Well, the best thing that I’ve got going for me right now is that the offerings on Netflix are so d*** boring that doing research is MUCH more interesting.
I’ve read and re-read all my light, fun reading around the house … about three to four times each, over these past two years. My fave authors are working hard, and their next releases are all about March. One just announced a three-month delay on her latest … just plain COVID-fatigue. Another has struggled on gamely, producing works as fast as her own COVID-fatigue will allow – and her latest has a five-month delay on getting into paper format; paper shortages. (COVID-induced.)
So … with distractions removed, it is SO much easier to get focused on research, which is what I’ve been doing for the past week or so. (Modulo life-maintenance.)
This blogpost is all about process. The actual process – just as an inventor/ scientist (or any creative person – an artist, playwright, composer, or author) gets really immersed in the creative project, and gets VERY close to having the central theme or core emerge.
It is NOT always as pretty as the final results would lead us to imagine.
More Christmas in Hawai’i
But first, a picture …
The Invention Process
I sometimes wonder what it was really like when a number of our major AI inventions got hashed out, and brought from initial scrambles and wild ideas expressed on the white board to the final polish of a published journal paper.
Word to my students (and to my creative colleagues): There are many, MANY steps involved. Most of them involve editing. Most involve banging your work against the tough-reality walls of your colleagues’ scrutiny. Not always fun. Absolutely essential.
So here’s the essence.
I’ve been working on the 2-D cluster variation method. In particular, I’m (finally) at the stage of taking a natural topography, creating a 2-D grid representation of it, and asking two hugely-important question:
- Is this topography at all near some sort of free energy equilibrium, and
- What are the enthalpy parameters of the “closest” free energy equilibrium system?
To be continued … (This Christmas season is all about catching up on life-maintenance chores, and it’s time to go feed feral cats, get a quick coffee-shop breakfast, and head towards Kona to get two car tires replaced and the alignment checked. It’s … just life.)
All my best, dears. I’ll be back w/ you soon, because the INTERESTING thing about asking those two research questions is just ahead.
Live free or die, my friend. (Attrib. Revolutionary War General John Stark.)
Alianna J. Maren, Ph.D.
Founder & Chief Scientist, Themesis, Inc.
P.S. Feral cats. Life is much better with breakfast.
Cats need to be fed pre-dawn because – hey – everyone else wants breakfast. Chickens. Mongooses. (Yes, that’s the correct plural for a mongoose.) Just life on the Big Island.
P.P.S. “Natural topographies.” Does anyone really care?
Honestly, I don’t. Not that much. There are some that do, and my guess is they are a VERY small sub-sub-fraction of humanity.
BUT … (this is the BIG “BUT”) … a lot of us care about the next steps in AI. I particularly care about variational Bayes and active inference, and will start talking/writing about them soon. And as far as I can tell, everyone is bored to tears with using a set of Gaussians as the “best” model available.
Not only that, a set of Gaussians might not be the right model for some very interesting cases.
Maybe we want a topographic model … (lead-in to the next blog post).