AI Abundance & Crypto Scarcity // #79
A Synergy for the Next Era of Tech
The Curious Cat is a journey where I aim to reconnect with my inner-child and explore my curiosity. I pinpoint themes & topics I’ve been interested in, ask myself questions about them and then write about them. I hope you find value within this issue and have a fantastic day doing what you love.
The #1 reason why people fear AI is because it creates abundance.
An abundance of creation, perspectives and possibilities; one that us humans cannot match. We don’t get anywhere close.
Now the beautiful thing about crypto is that it can regulate that abundance; as it is the only technology in human history offering digital scarcity.
“AI generates. Crypto authenticates.
AI is probabilistic. Crypto is deterministic.
AI busts captchas. Crypto rebuilds them.
AI is digital abundance. Crypto is digital scarcity”
Let’s lay some foundations: where are we currently at with AI and Crypto?
AI is an immensely fast growing technology. It is expected to transform every corner of society, with centralised AI systems currently dominating the landscape.
Crypto offers immense value to everyone in the world, but is lacking in use cases & UX.
Crypto could help AI become more decentralised. While AI could help crypto leapfrog over it’s two main hurdles: UX & mainstream adoption. How? Let’s dig.
There are 2 ways in which AI and Crypto fuse together:
(1) Crypto that helps AI
(2) AI that helps crypto
Today we will solely focus on (1); and leave (2) for another day.
To start, let’s look at what crypto is good at:
It’s good at decentralising; by removing middlemen and establishing p2p marketplaces
It has an unparalleled ability to store value and transfer value within digital realms
And last, it’s the best technology in the world for verification
So Crypto has a bunch of powerful features that could bolster AI. But the real question is: why does AI need crypto? What limitations does it currently have?
Decentralisation
AI in 2024 is extremely centralised - crypto decentralises itThe AI ecosystem is currently controlled by a few giants including Nvidia, OpenAI (thus Microsoft) and Google; very much centralised. So we would be fools to not foresee selfishness, biases, restrictive regulation, data privacy concerns, restricted innovation and monopolistic control following suit.
chatGPT has centralised training and centralised deployment.
Llama 3 has centralised training, but decentralised deployment.
Crypto could ‘forge’ decentralised training and decentralised deployment.
The key problem with centralised training is that it’s an opaque process. We don’t know what data was used to train GPT4 or Llama 3, and how that data produces the generated outputs. Not only that, but centralised training means that one has to build an immensely large cluster of GPUs and concentrate them to one venue - reserved only for those with billions of dollars.
Crypto offers a more geographically decentralised alternative for training. Did you know 7/10 of the highest-funded crowdfunding projects ever are crypto related; ie crypto is VERY good at raising money in a distributed manner. Now imagine if you decentralise the funding of AI training models. We would see the birth of a free market, where teams build and raise money for their new model. They battle to theorise the best AIs to get the highest rewards (a positive sum game where incentives for both builders and investors are aligned). Those with conviction can invest in the development of any model; and even potentially contribute to the training of the model themselves ‘Napster style’ where nodes combine forces to form a distributed network of GPUs, needed to train the AI. Instead of solving math problems like crypto miners currently do, you have them build AIs.
This is the vision of companies like IO.net and Bittensor that aim to connect non collocated GPUs, so they can work & train models together. On one side, you have the Machine Learning engineers and companies that are desperate for compute. And on the other side you have the data centers, idle mining rigs, and hobbyists with inactive GPUs. This global supply is massive - but uncoordinated.
Napster, although released +20 years ago, was a much better product than Spotify and Apple Music today. Napster was free unlimited music. Then came regulators, and now we have a product that we have to pay for. A product where songs can be made unavailable at whim, while also having to blindly follow the rules of a company’s business model. The unregulated version was unequivocally better. But it didn’t survive as it wasn’t decentralised enough. Now with crypto, we can decentralise significantly better than in the year 2000.
To close this point, by pushing for decentralisation, we protect the industry from monopolistic control and censorship vulnerabilities, while simultaneously constructing a more cooperative and open AI ecosystem.
Verification
We cannot verify what is AI or human generated - crypto verifies itThe problem we currently face is that AI can fake content (deep fake videos), AI can fake transactions (bots to inorganically inflate activity) and AI can fake human activity (by passing captcha tests). We don’t know if something was done by a conscious human; or an unconscious AI.
Crypto has suffered from mis-direction. A mis-direction where the primary focus of the industry is on money; not the innovative tech. One thing that is severely underappreciated is how nothing in human history has been recorded more accurately than the Bitcoin blockchain; it is the most true thing ever known.
Thanks to the immutable chain of blocks and cryptographic signatures within them, we are absolutely certain that ‘on XYZ date, ABC person transferred DEF units of BTC to GHI person.’
This is also made possible by private keys. In order to cement a transaction in the blockchain forever, you need to stamp your private key. AI can’t falsify that stamp or pretend to be someone, unless their private key (which is encrypted and highly secure) is stolen. Rather than signing BTC transfers, we could have cryptographic signatures assure us that:
“this piece of content was created by Gary Winthorpe - and we verify this by matching it to his public key.”
“this AI model used the following data for training - and we verify this by cryptographically signing the model.”
“the generated content that an AI produces did indeed come from XYZ LLM - and we verify this by levying a cryptographic signature to every output the AI generates.”
“The most important form of scarcity in the AI age are the private keys to control the robots. Those too will be crypto, because web3 backends like Bitcoin and Ethereum have far higher levels of security than any web2 system.”
Zero-Knowledge Proofs, another crypto-born technology that flies under the radar, can offer further privacy, while maintaining undeniable verification.
Zero Knowledge Proofs →ZK Proofs are a way to prove that you know something without revealing the actual information itself. It's like showing you have the correct password, without ever revealing the password.
Cryptographic signatures (like Schnorr) ensure that even as AI makes some resources abundant, selected resources remain secure and valuable, managed through blockchain technology and subsequent verification. Our AI world needs more robust and verifiable ways of proving identity. In a world where AI makes everything fake, crypto can make it real again.
Value Storage and Transfer
Our current money infrastructure doesn't accommodate AI payments - crypto is natively digital and can be transferred across bordersOur journey is leading us to a world filled with billions of IoT and AI devices operating and transacting among themselves. Trying to embed our current financial infrastructure into this world will be like trying to put a square in a round hole…it won’t fit. This is akin to early Tesla models where they attempted to convert combustion-engine Lotus cars into EVs; only to realise that they were much better off inventing their own new chassis, exclusively designed for EVs. Sometimes it’s both easier and more effective to start afresh than to renovate. In this instance, we give-up trying to make 20th money rails fit in the 21st century, and focus our efforts on making 21st century technologies as good as they possibly can be.
The instant settlement, tamper-proof auditable history and natively digital nature of crypto payments make them ideally suited to seamlessly integrate into a digital-first world powered by superintelligent machines. Digital money makes more sense when you have digital minds.
Closing Thoughts
As Crypto begins to support AI, Crypto will transition from an asset class, to a utility.
And with the help of Crypto, AI transitions from a dangerously monopolised industry, to a more democratic and aligned industry.