John Carmack has a reputation for frank, no holds barred technical discussions in his publish speaking, but he’s also known for his generosity with his time and seemingly endless patience. At last year’s Oculus Connect, the programming legend who co-founded id, the company behind Doom, was mobbed wherever he went with impromptu Q&A sessions developing in foyers and corridors. This is a video of just such an occasion, at Oculus Connect 2015.
It seems this year will be no different as, in this video, Carmack answers questions on a variety of different topics to a number of assembled developers. The subjects range quite widely, from the latest in vision correcting computer vision to the fractious relationship between Facebook and Google. I’ll dive into his discussion of mobile positional tracking below as it’s probably the hottest topic, but you should watch the entire video for many more interesting subjects.
Carmack on Gear VR Positional Tracking
Yesterday Oculus Founder Palmer Luckey quelled expectations that had been building around new Samsung mobile hardware (including possible depth sensor cameras) that may open the door to an inside-out (i.e. information gathered from the camera is used to extrapolate head position) positional tracking solution, currently lacking on Gear VR.
When asked about this, Carmack responded “I spent a while on that and I’m confident I can do a good job with stereo cameras, so we’re left with the problem there of; if you put them in the headset then either you need to have processing in the headset and build a whole ‘nother system there, or you need to push all the data through USB3 to the phone which is going to take up a lot of power … so it does not look good for a system that doesn’t use a whole lot of battery power,”.
That’s a little disappointing to hear, but the technical imitations are clear. However, rather than those technical limitations being related to the emerging and hugely complex world of computer vision processing, it’s down to frustratingly fundamental limitations inherent in mobile technology and the platform itself such as power draw and connectivity bandwidth.
Carmack continues to elucidate on a possible technical solution, but what’s interesting is to hear him state that stereo cameras are possibly a ‘good enough’ data source for extrapolating head position for tracking of this nature. Carmack goes on to vent his frustrations that more time and expertise at Oculus hadn’t been burned on solving the problem, “It bugs me a little bit, we have like 30 computer vision experts at Oculus form the different companies we’ve acquired and none of them just wanna go solve this problem, they’re all interested in their esoteric kinda researchy things while this is a problem I want solved right now.” he says “I wish someone had spent all last year on it.” Camrack thinks it’s solvable, but he says that “I still don’t have a ton of support within the company.”
Carmack on Facebook and Google Relationship
One interesting thread spurred from a discussion over 360 video and content production started with some honest assertions from Carmack “Right now producing VR content is terrible! It’s just a lousy experience, we can’t even tell people ‘Just go buy this camera’, it’s like ‘buy some crazy multi go-pro rig’ and that’s no way to be doing creative work.” But the subject of Facebook’s scaled up cloud-based 360 video technology brought about an interesting tangent shedding light on the relationship between Facebook (who of course owns Oculus) and Google, a major potential competitor in the immersive technology space.
This was just the first day at Oculus Connect 2 and it’ll be interesting to see if subjects touched upon in this impromptu, informal discussion make it into Carmack’s formal keynote later today. Certainly, last year’s talk was a must-see for anyone with interest in VR.
Thanks to the folks over at UploadVR for upping the video and for asking the asking the right questions.