Tom Lieberum comments on Which singularity schools plus the no singularity school was right?

[Tom Lieberum]

This assumes a fixed scaling law. One possible way of improving oneself could be to design a better architecture with a better scaling exponent.