In mid-September Meta (FB) announced the release of its first commercial Ray-Ban AR glasses that contained an integrated display.  The company has shown a number of prototypes of its more advanced ‘Orion’ AR glasses but continues to develop and refine those prototypes before releasing them to consumers.  Meta took a different path for the design of the Ray-Ban Display Smart Glasses, diverging from the typical  display and optical track that most AR developers pursue, for what we believe are two reasons, cost and practicality..
Cost and Practicality: Comparing Diffractive vs. Geometric Waveguides for Augmented Reality:
When evaluating smart glasses and augmented reality (AR) wearables for consumer, applications, the type of display and waveguide chosen directly influences cost, design flexibility, and practicality. While this might seem an overly technical focus from an investor standpoint, we believe Meta’s choices here reflect a conscious focus on avoiding the common mistake made by many CE firms, that of overengineering products to prove their expertise in a particular technology or design category.  Apple’s (AAPL) Vision Pro is the perfect example of such, offering exemplary hardware, but over-promising on software, all at a price that is far above what the average consumer might pay especially for the first iteration of a new product category.
Given that Meta’s general philosophy has been to dominate the AR/VR/XR category by providing consumer oriented devices at a reasonable price, we expect that idiom carried through to the RBD (Ray-Ban Display) project by focusing on creating a device that balances mature technology against practical technology.  Simply, Meta used a display technology (LCoS) that is mature and cost effective to offset the cost and complexity of a sophisticated waveguide system that avoids some of the issues that befall other AR devices. Here is that balance:
AR/VR/XR devices typically try to use the latest display technology to generate a high quality image that displays the most realistic image possible.  This is essential in VR and XR, where a totally immersive experience is the goal, but AR is a different animal and does not require the high resolution rates that VR does.  Meta chose a 600x600 LCoS (Liquid Crystal on Silicon)  as an illumination source.  LCoS has been available for many years and was the light source for the Google (GOOG) Glass series released in 2013, so it has become a mature and cost-effective near-eye display system.  LCoS is exceptionally small, a very necessary feature for AR glasses which must be light and unobtrusive.  LCoS is bright as it is driven by LEDs, necessary as AR images often compete with bright ambient daylight.  LCoS is inexpensive relative to most other near-eye display technologies. 
Cost Comparison of AR Display Systems
​

LCoS is a sequential system.  Using three (Red, green, blue) LEDs, the system turns on the red LED and the liquid crystal acts as a gate (modulator) turning each pixel in the frame on or off depending on the image.  When each pixel in the frame is completed, the Red LED is turned off and the green LED is turned on.  In the same way, the liquid crystal turns each pixel in the frame on or off.  This is followed by the blue LED.  As this process happens very quickly (~3.7 milliseconds for each color), the user sees a blend of colors for each pixel, as shown below.    Note that the liquid crystal allows for a number of variations between on and off (between 256 and 1024 for each color) so the number of colors that LCoS can generate is similar to other display modalities, between 16.7 million and 1 billion.
The LCoS system is more complex than we illustrate here, but the image it generates must be directed to the viewer’s eye and that is another characteristic that sets the Meta Ray-Ban Display glasses apart from others.  Waveguides allow the image to move from the display or projector to the user’s eye through optical means, however the typical AR choice is a diffractive waveguide system, while the Ray-Ban Display Smart Glasses use a geometric waveguide.  Here are the details:
What Are AR Waveguides?
AR waveguides are specialized optical components that control how digital images are projected into the user’s eyes while maintaining see-through transparency. This approach allows users to interact with both the digital and physical worlds, forming the backbone of leading AR and XR glasses. Waveguide technology directly affects field of view (FoV), image brightness, weight, and overall device comfort, all of which are key features that drive consumer adoption and competitive advantage in the smart glasses space.
Diffractive Waveguides: Cost-Effective for Mass Market
Diffractive waveguides utilize holographic gratings or nano-imprinted patterns etched onto glass or polymer substrates. These patterns bend light to precise angles, creating a virtual display effect with minimal material. Thanks to scalable lithography and nano-imprint processes, diffractive waveguides now offer some of the lowest per-unit costs in the AR optics field once initial R&D and master tool investments are amortized. High-volume manufacturing enables AR brands to deliver lightweight, thin-profile glasses at prices suitable for mass-market consumers and logistics workers.
However, the practicality of diffractive waveguides is tempered by technical challenges. As diffractive structures lose a significant percentage of light, manufacturers must use extremely bright micro displays, commonly MicroLED or LCOS. While affordable, these displays can drive up device power requirements and impact battery life, and  diffractive waveguides are known to produce color dispersion artifacts, sometimes termed the “rainbow effect” adding to technical challenges that can distract consumers.  Most diffractive waveguide-based AR glasses today achieve a field of view between 40°–50°, which is enough for notifications, light navigation, and basic media playback but may fall short for immersive enterprise or gaming scenarios.
Geometric Waveguides: High-Quality but Costly
Geometric waveguides use carefully crafted multi-layered glass structures with partial and total internal reflection to inject and guide images from micro displays directly to the wearer’s eyes. This approach minimizes light loss, creating higher brightness, sharper color reproduction, and a larger field of view, often 60°–70° or greater, which is ideal for immersive training, mixed reality, medical overlays, and advanced consumer entertainment.
The practical downside of geometric waveguides stems from their manufacturing complexity. Precision glass bonding and multi-layer assembly require advanced facilities, and have higher raw material costs, along with some significant yield challenges. Defect rates are higher, especially as manufacturers push for ultra-wide FoV (Field of View) designs. As a result, geometric waveguides often cost 30–50% more per unit than diffractive counterparts and are most common in premium AR headsets, industrial applications, and military devices where fidelity cannot be compromised. The slightly thicker profile can be less comfortable for all-day consumer use, but users benefit from lower power requirements and longer battery life thanks to higher optical efficiency.
Meta Ray-Ban Display Smart Glasses – Decisions
The two choices that Meta made concerning the  display and the optics for the Ray-Ban Display smart glasses indicate a deep understanding of the current AR customer base and the potential future for AR.  There seems to be little or no thought of ignoring costs and using both the highest cost and quality display with the highest cost and quality optics.  Rather, a balance between quality and practical need seem to have guided the multitude of decisions needed to create a successful consumer device:

 Conclusion: Why are Meta’s Ray-Ban Display Smart Glasses Different?
The Meta Ray-Ban Display glasses represent a decisive and commercially strategic divergence from the common paradigm driving AR/XR competitors like the Apple Vision Pro. They are fundamentally different because they prioritize practicality, style, and mass-market readiness over pure technological prowess.
The Meta Difference is a Calculated Trade-off:

• Strategic Trade-Off (LCoS & Geometric Waveguide): Instead of pursuing the cutting-edge (and expensive) Micro-OLED/Micro-LED displays, paired with complex Diffractive Waveguides, Meta intentionally selected a more balanced, mature stack. By using the cost-effective and power-efficient LCoS display with a high-fidelity Geometric Waveguide, Meta achieves bright, full-color visuals necessary for AR notifications without the high cost, bulk, and power drain of other systems. This choice directly translates to an accessible price point for a first-generation product.
• Focus on Form Factor (Style over Specs): The Meta Ray-Ban Display is, first and foremost, a fashionable accessory. The partnership with Luxottica and the use of the Geometric Waveguide (which enables a thinner, more familiar lens profile than other AR optics) ensure the glasses resemble classic Ray-Bans. This design decision—prioritizing all-day comfort, style, and social acceptability—is the greatest barrier to adoption that Meta is actively trying to break.
• The New Input Paradigm (Meta Neural Band): While other smart glasses rely on voice or clumsy touch controls, the integration of the optional Meta Neural Band (EMG wristband) offers a novel, discreet, and hands-free control layer. This quiet interaction method is critical for public use, making the experience private and seamless—a necessary step for making smart glasses truly practical in social environments. We will evaluate the Neural Band at a later date.

In summary, the Meta Ray-Ban Display is not trying to be a "full AR" computer like the future "Orion" or the current Vision Pro. It is a mass-market, AI-powered overlay device—a calculated "Phase 1" designed to establish the core form factor, build user confidence, and integrate Meta AI into the consumer's daily life at a price point that avoids the "over-engineering mistake" of its competitors.
Note:  We have no connection with Meta or any CE company we mention.  We have no investments in any CE company and make no equity investments, other than ETFs.  We receive no compensation for mentioning companies in our notes and provide not investment banking  advice or receive participation compensation
​

While VR devices languish, AR glasses seem to have become the focus of those looking to capitalize on a “VR Lite” world that can provide additional visual functionality without the massive headgear that keeps VR out of the mainstream.  That said, even lower on the VR/AR scale are the latest category of ‘glasses’, AI glasses.  AI glasses are just what they seem to be, relatively standard glasses with the ability to connect to an AI system that can be used for a number of functions.  They tend to be inexpensive relative to AR and VR devices, with functionality varying with each new device.
The category is ruled by Meta’s (FB) Ray-Ban glasses that were released in October 2023.  They have sold over 2 million units since and are the standard against which competitors are marked or mark themselves.  Yet, when CE brands see unit volumes in the millions it is hard to resist sticking a bunch of engineers into a facility somewhere and seeing what they come up with in a few months.  While Meta basks in the glory of being the best by far in a ‘new’ category, the competition is beginning to come from other substantial players, with China’s Xiaomi (1810.HK) the latest to release its own branded AI glasses.  We note that while AI glasses are a standalone category today, we expect the category to merge with AR glasses at some point in the near future as consumers will see AI glasses as the ‘low-end’ of the visual headset spectrum, with AR being the mid-tier and VR being the upper tier.
Given the success of the Meta Ray-Ban’s Xiaomi did not stray far from both the Ray-Ban look and functionality, and we already get the feel that while AI glasses are reasonably inexpensive, it’s going to be a knock-down, drag-out battle on price once the basics are established.  Just like the smartphone world, where a new feature or application can sustain a line for a year or two,  AI glasses and eventually AR glasses will be one and the same, although at a lower price point, and for AI glasses that become available under $300 (currently) and eventually under $200, it is easy to see how volumes can become attractive to CE brands that deal in millions of units regularly.

As can be seen in the images above, there is very little physical difference between the Xiaomi Ai glasses and the Meta Ray-bans, and when it gets down to components that similarity remains.  Here is a comparison of the feature sets of both.  These are for the basic models of each.

So, what can these glasses do, considering they are only a bit more expensive than the average pair of glasses at a discount retailer (avg. ~$184).?

• Translation - They have the ability to translate conversation in real time, with the number of languages available dependent on the model or brand.  The Xiaomi glasses include 10 languages including Chinese.
• Text recognition – AIs have made text recognition a much easier task then having to run documents through a physical OCR or digital documents through software to break down image-based text into actual editable text.  AI systems can do this easily using the camera that is built-in.
• Object Recognition – Object recognition requires either a massive database and a matching system, or a model that has been trained on large amounts of image and video data.  Again, using the camera, one can have the AI examine an object that the user is looking at and classify or describe it.  Taking it further, in theory, the user can then ask the AI about the object, such as where it might be sold and in what colors.
• Voice commands and Assistants – Multiple microphones allow voice commands to be used to initiate processes or systems, such as video recording.  This allows the user to initiate the command without having a free hand.
• QR Codes – The glasses can scan mobile QR codes for payments , avoiding having to take out your phone to scan QRs.
• Transcription – The glasses are able to record meetings, conferences, and  conversations and create transcripts and translations.
• Video and Image Capture – Using the embedded camera, the user can capture images and video (30 fps), take video calls, or livestream.
• Audio – The glasses have a 5 microphone array that can be used for voice calls and built-in open-ear speakers in the frames so no earbuds are necessary.
• Integration – This is where things can get a bit more difficult as applications need to be able to interact with the OS and, potentially, applications on a smartphone.  , so things like livestreaming might be limited to only those live streaming applications that have APIs or other necessary interfaces to work with the glass’s hardware.  This is where proprietary systems can cause AI glasses users to become frustrated and be unable to use all of the features.  However, as glasses become more popular the interface issue should fade as both applications and hardware modules become more standardized.

One question that comes to mind at this relatively early product stage is whether the direction will be whether brands will take the route of maintaining a price range in the $200- $400 range and compete by adding features or will the direction be to keep hardware costs low and keep reducing the price to attract more users?  We expect the former.

​It has been well-documented that Meta (FB) has been losing considerable sums by seeding the VR market with its Quest VR headsets, and while they are certainly the best-selling and most popular VR devices available to the general public, even the biggest proponent of the Metaverse, seems to have begun to understand that there is some point at which selling items below cost is not quite the best solution to developing a new market.  Meta’s Quest Pro XR headset is the company’s first attempt to create a device that has appeal outside of the traditional VR gaming market and while its considerably higher price ($1,500 compared to the $300 Quest 2) will certainly reduce unit volumes, we expect the company will actually be able to sell the Quest Pro above cash cost.
 
That said, while other VR brands struggled to produce VR headsets with similar features and such a low price, without bankrupting their companies, this new and more realistic pricing tier also gives others VR brands the opportunity to take on Meta’s unit volume domination and challenge the company for leadership in the VR space.  One such company that seems to have made such a challenge is Taiwan’s HTC (2498.TT), who announced their Vive XR Elite headset on January 5 and expects to ship units toward the end of February.  HTC generated ~$145m in sales last year, between it headsets, notebooks and smartphones, so they are no match for the seemingly endless resources of Meta, but have made the challenge regardless, despite Meta’s dominance of the market over the last few years. 
 
That said, it was not a great year for RA/VR sales in 2022, and unlike some who saw unit volume growth essentially flat, we estimate shipments were up ~14.8% y/y and revenue was up 24.1% based on our composite sources.  Regardless, those brands, and there seem to be ne ones popping up monthly, that are serious about the XR business, continue to churn out new and more sophisticated headsets on a regular basis.  Right now, we count about 18 announced but unreleased VR headsets and a similar number of AR devices from a variety of companies, both large and small, that are expected to be released this year.  Some, particularly those from smaller companies, could fall into 2024, but most are pushing for a release before Apple (AAPL) might announce some sort of XR device.  As the average time between announcement and product release was 193 days last year, in theory, all should be released before the end of 2023.
 
Back to the HTC/Meta battle…  based on what we know thus far here are some of the comparisons between the two devices:
 
·       Both are standalone devices, meaning they are not required to be tethered to a PC or smartphone but both can be connected to a PC via USB or Wi-Fi 6E
 
·       The HTC device has a retail price (including controllers) of $1,100, while the Meta device sells for $1,500
 
·       The Meta deice is based on the Qualcomm (QCOM) XR2+ chipset while the HTC device is based on the Qualcomm XR2, while both have a Kryo 585 CPU and Adreno 650 GPU.  The XR2+ is said to have 50% higher sustained power and 30% improved thermal performance (Company data), which, in theory, should allow for faster image processing and lower latency, while leaving room for camera image processing without raising latency.  As the XR2+ is quite new, real world usage metrics are still being developed, so we must take Qualcomm’s word for it for the time being.
 
·       Both use pancake optics, a system that allows lenses to be closer together, reducing the thickness of the device
 
·       We do not know what type of display the HTC device uses, although the 1920 x 2160 resolution is slightly higher than the LCD display used in the Meta device, with both having a 90Hz refresh rate
·       The diagonal field of view, meaning the width and height of the visual image field is 110° for the HTC system and 95° for the Meta, so close to the same, but every little gain in FOV helps to convince the brain that it is not seeing something that is confusing, which helps reduce fatigue and motion sickness for some.
 
·       The Meta device (inclusive of battery pack, etc.) is slightly heavier at 722 grams, while the HTC is 625 grams, which comes to 1.59 lbs. and 1.38 lbs. respectively
 
·       The Meta device does come out ahead when it comes to tracking, as it is able to track eye, face, and hand movements with dedicated tracking cameras, while the HTC can track only hand movements, and important factor in game play.
 
According to HTC the battery is expected to last15 hours, while the Meta battery is expected to last 2 hours, but we seriously doubt the HTC metric on an apples-to-apples basis.  Some brands give battery life metrics using hot swappable battery packs or low-level usage, so we do not see this as a valid comparison
All in. as these two devices are quite similar, real world performance and price will be the determining factor as to who will win the VR race in 2023, although with Apple as a major wild card, even a whiff that Apple has a product that will actually be released in 2023 (other than speculation) could push potential XR adopters to wait and see what Apple might release. 
 
While we expect unit volumes and sales of XR devices will rise this year, macro economics are not going to make for a stellar 1Q for the XR space, but there is some hope for the 2nd half, especially if Apple is ready to enter the space.  Even with Apple, the industry is still looking for practical applications that can drive hardware sales, and the Metaverse seems to have returned to a more sedentary state, rather than the hype it received earlier last year, so it is up to hardware brands to find those applications that will appeal to the public.  While there are many niche applications, we would hope that someone will create an inexpensive AR headset that could function as a dedicated translation device that could generate translated text for the wearer.  An application such as that, especially if priced right, would at least have the potential to become ubiquitous and help the public become more familiar with the potential that AR can bring.
 

​Qualcomm (QCOM) has released what we believe is the 1st chip set platform specifically designed for AR headsets/glasses.  Previously most AR glasses used Qualcomm Snapdragon 670 or 662 chipsets, which were designed as smartphone chipsets using typical X12 modems, Adreno 615 GPUs, and Kryo 360 CPUs.  Qualcomm saw VR as the more important device, having designed the XR1 (5/18) and the XR2 (12/19) specifically for VR headsets, although the XR1 had some features that made it at least a stepping off point for the Ar2 – Gen 1 noted here.
The AR2 – Gen 2 has been specifically designed for AR in that it uses a distributed architecture that processes latency sensitive information on the glasses themselves, while high-computational processing is transferred to the Snapdragon chipset on a smartphone, PC, or similar device with less than 2ms of latency.  The chipset uses ~50% less power than the XR2 chipset and, according to the company, has 2.5x better AI performance than the XR2, with the main processor being 40% smaller (PCB area).  The chipset supports 9 cameras, which goes toward the trend in AR for a shift from 3 degrees of freedom systems (3DoF)  to 6DoF systems that are common in VR, along with hand, eye, face, and body movement tracking.
Most significant is that Qualcomm has designed this chipset specifically for AR, which legitimizes the category further, especially as the platform is the basis for the Meta Quest Pro XR headset released last month by Meta (FB).  Qualcomm has also indicated that it is working with a number of OEMs to include the chipset in upcoming devices.  Qualcomm specified Lenovo (992.HK), LG (066570.KS), Nreal (pvt), Oppo (pvt), Rokid (pvt), Sharp (6753.JP), TCL (000100.CH), Vuzix (VUZI), and Xiaomi (1810.HK), among others as those developing platforms using the AR2 – Gen 1.  This comes in what is a particularly weak year for AR/VR after a strong growth year (units) in 2021. 
While we do see the AR/VR space as one that is growing, the industry and forecasts for the industry seem very oriented toward 2023 and 2024, likely based on expectations that Apple (AAPL) will release an XR product.  There is already some very significant differences in forecasts for AR/VR in forward years, 29.1% between high & low forecasts for 2022, 46.6% for 2023, and 105.1% for 2024, so we put little faith in any forward estimates after this year’s much revised forecasts, but we compile as many as possible to give a composite of group, and begin to map out our own set of unit volume estimates early next year.

​As of Tuesday a Federal judge in California is wrestling with a decision as to agree to a motion that would block Meta (FB) from collecting or disseminating information that the company collects from hospital, provider, or patient portal websites using pixel tracking.  The motion was made as part of a number of consolidated individual and consolidated class action lawsuits against Meta concerning the collection and potential sharing of medical information that was allegedly collected without the permission of the user.  Last year a study found that 33 of the top 100 hospitals used the MetaPixel application to collect information about ads that ran on their sites, but the application (allegedly) also collected information about clicks and form information.
While on-line privacy and security is a relatively new area for litigation, we know that such information can be collected and shared as long as the information is anonymous, but studies have found that when a patient communicates with a health care provider’s website where the application is present on the login page, the code redirects the exact content of the patient’s communication with the health care provider in such a way as to identify them as a patient.  According to court documents, the data that the Facebook app records or sends is:

• What method the user used to communicate with the provider
• The fact that the patient logged in to the site and the page name of that login
• Whether the patient had previously been to the site for similar specific health information
• The patient’s IP address
• Cookie and browser information and identifiers

Patient status is protected by HIPAA laws which require a compliant authorization before it can be collected and according to the suit neither the hospitals nor Facebook had any such authorization, although Meta says that the term sheet for Meta Business Tools requires users to ‘obtain all necessary consents’, and that ‘advertisers should not share data with Meta for any child under the age of 13, or that includes health or financial information or other categories of sensitive information.’  The suits go further in that it alleges that Meta has not taken any action to enforce or validate its requirements and uses the information to generate targeted advertising on-and-off Facebook, specifically targeting ads that relate to the patient’s information and page views on the provider portal.
HIPAA does not consider personal names, addresses or phone numbers as protected as they are typically available from public sources, but if that information is listed along with health condition, health care payment data, especially if it contains specifics as to treatment location or treatment location type (XYZ Cancer Center, etc.), individual authorization must be given before that information can be used for marketing. 
The Meta Pixel application places a small bit of JavaScript code on customer websites that works with Facebook cookies to match website visitors to their Facebook accounts (Meta’s own words), and given that the identifier is one pixel by one pixel, it is essentially invisible to the site visitor.  When the user takes any action on the site, the Meta system tells the user’s computer to redirect that information to Meta as it happens, even before the provider gets the information, and while Meta has stated that it (plans top) remove ad targeting options that refer to ‘sensitive’ topics, but qualified it for only the Facebook platform, telling advertisers they could still use “website custom audiences and lokkalikes” to “healp reach people who have already engaged with a business or group’s website or products, which, according to the suit, means people who are already engaged, such as patients.
Yes, this is complicated stuff, especially given HIPAA’s ridiculously complex language, and the legal limitations on who can sue for privacy violations (the government can but you cannot unless its part of a class action suit), but on-line data collection is an area with few hard regulations and is technical enough that legislators tend to have the attitude that such regulations mean little unless there is some political capital or saber-rattling that can get some media time, deferring much of the research to ‘the kids’ on their staff.  But it is a serious topic and the judge’s outrage over the sensitive data collection mentioned in these suits is interesting, if he finds that the suit has merit.  Otherwise Meta will know every time you make an appointment with a doctor, visit an ER or urgent care, or get a vaccination, and that virtual folder of information that Facebook has on almost everybody will get larger every day.

In the pre-GPS world, day-to-day navigation was done with a combination of landmarks, maps, signs, and a good sense of direction, a struggle for many who do not have that last item, but with the advent of GPS systems, the necessity for a good sense of direction has become almost moot, other than keeping you from driving into a lake when the navigation system is incorrect.  Life without GPS is difficult for those that have grown up with it, and the fact of not having your smartphone with you when traveling almost anywhere is testament to our dependency on GPS navigation, but there are alternatives on the horizon.
In VR systems it is absolutely necessary for the user (and the system) to know where they are spatially, not only for the safety of the VR user but for others that might be negatively affected by wildly flailing arms or other body parts, or for the possible destruction of nearby furniture or pets.  Kidding aside, there are a number of systems that are used to gather position information for VR systems, primarily the requirement to mark absolute boundaries before a VR session, or the system’s ability to sense object information around the user using RF, although more sophisticated sensing systems are needed for the advancement of VR into society.  That said, the characteristics needed for positioning in AR systems are far more related to everyday location information, as the evolution of AR systems into daily life will necessitate precise positioning information in addition to the visual information seen by the user.
Without location data the AR system will not be able to point you toward a particular destination or help you pick out your rideshare in an airport waiting line, and more importantly, it could point you in the wrong direction if being used in a driving situation.  We have noted that Google (GOOG) has been collecting visual data for its ‘Street View’ database since 2007, which it incorporates in Google Maps and Google Earth, and more recently made the 20+m GB of data and 10+m miles of roadway imagery available to developers under the “Live View” API, and while Google is the leader in non-GPS location data, they are certainly not alone.  Apple’s geo-anchors use the company’s “Look Around” data, movement indicators, and user imagery to develop a global 3D map, while Facebook (FB) focuses on “Live Maps”, a collection of physical and geometric information, along with a host of other social media oriented companies that are looking for ways to generate location data without using GPS.
The problem stems from the fact that the GPS system relies on signals from at least 4 of the 30+ GPS satellites that orbit our world.  There are instances when atmospheric conditions or signal blockage can compromise GPS signals and smartphone GPS data has a 4.9m (16 feet) radius, which could make it a bit difficult to pinpoint a specific parked vehicle or an item in a large warehouse.  Niantic (pvt) a small company that was spun out of Google, and financed by Google, Nintendo (7974.JP), and The Pokemon Company (pvt), recently bought 8th Wall (pvt), a company that created an interactive AR development platform that is browser based rather than a standalone application.  Niantic’s system is based on geometrics and the system’s understanding of its surroundings along with an understanding of objects themselves, essentially ‘does that object look like something in my database that is defined as a…’, with some of its data collected from Pokemon Go users, a vast network of users who play individually but are now able to play ‘in-network’.  As much of the game is based on finding Pokémon hidden in various locations, the visual data that is collected during the games is added to the Niantic database to build out the 3D maps.
While all of the players in the non-GPS location space are approaching it from different angles, the important factor is that the data is robust, giving object recognition and spatially oriented systems more information on which to rely, making their ability to recognize an exact location more precise.  As noted Google has a huge amount of data from which its systems can match,  and certainly has an advantage over smaller data sets, but tapping into social media or other image sources can build maps quickly with sophisticated algorithms, so there is still no foregone conclusion that one system will rule.  The good news is that will all of this data collection, and more socially acceptable hardware for AR, the idea that you could walk down a street wearing AR glasses that tell you how to get to your destination by painting a red dotted line on the sidewalk or indicating which way to go to find that food truck that used to be nearby, is getting to be more of a reality.

​There is a war going on that rarely makes it into the press as the battlefield is not on the ground, in the air, or in space, but inside the guts of massive processing nodes that are used to understand the nuance of language.  The algorithms that direct these processor are based on a subset of Ai that deals directly with syntax, expression, and a host of other language variables that make understanding other languages a challenge for humans and a monumental task for digital entities.  The opponents here are companies rather than political entities, and well-known ones at that, with Microsoft (MSFT), Google (GOOG) and Meta (FB) all pitted against each other with the focus of being the dominant force in the digital translation market.
There are no casualties or battle lines in this war, as it is fought with processing metrics and advertising, and that makes it hard to know who is winning, but the participants seem to have taken on a particular metric to make the public aware of who might be in the lead.  That metric is the number of languages a system can translate, which seems to be used a s a gauge toward whose system is the most advanced.  While the number of languages that a system can translate is certainly important, especially to those languages that might be considered secondary, but by far the most important metric for translation services is accuracy. 
Recently Meta indicated that its NLLB-200 Ai model has increased the number of languages it can translate to 200, which it accomplished in two years, while Google Translate is only able to work with 133 languages and Microsoft’s system translates only 111 languages, although that includes two Klingon dialects, and is clearly staking a claim as the world’s most advanced translation tool.  While the number of languages a system is able to detect is certainly easily understood by the general public, the quality of the translation, which is based on the algorithm and the sample base, is far more important and there are two ways in which that can be evaluated, by humans or by machines.  Using humans to evaluate translation quality throws subjectivity into the mix, while automated machine evaluation does not, but a machine evaluates a translation by averaging words and sentences against a human evaluation, with the idea that the closer the machine score is to a human score, the better the translation is.
With all of this being beyond the scope or desire of the general population, translation giants will continue to use the simplest metrics to give credence to their systems, but will have little correlation to real world results, as the ability of the AI to understand nuance and what to do with that understanding is really the key.  All three companies mentioned here have access to vast stores of speech, which certainly goes toward the ability of an AI to learn, but the algorithm is the key and that is something that not only grows with more resources but must change as humans better learn how to convert those subjective views into language that a machine can understand.  So the question is, does the AI need more language to read or does it need more human evaluations in order to take the subjectivity out of their evaluation?  The only way to know is…
E ninau i ka Mīkini - Hawaiian
Spurðu vélina - Icelandic
Faighnich dhan Inneal – Scottish Gaelic
quaeritur machina – Latin
Ask the Machine - English
 
 
 
 
 
 
 

​Facebook (FB), aka Meta, did a nice thing, they broke out the yearly revenue and operating income for Reality Labs, the division of Facebook that has been developing and selling the Quest line of VR headsets, along with content and associated software.  The Oculus brand, starting with the Oculus Rift DK1 in 2013 and continuing through the Oculus Quest 2 released in 2020, is known as the best-selling VR headset, primarily due to its low price and reasonable specifications.  There are much higher quality VR headsets that touch upon newer technology and more competitive specs, but for what comes down to ~$400 the Oculus headsets have been hard to beat.
Facebook has seen big increases in its Reality Lab sales over the last three years, with a 127.3% increase in 2020 and a 99.6% increase last year, but at the same time seeing a 47.1% increase in operating income in 2020 and a 53.9% increase in same last year.  So Facebook is losing lots of money to maintain marjet share in the headset game, or is there another reason why the company continues to sell headsets at what is below cost?  Given that Facebook is one if not the biggest proponent of the Metaverse, a system that will feed it tons more data about users than they already garner, they are encouraging the global population to adopt VR as the means to  what they would like to be the new social platform.  Rather than compete with other headset producers who must look closely at profitability, Facebook has the ability to continue to lose money while seeding the VR market and will likely continue to do so until they either find that the Metaverse is not developing at expected, or they can really sell high quality headsets at such prices, which, at best, will take a few more years.
While we expect that little will change in Facebook’s attitude toward VR headsets in the near term, given the current interest in the space, we continue to look at the VR space and the Metaverse as both fertile ground for the hyperbole that technology marketing thrives on, the ability to capitalize on that hyperbole, and a lack of overall consumer demand until the technology takes at least another leap forward.  All in it is fertile ground for examination, which we continue to do, but the little glimpse into Facebook’s VR losses gives us even more fodder for taking the VR space apart.  More to come.
 

​Facebook (FB), now Meta, is a logical supporter of the Metaverse as an opportune way to create new data channels and stickier users, but as part of the Facebook/Meta Metaverse dream, the company had been ‘developing’ an in-house cryptocurrency since 2018 through a subsidiary named Calibra, under the crypto name of Libra, which along with the name of the subsidiary (changed to Novi), was changed to Diem in 2020.  The concept was to offer Diem through Messenger and WhatsApp to the “half of the adults in the world that don’t have an active bank account” and to the “70% of small businesses in developing countries that lack access to credit.”
Under the Facebook plan, Diem was to be a safe and private currency that can be used for any transaction, and was to guarantee that no data (except in limited circumstances) would be shared with Facebook or any 3rd party, although “if a Calibra product feature can be personalized or improved with data from Facebook”, customer permission will be requested, adding “We may also use customer data to conduct research projects related to financial inclusion and economic opportunity with, for example, academic institutions and NGOs, though any published results will only contain aggregated statistics.”
Facebook created the Libra Association  (changed to the Diem Association) to oversee the Diem network and was intending for the Diem to be a stablecoin, but not pegged to a single currency but rather a basket of currencies, which means that the value in local currency would vary, albeit likely less than Bitcoin or other unpegged currency.   The assets backing the currency were to be held by a ‘geographically distributed network of custodians with investment grade credit ratings’ in order to increase security, and Diem was to beinflation neutral as the number of Diem coins will be adjusted to the value of the exchanged currency, either by adding or destroying Diem coins, with Facebook and 3rd parties collecting interest on Diem reserves.
Unfortunately, Libra/Diem faced considerable opposition, particularly in Europe where Facebook and Google (GOOG) were facing substantial fines for anti-trust violations, with the French foreign Minister stating “It is out of the question” that Libra “become a sovereign currency, It can’t and it must not happen,” and the German European Parliament representative posting on his Facebook account that the company could become a “shadow bank” and that regulators should be on high alert.  With the scrutiny that Facebook’s original Libra announcement caused, and the added warnings from European regulators, Facebook postponed the launch of the currency and renamed it Diem, which seems to be a go-to concept that Facebook uses to deflect bad press or increased scrutiny. 
However on January 31, the CEO of Diem stated that the Diem Group’s assets and IP were being sold to Silvergate Capital Corp (SI) for $182m (+ $30 in associated costs), essentially selling the rights to the Diem stablecoin and the blockchain network associated with the currency, giving Silvergate its own stablecoin, although the press release seemed to indicate that the current Diem Association would be wound down, and the operations melded into Silvergate’s Federal Reserve Bank organization and the Silvergate Exchange Network, a 24/7 digital network that allows SEN network customers to move dollars to and from any other Silvergate customer account using a stablecoin. 
With the assets behind Diem, and the negativity of its association with Facebook, Silvergate has bought a way to provide a stablecoin intermediate for its customers, which is especially attractive given the release of the Fed’s “Money and Payments: The US Dollar in the Age of Digital Transformation” report which we highlighted in our 01/21/22 note.  The report favored stablecoins over other cryptocurrencies, although we expect the Fed is still a bit away from adopting a CBDC  (Central Bank Digital Currency) and has put much of the onus on Congress to pass legislation that would put such currencies into a federal regulatory framework that would give the public an understanding that such a currency would not require deposit insurance, but would be a direct liability of the Federal Reserve, with no credit or liquidity risk, something Facebook was not able to provide.

We mentioned in our note yesterday that the hardware focus on Apple’s (AAPL) potential AR/VR products is missing the big picture surrounding the long-term viability of such a product’s ecosystem.  For others, especially smaller companies, AR/VR is a hardware game, and there is considerable competition between VR development teams and companies to provide the latest technology to differentiate their products and justify entering a smaller and likely less robust AR/VR environment.  As we have noted previously, we keep an AR/VR database, separating out those products that have been released from those that have been announced, given that some announced products are still unreleased more than a year after being announced.
One AR/VR product from a Prague based company called VRgineers (pvt), who showed their XTAL 3 Mixed Reality and Virtual Reality headsets at CES, and while both are still in pre-order mode, they have a release date of April 2022, only a few weeks away.  These headsets are the next generation following the company’s XTAL 8K headset released in 2020, with much of the design oriented toward the use of VR headset in pilot training.  While gaming VR headsets do have to meet demanding specifications, VR headsets used in pilot training are a bit more specific in use but more demanding in terms of their ability to interface with physical cockpit training systems, and in the case of the XTAL 8K and XTAL 3 headsets, they were designed in cooperation with the USAF and the Royal Air Force and interface with a wide variety of cockpit simulation hardware and software, something not part of most VR headset specifications.
These are not your typical $500 headsets, as the older XTAL 8K (currently on sale) model sells for $4,800, while the two newer versions sell for $8,900 and $11,500, so they are really at the top end of the AR/VR universe, but also not out of the range of some of the AR headsets that have been developed for industrial use.  That said, they have some interesting features that make them a bit different than other VR headsets and justify the price in the right environment, particularly eye and hand movement tracking and 4K resolution.
A number of VR headsets employ eye tracking, a technique that uses internally mounted cameras to measures the position of a reflection on the cornea of the eye (Fig. 1) (red arrow) against the center of the pupil (blue arrow) and calculates where the user is looking, regardless of head movement.  In most VR systems the data is used to move the user’s field of view in game software, so when the user looks to the side, the game view shifts the same way.  In the VR headsets mentioned above, not only does the eye tracking data re-locate the FOV, but the data is recorded and used to measure how long it takes a pilot to notice something appearing in the periphery, or how often they look at controls or other external objects.  Similar data is collected from controllers that evaluates hand motions and can give insight into how quickly a pilot reacts physically to visual stimuli.
Taking eye tracking out of the aerospace environment and into the Metaverse, eye tracking information can be used to give game developers ways to help you improve your gaming ability.  By tracking where you are looking during a game, the eye tracking information can adjust where you are throwing or shooting to more accurately align the shot.  But eye tracking information also gives clues as to emotion and reaction to various situations, which is the kind of data that can help data collectors build a more accurate model of you in the Metaverse, although when we say model, we don’t mean your avatar but more things like your level of excitement when viewing a new smartphone or piece of clothing, data that helps them ‘improve the user experience’ or in real terms produce a better selling environment.
This is just one small aspect of why on-line data collectors like Facebook (FB) and Google (GOOG) are excited about and promoting the Metaverse.  By increasing the amount of information a user generates, the value of the data is also increased, and while there will be much said about selling virtual real estate and other virtual items that don’t exist in the real world, the game remains the same as it is in the 2 dimensional internet, collect more data and sell it to folks so they can sell more stuff, whether its virtual or physical.