[tmp538394722]

> Petzold is considering using his rigs for an aspect of digital video production known as rendering, which can require significant computing resources. “There are other uses for the cards”

This is an incredible quote.

People have realized that there are other uses for graphics cards than mining crypto currency - like, oh I don’t know, I guess graphics.

[pcthrowaway]

That's pretty funny. I assume this is in reference to the render project: https://rendertoken.com/

Basically a distributed rendering blockchain, where participants perform actual useful rendering in order to earn the token.

I don't know too much about it though, or whether there's actually much demand for on-demand rendering. I'd assume most potential "customers" would have concerns about their workloads being accessible to the public if they're working on something proprietary.

[bergenty]

Is it possible to render an encrypted version of pre rendered data that outputs the rendered file in that same encrypted format that can only be decoded using the original private key it was encrypted with?

[Anunayj]

actually.. it should be possible! Search Homomorphic encryption, I don't know much about it, so I hope someone here can expand on this.

[1] https://en.wikipedia.org/wiki/Homomorphic_encryption

[pcthrowaway]

It seems like you'd have to know who is going to be rendering your pre-rendered source in order to encrypt it to them.

Also, the render nodes would have to be trusted (to not redistribute the content); I doubt this is what render token is doing.

[bergenty]

No I mean the renderer has no ability to decrypt. Can you turn encrypted pre rendered data into finish encrypted rendered data without decrypting it.

Then only the owner could decrypt it back to a useable form.

[pcthrowaway]

I'm way out of my depth here, but my guess: no, and if it was even possible it would probably need to be handled at the GPU level.

[gte525u]

the parent comment means something like render based off of homomorphic encryption.

[m-p-3]

I'm wondering if something similar could be done but for ML, using a distributed network of GPU to train a ML model in exchange of a payment for using the resources.