User blog comment:Nnn6nnn/Where Will Googology Go?/@comment-25337554-20180213032340/@comment-30754445-20180213091140

I'm afraid that in this day and age, "creating new googolisms" is the least interesting part of googology.

In the early days (say 10-15 years ago), where the large numbers realm was an unexplored frontier, it was a great challange. But today, any notation you might come up with is going to be an FGH-clone. All recursive notations are basically isomorphic to one another: You can easily write a computer program (with no AI what-so-ever) that converts Bower Arrays to Saibian Hyper E's to any of the myriad other notations that currently exist.

So while it is no doubt possible to program an AI that creates another FGH-clone with a more-or-less original style, I fail to see the point in it. Unless we could make an AI that actually enjoys the creation of these whimsical notations and names (which is the chief reason that human googolists continue to create new googolisms) :-)

TL;DR I too would love to see an AI googologist one day, but I also hope it will be do more  interesting things than merely "creating new googolisms".