Can Replicator be considered Generative AI?

@rthaker (sorry for the tag, I just really wanted to hear from one of the developers)


I have been getting into this debate at work about whether the “Replicator” engine for synthetic data generation in Isaac Sim can be considered “Generative AI” and I need the opinions from the developers at Nvidia to put an end to this debate.

We know that the Replicator is a tool/engine for generating(replicating) photo-realistic and physically accurate synthetic data, but can it be considered as Gen AI, if we use the term very loosely, since it generates synthetic data, similar to GAN or VAE???

Are there any components in Replicator that makes it “Gen AI” (again, very loosely)??

Id like hear from the Nvidia developers and other AI engineers on this forum!


Hi @jaeyeun - I will forward your question to right devs to give you accurate answer.

Disclosure: I am on the Replicator team

I would not currently consider Replicator a form of Generative AI.

Similar data can be produced by other Digital Content Creation (DCC) tools such as Blender, Maya, Houdini, Unreal, etc. - and many researchers and companies have used them for synthetic data generation.

But those tools, and Replicator, currently do not feature the use of AI (as far as I know). Replicator is currently similar to procedural content creation that has been done in film VFX (crowds, vegetation, city generation) and video games.

I consider the main distinction to be that randomizations/distributions are currently still input by users. This still leads to amazing and useful results, but there are no neural networks driving these inputs.


1 Like