Magic Leap: When Reality Hits the Fan
kguttag.com
kguttag.com
1. Not achieve a form factor acceptable to the general consumer for a while.
2. Be more immediately useful to companies looking to improve worker productivity (like on assembly lines), where form factor is a lesser concern.
As a consequence of (1), we will probably start seeing more robust applications of AR tech using the smart phone and camera (like Pokemon GO and Snapchat filters) before we see any reasonable hardware device.
For the enterprise case, AR capabilities on the phone are a great and cheap place to start in order to get a HUD view around them. Kludgy solutions like reading QR codes also work fine for their expected use cases - i.e. reading machine states, maintenance records, etc. These solutions solve a mobile interface problem that helps align information retrieval with the context of being out the field looking at a physical object.
My personal favorite AR application would be drone operation, though. AR goggle can do two things. They can assist the Remote pilot in command in maintaining a computer aided visual line of sight to the drone. (circling it in red, in your field of view, for example). It also can provide a HUD of the drone's status (battery, altitude, attitude, settings, perceived wind speed, etc.), and finally it can allow you to view an overlay of FPV from the drone to aid in your both maintaining direct LOS to meet Part 107 requirements, and also to ensure that the pilot can be inspecting, aiming, orienting and piloting the drone successfully by relying of FPV input.
Those capabilities don't quite exist yet, but I wouldn't doubt that it will be my first AR use case.
Mostly because the numbers just are not there.
Specific industries will want specific versions, specific apps - and without the scale ... it's hard to make money.
And although it may 'seem' rational to want to production line workers into a HUD ... the actual, tangible and measurably improvement in productivity may not be that much. Today, if workers need 'data on the fly' they can ref their tablet or mobile. Does a HUD really make such a huge difference that it's worth the cost? That's a real question.
So - I think you're right, essentially about the application opportunities, but the business case is really hard for that. The numbers just don't work out very well.
Already the biggest data/reality hurdle is not visualization but record accuracy - Amazon pickers get quick-and-easy guides to where products are, but face massive delays when something is mislabeled and they have to scan every item in the bin they're directed to.
Fancy HUD systems may be worthwhile for a handful of high-end industrial tasks that value time very highly and can justify custom purchases (power plants?), but for most applications boosting accuracy and system usability sounds better than investing in new methods of data access.
Source: Early Daqri employee (when marketing apps were the thing), and industrial solutions architect at the same while they had any. (No longer work there)
There is a bejeebus amount of money in the industrial sector for these kinds of solutions, although at first glance it might not be clearly "AR", although you would use pretty much exactly the same technology.
First up - although there's not a lot of information, the information that IS present is staggeringly decisive:
https://www.youtube.com/watch?v=P-qJ6U-ixX0&feature=youtu.be... http://www.recode.net/2015/6/8/11563374/boeing-says-augmente...
TL;DR - ~70% time spend with <10% errors made on a first-time assembly task, when using AR instructions vs classic instructions. Gains are (mostly) kept even if you stop using AR on subsequent assemblies.
Second up -
My favorite anecdote from a boss, about building satellites: Once, while building this $100m+ object, they got to the end and found a left over bolt. They then had to spend a giant chunk of money taking the thing back apart, first to find where that bolt was supposed to go (and put it there), and then to prove that nothing else was missing.
If you're doing AR-instructions, not only are you (probably) not going to make that mistake in the first place (>90% error reduction), but you're going to have video records (ideally already in a reality-capture format, rather than just raw feeds) to review instead of an expensive disassembly, to prove nothing else is wrong.
This mistake was worth tens of millions, and mistakes of this cost-magnitude are so common they're part of the costing structure. You expect them.
As a counter examples to my points, we once looked at document pick-and-place for a document storage company. The value they'd get from the AR, after development and equipment costs were factored in, just didn't make sense. Another company wanted to use AR tech as a replacement for existing industrial quality checks (parts fly by on a conveyer, how do you know if they'd been made correctly?). AFAIK, what they wanted to do is still basically impossible. Finally, there was an ongoing debate as to whether it'd make sense for people building buildings to use industrial AR (everyone? only managers / shift leads / etc?). AFAIK, there are still issues with making that happen, although people are making solid progress.
Finally, you're forgetting about the value of a hands-free display; which is an AR headset minus the computer vision; if you've ever spent time in a facility (which is so totally awesome and I miss those trips) it's pretty clear why that's valuable.
So you're right about a couple of things - it's still hard, and you still have to run the numbers to see if makes sense for your situation (or which people in your situation) - but you're essentially wrong about the application opportunities, and the numbers usually work out amazingly well.
Any questions? Happy to answer what I think I'm allowed to.
Google was keen on using Google glass for all sorts of relevant things, and so far, they've failed.
In the first link, I see someone saying 'this could be huge'. Emphasis on 'could'.
The second link is similar. Some nice feasibility/test studies. Have they made a purchase order yet? No? Why not?
Is anyone using this in production yet? No?
"It's tough because industrial giants are slow moving, and very particular about certain hard things"
Yes, exactly. Slow moving, very specific - and even then, the number of such workers is not that big. If Airbus and Boeing both make orders for 50K units each, that's still not enough to make a company work - and those would be the biggest clients imaginable.
i.e. bad business. They're not going to get a $5 Billion dollar valuation by selling to these clients, at least not for a very long time.
(Intermission: Unless "they" is Chinese. While pitching the tech in front of a Chinese delegation, the audience went quite literally wild. The enthusiasm about gimmick-y tech radiated by Chinese business (wo)men is the exact opposite of the stone-cold non-interest of Germans)
Where ARA shine however is, as you predicted, industrial tech. The implementations I witnessed were used by Airbus, Siemens, Bosch, Audi et al. for exactly two purposes:
1. Assembly line education or 2. Massive scale marketing (in a physical sense) like projecting actual buildings for architects
At least this is my experience.
There's still no killer app for AR, but that may be a price point problem. If the HoloLens was cheap enough for Pokemon Go, it would sell quite well.
VR may be stuck in the FPS game niche forever.
I increasingly feel like it will take a fashion-forward approach like Apple and Snapchat have done in their respective areas to make these devices hip enough to truly gain mass market acceptance.
The wires on the current VR setups are a nightmare, as is the requirement for a $2K PC setup.
The HoloLens is great, but it doesn't really seem to be designed for gaming. Games are cool to demo in AR, but less practical I feel, given the requirement of meshing with the real world. Meanwhile, gaming in VR is taking off decently, and the FOV in the current generation of VR devices is much better than what the HoloLens has, and that's a pretty big deal for gaming.
Also, wireless is almost here for VR headsets: http://www.roadtovr.com/tpcast-wireless-vr-htc-vive-relased-...
The wires are annoying I agree, but not really a "nightmare".
That was my experience when messing with the Rift dev unit I got a couple years ago. It was more an excuse to buy a new video card than a need for a whole new computer.
It takes forever to get new HOV lanes added to big interstates - I somehow doubt that 'dedicated' lanes are going to appear faster than self-driving tech takes hold.
It also made the point that in addition to specific issues around form factor (i.e. size/weight/looks), AR in glasses brings with it the whole social acceptance factor. This can be ameliorated to some degree with decreased size and less obtrusive design but there will still be the potential for a "Glasshole" factor in a consumer environment until/unless societal expectations change.
Someone taking pictures of you and having data displayed in real-time is just not something most people would be comfortable with today.
The article was also more generally bullish on AR than VR. Which seems right. Even given another generation or two of development, VR seems like a niche for hard-core gamers and maybe a few other uses.
I used the HoloLens at a demo and it was amazing, but after 3 demos of 5 minutes each over the course of 30 minutes, my eyes felt tired for the next hour. And the limited field of view did detract from the experience.
I've been wanting a heads up display in my glasses for years ever since the first "wearable computing" wave. Eventually it will happen, but all the hardware breakthroughs are so much slower than software.
The Virtualboy was objectively, empirically utterly uncool the second it came out. There are multiple peer-reviewed papers on this. The µVb was a widely used unit for uncoolness for a while at the time.
People were getting their first taste of the 3D rendering capabilities of the Playstation and Saturn around that time. Given a choice between asking their parents to spend their hard earned money on what was basically a new monochrome vector based, single player VirtualBoy and the new, flashy 32-bit systems that supported multiplayer, it's pretty clear which systems were going to win.
Add the fact that Nintendo fans were also saving up their money for the Nintendo 64, and the Virtual Boy all of a sudden becomes a very hard sell.
I recall reading (unconfirmed) stories in the gaming press that after the Virtual Boy failed, Yokoi was basically relegated to an office where he basically did nothing.
I do wonder how many people give up on VR before the comfort level sets in.
http://webcache.googleusercontent.com/search?q=cache:oh3mPHO...
I really, really want VR to improve and become a commodity in the near future, but most people I talk to think it's dumb even after they try it. I think VR is at least 4 years away from being a serious part of gaming (and that assumes a lot of things go right), and I honestly don't know how many years before it finds other consumer applications.
They have hundreds of thousands of units out, developers of all sizes working on content, and an enthusiastic (and paying) community of users. That's a pretty good launch for a new consumer tech category.
Then the devices came out, and you hear nothing anymore. Sure, there are a bunch of people who bought it, but when none of the vendors are even releasing sales figures, you know it's disappointing. With the recent news that Oculus is closing their demo stations because no one is using them suggests that things aren't going too well.
VR is probably not as bad as 3D TV in terms of a flop, but I don't think anyone can call it a "success" unless your expectations are very different from that of most people.
https://www.cnet.com/news/vr-disappoints-at-ces-2017-lenovo-...
Selling hundreds of thousands of $800 units is $80M in revenue, not a bad start for a new technology starting out in the "consumer goods, luxury entertainment" market.
But one 'killer app' could change it all.
VR needs it's 'Halo' :)
I suppose Superhot could have sufficed without a non-VR alternative, but I rather doubt it.
The worst part? The Oculus version of the game works perfectly on the Vive through ReVive so it's not like they had to do much extra work to get the game working.
> We're also pretty comfortable with the idea that it will turn out to be a complete failure.
About Vive he said:
> Vive is the most expensive device on the market. It's barely capable of doing a marginally adequate job of delivering a VR experience.
Not sure if he's underselling, but he doesn't sound that enthusiastic about it. It sounds more like the Vive investment is an insurance policy.
Per your quote, he apparently also is open to the possibility that VR could turn out to be a dead end but that's not so much a lack of enthusiasm as not being a mindless cheerleader.
[1] http://www.polygon.com/features/2017/2/15/14616192/gabe-newe...
Something like a spreadsheet or communication would be just the thing.
Here is an alternative cached view:
http://webcache.googleusercontent.com/search?q=cache:http://...
"Abovitz says that Magic Leap's "light-field" technology essentially mimics the brain's visual-perception mechanisms to create objects and even people who look and behave just the way they would in the real world, and interact with that world seamlessly."
If Magic Leap's use of the term light-field is anything like its meaning in the literature, and it does sound like it, from one other source that I googled, then it is very much not to do with "the brain's [...] mechanisms", it's a totally geometric/optical concept that any optically equipped device/creature could enjoy. It could be tailored for the typically horizontal human eye arrangement, like some previous autostereo displays have been. That is an anatomical/geometrical shortcut, though. If there's real neuroscience-style brain science here I'd be interested to hear more about it.
http://webcache.googleusercontent.com/search?q=cache:oh3mPHO...
At that time (>2 years ago), Magic Leap seemed to have achieved the holy grail of Mixed Reality: optics that selectively block out light in real-time. Combined with demos, mockups, and sketches of real-world applications, this tech provided compelling evidence that ML was going to unlock true Mixed Reality real soon.
As erikpukinskis commented, even if Magic Leap fails to deliver, it will likely end up with an extremely valuable IP in an emerging space. Magic Leap did not invent Mixed Reality, but they filed patents on many first-order applications of the tech (there are countless others).
The most I've seen done in this space is using very very bulky electrochromatic glass of which the 'pixels' are ~32x the size of actual pixels on the AR display, and have extremely "fuzzy" borders