Paper Belt on Fire: How Renegade Investors Sparked a Revolt Against the University by Michael Gibson

Author:Michael Gibson
Language: eng
Format: epub
Publisher: Encounter Books

Recall that Austin was seventeen when he wrote the phrases “years of advanced research” and “a necessary solid technical background” and then imagine him laughing about it as though he just shredded a 360 on a ten-foot wave surfing down in La Jolla. Like anyone reading this, we wondered if Austin’s ideas were crazy, or crazy awesome. And truth be told, we couldn’t tell at first.

But after some intense due diligence—we had one of our physicist friends dig into Austin’s claims—we awarded Austin a Thiel Fellowship in the spring of 2013. His work was speculative, but not bananas, and that was good enough for us. Peter’s $100,000 grant went a long way for Austin to continue his research and development on his technology. Across our first few quarterly meetings, he thought he’d develop its use for cameras in Hollywood. The core ideas powering his prototype could be repurposed for special effects so that green screens would become a thing of the past. But by then another trend had begun to build momentum and capture the public’s imagination.

Google’s self-driving car program began in 2009, and then the buzz, the mania, and the bold pronouncements crescendoed over the decade that followed. By 2014, Austin saw an enormous gap in all the hype. Despite companies claiming they were going to deliver full self-driving cars soon—even companies as big as Google and Tesla—the inadequacy of the current technology presented tremendous obstacles to their claims. The robots driving the cars were much worse than humans.

The problem came down to perception, the ability to learn, and cost. Autonomous vehicles need eyes to see and a brain to process the information and learn from experience. Throughout the 2010s, the overwhelming faith of the industry was in the brain, not the eyes. The orthodox view was that both the algorithms for object detection (the visual cortex) and the chips running the algorithms (the neurons) would become so fast that it didn’t matter all that much what types of sensors a vehicle had. A simple video camera might do because, in the end, a system’s increasingly powerful computation would compensate for the faint and noisy signal coming in from the world. For the most part, though they had big brains, many of the early experimental cars used weak lidar systems, radar, and cameras for eyes. These old lidar systems looked like a giant spinning Kentucky Fried Chicken bucket. But even with all the training in the world—and I saw fleets of these bucket-head cars endlessly circling the streets of San Francisco as though in some robotic purgatory—the algorithms couldn’t learn to recognize the rare, but important edge cases: a plastic bag blowing in the wind, an open car door, a jaywalker looking the wrong way, a bike on a rack, a blind left turn. To add to these difficulties, the cost of the stronger fried chicken buckets was expensive, as high as $120,000 per unit.

Austin rejected the big data software-based approach and joined forces with another photonics ace, his mentor Jason Eichenholz of Open Photonics.

Download

Copyright Disclaimer:
This site does not store any files on its server. We only index and link to content provided by other sites. Please contact the content providers to delete copyright contents if any and email us, we'll remove relevant links or contents immediately.