Decentralized AI: How Web3 AI Agents are becoming part of our Digital Economy

In the upcoming months I will publish a range of AI-related articles in which I am to dive deeper into topics related to AI, Blockchain and Web3.

As we are moving into an era that is more and more influenced by artificial intelligence and blockchain technologies, we can see a new and somewhat unexpected economy emerge. One where AI agents interact autonomously, where they trade and negotiate access to resources and one that is powered by Decentralized Physical Infrastructure Networks, such as Aethir 's decentralized compute network or peaq 's decentralized network for connected machines.

The converge of AI and Web3 transforms how digital economies operate, creating new opportunities for businesses and individuals alike. Before we jump in, lets briefly touch base on the state of AI today, and where we are going to.

The state of AI

Just to put the record straight, at the moment of writing (September 2024), when we speak about AI we really speak about Generative AI. Something as wonderful as ChatGPT in principle is not yet 'self aware' or can not always sufficiently reason why it makes certain predictions. It has been trained on loads (and loads) of content and data, and uses that data to generate answers to questions (inference).

Though the AI gurus do not agree on the exact 'when', it is generally accepted that we will see Artificial General Intelligence (AGI) anywhere between 2027 and 2040, which hopefully for is is within our lifetimes. This next phase of AI, which can be explained as an AI that can think and act autonomously at the level of a University student, will be the next big game changer. It will be able to create on its own, negotiate on its own and 'live' on its own.

I foresee that we will have many different AGI-level AI agents, which will create a whole new depth in our society, as humans and AI interact, trade, do business and work together.

Do you think that this is a crazy and big vision for the future? Well, two years ago you probably also did not even consider the idea that in 2024 almost everyone uses AI in their daily jobs.

Decentralization meets AI: A New Economic Reality

Traditionally, central authorities manage the allocation, access and distribution of vital technologies, data, knowledge and networks. We hand over everything we have to "The Google Gods". Those who control the network and resources control everything, including Artificial Intelligence.

It is their role to create shareholder value, and it is in their interest to remain in power over as many resources as possible. And herein lies one problem, this skewed power distribution results in unfair advantages and as Aethir rightfully point out in their white-paper, leads to an AI wealth gap.

Decentralized systems have different starting points, such as a) generating stakeholder value (opposed to shareholder value) and b) distribution of power. In my Thesis of Data Economy 3.0, I outline where this need comes from and how this relates to something as important as the Fourth Industrial Revolution.

These decentralized systems will exist alongside centralized systems and I would not be surprised when hybrid models appear. With centralized systems, you cannot really talk about the rise of a new AI-economy, as in the end it will be just a handful of entities that will really be driving this new 'economy'.

Decentralized systems, where many different parties collaborate together, do have everything you need in order to create a new economy. This is also the reason I prefer to focus on decentralization in this article.

With the rise of Web3 and blockchain, we are witnessing a paradigm shift, one that allows AI agents to operate without centralized oversight.

These AI agents are autonomous entities that can manage assets, execute smart contracts and participate in governance decisions. What happens, for example, when you have an AI-agent that negotiates the best price to charge your electric vehicle, with the AI of a community owned charging station? What happens when your farm AI-agent advises you on what seeds to plant next, and when to do that? And what if that agent negotiates the best price for you with its counterpart from the 'seed grower'?

Do you think this is something from a time far far away from now?

Well, in the Web3/Crypto space much of this is already happening (in its infancy). AI's are, for example, being trained on blockchain data in order to make the best crypto trades on behalf of the people paying for these AI agents. The crypto space is actually a space where a lot of these 'on-the-edge' innovations are taking place.

Taking this a step further, what if an AI 'lives' by itself, and therefore becomes truly autonomous? It will need storage, compute power, bandwidth, new training-data etcetera. Without those resources, it will not function. This means that it needs to be able to 'sell' its services, earn an income and use that to pay for stuff. Does this sound familiar? Well, it is pretty much the foundation of todays economies as well, with one small obvious difference.

These agents will soon be negotiating with AI agents that represent businesses or even you! It can, for example, process data and manage a company's cloud resources, it can make trades in DeFi and it can advise on the governance in DAOs.

DePINs: The Backbone of the AI-Driven Economy

I foresee that a vital element making this future possible are DePINs (Decentralized Physical Infrastructure Networks). DePIN decentralize access to hardware such as compute resources (needed to train AI), and machine to machine interactions.

As AI agents grow in complexity, the need for scalable, affordable compute infrastructure becomes essential. This is where decentralized ecosystems such as Aethir come in, offering a decentralized cloud compute network that powers AI agents by aggregating underutilized GPU resources.

Aethir enables smaller players, including startups and AI developers, to compete with larger corporations by providing on-demand access to enterprise-grade compute resources. The tokenized model of Aethir further aligns with the Web3 ethos of decentralization, rewarding resource contributors and ensuring transparent access to compute power.

Another example of a DePIN network I know well is Peaq, which already has 1m+ devices that are connected. AI + connected devices mean that they can interact, trade, share resources etcetera. Now I think the potential to combine these two becomes clear, as AI-startups can build Machine AI's on Peaq, powered by Aethir.

AI-Agents and the AI-Driven Economy

As all these AI agents interact, with people, machines and with each other, we see an AI-driven economy emerge. I believe in the idea that we will see thousands of AGI-level AI agents, each with specific capabilities and expertises.

This also means that within this economy, new business opportunities will emerge. Will you invest in an autonomous AI agent that aims to optimize, say, P2P machine driven supply chains? Or perhaps you prefer to invest in AI-agents that execute the best crypto trades on people's behalf?

Tokenized AI agents will own and manage digital assets, making decisions on trading, investments, and resource allocation, all without human intervention.

As, AI agents perform tasks previously handled by us (simple humans), without the limitations by centralized intermediaries, new levels of efficiency and scalability are unlocked.

This promise to democratize access to technology, and therefore level the playing field, is at the core of the 'Web3 movement'. A smart engineer who lives in a small town in a remote area, will have the same chances of building a successful AI-business as the big players. This is one reason why I, a long time ago, entered the Web3 space.

As we move forward, the integration of AI and Web3 will unlock unprecedented potential for innovation, efficiency, and accessibility, enabling businesses of all sizes to thrive in this emerging digital landscape.

Share your thoughts

I am aware that there is much more to say and think about this topic. How do we make sure AI is owned by the people? How do we ensure it is governed in the right way? What about ethics, and who decides what training data to use, and what not?

These are just a few questions we should answer, please do leave your thoughts in the comments ALSO in case you disagree with me or saw some mistakes!

Thanks!

Jochem