April 26 2024
Deep Dive: Symbiotic AI-Crypto Nexus
AI-Crypto Intersection Enables Decentralised Trust: AI and blockchain synergy promotes decentralised AI development, addressing concerns of centralisation, data privacy, and disinformation. This intersection enhances transparency, verifies content, and protects intellectual property, fostering secure AI ecosystems
AI Empowering Blockchain: AI enhances Blockchain by addressing limitations like inefficient data management, high energy consumption, scalability challenges, redundant tasks, and security vulnerabilities, ultimately improving data access, analysis, and security in decentralised AI systems
Integrating AI and Crypto for Tangible Value: AI and crypto must work together to create tangible value for a project to qualify as AI-Crypto, as superficial combinations are common; a market map can illustrate use cases related to product-market fit that mutually combine these technologies.
Decentralisation Networks Alleviate Resource Intensive AI: Decentralised networks providecomputational resources in training generative AI models at a time when there is a long waiting list for high-performance GPUs and increasing costs associated with their utilisation
Regulating AI with Blockchain: Blockchain-based AI regulation utilises decentralisation, transparency, and immutability for licensing, enforcement, traceability, and adaptability, setting new standards for data protection and responsible AI development and innovation
Unlocking the Synergy between AI and Crypto: AI and crypto convergence revolutionises industries, enhancing security, transparency, and user empowerment. Tagus Capital invests in this convergence to drive innovation and disruption, supporting Web3 entrepreneurs in the UK, USA, and Europe.
AI and Blockchain - A Decentralised Future: Artificial Intelligence, particularly generative AI (GenAI), with advanced technologies like large language models (LLMs) and neural networks, simulates human intelligence through machine learning and natural language processing. This is set to revolutionise company operations and people's lives by automating processes, providing predictive analysis, and enhancing customer experiences. As AI integration increases, the AI-based solutions market is set for significant growth in the coming years. However, the development of GenAI is becoming increasingly centralised in the hands of big tech companies, which have exclusive access to vast computational resources and data, and can decide which users or groups of users have a say (or not) in the digital world. This concentration of power raises concerns about their ability to dictate the norms and values that govern AI models and potentially exert outsized influence on the AI industry. Centralised AI systems also pose concerns about disinformation, data privacy and ownership, and fraud, which could lead to increased litigation and operational complexity.
To address these challenges, blockchain technology and crypto protocols offer a decentralised solution. Blockchain, a decentralised and tamper-proof ledger technology, is transforming data security and transparency, initially for cryptocurrencies but now across various industries including finance, supply chain, and healthcare. Its adoption in these sectors drives market growth, offering secure, transparent, and traceable transactions. In short, Blockchain will become a crucial component of digital transformation. Blockchain can verify the authenticity of online content, prevent the misuse of intellectual property, and create an auditable track record of decision-making processes used by opaque AI algorithms. By capturing the lifecycle of content creation on a blockchain, it becomes possible to differentiate between content generated by AI and that created by humans. This enhances transparency and scrutiny in the centralised AI landscape.
By identifying opportunities for synergy between centralised AI and decentralised crypto networks, both technologies can address their respective weaknesses. Crypto networks, while sacrificing raw computational power for user control, face challenges in scalability, accessibility, governance, and practical adoption. The symbiotic relationship between AI and crypto can lead to the development of AI that evolves safely, providing powerful and beneficial features for humanity without being captured by special interests. Moreover, integrating AI benefits into crypto can help it transcend its niche status, enabling it to become a practical everyday tool for equitable and democratic participation in technology, finance, and other domains, ultimately fostering a more balanced and inclusive technological landscape.
Decentralised computing can also become a significant resource solution for addressing the increasing computational demands of AI. Blockchain technology enables AI agents to form agreements, and collaborate. As these agents become integral to every level of the stack, the AI Agent to AI Agent economy is expected to grow substantially.
In this report, "crypto" is used to encompass both blockchain technologies and cryptographic solutions(Baioumy and Cheema, 2024), i.e. the Web 3 ecosystem.
Blockchain Offers Decentralised Solution for Transparent and Secure AI Ecosystems: According to the AIAAIC database there has been a substantial increase in AI-related ethical misuse incidents, with a 26-fold rise from 2012 to 2021, highlighting growing concerns about the misuse of artificial intelligence. Since then, multiple incidents related to the misuse of AI have continued to emerge and more recently Google has been fined €250mn by France's competition authority for training Gemini AI models using news content without consent. Other events include OpenAI and Google’s Gemini demonstrable biases and ongoing regulatory assessments in the United States and the European Union, emphasise the importance of permissionless, censorship-resistant, and decentralised AI networks. Blockchain technology can enable AI systems to operate without centralised control or restriction, offering a promising alternative to centralised AI systems. By combining the strengths of AI and blockchain, it is possible to create more transparent, secure, and equitable AI ecosystems. As the AI industry continues to evolve, the integration of blockchain technology will become increasingly important to ensure the responsible development and deployment of AI.
AI Incidents on the Rise
Source: AIAAIC. Note: The figure omits 2022 AI incidents because of the extensive vetting process needed for inclusion in the AIAAIC database. Latest data available.
The Rise of Integrated AI and Crypto: Currently, the intersection of AI and crypto is expansive. However, for a project to qualify as AI-Crypto, the technologies must work together to create the solution, rather than just coexisting (Baioumy and Cheema, 2024). AI and crypto are often superficially combined without tangible value. Core frameworks for AI-Crypto involve a direct integration that enhances the overall project by leveraging the strengths of both technologies. This can be illustrated by a market map of use cases related to product-market fit that mutually combine these technologies.
Web 3/AI Market Map
Source: Tagus Capital, FLock.io. Respective company websites for logos. Note: The product market map presents a selection of companies and protocols active in the intersection of Web 3/crypto and AI, serving as an example and not an exhaustive list.
AI tokens are tokens designed to support AI-based projects, applications, and services. They can be used as payment for transactions on AI platforms or to grant governance rights to holders. These tokens power various AI-driven initiatives, such as decentralised AI marketplaces, AI trading algorithms, and autonomous organisations. Crypto projects directly involved in AI constitute a market capitalisation of above $25 billion, representing 0.09% of the total crypto market cap, and nearly tripling in market cap year-to-date.
Top 10 AI-Crypto Tokens by Market Cap
Source: Tagus Capital, Messari, Forbes, CoinGecko.
Tradeable tokens associated with AI have surfaced as stand-ins for advancements in AI technology. These tokens often gain value based on market sentiment, but may not always reflect the actual progress or substance of the AI projects they represent. For instance, the rally of AI-related crypto coins during March 2024, such as Worldcoin, coincided with events such as OpenAI's Sora AI video preview and NVIDIA's earnings release. According to Google Trends data, worldwide searches for "AI Crypto" reached new highs during this period. Therefore, many AI-related tokens could continue to be traded as a general proxy for AI progress and decentralised solutions. However as mentioned previously, at Tagus Capital we prefer screening for viable product-fits as indicated in our earlier framework and Market Map.
The Rapid Rise in AI Crypto Interest*
Source: Google Trends, Tagus Capital.*Note: The term's popularity on Google Trends is rated on a scale of 0 to 100, with 100 being the peak popularity. A value of 50 indicates that the term is half as popular as its peak. These values represent the term's search volume relative to the total number of searches on Google Trends.
Decentralisation and Democratisation: Blockchain has the potential to democratise the development and utilisation of AI systems, providing opportunities for global contributions. Currently, a few large organisations, such as OpenAI and Stability AI, dominate the generative AI market, creating barriers for new developers in terms of economics, culture, and geography. The centralised control over artificial intelligence, such as large language models (LLMs), which is concentrated in the hands of Microsoft and Google, is stifling innovation and raising concerns about privacy, governance, and data ownership. Closed-source LLM providers such as OpenAI can monitor all user interactions with the model, compromising privacy. Centralised institutions have complete control over the ChatGPT model, resulting in amplified biases and inaccuracies. Furthermore, every user of a centralised LLM is essentially an accessible data contributor to the large corporations that own the model and the data, highlighting the need for fairer contribution incentives and data value assessments. Blockchain can address these challenges by fostering a more diverse range of contributions and redistributing ownership, enabling broader utilisation of AI benefits.
Microsoft, Google Control of LLMs Raises Centralisation Concerns
Source: Flock.io. Tagus Capital1.
1. Tagus Capital holds a minority stake in Flock.io.
AI Training and Optimal Decentralised Resources: The need for computational resources in training generative AI models makes them costly and there will be a growing demand for additional resources in the future, given the limited availability of high-powered GPUs. Utilising decentralised networks of computing power allows users to be rewarded for offering their idle GPUs, easing the computational burden and potentially lowering costs. This approach is already applied to other computationally intensive tasks, such as video rendering for video games and movies.
Large-scale Training Compute Costs (USD): Using Price of Actual GPU Used for Training
Source: Epoch AI. Notes: this chart illustrates the rising expenses associated with the training of machine learning ("ML") models. The significant costs of these models provide an advantage to established entities, which can be disadvantageous for new developers.
AI and crypto projects are still developing the necessary infrastructure for large-scale, on-chain AI interactions. Decentralised compute marketplaces are emerging to provide the required physical hardware, mainly GPUs and TPUs, for AI model training and inferencing. These marketplaces facilitate connections between compute providers and seekers, enabling the seamless transfer of value and verification of compute resources. Within the decentralised compute landscape, various subcategories emerge, offering specialised functionality, including machine learning training providers and initiatives that bridge compute and model generation for artificial general intelligence.
Verifiable Decentralised Data for AI:At the moment, Generative AI relies on extensive, generalised (not specialised) high-quality databases to address a wide range of queries. The availability of data for training these models is a significant limitation to their ongoing enhancement. Decentralised data marketplaces could provide these models with access to verified, diverse, and reliable data through agents that interact and transact. A system of token incentives and penalties could ensure that the data meets the required quality standards for these models.
Zero Knowledge Machine Learning (zkML) is gaining traction as a solution for projects seeking confidentiality and verifiable model outputs on-chain. It allows applications to handle heavy compute requests off-chain and post a verifiable output on-chain to prove the off-chain workload's completion and accuracy. Despite being expensive and time-consuming, zkML's popularity is growing, as seen in the increasing integrations between zkML providers and DeFi/Gaming applications leveraging AI models.
Developments in AI and Web3: The AI data storage,modelling and inference ecosystem is transitioning from Web2's centralised services, driven by Amazon Web Services (AWS), OpenAI, Google, and Meta, to Web3's decentralised networks, driven by innovators such as Render Network, I/O Net and Flock.io2. This shift represents a significant leap towards a more inclusive computing and data storage ecosystem, where resources are distributed across a network of nodes rather than being controlled by a single entity. Decentralised networks offer several benefits, including greater security, privacy, and transparency, as well as the potential for more equitable distribution of resources and value. As the AI ecosystem continues to evolve, this transition towards decentralised networks is likely to play a key role in shaping its future.
2. Tagus Capital holds a minority stake in Flock.io.
Centralised platforms like OpenAI have enhanced data processing and model training, while decentralised initiatives such as Bittensor and startups like Flock.io offer decentralised data infrastructure for Web3 applications. These platforms enable the public to contribute knowledge, improve AI models, and preserve privacy by integrating on-chain machine learning and redefining model training, finetuning, and inferencing processes. This platform technology enables and encourages communities to contribute their idle resources while ensuring privacy, creating a more efficient and inclusive AI ecosystem.
AI-focused tokens such as Ocean Protocol (OCEAN), The Graph (GRT), Fetch.ai (FET), Bittensor (TAO), Numeraire (NMR), Oraichain (ORAI), and Cortex (CTXC) have seen a significant increase in development activities, indicating potential for greater adoption and impact for AI in the cryptocurrency market. Ocean Protocol's 2024 roadmap, including AI-powered prediction bots and securing private data, has made headlines.
Identity: Generative-AI has made it easier to create content, but it has also led to an increase in misinformation and deepfakes. Blockchains and public key cryptography can address deepfakes, which are AI-generated content that impersonates people and events, posing security threats. Deepfakes can cause misinformation, fake news, IP theft, privacy concerns, security threats, and issues with trust and credibility. Current deepfake detection technologies are falling short due to accuracy, longevity, and accountability challenges. Blockchains can help by providing an immutable ledger, timestamping, decentralised verification, smart contracts for authentication, and content ownership and attribution. A practical solution involves leveraging public key cryptography, associating public keys with verified identities, and establishing a decentralised identity registry.
Emerging players in blockchain-based content authentication include, Story Protocol, Numbers Protocol, Atem, and Worldcoin. Story Protocol offers an open IP repository and modules for seamless interaction, with a focus on content ownership and cryptographic signing. Leveraging existing cryptographic hardware, See3 can prove whether a media was taken on an honest device or created by AI. By building a new ZK content standard, its privacy-preserving device reputation layer allows it to blacklist devices that misbehave, finally creating a cost for deepfake disinformation. Numbers Protocol provides content verification services, leveraging blockchain infrastructure for trust and monetisation. Atem is a decentralised content creation protocol using soul bound tokens and NFTs for content ownership and asset record-keeping. Worldcoin is a blockchain-based identity protocol that uses retinal scans to authenticate human users and a digital identification system that verifies personhood and authenticity of online content or interactions while preserving privacy. These companies are making progress in addressing deep fakes and authenticating content and individuals.
AI Enhancing Blockchain: While Blockchain has some limitations, AI can help overcome them, creating a perfect evolution. AI benefits in the Blockchain ecosystem include streamlined and innovative data management with AI algorithms and controlled access, optimised energy consumption, improved scalability, enhanced transaction efficiency, and augmented security. AI-enabled applications face a challenge in limited centralised data access, often unable to access data controlled by other entities or verify data authentication, resulting in low-quality data being used for predictions. Blockchain-authenticated, shared information enables AI systems to analyse big data patterns, unveiling new insights that deep learning algorithms use for accurate decisions and improved predictions. This approach addresses data centralisation issues in AI-powered business models, effectively providing decentralised AI
.
Advantages of Integrating AI within Blockchain Networks
Source: Tagus Capital.
AI can essentially address several limitations of blockchain, including:
Inefficient data management: AI streamlines data management processes and reduces complexity associated with existing methods, such as hashing algorithms and brute force techniques.
High energy consumption: AI refines data mining processes, optimising operations and reducing time, effort, and energy invested in data mining.
Scalability challenges: AI introduces advanced decentralised learning systems and novel data-sharing techniques, improving efficiency and creating opportunities for startups and enterprises in the Blockchain ecosystem.
Redundant tasks and transaction inefficiency: AI identifies the node likely to deliver solutions first, allowing others to cease their efforts, thus reducing costs and boosting system efficiency.
Security vulnerabilities: AI enhances security by integrating natural language processing, image recognition, and real-time data transformation capabilities into Blockchain's peer-to-peer linking.
AI can be leveraged to improve the security of DeFi platforms by identifying vulnerabilities in smart contracts and incorporating an AI-powered security layer into their blockchain infrastructure (Baioumy and Cheema, 2024). DeFi's protocol risks persist with $1 billion lost in 2023, representing 2% of total value locked, despite a drop from $54 billion in 2022 (IntoTheBlock, 2023). Whilst this is much lower than 2022 – the highest on record, this has been on the rise so far in 2024. The industry's struggle to attract mainstream investors is due to increasingly sophisticated attacks and users' lack of knowledge to protect themselves. Institutional investors are reluctant to invest in DeFi with such high risks, hindering the ecosystem's growth. AI can be used to reduce smart contracts vulnerabilities.
DeFi Total Value Lost due to Exploits
Source: IntoTheBlock.
AI can significantly contribute to crypto coding, creating a mutually beneficial relationship. By automating code reviews, identifying vulnerabilities, and optimising code performance, AI enhances smart contract security and uncovers potential threats in blockchain data. This collaboration leads to improved security, functionality, and reliability of cryptocurrency systems, ultimately benefiting both AI and crypto developers. Indeed, Ethereum co-founder Vitalik Buterin expressed his enthusiasm for using AI to identify potential flaws in Ethereum's codebase, which he considers the network's "biggest technical risk." However, some experiments have shown that AI can sometimes invent vulnerabilities or create more security issues than it solves. While AI can be a helpful tool for experienced coders, it should be used with caution, especially when deployed alongside high-risk applications like oracles.
Blockchain can, in turn, safeguard intellectual property (IP) in an AI-enabled world. For example, a company with proprietary data wants to use ChatGPT for various purposes without risking unfair use by competitors or violating confidentiality agreements. Transforming IP into NFTs with embedded smart contracts and storing them on a blockchain can enable the company to flag each piece with codes indicating usage permissions, potentially reducing infringement risks. Getty Images sued Stability AI over alleged copyright violations related to AI-based image generation, highlighting the legal uncertainties surrounding generative AI.
Regulatory Challenges: Gen AI is facing increasing scrutiny and a greater chance of regulation due to controversies surrounding AI language models. For instance, AI Gemini's comments on political figures, such as Indian PM Narendra Modi, have caused debates on its handling of sensitive matters. ChatGPT has also faced glitches generating confusing and nonsensical responses highlighting the potential risks of generative AI systems producing unexpected or inconsistent outputs. However, it is the centralised collection of data and allegations of copyright infringements that are at the heart of the challenges facing the AI industry. Regulatory issues, lack of standardisation, and interoperability pose hurdles to widespread development and adoption of AI. Nevertheless, progress has been made in addressing regulatory issues for both crypto and AI and improving interoperability at the intersection of the two. Such as the establishment of regulatory frameworks like Markets in Crypto-Assets Regulation (MiCA), which came into force in June 2023 (European Commission, 2024a), and the EU AI Act, which is the first-ever comprehensive legal framework on AI, due to be introduced during Q2-Q3 2024 (European Commission, 2024b).
Blockchain's unalterable ledgers and transparent operations also offer a new approach to AI regulation, balancing responsible AI development with continuous innovation. This union of blockchain and AI could change the regulation and interact with AI, setting a new standard for the digital future. Blockchain's role in AI regulation leverages its decentralisation, transparency, and immutability for monitoring and controlling AI systems. Key implementation aspects include:
Licensing: Blockchain can issue unique digital licenses for AI systems, storing capabilities, limitations, and ethical compliance data.
Regulation: Intelligent contracts automatically enforce regulatory adherence, ensuring AI operates within its intended use.
Traceability: AI actions are recorded on the blockchain, creating a transparent and unalterable audit trail for accountability.
Adaptability: Smart contracts adjust to new AI advancements and standards, maintaining regulatory compliance
In the AI industry, OpenAI is currently the dominant player, backed by Microsoft and other tech giants. Despite OpenAI's dominance, there are compelling reasons to suggest that open source AI has the potential to challenge and even surpass it in the market (Baioumy and Cheema, 2024), particularly due to the impending regulations on the AI industry. Governments worldwide are rushing to regulate and restrict the emerging AI industry, primarily targeting the traditional, centralised AI used by OpenAI. However, open source AI is decentralised and without a single owner, making it difficult for governments to regulate (Baioumy and Cheema, 2024).
Environmental Concerns: Deep learning costs for AI are escalating rapidly, with computational resources doubling every three to four months (OpenAI, 2018), and developing large-scale natural-language processing models can produce a significant carbon footprint. The focus on advancing technology without considering costs is driving this trend, and linear performance improvements often require exponential resource increases. AI is predicted to account for up to one-tenth of the world's electricity use by 2025 (Giles, 2019). Cryptocurrencies, particularly Bitcoin, also consume a considerable amount of electricity, impact the water ecosystem, and generate e-waste. However, newer blockchain networks are adopting eco-friendly consensus mechanisms, with Ethereum's proof-of-stake model resulting in a 99.97% reduction in emissions and operating on nearly 48% sustainable energy (Cambridge Centre for Alternative Finance, 2023). Further progress is needed though, particularly in AI, to promote renewable energy use and reduce emissions.
A balance between technological progress and environmental sustainability is essential for a sustainable future but the intersection of AI and Blockchain can also help. The partnership between AI and blockchain technology can transform the fight against climate change by providing transparency, accountability, and efficiency. This collaboration can create sustainable supply chains, optimise energy consumption, improves waste management and recycling, and enhances natural resource monitoring and conservation. AI and blockchain enable data-driven decisions and transparency, driving the establishment of sustainable practices and significant advancements for the environment (BSV Blockchain, 2022).
Supply chain transparency and sustainable sourcing: Blockchain and AI provide consumers with trust and transparency in sustainability by offering tangible proof of a product's eco-friendliness.
Energy efficiency and optimisation: AI's predictive capabilities can enhance power grids by optimising energy consumption across industries, buildings, and transportation systems. Blockchain ensures transparent and secure energy transactions.
Waste management and recycling: Blockchain serves as an unchangeable ledger for waste disposal and recycling, while AI analyses data to enhance waste management and reduce illegal dumping.
Natural resource monitoring and conservation: AI and blockchain make monitoring and managing natural resources more efficient by leveraging IoT data and maintaining data integrity on a secure, collaborative platform.
Decentralisation networks alleviate resource intensive AI: Decentralised networks providecomputational resources in training generative AI models at a time when there is a long waiting list for high-performance GPUs and increasing costs associated with their utilisation.
The combination of AI and blockchain therefore can facilitate well informed decision-making across several sectors, presenting significant advancements for the environment and driving the establishment of sustainable practices.
Unleashing Untapped Potential: The synergy between AI and crypto holds immense potential. Generative AI and blockchain offer unique benefits, such as endless content creation, disintermediation and users’ data empowerment. Blockchain enables transparent governance of AI models, while AI can improve smart contract security standards and network governance. AI potentially enhances and automates financial decision-making, and blockchain provides a transparent and secure transaction layer. As early-stage venture capitalists, Tagus Capital has explored the intersection of these two innovations that create new use cases and add dimensions to the traditional market map related to product-market fit. These advancements give rise to novel applications and are reflected in the evolving market landscape.
AI and crypto technologies are already making a significant impact, even in their early stages, but the synergies and potential are just starting to emerge as powerful catalysts for innovation and disruption across multiple industries, especially the heavily data-dependent ones, such as financial sector, pharma & healthcare, education, and media. As these technologies continue to mature and further integrate, one can anticipate a future where secure, decentralised, and intelligent systems become the norm, transforming the way people live, work, and interact. While current AI's large-scale models concentrate power and compromise privacy, crypto's decentralisation emphasises user control and empowerment. Strategically combining their strengths can create AI that evolves safely, serving humanity without undue influence.
While AI-linked crypto-assets have outperformed non-AI crypto-assets, some of this may be due to hype surrounding AI sentiment rather than actual substance. That said, as AI systems evolve, challenges and solutions will deepen the intersection between crypto and AI. At Tagus Capital, we have long been advocates for the transformative power of blockchain and the crypto ecosystem, as well as AI and other cutting-edge technologies such as robotics, quantum computing, machine learning, and reverse computing. Our purpose statement reflects the commitment to investing in these rapid evolving developments, supporting entrepreneurs at the forefront of Web3 innovation in the UK and Europe.
References
Baioumy, M., and Cheema, A. (2024). AI x Crypto Primer. Available:https://alexcheema.github.io/AIxCryptoPrimer.pdf (Accessed: March 22, 2024).
BSV Blockchain. (2022). The positive impact blockchain technology is having on the environment. Available: https://www.bsvblockchain.org/news/the-positive-impact-blockchain-technology-is-having-on-the-environment (Accessed: March 23, 2024).
Cambridge Centre for Alternative Finance. (2023). Ethereum’s climate impact: a contemporary and historical perspective. (Accessed: April 2, 2024) https://www.jbs.cam.ac.uk/2023/ethereums-climate-impact-a-contemporary-and-historical-perspective/#:~:text=This%20novel%20update%20released%20by,impact%20of%20The%20Merge%20from (Accessed: March 15, 2024).
European Commission. (2024a). Crypto-Assets: What the EU is doing and why? Available at: https://finance.ec.europa.eu/digital-finance/crypto-assets_en (Accessed: April 24, 2024).
European Commission. (2024b). Shaping Europe's digital future. Available at: https://digital-strategy.ec.europa.eu/en/policies/regulatory-framework-ai (Accessed: April 24, 2024).
FLock.io. (2024). Decoding Web3 AI: The Next-Gen Tech Stack. Available at: https://flock-io.medium.com/decoding-web3-ai-the-next-gen-tech-stack-24f56f0052fa (Accessed: April 23, 2024)
Giles, M. (2019). Is AI the next big climate-change threat? We haven’t a clue. MIT Technology Review. Available: https://www.technologyreview.com/2019/07/29/135905/is-ai-the-next-big-climate-change-threat-we-havent-a-clue/ (Accessed: March 25, 2024).
IntoTheBlock. (2023). DeFi exploits. Available at: https://app.intotheblock.com/perspectives/defi_exploits (Accessed: March 22, 2024).
OpenAI. (2018). AI and Compute. Available: https://openai.com/research/ai-and-compute/ (Accessed: March 18, 2024).
Related Posts
