He started with a noisy desk. The desktop was a wooden cubicle in a laboratory at the University of Northumbria, in northern England, where a young A -researcher began his doctoral court. This was in 2015. The researcher was Ben Fielding, who had built a large machine full of early GPUs to develop AI. The machine was so strong that it bothered Fielding laboratory partners. Fielding crowded the machine under the desk, but it was so large that he had to hit his legs aside.
Fielding had some unorthodox ideas. He explored how the “swarms” of AI, groups of many different models, could speak with each other and learn from each other, which could improve the group. There was only one problem: he was handcuffed by the realities of that noisy machine under his desk. And I knew I was overcome. “Google was also doing this research,” says Fielding now. “And they had thousands [of GPUs] In a data center. The things they were doing were not crazy. I knew the methods … I had many proposals, but I couldn’t execute them. ”
Ben Fielding, CEO of Gensyn, is a speaker at consensus 2025 in Toronto.
Jeff Wilser is the host of The People’s AI: the decentralized podcast of AI and will be the host of the AI summit in consensus 2025.
Then, a decade ago, Fielding realized: The limitations of calculating would always be a problem. In 2015, he knew that if he computes it was a difficult restriction in the academy, it would be absolutely a hard restriction when AI became the mainstream.
The solution?
Ai decentralized.
Fielding co -founded Gensyn (along with Harry Grieve) in 2020, or years before the decentralized AI became fashionable. The project was initially known to build a decentralized calculation, and I have spoken with Fielding about this for Coendesk and in panel after panel in conferences, but the vision is actually more broad: “The network for artificial intelligence.” They are building solutions in the technological battery.
And now, a decade after the noisy Fielding desk brought to his laboratory partners, Gensyn’s first tools are in nature. Gensyn recently launched its “RL Swarms” protocol (a descendant of Fielding Doctoral work) and has just launched its Netnet, which takes blockchain to the fold.
In this conversation prior to the Summit of AI, the consensus in Toronto, Fielding offers an introduction to the swarms of AI, explains how Blockchain fits the puzzle and shares why all innovators, not only technological giants, “should have the right to build automatic learning technologies.”
This interview has been condensed and slightly edited for clarity.
Congratulations on the launch of Testnet. What is the essence of what it is?
Ben Fielding: It is the addition of the first MVP characteristics of blockchain integration with what we have launched so far.
What were those original characteristics, pre-Blockchain?
Then we launched RL [Reinforcement Learning] Swarm a few weeks ago, which is reinforcement learning, post-training, such as a peer network.
This is the easiest way to think about it. When a previously trained model goes through reasoning training, such as Deepseek-R1, learns to criticize your own thinking and improve recursively against the task. Then you can improve your own answer.
We take that process one step further and say: “It’s great for models to criticize their own thought and improve recursively. What happens if they can talk to other models and criticize the thought of others?” If you gather many models in a group that can talk to each other, you can start learning to send information to the other models … with the general objective of improving all swarm itself.
Gotcha, which explains the name “Swarm”.
Good. It is this training method that allows many models to combine, in parallel, to improve the result of a final model goal that could create from those models. But at the same time, you have each individual model only improving on its own. So, if I came along with a model in a MacBook, unite a swarm for an hour and then leave again, you will have an improved local model based on knowledge in the swarm, and the other models in the swarm would have also improved. It is this collaborative training process that any model can join and any model can do. So that’s what RL Swarm is.
Well, that’s what you launched a few weeks ago. Now where Blockchain enters?
Then, the block chain is moving some of the primitive level lower than the system.
Let’s pretend that someone does not understand the phrase “lower level primitive.” What do you mean by that?
Yes, I mean, very close to the resource itself. So, if you think about the software battery, you have a GPU stack in a data center. You have drivers on top of the GPU. You have operating systems, virtual machines. You have all these things up.
Therefore, a lower level primitive is the closest to the lower base in the technological battery. Am I doing that right?
Yes, exactly. And the swarm RL is a demonstration of what is possible, basically. It is just a something Harkky to make a really interesting and scalable automatic learning. But what Gensyn has been doing during the last four years, in a realistic way, is to build infrastructure. And then we are in this period now where the infrastructure is everything in that type of beta v0.1 level. Everything is done. You are ready to start. We have to discover how to show the world what is possible when it is a great change in the way people think of automatic learning.
Do you seem that you are doing much more than decentralized computing or even infrastructure?
We have three main components that are under our infrastructure. Execution: We have consistent execution libraries. We have our own compiler. We have reproducible libraries for any hardware objective.
The second piece is communication. So, suppose you can execute a model on any device in the world that is compatible, can they make them talk? If everyone opts for the same standard, everyone can communicate as TCP/IP from the Internet, basically. So we build those libraries and RL Swarm is an example of that communication.
And then, finally, verification.
Ah, and I suppose this is where blockchain enters …
Imagine a scenario in which each device in the world is being executed consistently. They could link together models. But can they trust each other? If I connected my macbook with yours, yes, they could execute the same tasks. Yes, they could send tensioners from one side to another, but do you know that what they send to the other device is really happening in the other device or not?
In today’s world, you and I would probably sign a contract to say, yes, we agree that we will make sure our devices do the right thing. In the world of machines, it must happen through programming. So that is the final piece we build, cryptographic tests, probabilistic tests, theoretical tests of the game to do that completely programmatic process.
So, that is where the block chain enters. He gives us all the benefits of blockchain that he can imagine, as persistent identity, payments, consensus, etc., so what we are doing with the teston is now taking RL Swarm and the primitive of the other infrastructure and we are adding in the blockchain components and saying: “Hey, when it joins a swarm now, since it has a persistent identity, what exists, what is in mind, what is in mind, what exists, what exists, what exists, what exists, what exists, what exists, what exists, what is in mind, what exists, what exists, what exists, what exists, what exists, what exists, what is in mind, what exists, what exists, what exists, what exists, what exists, what exists, what exists, what exists, what is in mind, what exists, what exists, what exists, what exists, what exists, what exists, what exists, what is in mind, what exists, what exists, what exists, what exists, what exists, what exists, what exists, what is in mind, what exists, what exists. It is into account, what is into account, what is into account, what is into account, what is into account, what is into account, what is into account, which is in one day, what is into account.
In the future you will have the ability to make payments, but at this time, it has that mechanism of consensus of trust where we can finish disputes. So, it is a kind of MVP of the future Gensyn infrastructure, where we are going to add components as we advance.
Give us a mockery of what will come in the pipe?
When we reach the main network, all software and infrastructure is live against Blockchain as the source of trust, payments, consensus, etc., identity. This is the first step of that. He is adding identity and saying that when he joins a swarm, he can register as the same person. Everyone knows who you are without having to consult a centralized server or website somewhere.
Now let’s go crazy and talk more in the future. How is this within a year, in two years, in five years? What is your star of the north?
Sure. The definitive vision is to take all the resources that are under automatic learning and make them instantly programmatically accessible for all. Automatic learning is very limited by its central resources. This creates this enormous pit for centralized AI companies, but it does not need to exist. It can be open source if we can build the correct software. So, our opinion is that Gensyn builds all low -level infrastructure to allow the most open source to approach as possible. People should have the right to build automatic learning technologies.
