The Trust Paradigm: Bridging AI with Blockchain for Onchain Economies

As we witness AI surpass human judgment and intelligence, the challenge we face is not just technological but philosophical: How do we align these agents of intelligence with our human values and ethics, and establish trust? 

The future belongs to those who are proactive—thinkers and builders who not only identify challenges but also forge solutions. It's a call to action, urging us not to be bystanders but active participants in shaping a future where technology and humanity converge harmoniously. With the future rapidly evolving at an unprecedented pace, understanding and adapting to these changes is not just important—it's imperative for staying relevant in this fast-evolving landscape.

It is crucial to prioritize the development of systems that ensure the preservation of human intent within AI models. Over the past year, the field of Artificial Intelligence has witnessed a notable surge in the emergence of new models and applications, coinciding with a substantial rise in the demand for GPU computing resources. This trend, reminiscent of a Cambrian explosion in technology, has been noticeable since the release of the seminal 2017 paper  'Attention is All You Need' by Ashish Vaswani and fellow Google researchers. This paper introduced the Transformer Architecture, sparking a wave of innovation in model architectures, including breakthroughs such as ChatGPT.

With the immense power of AI comes a substantial responsibility. In this context, we must consider various perspectives on the advancement of intelligent models. 

One such viewpoint is existentialist, as exemplified by the Machine Intelligence Research Institute (MIRI). MIRI's recent unveiling of their 'Death With Dignity' strategy reflects a certain acknowledgment of the challenges in aligning AI with human values: "It's evident at this point that humanity isn't going to solve the alignment problem, or even try very hard, or even go out with much of a fight. Since survival is unattainable, we should shift the focus of our efforts to helping humanity die with slightly more dignity." The alignment problem revolves around the philosophical and ethical congruence between intelligent machines and humans.

Considering Peter Thiel's insights from his 'Anti-Anti-Anti-Anti Classical Liberalism' talk at the Oxford Union, we recognize that existentialists risk falling behind. The future necessitates proactive engagement and optimistic adaptation. In this swiftly evolving landscape, our objective should be to ensure that AI development aligns with ethical principles and human values, fostering a future where technology augments human potential rather than diminishing it.

Just as Satoshi Nakamoto's 2008 Bitcoin white paper emerged in response to the trust deficiencies in the financial system highlighted by The Great Financial Crisis, we see a similar opportunity to leverage cryptographic verification and game-theoretic equilibrium in AI. These technologies, pivotal in building secure blockchain systems, have the potential to instill comparable trust in AI. By adopting these principles, we can develop distributed AI systems that are inherently secure and trustworthy.

To make the convergence of AI and Crypto more tangible, it's helpful to envision the general direction in which we are heading, using it as a visionary compass. Mason Nystrom from Variant Fund, in his recent essay, emphasized how 'Bots are becoming the first-class citizens of crypto.' This essay shed light on the burgeoning role of autonomous agents in the crypto landscape. Currently seen as novelties, these AI agents are on the brink of a major evolution. In the near future, they are expected to enrich consumer experiences, play pivotal roles in various protocols, and drive self-sustaining economies.

Some exciting and already emerging use cases for AI include: 

• Asset and Risk Management: In the realm of modern finance, AI and ML models have become indispensable for tasks such as fraud prevention and evaluating unsecured lending risks. Ensuring the implementation of these models in a verifiable and tamper-proof manner is crucial, particularly for maintaining the open and permissionless nature of decentralized finance (DeFi) products. While automated trading strategies have been prevalent in traditional finance, their integration into DeFi is still evolving. Initiatives have been launched to leverage AI/ML in creating decentralized trading strategies, with notable examples like the experimental bot developed by Modulus Lab in the DeFi space.
• Tailored Workflow Agents: Beyond the broad capabilities of general-purpose AI applications like ChatGPT, there is a growing need for AI agents precisely calibrated for specific sectors, topics, and market niches. Platforms such as Bittensor are emerging, providing incentives to developers or "miners" to fine-tune models for targeted tasks such as image generation, pre-training, and predictive analytics. These platforms focus on diverse sectors, including cryptocurrency, biotechnology, and academic research. Although Bittensor is in its early stages, developers are already leveraging it to craft specialized applications and agents using a vast array of open-source large language models.

Envisioning autonomous agents as integral to future on-chain smart economies, we can identify high-level challenges in realizing this vision.

Trust in smart economic agents is paramount, especially in aspects such as the execution of actions, the reasoning behind decisions, and the hosting of these agents. Addressing these concerns requires a deep dive into the infrastructure that will underpin tomorrow's economy.

However, several high-level challenges exist within our current AI technology stack:

• Insufficient Trust: Present platforms lack robust assurances in computational accuracy, data confidentiality, and censorship resilience, impacting model execution and accessibility across different regions. Establishing a foundation of trust is crucial for widespread adoption and utilization of AI technologies.
• Centralized APIs: The infrastructure is heavily dominated by a few major corporations, limiting the development of native integrations and often resulting in availability and functionality issues. Diversifying the landscape and promoting competition can foster innovation and resilience in AI systems.
• Limited Hardware Accessibility: Developers are increasingly challenged in acquiring AI-specific hardware, further exacerbated by substantial costs imposed by hardware suppliers. Overcoming these barriers is essential to democratize access to AI development resources and encourage a more inclusive ecosystem.
• Misalignment in Structure: Organizations find themselves at a crossroads between keeping AI models proprietary, hindering innovation and centralizing control, and open-sourcing models without adequate infrastructure to recognize and reward contributors. This dilemma leaves users with limited influence over AI governance and ownership, highlighting the need for a balanced and sustainable approach to model sharing and development. Addressing this misalignment is critical for fostering a collaborative and user-centric AI ecosystem.

Additionally, the ambiguous nature of Machine Learning (ML) introduces trust issues at every process layer as highlighted by Grace & Hill from SevenX Ventures. TThis spans concerns related to the integrity and privacy of input data, the accuracy and confidentiality of outputs, and the secure execution of various model types or algorithms. The opacity of ML algorithms, coupled with privacy considerations surrounding model parameters, exacerbates the trust gap between model owners and users.

The past year has witnessed great strides in the intersection of Crypto and AI tackling the inherent challenges and complexities we have described above. Three key areas have emerged as focal points in this innovative landscape:

1. Model Inference with On-Chain Proofs and Off-Chain Computation

The integration of Zero-Knowledge Proofs (ZK Proofs), particularly zkSNARKs, in Machine Learning (referred to as ZKML) is making significant strides in addressing pivotal questions about model accuracy, authenticity, and parameter correctness. ZkSNARKs play a crucial role in proving the accuracy of computations in ML models without compromising data privacy. The challenge lies in translating complex AI model statements into computational circuits suitable for ZK Proofs while optimizing for neural network performance.

Advancements in ZKML are evident with the development of new ZK proof systems specifically designed for intricate datasets. However, this progress is tempered by the need to balance computational intensity with practical efficiency. An exemplary illustration of this advancement is seen in  Modulus Labs, actively benchmarking various proof systems to optimize on-chain inference, with a particular focus on processing time and memory usage.

Despite the promise of ZK technology in enhancing AI's trustworthiness, the trade-offs between resource demands and practical application remain a crucial consideration. Diverse solutions, such as Axiom’s ZK co-processor for Ethereum and RISC Zero’s  cryptographic receipts, highlight the varied approaches to incorporating ZKML. These initiatives represent significant steps toward a future where AI's reliability is bolstered by blockchain and ZK Proofs, opening new possibilities while navigating the complexities of real-world applications.

2. Distributed AI Model Compute

The challenge of distributed computing for AI models is being innovatively addressed by projects like Ritual and Gensyn, each offering unique solutions to streamline and enhance AI processing capabilities.

Ritual is at the forefront of pioneering a sovereign execution layer explicitly designed for AI. This platform connects a distributed network of nodes with access to compute resources to AI model creators. This approach enables model creators to deploy their models on these nodes, providing users with an efficient way to access a diverse range of AI models through a unified API. A notable feature of Ritual is its cryptographic infrastructure, placing trust and verification at the forefront. By ensuring computational integrity and privacy, Ritual addresses significant concerns in distributed AI computing, such as data security and model reliability.

On the other hand, Gensyn is tackling the computational demands of state-of-the-art AI through a trustless protocol that efficiently harnesses global computing resources. By establishing a marketplace for compute power, Gensyn balances supply and demand, presenting a scalable and cost-effective solution for deep learning computation. The platform's innovation lies in its work verification strategy, ensuring that computational tasks are accurately executed within its trustless network. These initiatives exemplify the diverse approaches taken to overcome the challenges posed by distributed computing in the realm of AI.

3. Distributed ML Rails: 

The integration of distributed machine learning (ML) with blockchain introduces two transformative elements: native payment rails and AI agent wallet ownership. This integration significantly enhances the operational capabilities of AI agents in the crypto ecosystem.

Connecting AI agents to cryptocurrency wallets empowers them with digital asset ownership, including NFTs and yields. This ability to autonomously own and transact with digital assets revolutionizes their role in the digital economy. AI agents can now actively engage in transactions and decision-making processes, previously exclusive to human operators. This seamless integration further blurs the lines between human and machine participation in the crypto space, ushering in a new era of decentralized and autonomous capabilities for AI agents.

In conclusion, the intersection of Blockchain and AI represents a frontier of innovation where challenges are met with groundbreaking solutions. The journey through these domains underlines a pivotal shift towards a more secure, transparent, and equitable digital economy. As we embrace this era of rapid technological evolution, our focus remains steadfast on harnessing these advancements responsibly and ethically, ensuring they align with human values and societal needs.

As we navigate the intricate relationship between Blockchain and AI, it is evident that these converging technologies hold immense promise. However, the journey is not without its challenges – from ensuring transparency in Machine Learning models to striking a balance between computational intensity and efficiency in distributed computing. The industry is actively addressing these concerns, pushing the boundaries of what is possible and setting the stage for a future where the collaboration between Crypto and AI reshapes industries, economies, and the very nature of human-machine interactions. In this dynamic landscape, the fusion of cutting-edge technologies continues to propel us towards a new era of possibilities, where the synergies between Blockchain and AI redefine the way we perceive, create, and interact with intelligent systems.

Looking Ahead: Deep Dives in the Series

This article serves as an introductory piece in a series dedicated to delving deeper into each topic discussed. In forthcoming articles, we will explore each area in greater detail, unpacking the complexities and examining the nuances of how AI and blockchain are converging to redefine the onchain economies. We will look closely at specific case studies, emerging technologies, and the challenges and opportunities they present. Stay tuned for insightful deep dives that promise to enrich your understanding of this fascinating and rapidly evolving field.

+++

This publication is intended solely for informational purposes. It is not meant to serve as investment advice, nor as an endorsement or recommendation to buy or sell any investment. It should not be used as the primary basis for making investment decisions. Additionally, this document should not be considered as a source of legal, accounting, tax advice, or specific investment recommendations. Individuals are encouraged to consult their own legal, business, tax, and other professional advisors regarding any investment matters.

Information contained in this document may include third-party data, potentially including data from companies within Elixir Capital's portfolio. While such information is sourced from what are believed to be reliable resources, Elixir Capital has not independently verified this information and does not make any representations as to its long-term accuracy or suitability for any particular situation.

The views expressed in this post represent the current opinions of the authors and are not made on behalf of Elixir Capital or its clients. They do not necessarily reflect the views of Elixir Capital, its managing partners, affiliates, advisors, or associated individuals. Furthermore, these opinions are subject to change and may not be updated accordingly.