"There's a saying in technology: It's often easier to predict what the world is going to be like 20 years from now, than it is to predict what the world is going to be like three years from now". That was a comment from Facebook's Mark Zuckerberg in the opening keynote at Oculus Connect 3, but Michael Abrash, Oculus chief scientist, was still there to give it a try.
When Michael Abrash talks about the more technical side of VR, people tend to listen. Before jumping to Oculus, he was Valve's VR specialist, and has worked on the underlying tech for many games, including Quake. Then there's the fact that his previous predictions on VR have been pretty much spot on. It's easy to get caught up in the hype of virtual reality and believe we're all going to be living in the holodeck before long. Abrash is here to give us a reality check.
But there's still plenty to be excited about, and if he's right, 2021 will make the current generation of headsets look Stone Age anyway.
"Talking about this in public wasn't an easy decision to make," he said at Oculus Connect 3. "However all of you are working on VR right now, out of faith that it will become very cool and important."
He wasn't going to talk about the games or experiences we'll be playing - just the technology that will shape them. So here's what he thinks...
Your field of view will improve, but it won't be 20/20 vision
Right now, VR is delivering around 15 pixels per degree, with a 90-degree field of view and fixed 2m depth of focus. But humans are capable of about 120 pixel per degree, between 220 and 230 degrees field of view, and, of course, a variable focus. "There's a long way to go," said Abrash. "It's certainly going to take a while to get there."
"I think we'll be around 4K x 4K per eye in five years," he predicted. Abrash added that we'll probably either chase better pixel density or a wider field of view. "It's my guess that a wider field of view will be very compelling, greatly increasing presence, and that VR will head towards the widest possible virtual image.
"Given that, I am predicting a 140-degree field of view resulting in approximately 30 pixels per degree. Not as sharp as 20/20 vision, but good enough to pass a drivers' licence test".
Sounds good to us. How do we get there?
"A wider field of view with higher resolution will require a breakthrough in optics. Fresnel lenses, of the sort currently used in the Rift, have fundamental limitations with respect to image quality, and both fresnels and normal fish eye lenses can't get much past 100 degrees without unacceptable distortion. So new optics technology will be required. I don't know what that enabling tech will be, but I am confident we will find a way to break past well past 100 degrees."
Eye tracking will improve because it's vital, but it might not be perfect
One of the most noticeable problems with virtual reality right now is focus. While so many of the virtual worlds we've explored have been rich in detail and character, currently VR headsets can't account for the way the human eye change shape to focus when looking at different distances - the result usually being blurry in VR.
As resolution increases in headsets, blurring will become more evident, says Abrash. But he also thinks the problem is fixable, even if it's not perfect by 2021. "I think one way or another VR will have good support for depth of focus in five years".
One answer to this is a technique called foveated rendering, which mimics human vision by only focusing on the small part of the image it's looking at. In VR, it would mean only rendering the pixels in that spot - maybe a tenth of what is currently rendered.
"Getting to extremely accurate, completely robust eye tracking may only require gathering a lot more data and doing a lot more engineering, or it may require real research and new tech," said Abrash. "But either way, great eye tracking is so central to the future of VR, I believe it will be solved five years from now. Although I have to admit it is the greatest single risk factor for my predictions."
A limited number of experiences will sound amazing, but only some
"You'll be able to quickly and easily generate a personalised head related transfer function - or HRTF - in the comfort of your own home".
HRTF describes how sound bounces off and is distorted by objects in the room around you. "Virtual rooms will feel much more convincing, even though you may not consciously know why," said Abrash.
Read next: All the big reveals from Oculus Connect 3
"While the theory behind audio propagation is well understood, the computational demands of working implementations are so high that only certain sorts of constrained virtual spaces, with limited movement of sound sources and listeners, will be practical," he said of his 2021 prediction.
"But those instances will be highly compelling and will point the way to steadily improving audio in the future".
Physical controllers will continue to rule
"It's quite possible that the Touch-like controllers could be the mouse of VR and still be the primary interaction technology 40 years from now," Abrash boldly stated.
He doesn't think we'll be able to use our hands as truly free manipulators in VR by 2021. "I don't see that happening in the next five years, because it involves haptic and kinematic technology that isn't even on the distant horizon today."
He does, however, believe that hand tracking and rendering will become a standard part of VR in the next five years, as we're starting to see with the HTC Vive and Oculus Touch controls.
"In five years we'll see good avatar hand tracking and gesture-based simple interface control, but Touch-like controls will still be the dominant mode for sophisticated VR interactions."
Oh, and he says Oculus isn't working on a real holodeck. Sorry.
Untethering from PCs won't be easy
At Connect, Facebook and Oculus announced they are working on a standalone headset that would bridge the gap between high-end and mobile VR, but Abrash thinks some important steps need to be taken before PC VR goes wire-free. "PC will provide the most sophisticated VR experience for a long time," he said.
Abrash believes that headsets in the next five years will progressively come with reduced weight, offer better weight distribution, and better handling of prescription correction. He does think we'll see more wireless headsets over the next five years, but getting high-end headsets away from the PC won't happen soon.
"The challenge here is developing a wireless link with enough bandwidth to meet the needs of VR. There's no existing consumer electronics link that's up to the task at current resolutions, let alone at the 4K x 4K resolution I expect in five years."
This is another reason why, again, Abrash believes foveated rendering will be so important.
No, not augmented reality - augmented virtual reality. What Abrash referred to wasn't like having glasses that overlay information onto the real world, but technology that mixes the two to the point where any part could be virtual or real without you really knowing.
Reconstructing the real world is challenging but doable. But as Abrash says, "Doing that with a consumer device in real time is another matter entirely, and yet that is what is needed to make augmented VR really useful."
You could, for example, be sat in a living room and completely change the colour of the walls convincingly, or conjure another sofa that's indistinguishable from the real one.
"My prediction is that five years from now augmented VR will be an integral part of virtual reality, and that it will transform VR into something that will be used for longer and for many more things than it can be today.
"While there are many unsolved problems, and a lot of research and engineering still needs to be done, augmented VR is so important that I'm confident the obstacles will be overcome and the boundary between virtual reality and real reality will progressively blur over the next five years."
So, it might not be entirely solved in five years time, but it will be well on its way.
Human uncanniness will be the toughest nut to crack
Abrash thinks humans and their thirst for social interactions will be what drives the more social elements of VR. He believes we'll see a number of systems providing a limited experience of being with another human.
"Other people are the most interesting thing in the world to most of us, and I believe that the development of virtual humans is going to be the single most important factor in making VR a part of our everyday lives, thanks to the social interaction it will enable."
But also the most challenging, he added. People are physically complex and not rigid; nuances in expressions are incredibly hard to replicate in a virtual environment.
"I think this area's so hard that in five years from now virtual humans will be widely used for social interaction and highly entertaining but will not be convincingly human, and the breakthrough will be yet to come."
So there you have it, Michael Abrash's predictions for 2021. We'll come back in five years time to see how well he did. We suggest you bookmark this page and come back and join us then too.
How we test