Post #4141363

@lagrangeinterpolator@awful.systems @flaviat@awful.systems As someone working in computational physics, this is the current trend and it is kindof depressing. We see a lot of mid/shit tier papers focused on developing new LLM based database analyses for materials discovery and they all seem to actually suck. More generally in my area of expertise people use Deep Neural Netwoks all around with bazillion parameters, and since all physicists know that "you can fit any data with a high enough polynomial", this is exactly what we were taught *not* to do. I really wonder how people will look at this in the future because this is all I don't like in science and will make me want to quit if it goes on for too long. For now I still trust that the bubble collapse will make it stop at some point.