The Future of AI Is Decentralized

The AI Summit at Consensus 2024 takes place Friday, May 31, in Austin, Texas.

If that comes across as an exaggeration, take note of the recently discovered weakness in Hugging Face, a prominent AI-as-a-Service provider. This flaw might enable maliciously modified models uploaded by users to run unintended code through their Inference API function, thus expanding their control. Thankfully, this issue was detected early on and appeared not to have caused significant harm to the users; however, experts emphasize that such vulnerabilities are unfortunately commonplace.

As an analyst, I believe the challenge isn’t rooted in AI technology itself but rather in the antiquated X-as-a-Service models that persist today. In these outdated systems, there’s no motivation to ensure security or create applications tailored to market demands and ordinary users. To usher in a future where AI is safe, secure, and can harness immense computational resources, we must upend the traditional cloud model and wholeheartedly adopt decentralization.

‘Big Cloud’ and the monopolization of AI

Large companies such as Microsoft, OpenAI, Google, and Amazon hold significant power in the AI industry due to their vast financial reserves, abundant human talent, and access to advanced computing resources that enable them to effectively implement AI technologies on a grand scale.

It’s detrimental to the progress of artificial intelligence when AI development is restricted to a select group of technologists at multibillion-dollar tech companies in California. Their biased perspectives can result in one-dimensional and subjective AI agents, impacting various sectors including finance, creativity, and human interactions.

The case against monopolizing the AI market is just as strong from a technical standpoint. During its development, AI requires a consistent influx of fresh data, including from various AI systems. However, the prevailing trend among Big AI companies is to keep their platforms and applications isolated, despite open-source models being available. This situation stifles innovation and creates opportunities for mistakes or malicious uses that could spread rapidly and potentially cause significant damage.

Additionally, the centralized model carries significant risks in terms of protecting user data, privacy, and often financial details. When a large amount of sensitive and crucial business information is managed by a single entity, it becomes a prime target for hackers and a potential source of vulnerabilities. Moreover, this setup grants the provider the power to restrict or deny services at their discretion without any recourse for users.

Democratization through decentralization

As a crypto investor with an interest in AI technology, I’ve come to realize that relying on the cloud model comes with significant risks. The demanding computational needs of AI are pushing the limits of even the largest centralized cloud platforms and the microchip industry that supports them. The chip shortage is so severe now that securing an H-100 server, which is crucial for advanced AI applications, can take as long as 52 weeks.

By implementing decentralization, we can effectively resolve this issue by establishing a distributed network of nodes that tap into vast amounts of underutilized computing power. This approach called Decentralized Physical Infrastructure (DePIN), boasts several advantages: it’s highly scalable with virtually no limits, significantly cheaper than setting up new servers via cloud services (cost savings are approximately 80%), and fosters parallel processing and AI desiloization, enabling applications to learn from one another more seamlessly. Furthermore, decentralized AI, fortified by blockchain technology, introduces groundbreaking methods for compensating creators of large language models (LLMs) via crypto tokens and smart contracts – creating a fair and sustainable model that incentivizes innovation in the rapidly evolving field of artificial intelligence.

The emergence of new economic frameworks, specifically those relying on digital tokens, significantly boosts the demand for robust decentralized systems. By constructing the AI environment around a token economy, developers are motivated to build more secure AI agents. This setup empowers them to distribute these models into crypto wallets, allowing users to maintain full ownership of their data. This arrangement guarantees users that their information remains private and under their control.

The token model in AI projects is significant because market demands and needs will ultimately determine what gets delivered, with costs serving as a reflection of supply-and-demand. However, the current monopolized state lacks incentives for AI to cater to real-life requirements. In contrast, decentralization empowers users to reward developers based on an AI agent’s popularity or societal impact. This stands in stark contrast to the dominant Big Tech oligarchy that will soon be challenged in the realm of AI.

Decentralization offers a solution to the vulnerabilities witnessed on platforms like Hugging Face. The advancement of blockchain technology, specifically zero-knowledge (ZK) proofs, presents us with effective tools to secure and authenticate AI applications. It’s easy to overlook the rapid pace and depth of this technological shift. Traditional cloud providers aren’t stubbornly clinging to outdated models; rather, they are still in the process of understanding how decentralization and ZK can benefit them and their clients.

As a researcher in the field of artificial intelligence, I believe it’s essential to emphasize the educational aspect of demonstrating the privacy and security benefits of decentralized AI architectures. When designed appropriately, these systems offer built-in privacy through encryption of on-chain data while maintaining functionality for interaction and collaboration between various projects, nodes, and parties.

Using AI, centralization fails on all fronts: technologically, philosophically, ethically, and commercially. Moreover, as people become increasingly skeptical (and cautious) towards Big Tech’s dominance – from developers to service providers to everyday users like us – it is clear that a revolution instigated by ourselves is long overdue.

As a researcher, I’d like to emphasize that the perspectives shared in this article are my own and may not align perfectly with those of CoinDesk, Inc. or its associated entities.