Ask HN: Why are there so many 3D Gaussian splatting posts?

It's a popular topic. Unless you'd prefer to see flamebait, cryptocurrency/AI submissions or a SaaS in it's place I'd not read too much into it.

2 years ago it was "that quirky idea from the 90s" but then it became "a newly viable way to do real time 3D rendering in a very atypical way". That gives a combination of novelty of approach with lots of room for interesting discoveries to be made when compared to more established methods. I.e. great hacker fodder.

The field is evolving quickly and it's pretty to see.

Another deeper reason to me is that it employs some neural network techniques -which have differentiability-. This means that you can backpropagate errors from a domain to another. In the future, we'll have videos as embeddings that can understand written signs, and interpret the voices and what they refer to in the scene.

This is not AGI, but will be much closer to the sensor fusion us humans do.

Part of it is that rendering with triangles has been the standard for so long that it’s interesting to see an approach that uses a different primitive.

And the topic hasn’t been fully explored yet so there’s a lot of low hanging fruit

[Maybe]

A few more users than average submit articles on a topic. Other people see them and find them interesting. Nobody is fatigued enough to flag the topic.

The front page doesn't happen in a vacuum. Other sites, conferences, youtube etc. all contribute to discovery, background knowledge, and saturation.

It recently started working. It's improving very quickly. The results look very cool. And it's mostly understandable how it works to nonexperts.

It might become the future of computer graphics for photorealistic rendering. Or it might not.

But it's advancing very quickly, and it definitely merits further study. People are interested to see how far we can take it.

There isn't that much genuine innovation going on in our industry. This is one of the few cases. No wonder people are interested.