HN Theater @HNTheaterMonth

Is it related to Tom's video?

"I asked an AI for video ideas, and they were actually good" https://www.youtube.com/watch?v=TfVYxnhuEdU

⬐ gdubs

This was an incredibly engaging video. I’ve said this before but if we are indeed at the bottom of the sigmoid curve, than the Star Trek holodeck seems more and more plausible [1]. Not the materialization part - that I’m aware of. But a machine that can generate a realistic story, improvise, dream up characters, and render them on the fly.

What seemed implausible in the 90s was the amount of 3D modeling that would be required. But now you can imagine a GAN that can generate 3D character animations from thin air, like StyleGAN does currently for faces. And another neural net that generates the narrative, and so on.

Nets within nets within nets.

Edit: 1: Actually, I’m reminded of something my college computer graphics professor used to say: good SciFi doesn’t have to seem possible — it just has to seem plausible. By that measure — Star Trek was good sciFi IMO — it would be more correct to say that it didn’t seem possible, or even probable, but it probably did seem plausible.

⬐ dragonwriter

> What seemed implausible in the 90s was the amount of 3D modeling that would be required.

No, what seemed implausible in the 1990s was:

* projection holography

* materialization

* kinetic feedback to fake moving through larger spaces

The 3D modelling was just “more computing power” and we hadn't even started bumping up hard against limits on Moore's law at the time, so that seemed easy.

⬐ gdubs

Rendering a complex 3D scene? Easy to imagine. But a 3D scene that the computer could generate with seemingly infinite variations? Given how manual the process of 3D animation was, that didn’t seem plausible*. After all — it’s only been in the past few years that GANs have started generating realistic scenes.

1: possible, not plausible (see my edit above)