Magic Leap Needs to Engineer a Miracle
To make its prototype augmented-reality goggles a product, Magic Leap will have to scale up silicon photonics—something heavyweights like Intel have struggled to do.
Technology that brings virtual objects into reality could have many applications in gaming, travel, fashion, and other industries.
Startup Magic Leap is developing a wearable display it claims will deliver on the promise of augmented reality: overlaying information or game characters on what we see naturally, without causing eye strain. Instead of transporting you to a virtual world, Magic Leap’s goggles make you see virtual objects in your own world. (See “10 Breakthrough Technologies 2015: Magic Leap.”)
But the company is still keeping its technology largely under wraps, and only a few people, including an MIT Technology Review senior editor, have seen prototypes. Company representatives and people who have experienced the prototypes sound, to the cranky and the uninitiated, like adherents of a cult that practices ritual intake of hallucinogens and whose sacred texts are Star Wars and Snow Crash (whose author, Neal Stephenson, is the company’s chief “futurist”).
Last week on the stage of MIT Technology Review’s EmTech Digital conference in San Francisco, Magic Leap executives indulged in the company’s increasingly common hand-waving, talking about “special” photons and wizard-training school. They also let slip a bit of information suggesting challenges ahead as the company moves from prototypes to products.
CEO and founder Rony Abovitz said Magic Leap is making a light-field chip that relies on silicon photonics. Interviewed by MIT Technology Review editor in chief Jason Pontin, Abovitz said that the company has developed novel fabrication techniques and is using them on a pilot manufacturing line in Florida. Abovitz also said that the company is now “out of the R&D phase and in the transition to product introduction.”
Whether Magic Leap can create that product will depend on whether it can scale up a new chip-making process for silicon photonics—something that’s a big undertaking even for semiconductor giants. The startup’s $592 million in funding is rich for an early-stage company, but it may need a lot more than that to make the leap to consumer products.
Silicon photonics is a broad term that refers to the semiconductor industry’s efforts to bring optical components into or closer to today’s silicon computer chips, which trade in electrons. Photonic components can carry more data farther and faster, without heating up and without degradation in the signal.
Integrating them with existing electronic components is proving to be an engineering challenge. In 2013, Intel announced with fanfare that it would mass-produce silicon photonics. Intel plans to make energy-efficient, high-speed optical links to ease data transfer in server farms like those found at Facebook. This February, however, Intel announced that it would delay shipment of its first silicon photonics products because of manufacturing troubles with one of the components. Making new hardware is incredibly challenging even for Intel, a company that’s synonymous with silicon.
Experts following the company say Magic Leap seems to be taking a gamble on silicon photonics because the technology would dramatically improve the augmented-reality display. Typical augmented-reality goggles use mirrors and beam splitters to reflect images from a microdisplay into the eye. These systems also let in light from the real world. They can achieve a 3-D effect by simultaneously showing slightly different images to the right and left eyes. This is called stereoscopic 3-D, and even though today it’s done with moving images produced by LCDs rather than the static photos used in the 19th century, it’s a technology with major limitations. Being confronted with left and right images that appear to be at slightly different distances can literally be a headache.
“Your brain gets conflicting views of what it’s supposed to see, and you feel sick,” says Gordon Wetzstein, a Stanford researcher who studies 3-D displays. “Magic Leap is promising to solve this problem.”
To eliminate these conflicting focus cues, Magic Leap must have figured out a way to simultaneously show not just a left and a right image but multiple images to the left eye and multiple images to the right eye, says Wetzstein. This should free the eyes to focus naturally. When company representatives talk about silicon photonics and stacked, nanostructured light-field chips, he speculates, they are probably referring to stacked silicon waveguides. Wetzstein, who has looked at Magic Leap’s patents, says these waveguides may be transparent to ambient light, like a clear glass window, while augmenting reality by transmitting complex and naturalistic virtual images to the pupils. And all the hardware that makes this possible must be relatively thin if Magic Leap is to live up to its promise of a wearable system.
Other imaging researchers think the company may be up to something even more complex. The company’s use of the terms “silicon photonics” and “nanomanufacturing,” and even of the word “chip” in describing its goggles, suggests it’s not about something passive, like a simple waveguide. Magic Leap may be using some kind of silicon-based modulators to tailor the image signal, says David Brady, a computational- imaging researcher at Duke University in Durham, North Carolina.
Those with more direct knowledge of the company’s technology will say very little. A Magic Leap spokesman declined to clarify Abovitz’s remarks, saying the company is still in stealth mode. Sanjay Banerjee, head of the Microelectronics Research Center at the University of Texas at Austin, confirmed that Magic Leap is using the center’s fabrication facilities. He would not comment on what kind of photonic devices it is building, but he did say it’s making polarizers—a component that’s common to the liquid-crystal displays used in microprojection systems like those in Google Glass. “I’m not at liberty to divulge what they’re doing but it’s very novel technology,” Banerjee says.
This technology may prove as difficult to manufacture as it is promising. “If you have to build a fab, $600 million is nothing,” says Wetzstein. He’s a fan of Magic Leap and of 3-D technology in general, but he notes that the costs of manufacturing lines dedicated to well-established technologies typically edge into the billions. Given these expensive realities, he says, “that money is going to be gone quickly.”
Keep up with the latest in intelligent machines at EmTech Digital.
The Countdown has begun.
March 25-26, 2019
San Francisco, CA