Hacker News Reader

Comment by debatem1 | Hacker News Reader

I've been hearing rumors about these guys for years and it never seems to be any more than what's in this article-- "we saw some special effects and our wallets flipped themselves out of our back pockets onto the table and began spewing money".

People so often underestimate the difficulty of productizing significant UI changes, not realizing that getting content to migrate to the new shiny will be at least as hard as getting rid of Flash on the web.

View on HN

John Carmack joining Oculus was a quick reason to think: "hm, this team has people who know how to get things done and launch."

Maybe Magic Leap has the same types of engineers [does Graeme have the same reputation as hyper productive dev?], but there are reasons Carmack has this reputation [1].

And I've known and worked with people like this. People who are utterly relentless. Insanely productive. And honestly? I would always bet on them. Every single time. It's just a major key that I don't know anything about on the Magic Leap team.

[1]: http://venturebeat.com/2015/01/27/how-john-carmack-pestered-...

By the time Carmack joined the Oculus Rift's CV1 technology and design was mostly finished. Carmack is reportedly focused on Oculus's mobile offerings which we have yet to see or is indeed the Santa Cruz prototype which, if rumors are to believed, is a mobile SoC powered inside-out markerless VR headset. When I saw the SC prototype I instantly thought this is what Carmack must has been working on. Not shoving cellphones into $20 goggles and working out the kinks like many think mobile VR would turn out.

I think there's some practical considerations here with the "magic engineer" who can do anything quickly and, of course, the laws of diminishing or even negative returns outlined in books like the Mythical Man-Month. Even the impressive Santa Cruz is years out from being sold. Graphically it cannot compete with the PC-based systems it would ultimately compete with.

The history between Carmack and Oculus goes much further back and predates the company itself:

https://en.wikipedia.org/wiki/Oculus_Rift#Initial_prototypes

It seems like every one of the guys who have gone to work at Id Software with Carmack have found themselves completely overwhelmed. They're all competent and brilliant people, but they're not Carmack.

"We're going to make augmented reality everything! Repair manuals, etc"

Bullshit. You're going to model and diagram every part and repair guide from every production model vehicle ever?

Magic Leap would need a software development industry unto itself to build the solutions being batted lightly about in the OP like pingpong balls

I think that having an (the) entire industry writing solutions for Magic Leap is exactly what they are envisioning. It would be a platform, i.e. Chilton would build an app for it.

Now, whether or not they are actually going to be the biggest thing in graphical interfaces since the flat screen is another question...

There's just no way.

They aren't going to beat Oculus/FB, or Sony, or HTC.

They don't have a product; they clearly must have some incredible tech demos, and a wizard working VC

Yeah, the simplest explanation is exactly that. What evidence is there that they're better than the current HMD race that is rapidly progressing?

Who else has plowed a billion into a waveguide and light engine factory?

I hope they can deliver what they are promising.

I remain highly skeptical.

I'm skeptical too, but I could envision a path where this gets a lot of the work done at a tooling level. All these products will have CAD models behind them already, AutoCad and SolidWorks, et al, could introduce tooling to pump out the assets necessary for this at not-terrible levels of effort, couldn't they?

One of the advisors to Magic Leap is currently the CEO of Onshape (and was also the creator of SolidWorks). There will surely be an Onshape app for Magic Leap at some point.

Back in the 90's when I was doing 3D on the web, we looked at this and the results weren't compelling. The models used for CAD/CAM aren't the same sort of thing you would want a GPU to try and render in a 3D scene. Just as one trivial example, none of them had normals so you'd need to do a lot of model cleanup simply to get them to render as a solid object.

CAD data is getting used more often in 3d graphics (although there's still some cleanup). It's used commonly for product shots and starting to be used for games. The most recent GDC had a McLaren pulled from CAD data running real time in Unreal Engine[1]. I honestly don't think much cleanup was done because of the timeline for projects like these (I thought I heard 6 weeks start to finish focusing on custom engine development for interaction and shading).

In my experience, the problem is more in how the model is organized; having ever screw or washer as a separate piece of geometry causes problems (3d tools have trouble with very large geometry counts even if the poly count isn't high). CAD data also doesn't work very well with non-hard materials like upholstery; stitching is likely missing, as well as things that make cloth/leather look normal like pinching or gravity pushing on things.

[1] https://www.youtube.com/watch?v=0oZfvoG6qYQ

Can't you just calculate the surface normals? What am I missing?

You had to be able to orient the normals correctly. For a hull this might be easy enough but for many objects it was a challenge.

Gouraud shading like it's 1989? :)

Sure, and there are already standards for exporting 3d geometry data.

The process isn't rocket science but the scale is.

Reply on news.ycombinator.com