The ultimate goal of AI is to pass what's known as the Turing test — to build a machine whose intelligence is indistinguishable from a human's. The emerging field of augmented reality (AR) is striving to pass its own "visual Turing test," in which virtual objects interact so seamlessly with their real-world surroundings that it's virtually impossible to distinguish what's digital from what's real.
The imaging technology startup Avegant says it has achieved a truly realistic and immersive AR experience with its newly introduced "light field" head-mounted display. The company's prototype headset appears to overcome one of the critical limitations of current AR devices like Microsoft's HoloLens — the inability to interact with objects at close range.
Journalist and tech industry analyst Ross Rubin was recently given a private demo of the Avegant Light Field prototype and compared it to the immersive experience promised by Magic Leap, a secretive AR startup that's received nearly $1.4 billion in funding from Google and Alibaba.
"One of [Magic Leap's] first signature images was of an elephant being cupped by a hand. That's what I experienced in the Avegant demo, to be able to hold a planet in my hands," Rubin told Seeker. "It's the difference between seeing animated objects out of your reach and being able to actually walk up to things, put your hands around them and navigate within the mixed reality space."
Avegant's light field headset works by mimicking the way that real-world objects emit light. In the real world, light enters the eye as a collection of rays, which are focused for the brain. Objects that are different distances from the eye emit different "light fields," requiring the eye to constantly adjust like the autofocus setting on a camera. In humans, our combination of binocular vision and the ability to switch between different focal planes gives real-world objects their depth and distance.
To display virtual objects with the same realistic sense of depth, Avegant needed to give each digital object its own light field. Instead of presenting the viewer with a flat projection of virtual objects at a set distance, each object now exists on its own focal plane. When the user turns his attention from object to object, his eyes refocus to accommodate the object's specific light field, including objects very close at hand.
According to Avegant CTO Edward Tang, this represents a critical technological leap from current AR or MR (mixed reality) headsets, including the $3,000 HoloLens.
"Mixed reality displays in the market today are fixed single focal plane displays which limit the area where images can be displayed and is particularly challenging in the near field distance, where most interactions occur in MR," Tang said in an email.
Jeremy Bailenson, founding director of the Stanford University Virtual Human Interaction Lab, was deeply impressed by his test of the Avegant Light Field prototype.
"It's not often you witness a technology that could disrupt an entire industry," Bailenson was quoted as saying in an Avegant press release. "Avegant Light Field is that kind of a breakthrough. I had long heard about the possibility of seeing close fields and far fields in the same scene, but experiencing it firsthand was believing. This is how the visual system was designed to work."
As Avegant works to turn its prototype headset into a viable commercial device over the next 12 to 18 months, there are still a lot of questions about how its patented light field technology actually works.
Gordon Wetzstein, director of the Stanford Computational Imaging Group, explained that there are several leading methods for producing the kind of light field effects advertised by Avegant, each with its own set of tradeoffs. In a recent paper, Wetzstein's group demonstrated the benefits of one such method, using a "focus-tunable" liquid crystal display that constantly refocuses on where the user is looking.
"It's a very practical approach that can be implemented right now, and we're working with a couple of consumer electronics companies to do that," Wetzstein told Seeker. "The tradeoff there is that you need eye tracking, and you need a few other technological components. All of them are already there, but you'd have to put them together in one headset."
According to a patent application by Avegant, this appears to be exactly what the company is doing, using some sort of eye tracking to follow the user's gaze and automatically adjust the focus with a tunable display. But when tech writer Karl Guttag asked Avegant about the patent, the company said it was not using that method and that the prototype had "no moving parts" or any kind of eye tracking.
How exactly the Avegant device works may be less important than whether or not anyone will buy it when it's available. The company's first device, a head-mounted cinema/headphone display called theGlyph, was launched from a Kickstarter campaign and has posted modest sales. In a video for the new light field display, Tang made the case that his company's tech is critical for the future of the AR space.
"The success of the entire mixed reality industry hinges upon having light field technologies," said Tang. "Light field is the missing link to having truly realistic and interactive mixed reality experiences."
Guttag expressed doubts that the augmented reality or mixed reality market would take off in the next five years, "with or without Avegant's technology." And Wetzstein at Stanford said that the focal issues addressed by Avegant are only one of a set of engineering and design challenges that need to be solved before AR displays have a chance of passing the visual Turing test.
A big challenge, explained Wetzstein, is the issue of "occlusion." Since AR overlays digital characters on real-world environments, the characters must interact with real objects in visually realistic ways. Wetzstein imagines a version of Super Mario Bros. that you can play on your desk.
"If Mario runs behind a physical object, you have to make sure that the physical object occludes [blocks the view of] Super Mario. Otherwise it doesn't create this consistency in the world," said Wetzstein, adding that the same is true for focal cues. "Your eyes will continually focus on either the digital content or the physical object. If those aren't consistent with one another, it doesn't provide a realistic experience."
And then there's the issue of form factor. The Avegant prototype had to be wired to a PC, greatly restricting freedom of movement. The final commercial device may be untethered, like the HoloLens, but all current VR and AR headsets still suffer from the same bulky and decidedly geeky form factor, which remains another barrier to consumer adoption.
Apple CEO Tim Cook has hinted that the design gurus behind the iPhone and iPad may be coming out with an AR device in the near future. And Magic Leap is still working on its own light field device which may very well prove a game-changer.
Until then, people hankering for an AR fix may have to settle for Pokémon Go.