This article is more than 1 year old
Pokémon Go was a 'success disaster' and Niantic is still chasing another hit
The past seven years have seen improved mapping, AR and AI, and the developer's gotta catch 'em all ahead of Monster Hunter launch
Niantic, the Google spin-out behind the smash success augmented reality game Pokémon Go, is set to release Monster Hunter Now – a game that even before launch is struggling in its predecessor's shadow.
The mobile game is a collaboration with Japanese video game publisher Capcom and has been soft launched in select markets where gamers are already joining forces with friends to take down monsters only visible through the screens on their smartphones.
It's hard for Niantic to live up to expectations after the success of Pokémon Go – the game that became both an instant phenomenon and nuisance in 2016. According to Business of Apps, at its peak the game drew in more than 250 million people per month. In 2021, it still brought in $1.21 billion in revenue.
Pokemon Go has had successors – like virtual pet game Peridot, another Nintendo team-up called Pikmin Bloom, Harry Potter tie-in Wizards Unite, and a few others – none of which have had anything remotely like Pokémon Go's success.
"Pokémon Go was the biggest success of any company from a ramp perspective. Going from nothing to that – nobody else has ever done that. None of the other games released have been in that category," Niantic senior vice president of engineering, Brian McClendon told The Register on the sidelines of Augmented World Expo (AWE) Asia 2023 in Singapore on Wednesday.
"If I had to predict, I would say Monster Hunter will be closer than anything else we've ever done," he added with a cheeky smile, as if the marketing department were to walk around the corner at any moment.
McClendon refers to Pokémon Go as a "success disaster" – a term he also uses to describe the launch of Google Earth. Both were products in which he and Niantic founder John Hanke were heavily involved. Google Earth's 2005 launch crashed the chocolate factory's servers when 100 million users very quickly gave the tool a try, using up more than half of the web giant's bandwidth.
It taught Hanke and McClendon lessons that came in handy when it came to Pokémon Go's sudden surge. Even so, the early days and weeks of the game were plagued with crashes, lags and limited functionality – such as the in-game tracking system that broke almost immediately under the strain – that took time to overcome.
"Building a server infrastructure that can handle geographic sharding of the world, was something that we got very, very good at," said McClendon.
"We've improved the infrastructure and rebuilt it. We're still built on Google Cloud services but, you know, we reimplemented things in a much more scalable way," he said.
The seven years between Pokémon Go and Monster Hunter Now is an eon in terms of technology development. Back in 2016 Microsoft had just begun shipment of its original HoloLens headset, Google hadn't yet released its ARCore software development kit, and International Data Corp (IDC) predicted 189.8 percent growth year-on-year in the AR headset market.
Since then Apple has unveiled its $3,500 Vision Pro and IDC has pegged growth for AR and VR headset shipments combined at 14 percent year on year.
- We're about to hit peak device count, says HTC veep, as AR takes over
- By 2026, total AR/VR goggle sales will trail a single quarter of current tablet shipments
- Metaverses are flopping – hard – says Gartner
- Metaverse? Apple thinks $3,500 AR ski goggles are the betterverse
Positioning and mapping technologies – the competitive edge behind Niantic's products – have also changed.
"We use the content from our developers and from our users – Ingress and Pokémon Go – to collect videos and scans that are then used to build these maps," McClendon told an audience at AWE. "And these maps are much higher resolution than any maps that I've built in the past – even in my ten years at Google. And we're able to build this without any LIDAR at all."
McClendon added that keeping data up to date was also a challenge. "The world is ever changing around us and as you get higher and higher in your resolution, the world changes even more. And so we're learning a lot about building accurate maps."
Developers can access that map technology to create their own AR through Niantic's Lightship Augmented Reality Developer Kit (ARDK) – and McClendon told The Register hundreds do so.
But many more – thousands, said McClendon – are using 8th Wall, the development platform it acquired in March.
"It's a lighter weight lift, it's in JavaScript, it's easier to develop and the experiences tend to be smaller. They're built by development groups and agencies who are typically doing a marketing or demonstration experience," explained McClendon.
"So I like to think of 8th Wall as sort of our Darwinian experiment where everybody's trying everything in AR and some of some of them are doing some amazing things with it," he added.
And because it's 2023, starting on Wednesday 8th Wall comes with generative AI developer tools.
"Being able to integrate one of these AR models into a web based platform like 8th Wall – one of the hardest problems is if you're a ChatGPT customer, you get a key, and that key enables you to do it. But if you try to combine it with somebody else's web surface, then you're showing your key to your end users," McClendon told The Reg
"So what we're allowing is for your developer keys to be hidden, yet still having all of the access to your ChatGPT or DALL-E or Inworld," he explained.
But for someone whose company has integrated generative AI into an arguably metaverse related product, McClendon seems uncharacteristically, yet refreshingly, cautious.
"If you're trying to deploy to this large physical world, you can't custom author content and experiences for all the geographic locations in the world. But a computer can," said McClendon.
"And so customized experiences for your particular environment – which is unique to both your time and place – is something I think Gen AI will eventually get very good at. But the path to that – that's more like a three to five year kind of future. We need to improve both image generation and language model understanding and semantic understanding,” added the engineering veep.
As for the metaverse, he said most people want to interact in the physical world.
"We can argue about how many hours a day we spend on the phone, but our phone is really allowing us to just peek through a pinhole into this virtual world," said McClendon. "We still spend our time face to face talking to people living our lives. And, even if we're wearing glasses, our goal is to see the real world." ®