It’s the weekend! Congrats on making it to Friday. Before you shut down for the next two days (and you are shutting down for some relaxation… right?), let’s review some of the big tech stories from the week that you might have missed, including the final demise of Google’s interesting VR project, as well as SpaceX’s kinda-terrible day.
Google Cardboard, We Hardly Knew Ye
A few years ago, Google released Google Cardboard, its attempt at the cheapest possible virtual reality (VR) headset. As the name implies, Google Cardboard was a VR headset made out of, well, cardboard, with a slot where you could insert your smartphone, which would act as the screen.
Google had big hopes for the effort, shipping more than 10 million Cardboard units by March 2017. In theory, that supply was supposed to ignite massive interest in VR apps. However, Cardboard had some disadvantages, such as a lack of active control: You could swing your headset around to view a virtual environment, but you couldn’t really play games or actively engage with apps.
After muddling along for a few more years, Google has made the decision to finally sunset Cardboard, which is no longer available in the Google Store. It’s also important to note that the search-engine giant has largely abandoned Daydream, its advanced VR platform that also held a lot of promise, once upon a time.
Although Google has killed its VR headsets and stripped support for VR from its Pixel phones, other tech firms are determined to conquer VR. Facebook continues to push its Oculus headsets, for example, and Apple is reportedly hard at work on a VR headset of some kind.
SpaceX Lands Its Starship… Kinda
SpaceX CEO Elon Musk has made no secret of his lifelong desire to start a colony on Mars. A big part of that quest is creating a suitable starship—a complicated endeavor, to put it mildly. This week, the company’s latest prototype, the SN10, had a successful test flight, followed by an equally successful landing on its Texas pad… only to explode moments later.
Let’s go to the tape. Yes, “Rapid Unplanned Disassembly” is a euphemism for “big explosion”:
An amazing shot of Starship SN10's post-landing Rapid Unplanned Disassembly (RUD) after Wednesday's test flight.— Chris B - NSF (@NASASpaceflight) March 3, 2021
If you’re interested in working as a technologist at SpaceX, you’ll likely face some complex, interesting challenges. SpaceX’s job postings make it clear that it’s interested in professionals skilled at debugging hardware and software—with good reason, as you can see in the above video.
Facebook Is Using Public Instagram Images to Train A.I.
If you’ve studied machine learning and artificial intelligence (A.I.), you know that it’s difficult to find a suitable database to train your algorithm. For Facebook, though, that’s not really a problem as it tries to improve its internal computer-vision tools—the company is relying on 1 billion public images from Instagram.
For this project, Facebook is engaging an A.I. learning technique called SEER (SElf-supERvised learning), which requires the system to learn using unlabeled images. That’s in stark contrast to most A.I. and machine learning projects, which rely on meticulously cleaned and labeled datasets. In theory, SEER will make an A.I. platform smarter at accurately analyzing whatever information you toss at it.
Facebook claims that SEER works wonderfully. From the company’s blog posting on the matter (which you should definitely read if you’re into A.I.):
“After pretraining on a billion random, unlabeled and uncurated public Instagram images, SEER outperformed the most advanced, state-of-the-art self-supervised systems, reaching 84.2 percent top-1 accuracy on ImageNet. SEER also outperformed state-of-the-art supervised models on downstream tasks, including low-shot, object detection, segmentation, and image classification. When trained with just 10 percent of the examples in the ImageNet data set, SEER still achieved 77.9 percent top-1 accuracy on the full data set. When trained with just 1 percent of the annotated ImageNet examples, SEER achieved 60.5 percent top-1 accuracy.”
Even if you’re not interested in A.I., just keep in mind that whatever you upload to Instagram could possibly end up used to train the next generation of A.I.
Have a great weekend, everyone! Stay safe!