Comment by arduanika

It's charitable of you to think that they might be qualified to talk about inference and model weights and such. They are AI safety researchers, not AI researchers. Basically, a bunch of doom bloggers, jerking each other in a circle, a few of whom were tolerated at one of the major labs for a few years, to do their jerking on company time.