[kguttag]

The GPU of The Brain author seems to be well meaning and nice guy but he don't not understand optics.

The parallel fiber idea is "optically silly" but it takes some understanding of light to prove it is impossible too. It is yet another example of trying to fool people. I did try and explain this before on my blog over a year ago. See http://www.kguttag.com/2016/11/20/magic-leap-separating-magi... and scroll down to the Appendix at the bottom.

Basically you have multiple fibers going in a circle each with their center of origin it becomes impossible to get them to act like a single image for use in near eye optic.

What the layman would not understand is that this is very different from image stitching on a projection screen that in part relies on the light being diffused/randomized by the screen. In the case of near eye optics, there is no way to get the multiple projector image collimated AND seamlessly put together.

It is yet another example of a good con has to seem believable.

[UncleEntity]

> Basically you have multiple fibers going in a circle each with their center of origin it becomes impossible to get them to act like a single image for use in near eye optic.

Why would they go in a circle when all they need to do is provide a single wavelength on a single axis? A single axis movement seems to be enough, no?

> In the case of near eye optics, there is no way to get the multiple projector image collimated AND seamlessly put together.

Unless they did something completely crazy and invented a method to do this. And maybe with this "invention" they built a demo and showed it to a few folks who were like "hey, this really works, I think I'll write an article about it and get published in Rolling Stone" or "Dude, take my money, please."

Reminds me of another story (definitely true and not about bumblebees). Henry Ford wanted a V8 engine block made out of a single casting and all the engineers said it wasn't possible. So what to do? Turns out throwing a shitton of money at a secret project lead to the answer and the flathead ford V8 engine was the result.

[yorwba]

> Why would they go in a circle when all they need to do is provide a single wavelength on a single axis?

Going in a circle (really a trochoid) is actually a pretty smart way to scan a 2D surface with a single light ray, you basically just rotate the vibrating fiber's plane of vibration around an axis to produce a cone of light. If you wanted to have only a single axis movement, you'd need many more fibers, one for each row of pixels.

>> In the case of near eye optics, there is no way to get the multiple projector image collimated AND seamlessly put together.

> Unless they did something completely crazy and invented a method to do this.

Did you read kguttag's argument on the matter? The reason this doesn't work is because the optics for collimation depend on the incoming angle of the light, so it must happen close enough to the light sources that the different light cones do not overlap. Then the light still expands within the optics before it gets straightened out, so there needs to be some gap for tolerance. So either they end up with some stray light rays that go into undesirable directions, or there is a gap at the border between the different regions.

Now maybe they showed a demo of this with visible artifacts to people and they were still blown away. Most likely they also told them that there was still room for improvement, while leaving open where exactly the fundamental limits are. Maybe they will release the imperfect version of their product than can actually be built, and people will still love it despite not living up to the marketing.

The problem is not that Magic Leap's ideas are completely useless or impossible to implement, just that they are significantly overhyping the expected capabilities of the finished product.