AR will be startup dominated, VR will not | Hacker News Reader

AR will be startup dominated, VR will not

reactionwheel.net

81 pointsallenleee50 comments

This article overlooks a critical difference between these two markets: AR is a much more complex and difficult software engineering problem than VR. The foundational spatial computing technology and infrastructure required to make non-trivial AR applications work well does not exist, and the level of expertise and effort to build it is very high. The article was written a few years ago but this is broadly recognized as a significant hurdle today by companies making serious investments in AR.

If a startup ultimately wins the AR race it will be because they cornered the market in spatial computing expertise and are extremely well-funded. There is already not enough of this core expertise for the existing big tech companies with AR programs to all build viable teams, which bids up the price of that expertise, never mind enough to go around for a dozen startups.

As a consequence, I expect AR to be something like a winner-take-all market relatively early on.

The difficulties you sight are all at the low level of platform development. The company that can create a viable AR platform will likely not have the expertise to create content, so I expect to see the platform to be open to indie creators early on in its lifetime.

In fact, I can imagine a future where this company decides that any consumer facing product is not their cup of tea, and sells/licenses chips/lenses/software/patents/etc to consumer facing companies that put the final product together.

I think this is broadly correct. The open question is how much of the revenue/profit in that market can such a platform capture. There is enormous revenue potential from such a platform that doesn't involve AR per se.

there's significant movement in the platforms, fwiw. 6d is one example. https://www.6d.ai/

Agreed. I'm not 100% sure about the winner-take-all aspect, but to make a serious AR product you essentially have to do all the work required to make a serious VR product, then add a whole lot of inherently fragile complexity on top of that.

I always figured that AR had much less of a latency problem. Because you still retain the actual world as a visual the mismatch between inertial and visual movements is less bad.

This means less nausea, better balance for users, and a much less visceral impact of glitches.

I think it’s the other way, with AR your virtual objects that interact with the environment are going to float around independent of the background unless you have extremely low latency rendering. At a guess people would find this both distracting and quickly get sick.

It’s only safe if all your doing is displaying a virtual clock floating around or whatever.

Give me a floating HUD please! And I don't just want a clock, I want a calendar and a map please. And that's just the lowest common denominator stuff. I imagine different groups could use other stuff.

Maybe blood sugar for diabetes sufferers? Maybe some stuff that integrates with sports/games?

I want ad block for billboards and street ads.

Or propaganda block for countries with dictators. Usually their face is plastered everywhere to remind everyone of their benevolent master.

But... on what? Something a la google glasses?

The latency problem for AR is way more about objects feeling grounded. As soon as you get any float as the registration fails so does the illusion. Latency also makes precise placement in the real world really hard so limits applications AR will work within.

But the failure mode is different. Float breaks the illusion, but an overlay showing e.g. tagged objects, or dimensions of a box, remains intelligible even with some 'float'.

Meanwhile, latency in VR gives the feeling that 'the entire world is slightly wrong'. I imagine working despite the first problem is a lot easier than working despite the second one.

The interesting challenge AR has that VR does not is the backend infrastructure requirements. For most AR applications you are synthesizing content from relationships in myriad external sensor systems continuously in real-time. VR requires a glorified game engine to a first approximation, a well-behaved Newtonian universe.

There is a strong implication in AR applications that there exists a single logical model of physical reality behind the scenes, continuously updated from multimodal sensor and other data sources. Not only does this enable strong interoperability of individual user contexts in real-time, but giving everyone their own narrow model of physical reality will not scale in any case, the infrastructure needs to be shared. Any single user is seeing a contextually relevant and very tiny slice of that backend spatial data model.

Companies cannot handle offline spatial processing of complex multimodal sensor data models at the required scales today, platforms don't exist with the necessary capabilities and architecture. AR is trying to do it on sensor feeds in real-time in a shared, event-driven environment at the same scales. The design and implementation of any such platform will be an extremely impressive engineering feat.

Autonomous systems have a similar platform engineering problem. Think of it as machines requiring a real-time contextual view of their local physical reality synthesized from diverse sensor feeds, many of which are external. Very similar to the AR problem at a computer science level.

What about Snapchat?

Optical-based AR is still very imprecise and requires advanced long post-processing to make it somewhat accurate (~1m precision). Usually, most startups quickly bring an 80% solution (a few weeks) and then get stuck forever to make it usable/precise. During that phase they trick their VCs/investors that a solution is right around the corner, internally the ones responsible for the initial phase gain promotions and block the ones that poke at holes in their solution that could never make it to 95%. I've seen it happening first hand...

Were you in NYC by any chance? I’ve heard of something similar happening twice here now

No, this was a SF Bay experience.

Lol, this is how 90% of startups with technical risk are funded

I concur. Saw it happen first hand as well...

As the founder of a startup creating an AR headset, I think it’s true but the AR market is not as mature as VR. In every tech product, first comes the hardware and then the software adds the magic at every update. In the AR world, we are still at the hardware stage. No one has the answer today for a compelling device and experience for B2C. There is no Moore law in optics and the compute power needed for SLAM, computer vision and stereo rendering is crazy if you want an untethered device (like hololens). You need more than the traditional CPU/GPU config, you need DSPs, ASICs and perhaps FPGA. Combined with the optics your NRE is high. We are far from a market that has the same ROI as VR now, even on the entreprise the adoption is slow. This game is hard but it will be worth it, and it’s the most exciting venture I’ve been a part of.

> There is no Moore law in optics and the compute power needed for SLAM

Are you sure? My instinct is that the number of probes required to reach the same level of accuracy would be halving regularly.

What makes you think it’s not? Are there studies?

The tiniest of creatures can do it very well with very little energy.

All open source state-of-art SLAM libraries fail on real-world data, often the same scene that was processed once successfully fails in another pass. The problem is with the computer vision algorithms themselves, the "classical" ones are super susceptible to per-scene constants tuning and randomization effect; the only hope IMO is Deep Learning at the moment, but that requires massive computational capability for real-time inference.

Yes, the best SLAM we have today based on visual inertial data (Camera+IMU) uses 500mW of power. It uses DSPs and ASICs. Also, the custom silicon team from MSFT did a great job for their SLAM and display engine, it's built on ASICs in what's called the HPU for Hololens: https://youtu.be/u0eBd2m_wEs?t=1641

I made my first AR demo for an expo 10 years ago with ARToolKit and IMO AR still hasn't gone beyond a gimmick. There are some popular examples like Pokemon Go but AR is not going anywhere else unless someone makes an affordable, comfortable, and capable AR headset/glasses. Holding a device in front of you gets old pretty quickly.

Some assumptions made in this article that are questionable:

- The dominant use of VR will be content consumption (I disagree personally, and think it will be communication and collaboration)

- making content for VR is expensive and difficult. (This has been trending downwards over time — game engine tech is nearly commoditized, creation tools are cheap or free, learning resources to create 3D assets and games are plentiful, and content libraries are cheap and free. VR content tends to be short form as well, so less need to have massive amounts of custom content.)

I also disagree about VR being mostly content consumption, but I imagine more along the lines of creation and problem solving. Imagine iron man like 3D modeling as an example.

But the collaboration angle is something I'll look into deeper now. You are likely right.

The current biggest holdup other than cost of hardware is that it's hard on your eyes, which I think is why the short form is so prevalent.

Do you see a market for VR tourism? There's some capturing stuff I'm working on without seeing a clear market (at the inherent technical/labor costs incurred during the capture and compute during post processing).

The economics always bite you in the end.

>8k VR + telepresence drones that you can tap into anywhere in the world would be sweet.

A lot of AR applications depend upon centralized map and street level info, which is susceptible to information siloing.

On conversely:

- making content for AR is inexpensive and easy

>>> The dominant use of VR will be content consumption (I disagree personally, and think it will be communication and collaboration)

Maybe not the dominant use, but certainly the dominant market revenue wise. Think of the movie and TV industry vs security cameras and employee training videos. I think it's naive to think that any new form of media won't be dominated by entertainment uses in the public eye.

I noticed the quest having inside-out tracking with cameras...point being that VR and AR are fuzzing a bit and it just might converge at some point as something that makes sense where you simply have more opacity or not depending on your app need. Then this type of argument is moot when say "oculus slim" comes out that looks like ready player one and is cool enough to wear everywhere and has a more contextual app model. (but i still think nueralink approach is the real future)

Yes, but these startups will be working for a platform which will take a percentage of the profit and will have full control over how apps are ranked etc. Participating is much like gambling, except for the bigger players (which usually are not startups).

Instead there should be an open standard, e.g. like VHS used to be (they probably took a small flat fee but at least they were not in control of the entire market).

This just sounds like its written in a manic flurry of excitement for AR. AR is so amazing! There are so many use cases that content will be easy and cheap to make!

The reality is not so simple.

If we're talking about apples to apples entertainment type content, its probably easier to make an indie VR game that runs on all platforms than an AR game or experience.

If we're talking about the computer vision aspects of AR like object tracking, those problems are hard enough that a creative novice has little hope of making progress at the moment. That said, this space is probably dominated by a bunch of tech heavy startups.

And then of course the hardware is dominated by a few players but with some steam's cross platform SDK it was more than I thought it would be in 2016.

mods please add (2016) to the title

Yes. This was written before the VR boom collapsed. "Peak VR" was holiday season 2017.

The most successful content in VR is rather simple. See "Beat Saber", the most popular VR game.[1] That's from Hyperbolic Magnetism, an indy game company in the Czech Republic.

On the other hand, "Facebook Spaces" is a total flop. You can browse Facebook in VR. Nobody wants to.

So the author was totally wrong about VR being dominated by big companies.

[1] https://www.youtube.com/watch?v=gV1sw4lfwFw

Did VR collapse? From all accounts I've heard, the Quest is incredibly successful.

They're "selling as fast as they are being made", but Facebook isn't releasing the numbers, so it's still impossible to say. The big question is whether there will be an ongoing demand as the months go by, or whether initial sales were all from a burst of interest that simply hasn't been satisfied yet by the number that have been made.

Gotta wonder why Facebook doesn't just buy one of the popular social VR apps like VR Chat.

Because VR Chat has about 5,000 users. Which is too tiny for Facebook.

I think this article makes very good points, it's very similar to something I've thought about VR vs AR and the internet analogy - which is that something that made (and still makes) the internet very approachable for indie hackers and startups is how simple the UI and UX are (or can be).

As a single person, with the right set of skills and enough care and attention, you can build a product on the web that looks and feels just as good, better even, than the big players. You're at no major disadvantage.

The same feels at least partially true for AR. It seems like AR UI is likely, at least in the near term, going to look like fairly simple overlay of images, boxes and text over physical objects. Not too different from a web page, basically. In fact I can even imagine that you could build such a UI with a markup language not dissimilar to HTML. You could even use CSS mostly as-is to style the components. Anyone who's capable of building a web page is going to be able to get going with this and build something useful really quickly.

VR feels obviously different in this regard. Simplistic UI is not going to look good or have good UX in VR, just as it doesn't in a 3D game. A solo engineer is not going to be able to compete with big company teams of artists, designers, etc to craft good looking VR spaces with comparable levels of quality.

I think the best we can hope for in VR is some kind of generic VR world toolkit, that's open source, within which one can overlay one's own UI using pre-designed models etc in a mostly cut and paste fashion (so kind of like how the browser gives you the DOM, and you just define what you want to display with simple markup instead of having to literally render pixel by pixel) - then it might be a bit more possible for solo engineers or small teams to actually build good looking and useful products. But still, it's a stretch to imagine that being competitive with custom designs that bigger teams and companies are capable of producing.

I guess we'll see. I like to be optimistic because usually with software and open source people find ways to surprise you with ingenuity and make things possible that you didn't even think of, and come up with usecases and products that you didn't think of in advance.

I think it depends on what VR ends up looking like. If it turns out to be just 3d immersive worlds, it will probably be dominated by the big players for the foreseeable future. We will probably have building blocks that individuals can use to make something acceptable in that catagory, but the fundamantal value proposistion of that seems to me to be the visual wow factor, which is already dominated by the big players and really does require a large investment of customized art.

On the other hand, I also see a use-case of "VR" as more of a light canceling monitor, to complement current noise canceling headphones. Think about being able to sit down on an airplane and pop on your VR goggles and a pair of noise cancelling headphones to watch an old-fashion 2D movie. Or to have a full giant-display that can replace your 3 monitor desk.

These types of VR as a monitor use-cases shouldn't be more difficult then their traditional counterparts (although the tooling and ecosystem isn't as good on VR yet), so there should be room for solo developers to compete.

That's a good example of a VR usecase I hadn't considered, but to be fair this kind of thing perhaps shouldn't be considered VR at all from the app developer's perspective. By which I mean, the developer wouldn't ever have to know or care that a user would use their regular 2D app in this VR setting, it should just work like any other monitor / web browser etc.

I was more talking about applications that are 'native' VR, for want of a better term. Apps that couldn't exist or wouldn't make sense in other formats than VR.

I certainly agree that this usecase isn't really "virtual reality" in concept, but it is "VR" as far as platforms go, which I think is what is relevent here.

I think that, at least in the near future, developers would need to be aware of the VR platform when working on such apps. Ignore toolchain and issues, developers would still need to design a VR first UI. There is probably a lot of room to innovate around how to take advantage of head gestures; how to take advantage of virtual vs absolute coordinate systems [0]. And, the more mundane fact that VR developers have far more screen real estate to work with then desktop developers, who typically assume that the entire monitor is visible at the same time.

[0] I'm not sure what the correct names for this are. Bassically, the fact that there can be an image that is always in the same spot in your vision as you move your head, as well as images that appear to be "static" and therefore whose location in your vision changes as you turn your head.

You're right, of course. In the same way that with mobile it's not just making the content fit into the screen but making sure it plays nice UX-wise with a touch interface.

You have convinced me - this monitors-in-VR thing could be an exciting area for development, and is definitely open to individuals and startups with little capital. Good shout.

What you're describing for VR already exists in numerous different platforms, many of which are "free" (e.g. - free until 6-7 figures of profit). My preferred would be the Unreal Engine which is also completely open source on top of having hoards of free content available to get started from nothing. The same is also true of Unity.

In my opinion getting a 'sound' VR project up and running is far more straight forward learning process (if not process) than getting e.g. a 'sound' single page application up and running.

> In my opinion getting a 'sound' VR project up and running is far more straight forward learning process (if not process) than getting e.g. a 'sound' single page application up and running.

I'm very glad to hear that, as someone with no experience of game / 3D graphics development I had honestly assumed it would be more difficult and/or involve many more people to build something basic. Really cool to hear that's not the case these days.

VR will be game studio dominated. AR is more appealing to startups because it's the ultimate answer to the question: How do we get more ads in front of more viewers? Simple: Shoot the ads straight into their eyeballs as they go about their day!

[2016]

Reply on news.ycombinator.com