Gibbs free energy and motivation

[sikefield3]

Thanks for opening the discussions, Linas !

From what I remember (or more accurately what I have forgotten) from thermodynamics, you have the relation:
dG = dH - T dS
with Gibbs energy G and enthalpy H, and the 2nd law of thermodynamics dG <= 0 and by dabbling with the enthalpy H you can decrease entropy.
Can you tell me a bit how this can be used for the motivation problem ?
I've found this paper Thermodynamic AI and the fluctuation frontier
and I think Marcus Hutter also uses entropy in his AIXI (but I only had a quick glance and it might be something different.)

[linas]

So most of the overall theory I've worked on is described in the README to https://github.com/opencog/learn and the main PDF's that it points too (some of which are quite long.)

I don't have any one particular "favorite theory of thermodynamics". There's vast quantities of ink that have been spilled on the topic of AI and thermo over the last decade or two -- books have been written on it. Mostly, I've been exploring the general question of "what is a structure?" and "how do structures arise in nature?" and "how do structures perceive, interact with, and act on their environment?"

A "structure" can be a bit of knowledge, or it can be an agent: at any rate, it is something that has complexity to it, and can be described algorithmically, so as to run on a computer.

These are very abstract questions; I've pondered them for decades, I have no easy or simple answers: rather, its the exploration and discovery that drives the work here. This might not be the answer you're looking for. The reality is that most of my work on the atomspace/opencog has been about finding answers to these questions.