Pokémon Go Players Unknowingly Train Niantic's Large Geospatial AI Model

Pokémon Go players have been training a world-faring AI model

Niantic, the developer behind Pokémon Go, has revealed that users have been contributing to the creation of an AI model. This Large Geospatial Model, as it is being referred to, is designed to traverse the physical world. It has used players' data in a similar way that ChatGPT uses vast quantities of text, according to a blog post from Niantic. Except, instead of learning language, the model instead will "enable computers not only to perceive and understand physical spaces, but also to interact with them in new ways, forming a critical component of AR glasses and fields beyond, including robotics, content creation and autonomous systems." While it might just be designed for smart glasses, robotics, and similar fields, some are concerned that it could be used for military applications. Of course, Niantic has no intention of using the model for such purposes now, and instead wishes to create a model for traversing the world.

Pokémon Go Players Have Been Unknowingly Training an AI to Auto-Complete the Real World - IGN

Pokémon Go developer Niantic is hard at work building and training an AI to essentially be able to auto-complete real-world locations with only a limited amount of information. And it's using data collected by Pokémon Go players to do it. In an official blog post spotted by Garbage Day and reported on by 404 Media, Niantic revealed that it's building something called a "Large Geospatial Model." You may already know what a "Large Language Model" is - it's Chat GPT. It's an AI that is trained on enormous amounts of existing text so that it can then produce text on its own that sounds normal, and conceivably like what a user might want to hear. A Large Geospatial Model is essentially the same idea, but applied to the physical world. It's trained on what real-world places look like (a church, a park, a house, etc), and then it can use that data to produce information on what actual places it hasn't seen yet might look like. Niantic claimed this will be useful for technology such as AR glasses, robotics, content creation, and other things. Or as Niantic put it: Imagine yourself standing behind a church. Let us assume the closest local model has seen only the front entrance of that church, and thus, it will not be able to tell you where you are. The model has never seen the back of that building. But on a global scale, we have seen a lot of churches, thousands of them, all captured by their respective local models at other places worldwide. No church is the same, but many share common characteristics. An LGM [Large Geospatial Model] is a way to access that distributed knowledge. But to make this work, Niantic needs lots of data to train that AI on, and it can only do so much on its own. Google's been collecting location data for years via Google Maps and those funny cars it uses to get street view info, but that's not sufficient in this case. Cars can only drive on roads, and Niantic needs pedestrian information from places cars can't go. Fortunately, Niantic has thousands of people globally pointing their phones at things and sending that information back via its various projects and apps, Pokémon Go included. Specifically, Niantic said in its post that it's been building something called a Visual Positioning System (VPS), a technology that uses an image from a phone to determine the position and orientation of a location on a 3D map. The technology is supposed to allow users to position themselves in the world with "centimeter-level accuracy," which allows them to then see digital content overlayed on the physical world "precisely and realistically." Again, from Niantic: This content is persistent in that it stays in a location after you've left, and it's then shareable with others. For example, we recently started rolling out an experimental feature in Pokémon GO, called Pokémon Playgrounds, where the user can place Pokémon at a specific location, and they will remain there for others to see and interact with. But all of this technology exists because users are constantly scanning the world with their phones while using Niantic's apps, including Pokémon Go, and have been for years now. Niantic said it currently has 10 million scanned locations around the world, one million usable with its VPS service, and gets one million new scans every single week containing hundreds of discrete images. That's a lot of data. For now, Niantic said it's using the data explicitly to develop its own technologies that it then turns around and implements into its existing products. However, in recent years there have been numerous concerns over how companies collect data, use it to train AI, and what those AI models might eventually be used for. While today Niantic's LGM work may just be limited to letting us drop cute Pokémon models in the world for other people to find, tomorrow its uses may grow increasingly complex. IGN has reached out to Niantic for comment.

If you played Pokémon Go you trained an AI without realizing

Pokémon Go players are unwittingly training an advanced AI system designed by Niantic to complete real-world locations. This initiative centers around a "Large Geospatial Model" (LGM), which relies on user-generated data to enhance augmented reality and robotics applications. Niantic's official blog outlines that the LGM functions similarly to a "Large Language Model," like ChatGPT, but pertains specifically to physical environments. The LGM is trained on extensive data points of real-world locations such as churches, parks, and homes. By utilizing this model, Niantic aims to predict the characteristics of locations it hasn't directly encountered. The company highlighted that while unique to their locale, many structures share common traits that make this model effective for understanding urban geography. To facilitate this, Niantic is developing a Visual Positioning System (VPS). This technology employs smartphone images to discern a user's position and orientation with high accuracy, allowing for precise digital overlays on the physical landscape. Niantic explained this will enable augmented reality content to remain at specific locations, contributing to a more intricate user experience. For instance, its recently rolled out "Pokémon Playgrounds" feature allows players to place Pokémon at pinpointed real-world locations, which remain accessible for other users. Palworld vs Pokemon debate could end in court due to copyright claims The sheer volume of data generated by Pokémon Go players has been foundational for this project. Niantic currently boasts about 10 million scanned locations, with 1 million being viable for its VPS service. The company collects approximately 1 million new scans each week, featuring hundreds of images each. This continuous influx of location data is essential for refining the geospatial AI functionality that Niantic is advancing. "Imagine yourself standing behind a church. Let us assume the closest local model has seen only the front entrance of that church, and thus, it will not be able to tell you where you are. The model has never seen the back of that building. But on a global scale, we have seen a lot of churches, thousands of them, all captured by their respective local models at other places worldwide. No church is the same, but many share common characteristics. An LGM [Large Geospatial Model] is a way to access that distributed knowledge." -Niantic Despite these optimistic applications, concerns over data privacy and the broader implications of AI training persist. As outlined by various commentators, including OSINT analyst Elise Thomas, the potential military applications of such technology raise ethical questions. The technology harnessed for gaming could evolve into tools with significant ramifications in various fields, beyond entertainment. The applications may start innocuously -- like creating digital Pokémon in specific real-world locales -- but investigations into the broader implications of this technology are likely to continue. As Niantic pushes forward with the LGM project, the balance between harnessing valuable data and ensuring user privacy remains a pivotal subject for ongoing discourse.

Pokémon Go Players Have Been Training an AI to Auto-Complete the Real World - IGN

Pokémon Go developer Niantic is hard at work building and training an AI to essentially be able to auto-complete real-world locations with only a limited amount of information. And it's using data collected by Pokémon Go players to do it. In an official blog post spotted by Garbage Day and reported on by 404 Media, Niantic revealed that it's building something called a "Large Geospatial Model." You may already know what a "Large Language Model" is - it's Chat GPT. It's an AI that is trained on enormous amounts of existing text so that it can then produce text on its own that sounds normal, and conceivably like what a user might want to hear. A Large Geospatial Model is essentially the same idea, but applied to the physical world. It's trained on what real-world places look like (a church, a park, a house, etc), and then it can use that data to produce information on what actual places it hasn't seen yet might look like. Niantic claimed this will be useful for technology such as AR glasses, robotics, content creation, and other things. Or as Niantic put it: Imagine yourself standing behind a church. Let us assume the closest local model has seen only the front entrance of that church, and thus, it will not be able to tell you where you are. The model has never seen the back of that building. But on a global scale, we have seen a lot of churches, thousands of them, all captured by their respective local models at other places worldwide. No church is the same, but many share common characteristics. An LGM [Large Geospatial Model] is a way to access that distributed knowledge. But to make this work, Niantic needs lots of data to train that AI on, and it can only do so much on its own. Google's been collecting location data for years via Google Maps and those funny cars it uses to get street view info, but that's not sufficient in this case. Cars can only drive on roads, and Niantic needs pedestrian information from places cars can't go. Fortunately, Niantic has thousands of people globally pointing their phones at things and sending that information back via its various projects and apps, Pokémon Go included. Specifically, Niantic said in its post that it's been building something called a Visual Positioning System (VPS), a technology that uses an image from a phone to determine the position and orientation of a location on a 3D map. The technology is supposed to allow users to position themselves in the world with "centimeter-level accuracy," which allows them to then see digital content overlayed on the physical world "precisely and realistically." Again, from Niantic: This content is persistent in that it stays in a location after you've left, and it's then shareable with others. For example, we recently started rolling out an experimental feature in Pokémon GO, called Pokémon Playgrounds, where the user can place Pokémon at a specific location, and they will remain there for others to see and interact with. But all of this technology exists because users are constantly scanning the world with their phones while using Niantic's apps, including Pokémon Go, and have been for years now. Niantic said it currently has 10 million scanned locations around the world, one million usable with its VPS service, and gets one million new scans every single week containing hundreds of discrete images. That's a lot of data. Update 11/22/2024: Niantic reached out to IGN post-publication to clarify that the scans building this model are entirely opt-in, and are currently focused on training the AI model to deliver a better player experience. As Niantic put it: We use player-contributed scans of public real-world locations to help build our Large Geospatial Model. This scanning feature is completely optional - people have to visit a specific publicly-accessible location and click to scan. This allows Niantic to deliver new types of AR experiences for people to enjoy. Merely walking around playing our games does not train an AI model. Original story continues: For now, Niantic said it's using the data explicitly to develop its own technologies that it then turns around and implements into its existing products. However, in recent years there have been numerous concerns over how companies collect data, use it to train AI, and what those AI models might eventually be used for. While today Niantic's LGM work may just be limited to letting us drop cute Pokémon models in the world for other people to find, tomorrow its uses may grow increasingly complex.

Pokemon Go players are actually training a giant AI model

In 2016, Pokemon Go was about as wholesome as video games got, as it largely involved walking around your neighborhood and meeting up with strangers. In 2024, nothing is allowed to be wholesome, apparently. Based on a blog post by developer Niantic as well as news reports from the likes of 404 Media and Garbage Day, it is now known that Pokemon Go players, whether they knew it or not, have been helping to train a large geospatial artificial intelligence. Tied into something called the Visual Positioning System (or VPS), Niantic's blog post said the idea is to help AI learn about complex three dimensional spaces, in a way that can be used for future augmented reality or even robotic applications. The technical details may fly over the head of many who read Niantic's blog post, but the easiest way to think of it is that certain player actions in Pokemon Go have been training this geospatial AI in the same way that written internet content trains things like ChatGPT. In particular, a feature called "Pokemon Playgrounds" that allows users to pin a Pokemon to a real-world location that they will persistently stay in for other players to see is apparently tied into these efforts. While Niantic may be telling the truth about the potential future applications of this technology, it's also worth remembering that AI data can be used for nefarious means, too. As OSINT analyst Elise Thomas pointed out, there's a decent likelihood that this technology also enters military use at some point.

Pokémon Go Players Have Unwittingly Trained AI to Navigate the World

Niantic says it is using data generated by Pokémon Go players to create a "Large Geospatial Model" that can navigate the real world and power robots. Niantic, the company behind the extremely popular augmented reality mobile games Pokémon Go and Ingress, announced that it is using data collected by its millions of players to create an AI model that can navigate the physical world. In a blog post published last week, first spotted by Garbage Day, Niantic says it is building a "Large Geospatial Model." This name, the company explains, is a direct reference to Large Language Models (LLMs) Like OpenAI's GPT, which are trained on vast quantities of text scraped from the internet in order to process and produce natural language. Niantic explains that a Large Geospatial Model, or LGM, aims to do the same for the physical world, a technology it says "will enable computers not only to perceive and understand physical spaces, but also to interact with them in new ways, forming a critical component of AR glasses and fields beyond, including robotics, content creation and autonomous systems. As we move from phones to wearable technology linked to the real world, spatial intelligence will become the world's future operating system." By training an AI model on millions of geolocated images from around the world, the model will be able to predict its immediate environment in the same way an LLM is able to produce coherent and convincing sentences by statistically determining what word is likely to follow another. "Large Geospatial Models will help computers perceive, comprehend, and navigate the physical world in a way that will seem equally advanced," Niantic said. I found the following explanation in Niantic's blog of how the LGM works to be the clearest: Niantic's LGM builds upon its Lightship Visual Positioning System (VPS), which allows players to pin virtual items to physical locations in the world with "centimeter-level accuracy." For example, Niantic recently introduced an experimental feature in Pokémon Go called Pokémon Playgrounds, where the user can place Pokémon at a specific location that will remain there for others to see and interact with. This feature, Niantic explains, is powered by massive amounts of data, and is unique because it is taken from a pedestrian perspective from locations inaccessible to cars. "Today we have 10 million scanned locations around the world, and over 1 million of those are activated and available for use with our VPS service," Niantic said. "We receive about 1 million fresh scans each week, each containing hundreds of discrete images." This data, Niantic's blog explains, is collected from its games and Scaniverse, Niantic's app for 3D scanning objects and locations. The AI gold rush of the last few years prompted a frenzied hunt for large datasets that can train generative AI models. We've seen companies scrape text from the internet, YouTube subtitles, YouTube videos, books, and more, with little consideration for the humans who created this data. In this case also, players of the incredibly viral Pokémon Go had no way of knowing that when they downloaded the game in 2016 that it would one day fuel this type of AI product. It should come as no surprise that Niantic would now try to leverage its data in the AI space. As the company says, data from Google Street View and various self-driving companies means that there's quite a bit of data from roads, but Niantic's games have created a huge dataset of where only pedestrians can go. At the moment, the company says the data can be useful in a few ways, like other augmented reality products, but the way this data might help robots navigate the world is the most interesting since robots that navigate the real world do anything from deliver food to carry automatic rifles. Niantic did not respond to a request to comment about whether it had any limitations on who and how it would allow people to leverage this data.

Gotta Catch 'Em All: How Pokémon Go covertly captured your data for years to train a massive AI model

Players of "Pokémon Go" -- an augmented reality (AR) mobile game released in 2016 that took the world by storm -- have been unwittingly training an artificial intelligence (AI) model to map the planet at street level. Niantic, the company behind the popular game, has revealed that it will use data scraped from its AR apps to construct a "large geospatial model" (LGM) that would enable robots and other devices to better navigate the physical world -- even if they only have limited information. The announcement, made Nov. 12 in a blog post on Niantic's website, reveals that the company has drawn data from more than 10 million scanned locations worldwide, with users adding around 1 million more new scans each week. This data has already been used to train 50 million local neural networks (collections of machine learning algorithms structured like the human brain) to operate in more than a million locations worldwide. "In our vision for a Large Geospatial Model (LGM), each of these local networks would contribute to a global large model, implementing a shared understanding of geographic locations, and comprehending places yet to be fully scanned," Niantic staff scientist Eric Brachmann and chief scientist Victor Adrian Prisacariu wrote in the post. "The LGM will enable computers not only to perceive and understand physical spaces, but also to interact with them in new ways, forming a critical component of AR glasses and fields beyond, including robotics, content creation and autonomous systems." Just as Large Language Models (LLMs) such as ChatGPT consume vast quantities of text to accurately guess the most probable words to complete a sentence, LGMs gorge on geodata to infer what buildings in physical space should look like. Related: 'I'd never seen such an audacious attack on anonymity before': Clearview AI and the creepy tech that can identify you with a single picture This might seem like a strange task. For humans, our time in the physical world has already exposed us to innumerable examples that helped us to build a robust spatial understanding. Sign up for the Live Science daily newsletter now Get the world's most fascinating discoveries delivered straight to your inbox. Contact me with news and offers from other Future brandsReceive email from us on behalf of our trusted partners or sponsorsBy submitting your information you agree to the Terms & Conditions and Privacy Policy and are aged 16 or over.RELATED STORIES -- Will language face a dystopian future? How 'Future of Language' author Philip Seargeant thinks AI will shape our communication -- 'Put glue on your pizza' embodies everything wrong with AI search -- is SearchGPT ready to change that? -- AI 'hallucinations' can lead to catastrophic mistakes, but a new approach makes automated decisions more reliable "But for machines, this task is extraordinarily difficult. Even the most advanced AI models today struggle to visualize and infer missing parts of a scene, or to imagine a place from a new angle," Niantic representatives wrote in the post. Niantic's LGM is built on its Visual Positioning System, which uses a single smartphone camera image to pinpoint an object's position and orientation down to the centimeter (0.4 inches). As for "Pokémon Go" fans, many seem largely unfazed and unsurprised that their data has been scraped for use by an AI system. Yet critics fear that some of the potential applications of Niantic's technology could be far from benign. "It's so incredibly 2020s coded that Pokémon Go is being used to build an AI system which will almost inevitably end up being used by automated weapons systems to kill people," Elise Thomas, a senior intelligence analyst at the Institute for Strategic Dialogue, a political advocacy organization, wrote on X.

Niantic uses Pokémon Go player data to build AI navigation system

Last week, Niantic announced plans to create an AI model for navigating the physical world using scans collected from players of its mobile games, such as Pokémon Go, and from users of its Scaniverse app, reports 404 Media. All AI models require training data. So far, companies have collected data from websites, YouTube videos, books, audio sources, and more, but this is perhaps the first we've heard of AI training data collected through a mobile gaming app. "Over the past five years, Niantic has focused on building our Visual Positioning System (VPS), which uses a single image from a phone to determine its position and orientation using a 3D map built from people scanning interesting locations in our games and Scaniverse," wrote Niantic in a company blog post. The company calls its creation a "Large Geospatial Model" (LGM), drawing parallels to large language models (LLMs) like the kind that power ChatGPT. Where language models process text, Niantic's model will process physical spaces using geolocated images collected through its apps. The scale of Niantic's data collection reveals the company's sizable presence in the AR space. The model draws from over 10 million scanned locations worldwide, with users capturing roughly 1 million new scans weekly through Pokémon Go and Scaniverse. These scans come from a pedestrian perspective, capturing areas inaccessible to cars and street-view cameras. The company reports it has trained more than 50 million neural networks, each one representing a specific location or viewing angle. These networks compress thousands of mapping images into digital representations of physical spaces. Together, they contain over 150 trillion parameters -- adjustable values that help the networks recognize and understand locations. Multiple networks can contribute to mapping a single location, and Niantic plans to combine its knowledge into one comprehensive model that can understand any location, even from unfamiliar angles.

Pokémon Go Players Unknowingly Help Train Niantic's AI Mapping System

Pokémon Go's developer, Niantic, is building an advanced large geospatial model out of the data gathered. This artificial intelligence system specializes in recognizing and mapping 3D spaces worldwide. Niantic recently published a blog post discussing this large-scale project, which will help the company improve space awareness and real-world engagement for AR experiences. LGM, a Niantic project, aims to map real-world locations with superior accuracy using inputs from millions of Pokémon Go players. This initiative is compared to Large Language Models (LLMs) like ChatGPT, but LGM researches space instead of processing text. When trained on information about landmarks and structures such as parks, churches, and buildings, the model can make likely predictions about unknown terrains. The company highlights a hypothetical scenario to explain its utility: for instance, if a model only knows the front of a church, it may have challenges mapping its back. Using global data collected by players, the model finds common traits of similar landmarks and can accurately place them even in locations not previously surveyed. Niantic plans to use this technology beyond gaming in areas such as augmented reality glasses, robotics, and digital content creation. These advancements could pave the way for a very accurate real-life experience.

It's not just a game. Your Pokemon Go player data is training AI map models.

Players of Pokémon Go may not have realized it, but they've been training more than their Pokémon. Niantic, the developer behind the popular mobile game Pokémon Go, announced last week it is building an AI model to map the physical world. This "large geospatial model" would utilize data collected from players to "achieve spatial intelligence," the company said in a blog post. Pokémon Go, first released in 2016, is an augmented reality game where players use their mobile phones to find and catch virtual Pokémon in the real world. In the game, which has had more than 600 million downloads since release, players can also collect items at PokéStops and battle at gyms, which are both located at real-world landmarks. Niantic's model is training and processing data using geolocation information from scans players submit of those real-world locations while playing Pokémon Go and other Niantic games. More:Why the new Lego Horizon Adventures game is perfect for the young (and old) "Over the past five years, Niantic has focused on building our Visual Positioning System (VPS), which uses a single image from a phone to determine its position and orientation using a 3D map built from people scanning interesting locations in our games and Scaniverse," the company said in the announcement. The company said it currently has 10 million scanned locations from around the world for use with its VPS, with about 1 million new scans each week. The model will process these geolocated images and create a 3D map, while also filling in information about geographic locations, "implementing a shared understanding of geographic locations, and comprehending places yet to be fully scanned," according to the blog post. Companies looking for more ways to utilize customer data is becoming the "new normal," Anton Dahbura, the executive director of the Information Security Institute at Johns Hopkins University, told USA TODAY. Niantic said the data is unique since it is taken from a "pedestrian perspective," unlike other mapping systems that rely on images captured by vehicles and may not include places inaccessible to cars. More:It's cozy gaming season! Video game updates you may have missed, including Stardew Valley According to Niantic's privacy policy, the company collects location data, and other personal data, such as name and email address. The privacy policy outlines what is and isn't shared with third-party vendors, but not what the company does with the data. Niantic has a separate privacy policy for children who play the developer's games, and a portal where parents can set up and manage their child's profile. "It's a typical problem with data privacy and the state of technology today," Dahbura said. "In fact, it's almost par for the course that companies are looking for ways to use their data, and it's even expected by investors." It's understandable for users to worry about how their collected personal data is being stored or shared. "Even with the best intentions, having troves of data that contain so much personal information can be dangerous," he said. "It can fall into the wrong hands, there can be a major data breach, and so on." For users who are worried about their privacy, Dahbura suggests players think carefully about their usage. "Use it exclusively in very public places, not places that you consider to be private, such as the interior of your home," he said. Users should also minimize having other people in images, "especially your loved ones," he said. Players should also be aware of location and how it is interwoven into so much personal data. "A lot of people really underestimate the importance of location data," Dahbura said. "Our critical infrastructure is much broader than people realize, including transportation systems, pharmaceutical, financial, food manufacturing and so on. If people with bad intentions figure out that you have access to these kinds of facilities, it can be used not only against you but also against national security."

Did you play Pokémon Go? You didn't know it, but you were training AI to map the world

You probably didn't know it, but if you played or are still playing Pokémon Go (there are more than half a million active players), you were helping train an AI-powered geospatial model that aims to map the world. A blog post from Niantic, the software developer behind the popular game, explains how it's working on "a large geospatial model to achieve spatial intelligence" and trying to build a "visual positioning system" to understand the world around us -- and it's using data from Pokémon Go. Also: AI transformation is the new digital transformation. Here's why that change matters To clarify, Niantic is saying that just like data on the web trains AI models, the AI model it's building needs to understand 3D spaces. An immense amount of data and photographs of 3D spaces are available thanks to Pokémon Go players creeping around the world. Niantic explains it like this: A local AI mapping model might understand that a church stands at a specific place, but it's likely only seen the front of that location and can't explain what the rest of the church looks like. With data from Pokémon Go players, who have likely walked around many churches and trekked areas that cars can't reach (and photographed those areas), the AI, now has a good guess at what a church generally looks like. Also: Traveling for the holidays? Google Maps uncovers 'hidden gems' to add to your route now The company also pointed out that it recently rolled out a new feature for the game called Pokémon Playground that lets you place a creature at a certain real-world spot for others to see. This means that placing the character and viewing it later conveniently involves using your camera, taking images from multiple angles, and sending the resulting image to Niantic. According to Niantic, it currently has 10 million scanned locations around the world, with one million of those activated and available for use in its VPS service. It added that it receives about 1 million fresh scans each week, each containing hundreds of images. Niantic says it will use this data for purposes like AR glasses, robotics, content creation, and autonomous systems. So, not only did the company make money selling in-game items to players, but is also going to make money on the maps those players helped make.

Ever Played Pokémon Go? You Helped Train an AI for 'Spatial Intelligence'

Pokémon Go developer Niantic is building a "large geospatial" AI model with the goal of giving the AI "spatial intelligence" by using real-world images collected from its games' players, several news outlets reported this week following a post from Niantic itself. Spatial intelligence is an AI's hypothetical ability to estimate the rest of an image in a 3D environment when given part of it, or to understand how 3D objects look when manipulated or rotated. Currently, as Niantic admits, AIs are terrible at this. They don't actually "know" what the 3D world is or how it works, so it's difficult for AI models to accurately predict what a location might look like from a different angle. This geospatial model is being trained on "billions of images of the world" tied to specific, precise locations to tackle this problem. This data has, at least in part, been harvested from the 10 million scanned locations from its Pokémon Go players around the world. Niantic says it gets a million "fresh scans" of locations every week with "hundreds" of images each to feed into its geospatial AI model. It's building off its existing Lightship Visual Positioning System, or VPS, which lets you move virtual items around a capture of the real world on a smartphone in its augmented reality games. Niantic believes its large geospatial model, however, will be able to take things one step further by anticipating what the rest of something might look like based on pre-fed internal "knowledge" of a specific type of building, for example. While a world covered in AR objects might sound like a cool, Bladerunner-esque future to some, others are already concerned that geo-mapping AI tools could be used by global militaries. "It's so incredibly 2020s coded that Pokémon Go is being used to build an AI system which will almost inevitably end up being used by automated weapons systems to kill people," tweeted Elise Thomas, OSINT analyst at the Institute for Strategic Dialogue in the UK. PCMag has reached out to Niantic for comment. Militaries are already using AI for war and intelligence purposes, though, and will likely continue to do so. A web page for the US National Geospatial Intelligence Agency calls AI "a revolution in warfare and intelligence creation" and says it's already using and will continue to use AI to provide information "to gain a competitive edge." The Department of Defense has also publicly advocated for military AI use if it's "responsible." In Ukraine, AI drone fleets are being built as part of its ongoing war with Russia. China has built at least one AI tool for its military, as well, but recently agreed with the US that humans, not AIs, should control nuclear weapons. When it comes to Pokémon Go, a Belarus military official tried to claim in September that the game is a tool for Western intelligence, presumably because Niantic is based in the US. When the game was first released, US military members were reportedly playing the mobile game on the job enough that one base had to tell its staff to "NOT chase Pokémon into controlled or restricted areas, office buildings, or homes on base."

Niantic is building a 'geospatial' AI model based on Pokémon Go player data

Niantic has announced that it's building a new "Large Geospatial Model" (LGM) that combines millions of scans taken from the smartphones of players of Pokémon Go and other Niantic products. This AI model could allow computers and robots to understand and interact with the world in new ways, the company said in a blog post spotted by Garbage Day. The LGM's "spatial intelligence" is built on the neural networks developed as part of Niantic's Visual Positioning System. The blog post explains that "Over the past five years, Niantic has focused on building our Visual Positioning System (VPS), which uses a single image from a phone to determine its position and orientation using a 3D map built from people scanning interesting locations in our games and Scaniverse," and "This data is unique because it is taken from a pedestrian perspective and includes places inaccessible to cars."

Niantic makes Geospatial models from Pokémon Go, Ingress data

Disclaimer: This content generated by AI & may have errors or hallucinations. Edit before use. Read our Terms of use Augmented reality video game company Niantic is working on a large geospatial model to help computers perceive spaces using data from its popular video game Pokémon Go. Geospatial models use billions of images of the world to get a location-based understanding of space, structures, and physical interactions. A person can see a physical structure from one side and imagine what it would be like from other angles because of countless similar structures we see in our daily lives, this is called 'spatial understanding'. "But for machines, this task is extraordinarily difficult. Even the most advanced AI models today struggle to visualize and infer missing parts of a scene, or to imagine a place from a new angle," Niantic explains. "Over the past five years, Niantic has focused on building our Visual Positioning System (VPS), which uses a single image from a phone to determine its position and orientation using a 3D map built from people scanning interesting locations in our games [like Pokémon Go and Ingress] and Scaniverse," the company says. In a 2022 Q&A, the company had more clearly said that it used the data its users upload when playing games like Ingress and Pokémon Go, to build high-fidelity 3D maps of the world. Instead of just relying on the phone's location through GPS, VPS uses gamers' phone cameras. VPS locations have 'centimeter-level' accuracy, which means that users of Niantic games and services can "see digital content placed against the physical environment precisely and realistically." The company mentioned that it recently started rolling out an experimental feature in Pokémon GO, called Pokémon Playgrounds, where the user can place Pokémon at a specific location, and they will remain there for others to see and interact with. "Today we have 10 million scanned locations around the world, and over 1 million of those are activated and available for use with our VPS service. We receive about 1 million fresh scans each week, each containing hundreds of discrete images," Niantic says. The company adds that it collects data from a pedestrian perspective, and as such includes places inaccessible to cars. This is unlike Google Street View, which uses cars to collect spatial data. Through VPS, the company has trained more than 50 million neural networks to date with more than 150 trillion parameters. "In our vision for a Large Geospatial Model (LGM), each of these local networks would contribute to a global large model, implementing a shared understanding of geographic locations, and comprehending places yet to be fully scanned," the company explains. Ninantic's VPS currently has local neural maps (or models) that can only work with what it has directly seen. So if you are standing in front of a church, and the closest local neural map available for your area has only seen the front entrance of that church, it will not be able to tell you where you are. "But on a global scale, we have seen a lot of churches, thousands of them, all captured by their respective local models at other places worldwide. No church is the same, but many share common characteristics. An LGM [large geospatial model] is a way to access that distributed knowledge," the company explains. As such, the LGM will function by enabling communication and data sharing across local models. Through the knowledge it gets from multiple local models, an LGM will be able to internalise the structure and common characteristics of a church. As such, even if Niantic's local model has only mapped the front of a specific church in your area, the LGM will be able to make an intelligent guess about where you are. Niantic explains that current AI research efforts are focused on 'modeling the 3D appearance of objects', called 3D Vision Models. However, geospatial models are a step above 3D Vision Models as well, they deal with real-life locations and deal with real metrics (accurate measurements). As such, geospatial models "represent next-generation maps, rather than arbitrary 3D assets", the company says. It adds that while 3D models can create and understand a 3D scene, a geospatial model can understand how that scene relates to millions of other scenes, geographically, around the world. "These models could guide users through the world, answer questions, provide personalized recommendations, help with navigation, and enhance real-world interactions. " Niantic mentions. It suggests that while AR glasses with 3D graphics are far from coming to the mass market, geospatial models could find utility in audio-only or 2D display glasses. Geospatial models can also find use in spatial planning and design, logistics, audience engagement, and remote collaboration. Developers could use geospatial models and large language models (LLMs) "so understanding and space come together, giving people the opportunity to be more informed and engaged with their surroundings and neighborhoods."