Saying “Moore’s Law or Apple” Does Not Make AR a Consumer Product

First, a Request for Input on Apple’s AR Effort

Before I get started on the article (although this is related), I would like to make a request for help in identifying what Apple is doing in AR. Apple has R&D groups working on all matter of things producing many patents about things Apple will never build. I’m hoping the use this information to give some technical analysis about what Apple will be doing in AR.

Personally, I think the reports like Bloomberg’s, “Apple’s Smart Glasses Could Make 2020 the Year of AR” are wildly optimistic, at least in terms of being what most people would consider AR glasses. As reported in CNBC and many other places, “Apple analyst Ming-Chi Kuo said Apple is planning a bunch of new hardware for the first half of 2020, including new augmented reality glasses that Apple will build with third-party brands” (my bold emphasis).

The “third-party brand” part of Kuo’s statement seems often overlooked or ignored, but this seems to suggest that Apple will be simply working with other companies making AR headsets. This frankly makes more sense than it being a full-fledged Apple device. It also means that it may be much less than people are wishing it to be. It could be something more like the Bose “Audio AR Glasses” or releasing an existing design by one of the known players in AR.

In spite of being “Apple,” they are still bound by the laws of physics. Yes, Apple has many smart people and a lot of money, but so do Microsoft, Google, and Amazon, not to mention unicorn startup Magic Leap.

Let me be clear, I’m NOT looking for confidential material but rather any information that you have found that may narrow my investigation. If you can, please leave what you think Apple will be doing in AR in 2020 in the comments section.

Introduction

Ok, this blog entry is, in part, a rant. As an engineer, it irritates me when I try to explain to “enthusiast” the technical issues related to designing an AR headset, only to have them chant, “its Moore’s Law” or “Apple/Steve Jobs” like just saying this is the answer to any technical problem.

If you ask them how they are going to solve a specific problem, they just repeat Moore’s Law or Apple can do anything chant. Sidney Harris back in the 1960s drew a cartoon that I call, “Insert Miracle Here” or more recently, a “Moore’s Law and Steve Jobs” argument.

The Possible with Effort and the Impossible

If you have to violate even one law of physics, it is never going to work. Some of the challenges AR faces are just a matter of smart people, time, and money. Some things people wish for AR likely require violating the laws of physics.

AR headsets are both figuratively and literally in a tie box. We have to fit the display and optic around a human and get the image to a place where the human can see it without blocking out the real-world. We are stuck with the wavelengths of light that humans can see. Often the pixels themselves are on the order of only 10 wavelengths of light across and sometimes less.

Factors like diffraction and etendue which are non-issues for larger displays, become major factors AR optics.

“No Moore’s Law for Optics”

I think even the most ardent evangelists for AR are being told regularly that Moore’s Law is the solution to all of AR’s problems. At some point even they are in effect saying, “enough with Moore’s law and optics.” Below are quotes from Michael Abash, Oculus Chief Scientist and Bernard Kress, Partner Optical Architect on Microsoft’s Hololens 1&2, about Moore’s Law concerning optics.

Abrash in his 2017 blog post Oculus Blog Inventing the Future wrote “there’s no Moores Law for Optics, batteries, weight, or thermal dissipation.” Kress, in Optics and Photonics news Sept, 2019 article “No Moores’ Law for Optics,” is reported to have said that “Moore’s law won’t hold you back—but the laws of etendue will.”

Abrash’s Blog Post (with my bold emphasis) additionally says, “The honest truth is that the laws of physics may make it impossible to ever build true all-day AR glasses; there’s no Moore’s Law for optics, batteries, weight, or thermal dissipation. My guess is that it is in fact possible (obviously, or I wouldn’t be trying to make it happen), and if it is possible, I think it’s highly likely that all-day AR glasses will happen within the next ten years, but it is an astonishingly difficult technical challenge on half a dozen axes, and a host of breakthroughs are going to be needed.

Every Technology Has a Learning Curve, But Some Are Slower than Others.

Moores Law is a special case of “The Learning Curve” (also known as the “Experience Curve). The basic premise of a learning curve is that for every doubling of volume, the cost should reduce by some factor. In the early days of a product, it is relatively easy to double the volume, but as the product matures, it becomes ever more difficult to double the number of units.

As any engineer or physicist should know, things in the real world are more apt to follow natural law (also known as an S-curve) than exponential (Moore’s law) growth. For shorter periods, exponential growth and natural growth look identical but as time goes on, exponential curves grow to infinity were natural curve roll-over. It has been said that Moores law has been kept going by a series of breakthroughs with their own S-curves, but even these are running out as transistor sizes start being measured in terms of how many atoms they are across.  

In reality, learning curves move by fits and starts as people make a series of breakthroughs and improvements. The improvements are also only made when part of the total effort included R&D and not just manufacturing. Early on, a relatively small effort can make significant improvements, but as time goes on it will take ever-larger expenditures to improve by the same percentage.

One of the most famous historical examples of the learning curve is the Ford Model T and how it cost dropped from $900 in 1910 to $260 (not adjusted for inflation) by 1925. There were many different improvements in efficiency along the way, including the mining of raw materials in large quantities and the use of new materials. The cost improvements in the Model T could not go on forever, and Ford’s emphasis on cost above all else led to him falling behind General Motors (maybe a lesson).

When people pick the learning curve(s) of semiconductors as in Moore’s Law, they are choosing a spectacular lottery ticket winner in terms of its rate of improvement. Other technologies also have learning curves, but they are slower and may be nearing the limits of physics. In the case of optics, they have been evolving since before the days of Galileo and his 1610 telescope, we are pretty far up the S-curve.

Addendum: The End of Moore’s “Law”

For 20 years, I designed CPUs, graphics Processors, and advanced DRAMs (including SDRAM) for Texas Instruments. I lived “Moores Law,” or more accurately as I think it was meant as a bit of a joke to call it a “law,” it is Moore’s “observation.” Someone as intelligent as Gordon Moore had to know that compounded growth of anything in physics cannot go on forever.

In 2009, Andrew Curry wrote an excellent article on the end of Moore’s Law. In this article, he used a figure from the Computer Measurement Group that gives a comparison between Moore’s Law and Natural Growth. Semiconductors have had a spectacular run, but even it is starting to succumb to the impossibility of exponentially growing forever. It has continued to improve, but just not exponentially.

Default image
Karl Guttag

14 Comments

  1. Hi Karl,
    I heard Apple is working on two devices. One of them is a pair of glasses with see-through AR, similar to the Nreal Light, and the second device is a “mixed reality visor” similar to the Varjo XR-1 (https://varjo.com/products/xr-1/) that connects to an Apple TV-like box wirelessly (WiGig?) and does pass-through AR (in addition to VR). This second device is the one that is going to feature the 8K foveated displays and sensors to track hands and facial expressions and it is an evolution of the totem from https://www.vrvana.com/#specifications. The glasses will require an iPhone to keep you tied/hostage to the Apple ecosystem, like the Apple Watch does, while the MR visor will be an indepent device on its own–like an iPad–and a new display to consume Apple Arcade and Apple TV+. I’ve been told monitor and TV manufacturers are really afraid Apple could wipe them out in a single move with this visor becoming “the display to rule them all.” Hope my comment helps. Keep up the great work!

  2. “…there’s no Moores Law for Optics, batteries, weight, or thermal dissipation.”

    Abrash is dead wrong.

    Moore’s Law is working for AR. Not in the manner he and the AR industry erroneously think it should be. But it in Edge Computing and 5G. This eliminates the need to put compute/rendering hardware on the device which creates all the problems Abrash mentions. But 5G will connect light weight smartglasses to servers which thrive on Moore’s Law and will stream content with low latency.

    The innovation will not come from Apple or similar slow moving giants. But from smaller innovative companies.

    • I like the way you think, Sanjit. Offloading functions and resources to accomplish them from the head worn hardware in the first place is going to be key to get where we all want to go, including having a new short distance wireless fat pipe from the phone to the headgear so we can leverage the phone resources a lot more.

    • 5G+edge could limit the electronic complexity of the headset, even though the radio is still likely to be expensive and high power consuming. In fact there has been a lot of work on short-range wireless links for VR headsets, something like WiGig is good on paper, but not realized for the mainstream.

      5G+edge does nothing for the optical problems that this blog is mainly concerned with. Not like the other things don’t matter (the Magic Leap SDK is not that bad) but if you don’t have good enough optics at a good enough price, you are stuck.

      I am unsure about the Edge. I am about 30ms in latency away from AMZN’s us-east-2 data center. That is practically across the street from a competitive Azure data center, and I can almost get any kind of resource I want there whenever I want it. I can set up a new server quickly because I have the whole resources of the data center physically close to each other.

      Maybe I can get 5ms of latency to a server which is located at the nearby tower. Instead of paying AWS prices (which startup people I know always complain about) you will be paying Verizon prices. You aren’t going to have as good of a choice of servers and if they are all busy at your tower, you are out of luck.

      Edge also defeats the purpose of mobility because when you move you will either face increased latency, service dropouts, long times to migrate servers, etc. So when you add it all up you might take the current cloud, maybe regionalized, where you can get consistent and fairly low latency as opposed to something a lot more expensive that has very low latency sometimes but various operational problems in return.

    • 5G may help with computing someday and this would help with the computing part and the battery power and weight associated with it. I’m concerned about how well in practice 5G will be able to compute and return information in real-time and how to deal with the lag (requiring significant compensation locally). Is there going to be a supercomputer just sitting there waiting for users in every base station?

      Also, how does 5G improve Optics? And there will still be problems associated with the display’s power and heat dissipation.

  3. This article resonates with men. I’m an optimist but the points you call out have jiggled at me for some time. Particularly the point that Microsoft is not going to release a consumer play anytime soon and they get this space. I think 5g and edge computing will disrupt a lot of the issues mentioned but we are still stuck with the physics of optics and wave guide displays. I don’t know enough about this but i’m leaning on other articles I’ve read. Snap CEO also reckons we are 10 years away from mass adoption all day wear. Apple do have the power to innovate but they had this advantage at the iPhone level. Have they really outpaced any other big player since the market caught up, that’s not as convincing.

    • Thanks, Yes I know Apple bought Akoni and that is one of the avenues of interest to consider.

      • Hello Karl, I can’t say much, but Akonia tech has some flaws. Their waveguides only works with a diode and not with a laser. Thus the image has holes and is not great at all, at least when I checked in 2017…

  4. Not quite sure about AR in a consumer context, but the guys over at inspectAR.com have built an actually useful AR application for interacting with circuit boards… I bet Apple themselves would be interested.

  5. Thanks for the link and the recommendation.
    An update by Mitchell Waldrop in Nature in 2016 confirmed that the end of Moore’s Law was nigh.
    https://www.nature.com/news/the-chips-are-down-for-moore-s-law-1.19338

  6. […] Just a quick update to my last article this time. I found another “No Moore’s Law” for optics reference that was the subject of my last article, “Saying “Moore’s Law or Apple” Does Not Make AR a Consumer Product.” […]

Leave a Reply

Physical Address

304 North Cardinal St.
Dorchester Center, MA 02124

%d bloggers like this: