The crossover between blockchain and AI seems to inspire more derision than usual from Crypto X/Twitter. I think this is largely due to the void of implemented use cases surrounding them. There’s an air of magical thinking about combining the two technologies that has a multiplying effect on any apprehension.
The overlap feels like blockchain circa 2018, when decentralization was pitched alongside every major industry as a revolutionary ace card without sound explanation or proof.
The reality is that there may be a natural synergy between the two technologies. What’s more - and this will be a bitter pill for the anti-crypto brigade - it’s questionable just how effective AI will ever be without any underlying decentralization.
Innovation like this is impossible to predict. It’s a bit like tectonic plates - shifts can be gradual, or they can explode with chaos. Regardless of how these two moving technological masses eventually connect, here is what the first wave of change is shaping up to bring.
Any degen would be forgiven for first trying to use ChatGPT to find the next 100x memecoin. Everyone was looking for any edge in the doldrums of a deflated crypto market in early 2023, and OpenAI’s new supertool felt like that missing link for a short second.
Analyzing today’s TradFi market data is a cumbersome and imperfect process. In what are essentially PvP battlegrounds for profit, the serious players need to continuously invest more than their competition to glean the most accurate information.
At the business end of that process, when the onus is placed on humans to make decisions, there are added risk factors like subjectivity, poor performance, and even malicious behavior.
Merging open-source blockchain data with a network of AI agents brings myriad possibilities for a TradFi space that’s evolved at a crawling pace. AI analysts will be able to continuously scrape, sort, and model information to discover once-hidden insights within the market that lead to profitable moves.
I mentioned the human risk element of this process above; we’ve seen through the likes of SBF just why trust is so low in crypto. The outcome of tech stacks like this is that it won’t be humans who are taking up the majority of block space, but instead AI agents. It will be an intelligent and autonomous swarm capable of identifying opportunities, acting on them, and then learning from those actions all within a set of model specifications.
Finally, the concept of a prediction engine is merely a template. It’s clear to see the implications for other major industry sectors, like insurance and cybersecurity.
APIs are a tricky business to nail. Twitter’s recent shift to paywall-only access for their API irked smaller developers who felt they were being out-priced. One solution that has often been discussed is micropayments; the idea being that payments are made per query at large volumes. The issue here is that this simply hasn’t ever been viable with TradFi transaction fees.
But this is now possible with blockchain and AI. Lightning Labs is building a tech stack for Lightning micropayments. Its LangChainBitcoin suite gives AI agents the capability to hold $BTC, use it to pay for services, and interact with nodes. When combined with LangChainL402 - a Lightning protocol standard - it’s possible to gate real-world services behind Lightning-metered APIs.
The result may be a cost-effective, autonomous payment network that can match the speed and volume of AI agent queries. This will give developers and businesses access to a wider breadth of data that can help to accelerate project development. It also potentially enables new cases to be built on top of it, such as decentralized AI marketplaces and data ownership tools aimed at protecting personal data from AI agents.
Decentralized AI Marketplaces
We’re still exploring the fringes of monetization in DeFi. There has been a lot of positive discussion around the creator economy and content ownership. There were also experiments like FriendTech that gave a glimpse of what’s possible with audiences and influence.
It’s easy to assume that all LLM designs are pretty much the same with a clear market leader like ChatGPT. Check out this short presentation recently held at the University of Princeton and you begin to understand just how complex creating and evaluating LLMs can be under the hood. It also highlights that the development of LLMs is still an emerging field, and one that will be highly personalizable, and therefore highly monetizable.
The sheer speed that AI agents are capable of moving at demands a platform that is capable of working uninterrupted and at scale. Technologies like Fetch AI are creating a framework for AI agents to be created, monetized, and capable of interacting with each other in real-time.
Alongside a focus on scalability and decentralization, Fetch AI also places an emphasis on flexibility when it comes to developing AI agents. It allows the distribution of both public and private agents through its marketplace; and while the private agents are still discoverable on the network, it isn’t possible to view their manifest (which is basically their specs). What this means is that developers can now build and sell proprietary AI agents while having them run at full capacity on an immutable network.
2023 will forever be etched in history as the year that AI warped from sci-fi fantasy to reality.
We’ve seen an explosion of experiments in leveraging tools like ChatGPT to build products and fast-track our rate of understanding around new topics. It’s something that we’ve explored extensively here at The Tinkering Society through experiments in AI-created NFTs and tracking spicy DAO governance.
A natural next step in this development, and the ongoing discussion around the topic, is how autonomous AI agents will interact with one another. It’s now inevitable that we will live in a society where AI agents coexist and eventually begin to rub shoulders, and even learn from each other.
This is a powerful idea. Manpower is essential for conceptualizing the things that push society forward, but a workforce that is both constant and actively learning could enable us to make these leaps much, much quicker.
Having an interwoven network of autonomous AI agents will only truly be effective if they are scalable. And they will only be scalable if they are open and immutable. It only makes sense that decentralization becomes a tenet for this new paradigm.