ensuring accountability for the content produced. Additionally, there are technical challenges related to training data biases, model interpretability, and the generation of diverse and realistic content across different domains.

• ImagineAI: An AI-powered art generator that also offers image editing (Image Remix) features. Soon, it’ll allow users to replace the background of photos, expand images, and remove unwanted objects.
• Botto: Botto is a decentralized autonomous artist that creates artwork based on input from the community. Botto's art engine generates over 4000 unique images weekly and the most popular one is minted as an NFT (non-fungible token) for auction.

The Metaverse

The metaverse is envisioned as a network of interconnected virtual worlds where users can interact, socialize, and participate in various activities through avatars. It represents a significant shift in how we interact with technology, blurring the lines between the physical and digital realms. In metaverse worlds, crypto tokens represent virtual assets like clothing, land, possessions, and other in-game assets. AI plays a crucial role in shaping the metaverse experience in several ways:

• Building Intelligent Environments: Imagine AI-powered landscapes that adapt to user behaviour or generate realistic weather patterns.
• Populating the Metaverse with Life: AI can contribute to creating non-player characters (NPCs) with believable personalities and the ability to engage in meaningful interactions with users.
• Personalization and Customization

Here are some projects launched in this area:

• SophiaVerse: a metaverse built into the SAIL game that aims to achieve what it calls tokenised sentience. SophiaVerse is powered by the Sophia robot which has made its mark in the generative AI space via NFTs.
• Alethea AI: it connects Generative AI and the blockchain. it launched CharacterGPT, a multimodal AI system that can generate interactive AI characters via natural language prompts. It is backed by Binance, Polygon, Crypto.com, and more.
• Futureverse: this is a platform that enables developers to create metaverse worlds powered by AI. It provides an AI gaming platform as well as SDKs and an AI protocol that allows anyone to own, train, and trade unique AI.
• The Root Network: in short, a modular development toolkit that enables asset interoperability across the open metaverse and enables users to have more seamless and interconnected experiences.

Risks

Reading up to this point, it might seem that the news is all rosy and positive. But that is not the case. Several people are already rightly concerned about the two independent technologies: AI and crypto. With the potential of both merging, there are even heightened concerns and it remains to be seen how these risks will affect the trajectory of the merger taking place. Let’s look at some of the major potential obstacles.

Regulatory Challenges:

Both AI and crypto are rapidly evolving technologies with a lack of established regulations. This creates uncertainty regarding how AI-powered crypto projects should (and will) be regulated. Moreover, while decentralisation is a core principle of crypto, many regulations aim to establish stringent control and oversight. This creates friction when applying traditional regulations to inherently decentralised systems based on blockchain technology.

Unforeseen Risks:

Merging two complex and emerging technologies such as AI and crypto can lead to difficult-to-predict consequences, which may be related to security vulnerabilities, unintended biases in AI algorithms, or unexpected interactions between the two technologies. One potential consequence is the emergence of "unstoppable AI" within crypto projects. If an AI application operates on a truly decentralized blockchain, it might become impossible to control or shut down even if it becomes harmful.

Malicious actors:

Crypto struggles with credibility problems in a lot of countries because the anonymity and ease of setting up crypto projects create opportunities for bad actors to launch scams and pump-and-dump schemes, potentially exploiting the hype surrounding AI integration. Many crypto projects have poorly designed tokenomics causing manipulation, market crashes, and ultimately harm to investors. Also, projects might use misleading marketing tactics like claiming to utilise advanced AI when they are simply wrapping existing functionalities that lack true AI integration.

Balancing Transparency and Confidentiality:

The blockchain is known for its transparency, but some AI applications require data confidentiality to function effectively. This creates a conflict when trying to balance the need for transparency in the blockchain with the need for data privacy and confidentiality in AI applications. Ensuring compliance with regulations like the GDPR, CCPA and other emerging regulations while maintaining the decentralized nature of blockchain poses a significant challenge. Innovative solutions are needed to navigate this complex space.

Security Concerns:

All AI models have inherent vulnerabilities, and even limited access to an AI model can be enough for malicious actors to identify and exploit these vulnerabilities to bypass security measures and potentially harm users. Integrating AI assistants into open-source wallets introduces potential security vulnerabilities. Malicious actors might exploit vulnerabilities in the AI assistant or the wallet itself to bypass security measures and steal user funds.

Closing Thoughts

As explored, the confluence of AI and crypto is a fascinating and potentially transformative development that could have far-reaching implications across various domains. Yet, it also raises important ethical and regulatory considerations. Issues such as algorithmic bias, data privacy, data security, and centralisation must be carefully addressed for the ecosystem to yield the full benefits of the merger.

Finally, AI and crypto experts and enthusiasts, institutions, organisations, regulations, and policymakers must collaborate to navigate these challenges and harness the full potential of this transformative convergence.