The maker of the wildly well-liked cell recreation Pokémon Go is making a pointy flip towards a brand new frontier in AI analysis.
Pokémon Go gamers have spent years voluntarily importing location knowledge from their smartphones coupled with photographs and video of their environment. Utilizing this uncooked knowledge, Niantic, which develops Pokémon Go and different video games, is aiming to create the primary ground-level geospatial mannequin of the earth, a mannequin that may see, motive, and work together with the bodily world round us.
Towards that aim, San-Francisco-based Niantic this week introduced an almost $4 billion deal that may spin off a lot of the gaming studio and return the remaining firm to its roots in mapmaking.
“People take spatial intelligence with no consideration. We stroll in a single route and we don’t must see the opposite route to have the ability to flip round and return, even for locations we’ve by no means been to earlier than,” Niantic senior VP engineering of the Niantic Spatial Platform, Brian McClendon, advised Barron’s in a December interview. “For a pc this can be a large problem.”
As of November, Niantic had 10 million scanned places worldwide, concentrated in North America, Europe, and Japan.
Niantic hopes to make a big geospatial mannequin that may be the muse for a spread of location companies which require detailed ground-level mapping.
In a Wednesday LinkedIn publish, Niantic CEO John Hanke mentioned, “At the moment’s LLMs characterize step one towards a future the place a wide range of skilled fashions collaborate to motive and perceive advanced issues, and plenty of of these issues would require deep and correct data of the bodily world.”
McClendon and Hanke led the groups that constructed Google Maps, Google Earth, Road View and different Google location-based companies.
They’re skilled mapmakers—Niantic spun out of Google in 2010 to take its mapping experience to gaming.
Cellular gaming studio Scopely will purchase most of Niantic’s video games, together with Pokémon Go, for $3.5 billion, leaving $350 million in money behind for Niantic shareholders whose haul would whole $3.85 billion if the deal is permitted by regulators.
The brand new firm, Niantic Spatial, might be capitalized with $200 million from its predecessor, plus one other $50 million funding from Scopely.
Niantic Spatial will want each penny of that for this challenge, and it nonetheless will not be sufficient: The quantity of knowledge collected and put into the mannequin are way over even the biggest language mannequin. Even in its early kind, if Niantic put all of its scanned places collectively, it might be about 100 instances the dimensions of the biggest language mannequin. For now, the corporate will restrict itself to small parts of the complete knowledge set. Niantic can have excessive and rising computing wants, and it might require technological breakthroughs that dramatically cut back AI computing prices for it to succeed.
And the mannequin won’t ever be completed. “One of many massive challenges with a 3-D map just like the one we’re constructing is it’s by no means full as a result of the world is dynamic and regularly modifications,” McClendon advised Barron’s.
The transaction is a barometer of the place tech has come within the 15 years since Niantic was based. Again then, cell gaming appeared like an enormous alternative. It was. At the moment, that chance is brewing in AI.
========================
AI, IT SOLUTIONS TECHTOKAI.NET
Leave a Reply