Recently, Bitcoin has taken center stage with the approval of spot Bitcoin ETFs in the United States, but the outperformance of AI-related crypto assets serves as a reminder of the broadening applicability and relevance of public blockchain use cases beyond a form of payment. Grayscale Research believes that developments in the intersection of AI and crypto could offer the potential to mitigate future AI-related societal issues, such as the rise of deepfakes, concerns around data privacy, and concentration of power. While many tokens may be simply riding the “AI hype wave,” crypto protocols associated with the development of AI projects have gained early signs of adoption. Specifically, the four largest AI-adjacent crypto tokens[1] This past November, OpenAI’s six-person Board announced company CEO Sam Altman was being replaced, which sent shock waves through the tech and business communities. Though this decision was later reversed and Altman was reinstalled as CEO, the conversation about Artificial Intelligence (AI) governance persists – and is even a large topic of conversation at this year’s World Economic Forum Annual Meeting in Davos, Switzerland.
The OpenAI incident underscores the potential dangers of centralized control over pivotal technologies. For Grayscale Research, this begs a critical question: how do we ensure that AI development is accessible, competitive, and transparent? Aren’t these concerns the core tenants of blockchain technology? Grayscale Research believes so, and others in the industry are starting to discuss similar themes. According to Sheila Warren at CCI[2] [3]
Despite the fact that many of these use cases are in their infancy, the market seems optimistic about the significance of this intersection. According to Coingecko web traffic, Artificial Intelligence was the most popular “crypto narrative[4] Utilities and Services Sector and the crypto ecosystem at large (Figure 1).
In this report we attempt to explain where progress has been made across the following areas in the overlap of AI and crypto: Verifying content authenticity, reducing model bias, and improving access and competition within AI development.
Figure 1: Large AI tokens outperformed each Crypto Sector in 2023
Verifying Content Authenticity
One major societal issue exacerbated by AI is the proliferation of bots and misinformation. This is particularly relevant in the coming months as top AI experts fear a flood of deepfakes attempting to influence the 2024 US presidential election.[5]
One major initiative working on this problem is a crypto protocol called Worldcoin. Co-founded by Sam Altman, Worldcoin aims to register every human on the planet via a biometric scan in order to verifiably distinguish humans from bots, all incentivised by a dedicated blockchain token. The Worldcoin team has been aggressive in tackling its ambitious pursuit. Since launching around six months ago, Worldcoin has signed up 2.9mm people around the world.[6] [7]
Another initiative tackling this problem is the Digital Content Provenance Record (DCPR) standard, pioneered by the teams at Arweave and Bundlr. The DCPR standard uses the Arweave blockchain to timestamp and verify digital content, providing dependable metadata to help users assess the trustworthiness of digital information.[8]
Reducing Bias in AI Models
As AI models become increasingly integrated into our daily lives, there’s a growing concern about overdependence on these systems and the inherent biases they may exhibit. Consider the scenario where an AI-powered chatbot could sway consumers’ choices by pushing them towards particular products or favoring specific political beliefs. Similarly, this technology could display bias in employment screenings by making decisions influenced by candidates' demographic features. The resulting breakdown in trust carries knock-on effects. According to one study, “AI detectors” may themselves be biased against natural writing by non-native English speakers.
Bittensor, a novel decentralized network, attempts to address AI bias by incentivizing diverse pre-trained models to vie for the best responses, as validators reward top performers and eliminate underperforming and biased ones. By fostering an open and collaborative environment for AI innovation across a diverse range of models and datasets, Bittensor could have the potential to propel AI development while attempting to mitigate the negative impact of bias.[9]
While development on Bittensor is still nascent, the decentralized network has shown initial progress with 32 subnetworks specialized for specific use cases including chat-bots, image-generation, price prediction, and language translation.[10]
Figure 2: AI-related crypto assets have performed well since major developments at OpenAI
Improving Access to AI Development, Increasing Competition
In addition to risks of model bias, another concern around AI is that development is particularly concentrated. As AI model sizes grow, associated high capital costs for compute and storage threaten to price out competition, leaving AI development largely in the hands of the few tech giants that can afford it.[11] [12] [13]
Decentralized compute marketplaces like Akash and Render are designed to address the inefficiency of underused GPU resources by connecting GPU owners with AI developers seeking compute power. This system allows individuals and organizations around the world to monetize their otherwise idle compute. At the same time, it provides AI developers with flexible access to computing resources. Because the blockchain cuts out intermediaries with a profit motive and overhead costs, these networks can offer services at sometimes a fraction of the rate ( around1/5 of the cost via Akash[14]
For instance, this past fall a Columbia student trying to work on AI development had difficulty gaining access to compute through Amazon Web Services; instead he rented GPUs through Akash for as little as $1.10 per hour.[15]
Recently, several of these decentralized marketplaces have gained preliminary traction. For example, since the launch of GPU deployments in September, Akash has grown to over 70 active GPU leases[16] [17] [18]
Figure 3: Increasing utilization of GPUs for decentralized marketplace Akash [19]
Conclusion
Today, the majority of the progress in this intersection has occurred within the context of crypto protocols helping democratize and accelerate AI development via decentralized GPU marketplaces. Other opportunities could lie in areas such as:
This synergy is still in its nascent stages, yet it shows promise of gathering momentum through 2024 and beyond, particularly if market participants continue to view these assets as a counterbalance against the future entrenchment of large centralized players like OpenAI. Whether or not AI and crypto are intrinsically linked, these two rapidly evolving technologies have the potential to mutually support each other’s growth, both in scope of use cases and relevance to the broader public.
[1]
[2] LinkedIn
[3] AVC.com
[4] Coingecko
[5] Fortune
[6] Worldcoin
[7] Reuters
[8] Github
[9] Bittensor and Plaintextcapital and Blockgeeks
[10] Messari
[11] CTECH
[12] Messari and The Information
[13] Tech HQ
[14]
[15] Semafor
[16] Akash.network
[17] Foundry
[18] Dune Analytics
[19]
[20] Worldcoin
[21] Substack
