Blackshark.ai, the Austrian firm behind the digital globe you fly over in Microsoft’s Flight Simulator, has secured $20 million in a round A funding round to continue developing and scaling its replica-Earth technology. The possibilities for a global “digital twin” are endless, and the business already has a leg up on mapping behemoths like Google.
Last year, the world saw a completely navigable and astonishingly (if not perfectly) realistic globe in Flight Simulator; we dubbed it a “technical wonder” and subsequently delved into depth about how and by whom it was built.
Blackshark.ai was split out of gaming company Bongfish with the goal of expanding its world-building technology beyond game settings, according to founder and CEO Michael Putz. Machine learning, a little clever judgement, and a lot of processing power are used to transform freely available 2D pictures into realistic 3D representations.
The specifics are here, but the Blackshark.ai system has a keen sense of how various structures seem from above, even in low-light conditions and with inadequate images. They developed a machine learning system that can extrapolate from incomplete outlines by taking into account the neighbourhood (residential versus commercial), roof type (slanted versus flat), and other criteria such as the presence of air conditioning units. It generates a credible 3D reconstruction of the structure using all of this information.
The difficult thing, of course, is figuring out how to accomplish it once, let alone a billion times on a daily basis, in order to produce a current 3D model of every structure on the globe. “Even if you could afford to acquire all the processing power for this, constructing the back end to service it is difficult!” Putz stated. We had to deal with a real-world problem.”
Their response was to optimise, as is typically the case with AI-powered systems. According to Putz, the process of computing the 3D model for every building on the earth used to take roughly a month, but now takes around three days, representing a 300x speedup.
Putz noted that having the capacity to update constantly based on fresh satellite images is critical to their business model. Photogrammetry, or aerial photography that combines several aerial photos and compares parallax data (as human eyes do) to infer size and depth, is used to create a lot of 3D map data, such as what you see on Google and Apple’s maps. This yields useful information… for when the shot is taken.
If you want your 3D map to show what a block in Chicago looked like last week, not two years ago, and you want to cover as much of the world as possible, satellite imagery is your only choice these days. However, this demands the previously indicated 2D-to-3D approach.
Although the Blackshark.ai 3D map and those from Google and Apple appear to be comparable, they are not rivals, according to Putz. All of them give a realistic “canvas,” but their goals are vastly different.
“Google Maps is the canvas for local businesses,” he explained, and what matters to both the corporation and its customers are things like locations, ratings, and directions. “For example, in the instance of flooding, which is a climate change use case, we supply 3D data for Seattle, and others who specialise in water physics and fluid simulation may utilise the real world as a canvas.” Our objective is to make the planet’s surface searchable.”
What is the total amount of flat rooftop space available in this San Diego neighbourhood? What regional airports feature a 4,000-square-meter facility that is available to the public? How do updated wind models and wildfire danger zones intersect? It’s not difficult to imagine how this may be beneficial.
“The more you think about it, the more use cases come to mind,” Putz added. “Of course, government applications, disaster relief, smart cities, and autonomous industries — driving and flying — are all possibilities. All of these sectors require artificial surroundings. This wasn’t just a case of, ‘Hey, we want to do this,’ but rather, it was required. And the only way to tackle this big difficulty is to use this 2D-3D technique.”
M12 (Microsoft’s venture fund) and Point72 Ventures lead the $20 million investment. Putz was ecstatic to have a few familiar personalities on board as advisors: Brian McClendon, co-founder of Google Earth, Dirk Hoke, former CEO of Airbus, and Qasar Younis, former Y Combinator COO and currently CEO of Applied Intuition. (These people are advising rather than joining the board, as this line formerly said.)
While additional engineers and researchers will be employed, the firm must quickly transition from “smart startup” to “global provider of 3D synthetic Earths,” or else it will discover another clever startup eating its lunch. As a result, a sales and support staff, as well as “the remaining parts of a hyperscaling organisation,” will be fleshed out, according to Putz.
There’s also the prospect of — you knew it was coming — metaverse applications, in addition to the more apparent use cases he described. However, in this case, it’s more about the concept that any intriguing AR/VR/etc. apps, from games to trip guides, may base their virtual experience on a newly produced representation of Earth. Not only that, but planets other than our own may be created using the same process, so if you wanted to muck up the planet’s layout and create a new one (and who could blame you? ), you could do it by the end of the week. Isn’t that appealing?
Expect to see “powered by Blackshark.ai” or something similar on a new generation of ever more comprehensive simulations of the complex markets and processes going place on the surface of our planet once the additional investment is put to use.