You can already run it on your browser with webgpu https://mlc.ai/web-llm/

so this is the metaverse... ok

After deforming the space with a sine function there is no problem in calculating the gradient. The problem is calculating the new correct distance to the nearest surface after a space transformation. You can see this problem in 2D, here I made an example https://www.shadertoy.com/view/NtsSRf

He is writing a book :)

Blender still does not support SDFs, but you can use MagicaCSG to play with SDFs https://ephtracy.github.io/index.html?page=magicacsg Also you can use my nodes at https://github.com/paulofalcao/MaterialMakerRayMarching for MaterialMaker for making procedural nodebased SDFs in Material Maker

SDFs are another tool. Very useful for content creation, for example, although there are still very few tools to create SDFs. You can't create SDFs in Blender for example. SDFs will be more used in the future, polygons are a mess... :) Games like Dreams (PS4 / PS5) make extensive use of SDFs

The opportunity for optimizing polygons in the rendering pipeline is extraordinary compared to SDFs. For that reason alone, it will not likely be economical to use them anytime in the near future for the majority of modern commercial 3D work, perhaps ever.

They are cool though!

5 stars reply! Well Done Linus! :)

"When I delete my account I'm going to lose old conversations, but that doesn't bother me anymore because I already recently lost them once when I switched from Android to iOS... WhatsApp has no way of keeping your history in that case." - YAP!!!! TRUE!!!!

MaterialMaker https://github.com/RodZill4/material-maker is also made in Godot :)

Will it be used? Probably the last one that does not use some sort of AI Compression.See this for image compression https://hific.github.io/ In the next 10 years AI Compression will be everywhere. The problem will be standartisation. Classic compression algoritms can't beat AI ones.

AI compression is super, super cool... but while standardization is certainly a major issue, isn't the model size a much larger one?

Given that model sizes for decoding seem like they'll be on the order of many gigabytes, it will be impossible to run AI decompression in software, but will need chips, and chips that are a lot more complex (expensive?) than today's.

I think AI compression has a good chance of coming eventually, but in 10 years it will still be in research labs. There is absolutely no way it will have made it into consumer chips by then.

"Isn't the model size a much larger one?" yap It will probably be different, and systems will have to download the weights and network model, as new models come in, I don't think that we will have a fixed model with fixed weights, the evolution is too fast. Decoding will take place using the AI chip on the device aka "AI accelerator"