作者：Will Ogden Moore
Compiled by: TechFlow
Artificial intelligence (AI) is one of the most promising emerging technologies of this century, with the potential to greatly increase human productivity and drive medical breakthroughs. While AI is already important today, its impact is only growing. PwC estimates that AI will become a $15 trillion industry by 2030.
However, this promising technology also faces challenges. As AI technology becomes more powerful, the AI industry has become highly centralized, with power concentrated in the hands of a few companies, which could have adverse effects on society. This has also raised serious concerns about deep fakes, embedded biases, and data privacy risks. Fortunately, the decentralized and transparent nature of cryptocurrency offers a potential solution.
Below, we explore the problems created by centralization and how decentralized AI can help solve them, and discuss the intersection of cryptocurrency and AI, highlighting some crypto applications that are showing early signs of adoption.
The Problem with Centralized AI
Currently, AI development faces several challenges and risks. AI’s network effects and high capital requirements make it difficult for many AI developers outside of large tech companies (such as small companies or academic researchers) to obtain the necessary resources or monetize their work. This limits overall AI competition and innovation.
As a result, influence over this critical technology is concentrated in the hands of a few companies, such as OpenAI and Google, leading to serious questions about AI governance. For example, in February this year, Google's AI image generator Gemini was exposed to racial bias and historical inaccuracies, showing that companies can manipulate their models. In addition, last November, the six-member board of directors decided to fire OpenAI CEO Sam Altman, exposing the control of a few people over the companies that develop these models.
As AI’s influence and importance continue to grow, many worry that a company with decision-making power over an AI model that could have a huge impact on society could operate behind the scenes, put in place safeguards, or manipulate the model for profit — but at the expense of the rest of society.
How decentralized AI can help
Decentralized AI refers to the use of blockchain technology to distribute ownership and governance of AI, thereby increasing transparency and accessibility. Grayscale Research believes that decentralized AI has the potential to bring these important decisions out of walled gardens and into the hands of the public.
Blockchain technology can help increase developer access to AI, lowering the barrier for independent developers to build and monetize their work. We believe this can help improve overall AI innovation and competition and create a counterbalance to models developed by tech giants.
Additionally, decentralized AI can help democratize access to AI investments. Currently, there are few ways to gain access to the financial gains associated with AI development other than through a handful of tech stocks. At the same time, a large amount of private capital has been allocated to AI startups and private companies ($47 billion in 2022 and $42 billion in 2023). As a result, the financial gains of these companies are only available to a handful of venture capitalists and accredited investors. In contrast, decentralized AI crypto assets are open to everyone, allowing everyone to own a part of the AI future.
The current situation at this intersection
At present, the intersection of cryptocurrencies and AI is still in its early stages in terms of maturity, but the market response has been encouraging. AI crypto assets have returned 20% in the year to May 2024 (Grayscale Research defines the AI Universe, with a minimum asset market cap of $500 million and quarterly rebalancing on April 1, 2024. Assets in the Universe include NEAR, FET, RNDR, FIL, TAO, THETA, AKT, AGIX, WLD, AIOZ, TFUEL, GLM, PRIME, OCEAN, ARKM, and LTP), outperforming all crypto sectors except the currency sector (see Figure 1). In addition, according to data provider Kaito, the AI theme currently accounts for the most "narrative mind share" on social platforms, surpassing other themes such as decentralized finance, Layer 2, memes, and real-world assets.
Recently, several prominent figures have embraced this emerging intersection, focusing on addressing the shortcomings of centralized AI. In March, Emad Mostaque, founder of prominent AI company Stability AI, left the company to pursue decentralized AI, citing the fact that “now is the time to ensure AI remains open and decentralized.” Additionally, crypto entrepreneur Erik Vorhees recently launched Venice.ai, a privacy-focused AI service with end-to-end encryption.
Figure 1: AI crypto assets outperform nearly all crypto sectors YTD
Today, we can categorize the intersection of cryptocurrencies and AI into three main subcategories (assets are illustrative examples, listed from largest to smallest by market cap):
Infrastructure layer: Networks that provide AI development platforms (such as NEAR, TAO, FET)
AI resources: assets that provide key resources (computing, storage, data) required for AI development (such as RNDR, AKT, LPT, FIL, AR, MASA)
Solving AI problems: Assets (e.g. WLD, TRAC, NUM) that attempt to solve AI-related problems such as bots, deepfakes, and model validation
Figure 2: AI and cryptocurrency market map
Source: Grayscale Investments. Protocols listed are illustrative examples.
A network that provides infrastructure for AI development
This category includes networks that provide permissionless, open architectures, purpose-built for the general development of AI services. These assets are not focused on a single AI product or service, but rather are dedicated to creating the underlying infrastructure and incentives for a variety of AI applications.
Near stands out in this category, founded by the co-creator of the "Transformer" architecture that powers AI systems like ChatGPT. It recently leveraged its AI expertise to launch an initiative to develop "user-owned AI" through its R&D division, which is led by a former OpenAI research engineer advisor. At the end of June 2024, Near launched its AI Incubator program to develop Near's native base models, a data platform for AI applications, an AI agent framework, and a computing marketplace.
Bittensor is another notable example. Bittensor is a platform that uses TAO Token to incentivize AI development. Bittensor serves as the underlying platform for 38 subnetworks (a subnetwork is a smaller segmented part of a larger network, designed to improve efficiency and security by isolating parts of the network for specific purposes or user groups. As of June 23, 2024), each subnetwork has different use cases such as chatbots, image generation, financial forecasting, language translation, model training, storage, and computing. The Bittensor network rewards the best performing miners and validators in each subnetwork with TAO Tokens, and provides developers with a permissionless API to build specific AI applications by querying miners in the Bittensor subnetwork.
This category also includes other protocols such as Fetch.ai and Allora Network. Fetch.ai is a platform for developers to create complex AI assistants (i.e. "AI agents"), which recently merged with AGIX and OCEAN, with a combined value of approximately $7.5 billion. Another is the Allora Network, a platform focused on applying AI to financial applications, including automated trading strategies and prediction markets for decentralized exchanges. Allora has not yet launched a token and conducted a strategic financing in June, raising a total of $35 million in private capital.
Resources needed for AI development
This category includes assets that provide the resources (compute, storage, or data) needed for AI development.
The rise of AI has created an unprecedented demand for computing resources such as GPUs. Decentralized GPU marketplaces such as Render (RNDR), Akash (AKT), and Livepeer (LPT) provide idle GPUs to developers who need computing resources for model training, model inference, or 3D generative AI. Today, Render offers about 10,000 GPUs, mainly for artists and generative AI, while Akash offers 400 GPUs, mainly for AI developers and researchers. Meanwhile, Livepeer recently announced plans to launch a new AI subnetwork in August 2024 for tasks such as text to image, text to video, and image to video.
In addition to requiring a large amount of computing resources, AI models also require massive amounts of data. As a result, the demand for data storage has increased significantly. Data storage solutions like Filecoin (FIL) and Arweave (AR) can serve as decentralized and secure network alternatives for storing AI data instead of centralized AWS servers. These solutions not only provide cost-effective and scalable storage, but also enhance data security and integrity by eliminating single points of failure and reducing the risk of data breaches.
Additionally, existing AI services like OpenAI and Gemini continuously fetch real-time data through Bing and Google searches. This puts other AI model developers at a disadvantage. However, data scraping services like Grass and Masa (MASA) can help balance the playing field because they allow individuals to monetize their data by providing application data for AI model training while maintaining control and privacy over their personal data.
Assets that solve AI-related problems
The third category includes assets that attempt to address AI-related issues such as bots, deepfakes, and content provenance.
AI is fueling the proliferation of bots and disinformation. AI-generated deepfakes have already influenced presidential elections in India and Europe, and experts fear that the upcoming presidential race will involve a flood of deepfake-driven disinformation. Assets that attempt to address the deepfake problem by establishing verifiable content provenance include Origin Trail (TRAC), Numbers Protocol (NUM), and Story Protocol. Additionally, Worldcoin (WLD) addresses the bot problem by proving a person’s identity through unique biometric identification.
Another risk with AI is ensuring trust in the model itself. How can we trust that the AI results we receive have not been tampered with or manipulated? Currently, several protocols are working to help solve this problem through cryptography, zero-knowledge proofs, and fully homomorphic encryption (FHE), including Modulus Labs and Zama.
in conclusion
While these decentralized AI assets have made initial progress, we are still in the early stages of this convergence. Earlier this year, well-known venture capitalist Fred Wilson said that AI and cryptocurrency are "two sides of the same coin" and that "web3 will help us trust AI." As the AI industry continues to mature, Grayscale Research believes that these AI-related crypto use cases will become increasingly important, and these two rapidly developing technologies have the potential to support each other's growth.
By many indications, AI is coming, and is poised to have profound impacts, both positive and negative. By leveraging the properties of blockchain technology, we believe that cryptocurrencies can ultimately help mitigate some of the dangers posed by AI.
