I have to wonder how much improvement you would get with a 100 trillion parameter model. There seems to be diminishing returns in model size. That effort coul

I have to wonder how much improvement you would get with a 100 trillion parameter model. There seems to be diminishing returns in model size. That effort could almost certainly be better spent.