I mean, maybe? Certainly scaling up compute will make better text predictors, I am entirely unconvinced that that's all which is needed, but eventually I'll be wrong about the hype I guess.