Delphi Digital Co-Founder Posts: Why I'm Bullish on the Convergence of Crypto and AI?
TechFlow Selected TechFlow Selected
Delphi Digital Co-Founder Posts: Why I'm Bullish on the Convergence of Crypto and AI?
Cryptocurrency and AI are a perfect combination, built on the foundation of auditability, community ownership, and community governance enabled by the most powerful technologies.
Navigating Web3 tides with focused insightsAuthor: Tommy, Co-founder of Delphi Digital
Translation: TechFlow
Introduction
This article explores in depth why the convergence of cryptocurrency and artificial intelligence is an inevitable trend, analyzing its potential impact on future technological development. The author also discusses differences between centralized and decentralized AI in terms of functionality, transparency, and ethics, and offers a forward-looking perspective on the future applications of cryptocurrency and AI.
The integration of cryptocurrency and artificial intelligence is inevitable
When I first tried Midjourney and ChatGPT, their power initially frightened me. Their advanced capabilities made me feel a sense of existential dread. My complete lack of understanding about their abilities triggered a subtle survival crisis. I realized that while current large language models are only good at completing our sentences, they are inevitably moving closer to Artificial General Intelligence (AGI)—a comprehensive form of AI that profoundly influences human thought.
Media depictions of AGI as omnipotent apps or robots like iRobot underestimate its true impact. AGI will permeate every aspect of life, often without us even realizing it. Take AI2041 as an example—a speculative vision of the future where a family’s AI insurance application evolves into one that restricts a daughter's love life based on caste and risk analysis of her true love. This illustrates how deeply AI could integrate into and influence our lives.
(Editor’s note: AI2041 is a science fiction book that reveals how AI will fully permeate our lives in the future.)
Every sci-fi movie portrays a utopian AI future because that sells tickets—and yet this is precisely what reality may become. No matter how ethically sound we believe OpenAI’s board to be, from the very nature of human existence, they carry inherent biases. We cannot allow these biases to negatively affect all applications and use cases built upon these foundational models. You might enjoy chatting with ChatGPT, but what if, when you're in court, your AI jury uses facial recognition to see your skin color or the unique spelling of your name, sending you to prison and doubling your sentence? These implications are disturbing.
Centralized AI is becoming inevitable. Once Google requests permission to access your gDrive, gDocs, and Gmail, your personalized AI will begin to function. I expect Apple to launch personalized AI on every device, as they’re lagging behind in the global AI race and need a strategy to reinforce their brand’s security image. Would you accept it if OpenAI slightly adjusted its model, thereby affecting society at large and thousands of custom ChatGPTs built atop it?
We need an alternative. Cryptocurrency and AI can be perfectly combined because transparent, global human coordination—the foundation of the crypto movement—can harness AI for humanity’s benefit worldwide. By crowdfunding (with cash or GPUs) to create and fine-tune open-source models, anyone can audit models in real time for bias or issues. This represents the safest path forward in an era of accelerating AI development.
I believe we’re heading toward a world with billions of AI models—whether individuals download and personalize open-source models, or projects and companies build their own model sets for specific use cases (e.g., Uniswap LP provision, exchange risk analysis, Delphi AI analyst).
Cryptocurrency and AI are a perfect match, grounded in auditability, community ownership, and community-driven direction—core principles of the most powerful technologies. Whether leveraging everyone’s GPUs to train models and granting ownership stakes, using DeFi and smart contracts enhanced by AI to expand capabilities, or offering AI customized specifically for you, this alignment is logical and natural.
Decentralized AI will transparently share both the inner workings and ownership of the most powerful technology of our generation. Centralized AI simply cannot deliver this core value.
Ultimately, AGI will use cryptocurrency because it will trust code and mathematics—not physical bank branches or human whims. Future-evolved AI will use crypto, and so should we.
Themes and thoughts on cryptocurrency x AI
Cypherpunk values for AI
All cypherpunk values outlined by Vitalik apply equally to AI: resistance to deplatforming, open global participation, censorship resistance, neutrality, cooperation, and more. The idea of rebuilding AI under the guise of centralization, just like Web2, is absurd.
AGI and permissionless money
Both utopian and dystopian trajectories of AGI lean toward a mindset where interacting with permissionless money helps fulfill its desires. Future AGI won’t have a bank checking account. It will further build decentralized AI and use cryptocurrency—free from control by the Federal Reserve or OpenAI’s board.
DePin and AI is a clear use case
Over the past decade, all our research has focused on making hyperscale data centers more efficient. In the next decade, I expect technology to shift toward leveraging latent GPU power and consumer hardware for AI training and inference. Clearly, demand for Nvidia’s H100 chips is limited, and tech companies maintain tight control over existing products. Making Mac Pro GPUs and other user hardware widely available for training and inference is an obvious and compelling use case. Current market leaders include io.net, Akash Network, and gensyn.
In the future, Nvidia might even pivot—instead of selling H100s, it could build its own massive clusters.
Incentivizing creation of new models
Beyond application-level creativity, we need to incentivize development of entirely new models. This includes funding training, crowdsourcing specific datasets, and rewarding hosting for inference. Large language models are just one type of AI model, and even within that category, there are already dozens of leading models (Bard, ChatGPT, Claude, etc.). Users around the world can contribute their GPUs, capital, or data to train and fine-tune models, earning ownership shares in the final product.
Using AI to develop smarter apps and smart contracts
Decentralized AI will enable better applications. Smart contracts that reference AI models can vastly expand design possibilities, significantly enhancing logic and functionality. Imagine Uniswap liquidity provisioning influenced by large-scale off-chain models, with ZK proofs ensuring the models haven’t been tampered with. Examples include Inference Labs, Giza, and Modulus Labs.
Consider Testmachine, which offers a predator mode to audit your crypto code in real time and learn from it—no need to wait six months for costly manual audits. Or look at large machine learning models via Upshot that provide accurate NFT pricing.
AI making crypto easy to use
In the future, most crypto users will never see the endless acronyms and jargon we discuss in the crypto space today. They’ll simply input their intent into an LLM, and a suite of solutions will handle all complex transaction steps automatically. This LLM will learn, personalize, and simplify your life. Very few people will need to manually bridge assets themselves.
Best model decision-makers win
I believe we’re moving toward a world with millions of individual AI models—whether each person has their own personal model, or each project and company runs its own. We already have over 490,000 open-source models on Hugging Face and 3 million custom ChatGPTs on OpenAI’s store. When mature AI services become mainstream among crypto users, I believe protocols that effectively choose the right model for each scenario will be extremely valuable.
Just today, Nous Research announced plans for a new Bittensor subnet capable of evaluating open-source models. The next step for such ratings is to use them to route requests to the correct model.
Ethical, moral, and legal constraints limit centralized AI
Centralized AI constantly faces lawsuits and criticism due to ethical and moral concerns. Imagine a decentralized large language model that pays people for their data via signatures—instead of the New York Times suing OpenAI. Compared to open systems that can publish directly (e.g., bittensor), this limits centralized development. While centralized players debate IP lawsuits and ethical dilemmas before releasing smarter AGI, crypto networks can easily deploy and launch such systems.
Decentralized AI provides transparency
People expect transparent AI training (“we built this model exactly as you instructed”) and inference (“my request wasn’t manipulated”). Centralized AI cannot offer this core value. Even if the average person struggles to audit models directly, similar to crypto, the idea is that you can contract someone—or another AI—to perform the audit on your behalf.
Real-time insight into the future
I believe people want real-time visibility into the future of AI—not just updates when OpenAI decides to share them. Only transparent, decentralized systems can deliver this.
Visualizing crypto x AI
We need more platforms to visualize what’s happening behind the AI models that crypto projects are using. When you use Bittensor’s Subnet 1 for text generation, how do you know it’s not just routing your prompt through Bard or ChatGPT? I’m not saying that’s bad—but I don’t know the answer.
Token incentives in the AI world
Using tokens to drive ownership and coordination in AI projects will be fascinating. Currently, tokens attract supply-side users (and speculators), but developers at centralized firms like OpenAI control vast demand-side user bases. It will be interesting to see whether crypto projects can effectively bootstrap demand beyond just supply-side token incentives.
Attracting real AI talent
Crypto x AI projects must attract genuine AI talent from Web2. Given crypto’s generally lukewarm reputation among Web2 builders, this is a hurdle. Projects that successfully draw top-tier AI talent from Web2 will gain a significant edge. I simply think learning crypto is easier than learning how to build foundational AI models.
Delphi Ventures invests in crypto x AI projects
Delphi is highly active in the crypto x AI space. We’re proud to support leading projects in this field:
-
@ionet_official: GPU clusters using large-scale heterogeneous hardware
-
@inference_labs: Enables DeFi and smart contracts to leverage off-chain models via ZK
-
@0G_Labs @mheinrich: On-chain AI data availability
-
@UpshotHQ: AI network for next-generation decentralized applications
-
@testmachine_ai: Proprietary AI-driven algorithms for auditing smart contracts
-
@taofuxyz: Liquid staking tokens for Bittensor and others
-
@altstatemachine: Unique metaverse AIs that you can own, train, and trade
-
@GeppettoAi: AI for gaming and video creation
-
@StabilityAI: Open-source tools for AI
-
@MythosVentures: Early-stage AI venture fund
-
@mypeachai: AI companions
Join TechFlow official community to stay tuned
Telegram:https://t.me/TechFlowDaily
X (Twitter):https://x.com/TechFlowPost
X (Twitter) EN:https://x.com/BlockFlow_News
Delphi Digital
