theschwa

Would you be willing to share about how much it cost to train a model like this?

Between the clear writing and the diagrams, this was a great write up. I had actually skipped reading up on mHC as it sounded like it was going to take some time to grok, but this made it immediately approachable. I hope you do more write ups like this in the future.

Yeah, looking at that example feels like jumping straight into the deep end of the pool. I think it helps going through a tutorial that breaks down the why of each piece. I really liked this tutorial on it: https://www.typeonce.dev/course/effect-beginners-complete-ge...

Some core things from Effect though that you can see in that Express example:

* Break things down into Services. Effect handles dependency injection, that's typed, for services so you can easily test them and have different versions running for testing, production, etc. * Fibers for threaded execution * Managing resources to make sure they're properly closed with scope

I think a lot of these things though often aren't truly appreciated until you've had something go wrong before or you've had to build a system to manage them yourself. *

I feel like this is particularly interesting in light of their Vision Pro. Being able to run models in a power efficient manner may not mean much to everyone on a laptop, but it's a huge benefit for an already power hungry headset.

I believe they are using the original Gaussian Splatting code from Inria under the hood: https://github.com/graphdeco-inria/gaussian-splatting

"How does “reflection” of the vase on the metal part of the table work, you might ask? The gaussians have “learned” the ages-old trick of duplicating and mirroring the geometry for reflection!"

Thank you! I'd seen people remove some of the SH coefficients while still getting reflections, and I couldn't understand how that worked. The old tricks sometimes really are the best tricks.

But,I believe this is being used for his Use.GPU project, which requires not just GPU support, but the latest WebGPU support.

Can someone help me understand why a CPU based method like EDT would be used instead of a more GPU friendly method like jump fill?

It seems like you could use the sun pixel distances like he did, but then use them with jump fill, but maybe I'm missing something?

We used to have some "disabled" checkboxes, but they caused a ton of confusion, and negative user feedback. We switched to toggles with On and Off written on either with all verbiage written in the positive, and we no longer get questions about it. e.g.: Use "blah blah" data source: Off (--o) On

Meta and Apple are both working on creating an environment in which they can succeed in AR. The hardware for AR is a good ways off, but I think the current market is more about building the momentum of software, hardware, and ecosystem that Meta and Apple need to succeed with AR. They are both willing to take losses to get there. In that regard, I don't think this is a make or break moment, but I do think that Apple is hoping to put out a product that gives people a better glimpse of their vision for mixed reality and AR.

I believe they've said that this was a choice to reduce the size and the weight, and it lets them bring the device closer to the users face. It also means that it can stay comfortably with just the one strap.

As an artist and shader geek, I just want to say that this looks awesome. I'd love to see that blog post on Swift + Electron.

Thank you so much for this! I've always wanted this capability, and I plan on making extensive use of it.

Since this uses an T2I model under the hood, would it be possible to build a version of this on top of Stable Diffusion?

Am I understanding this correctly that this would be a push and rebranding of HDR10+?

I think you can do a lot of this with React-three-fiber:https://docs.pmnd.rs/react-three-fiber/getting-started/intro... in combination with React-three-flex: https://github.com/pmndrs/react-three-flex . React-three-fiber lets you make three.js components in react, and react-three-flex will let you do flexbox style layouts to make interesting UIs. You still get all of the capabilities of three.js with a lot of the plumbing done for you. Check out the examples from react-three-flex for some examples close to what it sounds like you want to achieve.

The interview with Norm from Adam Savage's Tested gives a lot of good extra details: https://www.youtube.com/watch?v=x6AOwDttBsc

It depends on the goal, but the whole family of NeRF techniques have been achieving some really interesting results.

They don't go into much detail of HCT here, but I'm curious on their thoughts of the criticisms of CAM16 that lead to oklab: https://bottosson.github.io/posts/oklab/

Would you say that Ads are still unethical if they're not paid for?

As just an anecdote, I'm always shocked by the amount of advertising at Burning Man. There's always people trying to get you to come to their event or join in their activity. Signs and street barkers are everywhere. There's no money and no bartering allowed, but advertising is everywhere, since ultimately attention is what is in limited supply.

If they aren't unethical in this scenario, it seems like what you're saying is that it's the underlying socio-economic system that's unethical, which would explain why all of the alternatives to ads also appear to be unethical.

No, the coin doesn't have a memory, so the chance of tails is still 40% making it still optimal to choose heads.

It's an interesting concept, but it doesn't give much information on how the display works. The display looks to be based on a dihedral corner reflector array. I found this image which shows it pretty well. [0]

[0] - https://www.semanticscholar.org/paper/Blurring-correction-fo...

Yeah, that makes sense. There's some mental gymnastics involved in writing shaders since you're writing the code from the perspective of an individual pixel. As opposed to other methods like using Processing or the HTML canvas where you can just say "draw a square from here to here and make the lines this thick". When writing a shader, I have think instead "Ok, if I want a line here, and I was a pixel on that line, what would I need to know to determine my color".

If you're finding that to be the obstacle with shaders, it may be helpful to start with just learning a library like Processing, or general graphics library in whatever programming language you're comfortable with, and then come back to shaders. It will still take changing how you think about things, but then you'll have some solid foundation.

Personally, I've found it the most helpful to have some kind of goal in mind. Some kind of image or style that I want to make, then figure out what techniques I need to learn to get there. Starting from the technical side can work, but then it can be easier to loose interest when things get challenging. When I start with an artistic vision, then I often find myself over my head in the technical things I need to learn, but I feel motivated to push through that to get to the end goal.

Graphical programming is a blast. I can totally see how it feels overwhelming, especially with a lot of ways to do the same thing. I will say though that once you get started a lot of the knowledge is transferable to different technologies and methods. I tend to recommend that people start with The Book of Shaders [0].

What kinds of things would you ultimately like to make?

[0] - https://thebookofshaders.com/

I totally hear that. Especially if people are commenting that they were thrown off. I just wanted to make sure people knew to use this term if you're looking to learn more about the topic of this project. If anyone is looking to implement this or another method to do indoor mapping, you'll probably want to familiarize yourself with the research field of Simultaneous Localization and Mapping or SLAM.

Localization is actually the common term, frequently used in robotics and autonomous vehicles. There's a really interesting field of research here [0].

[0] - https://en.wikipedia.org/wiki/Simultaneous_localization_and_...

I work in an area like this. I'm in the automotive space where there's just a whole slew of industry knowledge to stay on top of; both business practices that change frequently as well as regulatory information.

It's pretty rare to see engineers that know the industry as well as the PMs, so we focus heavily on having good communication and dedicating time to make sure features are well understood before any work starts. Our structure works well for us, but I can definitely see how it would be advantageous, where possible, to have engineers who knew more about the business.

With the Quest being able to be used as both a PC VR and and standalone, why not Quest? I'm saying this as someone who owns 4 different VR headsets, and I find myself using my Quest 1 the most.

How would I use contravariants to improve the architecture of some code?

I geek out on articles like this, but I frequently struggle to see where I would actually apply it.

I'd also add reducing friction. It seems the goal here is to reduce the amount of time it takes to get the user to the exact functionality they need in the shortest amount of time with a user experience that makes it feel seamless.