Analyst: Vision Pro Demand Fell “sharply beyond expectations,” Leading Apple to Reduce Shipments for International Debut

I never said a “few years”, instead went with “several/a couple more”. I avoid making predictions about technological progress for more than a decade out, because things occasionally change rather quickly. But it if turns out that we are in fact still several decades away from wave guides being usable for HMDs (which I doubt), I will simply claim that with “a couple more years” I was referring to geological time, where something taking a thousand years is considered very fast.

Even if I would consider myself competent enough to make reasonable predictions in all these areas (which I do not), that would basically be the content of a book. So no.

The superficial, unspecific, ultrashort version: I still mostly ignore Quantum computing due to the experimental state, have minimal knowledge about chip production (it will get better, get more stacked/3D and modular like chiplets to compensate for physical limits). The capabilities of LLM surprised me. They are still technically only predictive/”stupid” and their answers non-transparent, limiting their autonomous utility, and I doubt we’ll get AGI to fix this anytime soon. Nonetheless they are amazing tools, and IMHO the most interesting developments are similarly capable small language models like those released by Microsoft or Apple, running locally on moderate devices, allowing for lots of weird applications without incurring the insane hardware and energy costs of e.g. GPT.

I regularly post predictions on VR/AR here, still expect XR to go mainstream, but like most, not before we have glasses-like HMDs, which will probably take another decade. Meta is in trouble, because their plan of growing to a self-sustained XR user base before Apple and Google enter failed. But I expect that we won’t see a continuation of the current mobile duopoly, because regulatory agencies will seriously clip the wings of large platform owners, leading to a more open market with more players than most expect, and giving content owners like Epic, Microsoft or Disney a much bigger weight than today.

Meta actually knew/knows this, and aimed to control central cross-platform services like user authentication and avatar/identity management instead of a platform, which was mostly needed to accumulate a user base in a not yet existing market out of their Quest user base. They will now struggle to prevent users just sticking to Apple/Google despite help from law makers, but has too much money and too much to loose to give up.

That’s 337 words not even scratching the surface, but already exceeding what is a reasonable length for this forum.

We’ll see. An optical waveguide is just something that allows to direct light somewhere else, including around the corner, by internal reflection. Glass fiber is a waveguide. The advantage for HMDs is that the actual display and the optics can be moved out of the field of view. And since the internal total reflection only happens at a certain angle, the material itself is transparent from the side, so the passing through light doesn’t interfere with the virtual image. This is why Hololens and Magic Leap use waveguides.

Other technologies are possible, like retinal projection or holographic lenses. Both require coherent laser light, with lasers still too large for HMDs. Technically the projection also requires a waveguide, and a holographic lens IS a waveguide with a very complex optical path.

I can’t tell what the main problems with increasing the FoV on HMDs using waveguides is, but my guess is the required magnification. On regular HMDs, the users is looking through a pair of lenses, which are necessary to allow the eye to focus on a display at close range, but at the same time distort the image, which has to be corrected in software for video passthrough. That’s not an option with actual see-through displays, as magnification would distort the view onto the world. So not only has the virtual image to be rendered pre-distorted, but the light has to enter the waveguide in a way that it will exit in front of the eye in the same way as it would if it had gone through a lens.

Higher FoV requires higher magnification, which probably makes the problem harder. At one point it would interfere with the angle required for the total internal reflection the wave guide is based on, and it is possible that there are physical limits that will prevent the waveguides we have from ever reaching the FoV of current HMDs. This is speculation on my side, but the reason why I think that waveguides are probably the tech that will enable see-through AR glasses, but could still turn out to be too limited.