It’s the year 2030 and you’ve sent your humanoid robot to go buy ketchup at the grocery store. It’s a capable walker, it has dexterous hands, and it can carry more groceries than any parent. High-resolution cameras, gyroscopes and pressure sensors help the robot move quietly through the aisles with such grace that you almost forget that it is hopelessly lost. Like its human counterparts, it is doomed to wander aimlessly around the store taking bets on if this is the kind of store that keeps the ketchup in the condiments or sauces aisle.
Nils Pihl, CEO and Founder ofAuki Labs, is an entrepreneur, behavioral engineer and social transhumanist specializing in the intersection of modern technology and human behavior.
Fully 65% of American grocery shoppers spend over half an hour in the store per shopping trip, and the average shopper also leaves without an item they couldn’t find every third visit. Without some pretty significant overhauls to how robots and computers understand and navigate the physical world, there’s not much reason to believe that your humanoid robot would fare any better — not without a decentralized machine perception network, perhaps the most vital DePIN for a future with independent robotic agents.
Like humans, machines either navigate by memory or guidance. For decades, the most common way for machines and humans to get guidance has been through geopositioning satellites like the GPS. As our cities have grown, however, the GPS is starting to show its age.
Although you rarely think about it, the GPS is a line-of-sight based technology that requires an uninterrupted path between you and several satellites. That’s why it works so poorly in big cities and indoor environments.
Several advancements have been made to try to fill in the gaps. One of the first, from the early days of mobile computing, was to quietly have cellphones measure the signal strength to every WiFi router they passed. Over time, thanks to complex triangulation, companies like Skyhook and Google have managed to create low-resolution maps showing the locations of many of the world’s WiFi routers. This is why navigation applications like Google Maps will ask you to turn on your WiFi to get better results.
In the last decade, critics raised many privacy concerns and lawsuits related to WiFi triangulation And it is fair to say that, unfortunately, privacy lost that battle. It is of some consolation, perhaps, that WiFi triangulation still only helps the average user position themselves to within a few meters — not enough for our ketchup-fetching robot to even correctly understand what aisle it’s in.
And so, big tech has turned to the next promising advance in geolocation: visual positioning systems. Spearheaded by companies like Niantic and Snap, visual positioning systems (VPS) compare the world as seen by an onboard camera to an external memory of what the world looks like, stored on their centrally controlled cloud. In a nutshell, a VPS is a trade where you tell big tech what you are looking at so that they can tell where you are.
Visual positioning systems are accurate down to the centimeter in ideal circumstances, and accurate to under a meter in many public urban spaces. It’s because of this unparalleled precision that big tech companies are betting on the VPS for the future of robotics and AR glasses.
But that should make us pause. Remembering the many privacy violations of the simpler bygone days of mobile social media — how will we fare when big tech companies can see the world through our very eyes, and our homes and private spaces through our machine companions?
If you walk into a grocery store and start filming the shelves you’ll quickly find yourself escorted out of the store. Products placed at eye level are more likely to be picked up and bought, and retailers put careful thought into how they place their products to maximize sales. As such, the visual merchandising layout of stores are carefully guarded competitive secrets.
Simply put, the stores are not interested in sharing the product layout of their stores with a central service. It would be unreasonable to expect that our robot could simply show up at the store and instantly know where every single product is, because it would undermine the intellectual property of the store.
Instead, the best we can hope for is that the store has its own self-hosted and secure system that can answer robot questions about individual products, and guide AI and AR glasses to where they need to go without compromising corporate security.
The attentive reader will already spot where DePIN is promising to outperform the giants of the Web2 era, and bring us our ketchup in a privacy-preserving way.
Unlike humans, robots and computers could exchange spatial data with each other, collaboratively perceiving the world together. Collaborative spatial computing would allow machines to navigate the world better by connecting to other external sources of information. In a Web3 DePIN paradigm, this exchange can both be financially incentivized and cryptographically secure.
Imagining that our grocery shopping robot could find the ketchup faster without compromising corporate security is a cute example. But the implications of decentralized machine perception are staggering. Once self-driving cars can coordinate with each other, and exchange live traffic information, traffic will be radically transformed.
In Beijing, where there are more cars on the road than there are people in Los Angeles, over 1,000 years of human productivity are lost in transit every single day. Decentralized machine perception would allow these cars to move faster in coordination with each other, unlocking hundreds of years of productivity daily.
Decentralized machine perception will one day allow for privacy-preserving AR glasses with a smaller form factor, as the glasses can offload some of the heavy spatial computing to local positioning servers, and change human communication in ways as profound as the invention of writing or the telephone. As our civilization grows to over 100 billion intelligent decision makers over the next twenty years, decentralized machine perception networks will help each and every one of them find their place in the world, both on Earth and beyond.
Note: The views expressed in this column are those of the author and do not necessarily reflect those of CoinDesk, Inc. or its owners and affiliates.
