Web3 x AI
Artificial Intelligence + Blockchain technology represent the most powerful technological symbiosis in our species’ history. Let's not screw it up.
Homo Extinctus
Humans are toolmakers. Technologists by nature. Crafting spears and language, antiseptics and jet engines, particle accelerators and smartphones. But the extent to which we fear the tools and tech we create depends upon the degree to which they evolve along the wrong substrate.
Money, for example, was a useful tool we created to help with accounting, but evolved along the wrong direction when it became untethered to gold. No longer constrained by a universal physical reality, it very quickly fell into the hands of a small few, ballooned out of control, and now, in a manner of speaking, enslaves us.
A.I. now threatens to do the same.
It’s an extraordinarily powerful tool at risk of falling into the hands of a small few (consider the recent OpenAI drama…), ballooning out of control, and the greatest AI fear of all: enslaving us.
But it’s up to us whether this tool evolves along that dystopian path. It’s our choice whether to build AI on a centralized, closed-source, “hegemonic-power-for-me-enslavement-for-thee” tech foundation, or instead on a decentralized, open-source, user-owned and human-first tech foundation.
Blockchain technology, not Amazon Web Services, is the substrate on which to build AI if we care about the future. Decentralized AI will resist falling inevitably into the hands of the few. Blockchain-based AI will have a computational backbone via mining that keeps it tethered to a universal physical reality: energy. And tokenized AI will more efficiently coordinate resources and more fairly distribute value to users.
Humanity’s newest tools, Blockchain & AI, likely represent the most powerful technological symbiosis in our species’ history. The tools to become Kardashev Type III if we get it right. Let’s not screw it up.
Move Fast & Please Don’t Break Anything
“AI” & “Blockchain” are more than just overhyped VC buzzwords, I believe they genuinely represent the two software technologies with the greatest potential to catapult our species into a prosperous future.
Their obvious similarities include immense computational requirements, the use of tokens to incentivize resource allocation, and the shared concepts of machines replacing humans.
Computational Intensity (Web3 ASICs & AI CPUs)
Token Incentives (Web3 user ownership & AI RLHF)
Machine Work (Web3 smart contracts & AI Agents/GPTs)
But their differences are just as glaring.
Anyone who has used ChatGPT knows it’s impossibly fast. GPT-4 can output a 30,000 line Gilgameshian epic in less time than it takes me to output a 3 line email. And those who understand Bitcoin know it is incomprehensibly secure. There are more Bitcoin key pairs than atoms in the known universe, and brute forcing just one - using the most powerful computer on Earth - would take somewhere on the order of a billion billion (not a typo) years.
AI is remarkable computational speed, Blockchain tech is remarkable computational security. And this is where their differences become opportunities.
AI is an unstoppable force.
Blockchain is an immovable object.
Work Smarter, Not Harder
Blockchain ledgers are the richest and fastest growing historical records we’ve ever created. The Bitcoin blockchain alone contains an impossible quantity of information for a human to parse through, but the perfect type of problem for AI.
Similarly, Blockchain tech offers a tamper-proof and transparent ledger able to solve AI’s growing problems like deepfakes, mis/disinformation, & IP infringements.
AI can help address crypto’s challenges…
Analytics: AI can parse through the vast amount of blockchain data and find patterns and signals invisible to us mortals
Volatility: AI agents can learn & trade entirely on their own, so can be put to work stabilizing the market, abitraging, rebalancing, and minimizing volatility
Security: AI enhanced smart contracts can be used to identify vulnerabilities, code bugs, and prevent potential exploit vectors
Blockchain tech can help address AI’s challenges…
Authenticity: Blockchain-based content verification can prove whether a video is an AI deepfake or if a news article was written by an actual human or not
Data Privacy: Blockchain concepts like zero-knowledge proofs (zkPs) & zero-knowledge machine learning (zkML) will be used by companies to cryptographically prove an AI model’s output was not tampered with, without exposing its confidential underlying data
Bias: Blockchain transparency can counteract centralized companies pre-training language models with bias (malicious or accidental), like below:
To De- or not to De-?
AI needs blockchain tech to solve inherent hardware limitations in Web2, to safeguard against monopolistic control of centralized AI, and to unlock the full potential of AI robots by letting them use robot money: BTC.
In other words: AI needs to decentralize.
Technological Reason: Hardware Limitations
AI hardware cannot scale to meet its growing demand because centralized compute can’t scale. There are only a finite number of servers in a finite number of data centers around the world trying to process a near-infinite demand for AI. The *only* way to meet AI’s exponential demand is with blockchain tech.
Right now, on the other side of the world, there is an ocean of unused CPUs & GPUs sitting idle inside billions and billions of iPhones, laptops, tablets, TVs, and other devices, doing absolutely nothing while that half of humanity sleeps.
Blockchain networks like Gensyn allow people around the world to connect their devices together to form a globally decentralized supercomputer that can enable AI to scale several orders of magnitude beyond what centralized data centers can achieve.
Ideological Reason: Anti-Authoritarianism
Largely because the required compute is centralized, the cost to train a competitive new AI model is over $100M/month. No individual AI developer on Earth can afford this, so they have to rely on OpenAI’s model. I would hope by now it’s obvious why we can’t have the world rely entirely on (1) AI model built by (1) company controlled by (1) guy... but if it’s not, take another look at the Joe Biden vs. Donald Trump image above and ask yourself how that bias got encoded into ChatGPT.
Decentralizing the compute required to train new models brings the price down to the point where independent devs can build more, new, and different AI models that compete with OpenAI’s hegemonic LLM. Decentralized AI prevents monopolistic control over one of the most powerful technologies ever built.
Practical Reason: Better Functionality
Decentralized AI will outcompete centralized AI for several reasons. Open models allow for permissionless contribution from millions of users around the world vs. just the handful working at a centralized closed-source Web2 company. This leads to exponential advances in the tech.
“For AI systems to provide the highest level of utility and functionality, they will have to securely interoperate across all user applications. This is only possible with decentralized architecture. In comparison, to build on top of OpenAI, you must seek permission from management, negotiate a business contract, and develop/deploy a specific API to connect two separate systems. This is a much slower and exclusive process that limits innovation & development.”
Token-based incentive mechanisms attract far more people to supply compute, provide data, and train models which ultimately lowers the cost & barrier to entry for AI devs, and means a more competitive landscape of higher-quality AI tools for users.
Lastly, it is obvious that AI Agents will transact using digital money, not dirty pieces of little green paper...
As AI tools increasingly interact with each other and take on tasks requiring them to transact, they will need a payment system that’s digital, 24/7, and completely deterministic. Bitcoin is the only currency on Earth that meets these requirements. Machines will use the only money made for machines.
The Landscape
Now some alpha… below are some of the top decentralized AI projects in the market. (M31 Capital is an early investor in Render, Gensyn, Bittensor, and Akash).
Conclusion
AI & Blockchain are the convergence of two deeply complimentary tools that together form a technological Yin & Yang. We are entering a new paradigm where machines can think, earn, spend, and build. They will use BTC as their native currency, marrying AI brains with Blockchain brawn, to the benefit of both.
It is exciting, but be careful. While Web3 is architected to be highly resistant to the allure of closed-source centralized monopoly, the Web2 AI community is not as guarded against these risks. But I promise you this: AI’s FTX moment is inevitable if it remains centralized, and the fallout will be worse. Let’s apply the wisdom of Satoshi to make sure AI remains open, pro-human, and as far away from the Effective Altruists as possible.
I am optimistic that we can. Decentralized models are going to unlock extraordinary technological potential for humanity, and the hard limits enforced by blockchain tech can prevent AI from spiraling out of control. Earth may get its Dyson Sphere yet.
What a time to be alive.
What We’re Reading
The Convergence of Crypto and AI (Multicoin)
AI & Crypto (Web3 with a16z)
Where Crypto & AI Meet (Distributed Compute Summit)
The Roaring 20s Are Back on Track (Noah Smith)
My Techno-Optimism (Vitalik Buterin)
It’s Time to Build for Good (Isaac Wilks)
Massa (Arthur Hayes)
Artificial Intelligence Can't Scale Without Artificial Brains (David Attermann)
About M31 Capital
M31 Capital is a global investment firm dedicated to cryptoassets and blockchain technologies that support individual sovereignty.
Website: https://www.m31.capital/
Twitter: https://twitter.com/M31Capital