As Congress continues to work through stablecoin and market structure legislation, this week, we turn to another issue at the forefront of tech policy: artificial intelligence.

In the interview below, Julie Stitzel—Senior Vice President of Policy for DCG—unpacks decentralized AI, how crypto ties in, and related policy considerations.

Q&A

Background/Decentralized AI 101

Q1: For readers who may be unfamiliar, could you provide some background on DCG?

A1: DCG was founded in 2015 by Barry Silbert. Barry played a key role in the development of the Bitcoin ecosystem, which has evolved into the nearly $3 trillion digital assets industry today. Over the years, DCG has built a reputation for uncovering early-stage ideas with massive potential, which has positioned us to become the most active investor in the digital assets space and expand across all blockchain technologies, including decentralized AI (“deAI”). At its core, DCG is a global investor, builder, and incubator advancing decentralized technologies, with a portfolio of 200+ crypto and blockchain startups, in addition to its five subsidiaries: Grayscale, Yuma, Luno, Foundry, and Fortitude Mining.

Q2: How is DCG involved in decentralized AI, and what are one or two projects you are particularly excited about?

A2: Much like our early engagement in the digital assets ecosystem, DCG has high conviction in decentralized AI, specifically the Bittensor network. Last November, we launched our new deAI-focused subsidiary, Yuma, which is also led by Barry Silbert as CEO, and focuses on accelerating Bittensor’s growth and development. Much like the internet opened access to knowledge and removed barriers to its development, Bittensor and deAI aim to democratize access to intelligence, unlocking the computing power of AI by making it community-owned and driven.

We see Bittensor as the “World Wide Web” of information: a platform for innovators to freely contribute and connect. Individuals can build projects on the network known as subnets and be rewarded for the quality of their contributions. As a subnet accelerator, Yuma provides access to necessary infrastructure and support for entrepreneurs, helping to bring world-changing ideas to life.

In addition to our work on Yuma and DCG’s direct investment in $TAO, Bittensor’s native token, DCG’s venture team also actively seeks other deAI opportunities and made 33% of their Q1 investments into deAI projects such as Masa. Masa facilitates AI development by transforming fragmented data into a unified access layer, providing permissionless access to quality AI training data while ensuring fair data attribution and compensation. Additionally, Masa has launched two Yuma-incubated subnets on Bittensor.

Q3: What are the fundamental differences between decentralized AI models, like the Bittensor network, and centralized models, like OpenAI?

A3: Decentralized AI models, like Bittensor, distribute model training, inference, and governance across a peer-to-peer network, whereas centralized models like OpenAI rely on centralized infrastructure and decision-making to control their models. The key differences lie in the accessibility, incentivization, and architecture of the models. Centralized AI models are typically constructed by a single organization that oversees all steps of the process, from training to deployment, and encompasses all the various inputs and outputs. These organizations determine who can access the models, how they are utilized, and what data they are trained on, thereby creating a closed ecosystem with limited transparency and participation.
​
Decentralized AI models enable open and transparent contributions, where anyone can contribute compute resources or machine learning models and receive rewards for their contributions. This creates a permissionless system in which value is distributed based on performance and utility, where users directly benefit from their data and contributions, rather than being monopolized.

Q4: How does crypto tie in? In other words, how does crypto enable or strengthen decentralized AI models?

A4: The incorporation of crypto is what differentiates Bittensor from some of the other decentralized protocols. Crypto is the incentivization mechanism that powers decentralized AI models to encourage participation; in Bittensor’s case, the platform incentivizes high-quality contributions and efficient applications by rewarding them with its native token, $TAO. As opposed to traditional centralized AI that is controlled by a single overarching company, Bittensor eradicates the need for a centralized power, using TAO to foster a competitive environment where better models receive better rewards. Anyone can join the network, train models, and earn TAO. In this case, crypto is what aligns incentives, opens access, and drives collaboration at scale.

Q5: What do you see as the main benefits of decentralizing AI development and infrastructure?

A5: Decentralizing this technology challenges the monopolies we see in AI’s current status quo and opens the door for researchers, academics, developers, and entrepreneurs to participate with lower barriers to entry. Currently, a select few dominate the space due to their access to data, compute, and proprietary models. Democratizing and decentralizing AI make it more resilient, in addition to distributing the benefits of AI more widely than centralized models.

Additionally, decentralization is a solution to consumers’ concerns about data privacy and security in AI. DCG’s recent survey, conducted in collaboration with The Harris Poll, found that 74% of consumers agree they’d be more comfortable using AI if they knew they could benefit from the use of their personal data. deAI addresses this concern by allowing users to retain control over and benefit from their own data and contributions.

Policy

Q6: What are the most pressing policy issues for AI today? Are there any specific bills or regulations you would like to see enacted (or not enacted)?

A6: One of the most critical issues for AI today is ensuring that it serves the public and not just the powerful, and this is where the regulatory framework plays a role. Like most things tech-related, it is essential to strike the correct balance between fostering innovation and protecting users. The current Administration has taken a refreshingly pragmatic approach to AI by prioritizing innovation and American competitiveness over heavy-handed regulation, and we urge Congress to do the same.

A federal framework that supports private sector leadership and decentralized development is crucial for building an open and resilient future for AI. This federal framework should incentivize innovation in open, decentralized ecosystems, empower individuals to benefit from their data, avoid codifying Big Tech’s dominance, and mitigate a fragmented, state-by-state patchwork of rules. Clear guidance will remain at the core of supporting a competitive and open AI ecosystem in the U.S.

Q7: In crypto policy, we’ve seen proposals in the past that risked undermining decentralization by attempting to apply rules designed for centralized intermediaries to decentralized networks.

Are there similar concerns with AI policy? In other words, what are some key considerations policymakers should keep in mind when crafting rules for centralized AI to avoid stifling decentralized innovation?

A7: There is definitely overlap between these concerns in terms of crypto and AI policy. Similarly to crypto, decentralized AI systems rely on incentives, open participation, and community governance. If policymakers apply uniform regulations to fit centralized players, these decentralized systems could be pushed out of compliance or innovation could be severely dampened. Key considerations should include distinguishing between centralized and decentralized architectures, preserving space for open-source development, and ensuring that compliance pathways exist.

Q8: Right now, many in the crypto policy community are primarily focused on market structure and stablecoin legislation. How, if it all, do you see these bills directly or indirectly impacting the future of decentralized AI?

A8: Although market structure and stablecoin legislation do not directly target decentralized AI, there are implications for systems such as Bittensor that utilize cryptocurrencies as incentives and are built on the same principles of distribution and decentralization that crypto is. It’s imperative for the future of both decentralized AI development and the digital asset space that clear regulatory guidelines are in place for protection without stifling innovation.

Q9: What resources do you recommend for readers who want to better understand decentralized AI?

A9: DCG and its subsidiaries have put together a variety of resources on deAI and Bittensor including DCG’s educational one-pager on deAI, information about Bittensor on Yuma’s website, and Grayscale’s research on Bittensor and the $TAO token. Additionally, Bittensor’s website outlines the project’s initial foundation and its implications for the world of AI.

***

Thank you for reading, and see you next Friday.
​
-GSL