Just one more doubling of parameters and AGI is here. Just ignore that fact that the training data approaches all released information available and then some and that still isn't enough to teach it robust reasoning.