In the search for new alpha information, we inevitably encounter a lot of junk information. When a project can quickly raise five or six figures with just a semi-clear brief and some decent branding, speculators will seize on every new narrative. And as traditional financial sectors have jumped on the AI bandwagon, the "crypto AI" narrative has exacerbated this problem.
There are two main problems with these projects:
1. Most crypto projects don’t need AI.
2. Most AI projects do not require cryptocurrency.
Not every decentralized exchange (DEX) needs an AI assistant built in, nor does every chatbot need an accompanying token to facilitate its adoption curve. This hard-wired marriage of AI and crypto almost broke me when I first dug into this narrative.
Real challenges and potential opportunities
The bad news is that if we continue on our current path and further centralize this technology, it will only end in failure, and the plethora of fake “AI x Crypto” projects will hinder our ability to turn things around. However, the good news is that there is light at the end of the tunnel. Sometimes AI can really benefit from cryptoeconomics. Likewise, there are real problems that AI can solve in some cryptocurrency use cases.
A high-level view of the AI stack
Here’s a look at the different verticals within the “Crypto AI” ecosystem. This is a very simplified view, but hopefully helps lay the groundwork.
At a high level, here’s how it all works together:
We will explore each of these areas, paying particular attention to how different cryptoeconomic designs can actually improve standard workflows.
Encryption gives open source a fighting chance
The debate over "closed source" vs. "open source" development approaches can be traced back to the Windows-Linux debate and Eric Raymond's famous "Cathedral and Bazaar" theory.
Although Linux is widely used among enthusiasts, about 90% of users choose Windows. The reason behind this is mainly incentives.
Open source development has many benefits, and it allows the largest number of people to participate in the development process and contribute to it. However, in this headless structure, there is no unified command. Open source projects are likely to evolve into a "chimera", splitting in different directions at every intersection of design philosophy.
The best way to align incentives
Build a system that rewards actions that further our goals. In other words, put money in the hands of actors who get us closer to our goals. With cryptocurrency, this can be hard-coded into law. We’ll look at some projects that are doing just that.
Decentralized Physical Infrastructure Networks (DePINs)
“Oh come on, this again?” Yes, the DePIN narrative is almost as overused as AI itself, but hang on for a moment. DePINs have a real chance to change the world.
What crypto is really good at is removing intermediaries and incentivizing activity, the original vision of Bitcoin was a peer-to-peer currency designed to take banks out of the equation. Modern DePINs also aim to take out centralized power and introduce provably fair market dynamics.
Data Network, the Case of Grass
GPT-3 was trained with 45TB of plain text data, and the amount of data required for GPT-4 and GPT-5 is staggering. Most people get data through web scraping, but this method can easily be limited. Grass connects more than two million devices, scrapes website data from users' IP addresses, collects, structures, and sells it to AI companies. Users participating in the Grass network can earn a steady income from AI companies that use their data. Future $GRASS tokens may further incentivize users to download their browser extensions or mobile apps.
GPU Networking, the Case of io.net
Computing power is essential for AI training. The global supply is huge but uncoordinated, and centralized solutions will create rent-seeking intermediaries. Cryptography can help create a market layer to connect buyers and sellers.
io.net introduced cluster stack technology, which enables GPUs distributed around the world to work together for model training. They worked with Ray to develop cluster middleware, which enables GPU DePINs to start up without permission in 90 seconds.
Use of incentive structures
Going back to the revelation of Bitcoin, miners keep calculating hashes quickly because that is how they are paid. The incentive structure built into the protocol determines its end product. Bitcoin miners and Ethereum stakers absorb participants in all of their native tokens because that is what the protocol wants to incentivize - participants to become miners and stakers.
AI builds networks and discusses Bittensor
Bittensor aims to create several experimental ecosystems for experimentation, with the goal of producing "commoditized intelligence" within each ecosystem. Miners compete with each other in a continuously reviewed "Thunderdome" to get the most rewards. Bittensor is a perpetual reward machine for AI development, and emerging machine learning engineers can earn large amounts of TAO by joining the Bittensor subnet.
Intelligent Agents, Exploring Morpheus
“Smart Agents” are AI models trained on smart contracts that understand the inner workings of top DeFi protocols and automate yield discovery, bridging, and detection of suspicious contracts.
Morpheus' platform facilitates the spread and prosperity of smart agents by incentivizing agent builders, communities, computation, and capital. Smart agents will be the primary way everyone interacts with blockchain in the next 5-10 years.
How to tell if a project is completely useless
Keep in mind our two broad categories of “Crypto x AI”:
1. Encryption helps AI
2. AI helps with encryption
A well-designed token system can lay the foundation for the success of the entire ecosystem. DePIN architecture can help kick-start the market, and creative token incentive structures can coordinate open source projects to work towards goals that were once difficult to achieve.
In the best-case scenario, centralized AI would not only control your finances, but would also impose biases on every piece of data we encounter in our daily lives. Cryptography has the power to coordinate decentralized individuals working together to achieve a common goal. However, this ability is now facing an enemy more powerful than central banks: centralized AI. We need to act quickly to resist the centralizing trend of AI.
