No. I think the need for adversarial losses in order to distill diffusion models into one-step forward passes has provided additional evidence that GANs were much more viable than diffusimaxis loudly insisted.

(Although I'm not really current on where image generation is these days or who is using GAN-like approaches under the hood or what are the current theoretical understandings of GAN vs AR vs diffusion, so if you have some specific reason I should have "caved", feel free to mention it - I may well just be unaware of it.)

Given it's an old article I'm sure someone else has pointed this out by now, but the article mixes up RSA modulus size expressed in base-10 (as in RSA-260) and base-2 (as in RSA-768 - The RSA challenge moduli are themselves inconsistently named)

The interesting thing about DMT is that it doesn't really seem to affect your mind/cognition (that is, if you can observe all the weirdness happening in a very detached way instead of being overwhelmed by it, it is an exceptionally clear-headed experience). Not exactly the same but I've drank ayahuasca hundreds of times (worked at retreat centers/apprenticed) and it's amazing how much work you can do on a light dose vs being blasted to the center of the universe (though, that is good once in a while to get over blockages). It can be hard to get a footing with dmt because it's so fast (and it's really hard to land in the "workable" zone compared to drinking ayahuasca).

"SotA diffusion uses adversarial methods anyways" seems like a bit of a departure from the case you make in the blog post.

edit: For what it's worth - I agree. At least some auto-encoders (which will produce latents for diffusion models) use some form of adversarial method.

Still, I'm curious if you think GAN models in their more familiar form are going to eventually take on LCM/diffusion models?

Silence says it all.