Crowdsourced Compute: Can DePIN Unlock AI Agents

Mawari is revolutionizing the intersection of AI, Spatial Computing, and Web3 by giving AI Agents a body through its Digital Humans as a Service.

During a recent tech conference in Tokyo, audience members watched an on-stage panel discussion. The moderator posed questions to on-stage speakers who conveyed their insights – a familiar scene to those who attend business conferences and tech meetups. But one thing was different in this case: the moderator was not there. She existed in bits rather than atoms. Her spatial stage presence was rendered as a hologram to audience members watching with AR headsets.

This experience is eye opening to the possibilities at the 3-way intersection of AI, spatial computing, and web 3. And relevant use cases expand well beyond business conferences. Dimensionally placeshifting individuals in virtually-embodied form not only offers benefits like reducing cost and carbon, but it has broad applicability. Musical performers can collaborate and play on stage together from across the globe. Political rallies can happen simultaneously in several cities while politicians broadcast from a single location – thus streamlining logistics and travel, freeing them up to work on policy rather than politics. Just to name a few potential use cases.

Mawari: What, Why, and How?

All the above just scratches the surface and addresses the what and why. But an equally-important question is how. Indeed, the spatial computing world – and the metaverse in particular – has gotten burned from overpromising and future-gazing without stopping to consider today’s underlying limitations. Any such future gazing requires a dose of practicality.  

That’s where Mawari comes in. Among other things, our technology unlocks these virtual-human scenarios. In fact, it powered the virtual panel moderator above. It’s all part of the company’s Digital Human as a Service (DHaaS) initiative, whose goal is to enable scalable and reliable embodied virtual avatars for a variety of use cases. 

The latest milestone in Mawari’s DHaaS is a collaboration with Meta Osaka to develop and deploy high-resolution, AI-powered digital human guides to interact with patrons at transportation hubs, commercial facilities and tourist destinations. Starting in Spring 2025, these digital human guides will offer multilingual automated response capabilities for use cases ranging from tourism guidance and facility navigation.

Mawari’s DHaaS also builds on the company’s past work with virtual avatars. For example, with Japanese telecom giant KDDI, it built Aiko in 2020. This was an embodied digital assistant to support onsite visitors in retail and gallery spaces, mostly through mobile AR interfaces. 

Now, the focus with DHaaS is to evolve, enliven and scale AI Agents to greater levels of interaction, including headworn form factors like Apple Vision Pro and future smart glasses. Most importantly, Mawari’s goal is to support this in ways that are scalable and reliable across the globe. That last part has been the biggest bottleneck in broader efforts to bring AI Agents to life.

XR’s Promise

What are those bottlenecks? They boil down to infrastructure shortcomings. For example, in its work with KDDI and Aiko, noted above, Mawari utilized KDDI’s 5G and edge servers. Such assets unlock high-throughput and low-latency requirements of immersive and bandwidth-intensive experiences. But the issue is that they aren’t prolific on a global scale. It’s a coverage issue.

Without getting too far into technological limitations (more on that in the next part of this series), it’s all about the distance to the closest server. Even super-scalers like AWS and Google Cloud don’t have the GPU coverage and performance at the edge to ensure Digital Humans can exist anywhere. (fun fact: this is one of the main reasons why Google Stadia failed). 

Given this bottleneck, Mawari has assembled a combination of its rendering engine, compression tech, and advanced network. That last part is the latest piece of the puzzle to unlock and scale DHaaS. And the key word is DePIN. Short for Decentralized physical infrastructure networks, this emerging Web 3 segment could be the key to unlock XR’s promise. 

DePIN does this by eschewing centralized servers – the traditional approach – instead relying on a decentralized network of GPU-powered and geographically distributed nodes. This not only can achieve load balancing and geographic coverage, but it also reduces bandwidth usage by about 80 percent – which directly impacts cost, quality, and reliability.

Nodeworthy Development 

The Mawari Network is built on this principle, utilizing far-flung nodes to comprise a network. For those familiar with web infrastructure, this acts like a traditional content delivery network (CDN), but specializes in bandwidth-intensive 3D content and live interaction. 

The other factor that separates the Mawari Network from a CDN is the nature of those “far-flung nodes” In short, they consist of you and me. Mawari’s Guardian Node license program crowdsources compute resources to anyone with a minimally-spec’d computer that wants to host nodes. These collectively power the decentralized network, while rewarding participants (more details about this will drop very soon). 

And to ensure reliability, Mawari is building on its longstanding collaboration with KDDI. The telecom giant will serve as the infrastructure management partner by hosting Guardian Nodes on its secure servers. This engenders more security, reliability and trust in Mawari’s network. 

But KDDI isn’t the only one, more partnership announcements are coming soon.

Altogether, this opportunity occupies the relatively-thin but tremendously-valuable slice of the Venn diagram between Web3 and XR. Tech pundits often talk about this convergence, which gets us back to lots of future-gazing around the metaverse. But the place where these two technologies practically come together today – to power digital humans and several other applications to come – is the Mawari Network. 

We’ll pick things up there in the next part of this series to go deeper into Mawari’s history, founding principles, and technological trajectory…