Unfortunately with the way scaling laws are working out, each order of magnitude increase in computer only makes models a little better.

Meaning they nobody will even bother to 10,000X GPT4.