Hi, I am a first poster and honestly was unaware of this community and am quite glad something like this exists as critical thinking is already rare as is, and assessments on the impact of AI on it at large worry me for the future of society.
I am Mec. Engineer by background and well stablished in my field for a bit over a decade now, but given that I have lived in different societal trust realities between a very low-trust city (in Latin America) and seen what a high-trust society can look like (e.g. Tokyo) this sparked me to constantly ask what is the difference, what is the reasoning behind life that could more accurately not fall to blatant exceptions like our current life preservation and survival of the fittest main pillars for morality.
Knowing I going out of my domain, I approached this from a purely engineering lens and arrived at some interesting observations of logical nature, which I am compiling in what I call the “Entropic Ultimatum”, which the TL;DR version is how to see the natural persistence of resiliency in nature and model for the best alignment with it while remaining substrate independent across scales of complexity, yes, including AI. Thus providing a meta-framework that could be used for AI Alignment and societal alignment alike post translations / projections of the model to make it measurable.
I already have a preprint on Zenodo and am more than welcome to constructive criticism or flat out invalidation of the theory if this is not something worthwhile pursuing. My intention here is to provide good grammar that can hopefully approach us to something closer to a harmony that does not take personal freedom away, while maximizing our survival at most if not all domains, realigning our sensory system that is currently living in tension with technological “hacks” (such as the very nice post regarding fitness-maximizers x adaptation-executioners in the new user guide (chocolate as a product design to profit based on evolutionary sensory work that was persisted to favor a very different ambient selection pressure)
I searched for “thermodynamics” and did not find much equivalent to this in the forum, but if there is a work that is similar in nature I am more than happy to continue this discussion elsewhere or in a new post if recommended to do so.
For a quick secondary example, if you understand more or less what the framework tries to define with its variables, ask yourself the following two questions to showcase its prediction power: 1 - If a feral cat that was recently adopted by a person is fed a bowl of food. Why does it hiss and scratches the person if they try to take the bowl away? What clauses between the Control Volumes (CVs) (Person and Cat) this seems to contradict? 2 - What does it take for the cat to no longer apply “penalties” to the person? And what effects this has to the combined predictive horizon H and metabolic rate represented here by the Ssys of the person, cat and Domain of both together? Therefore, what impact it has for the Encap (Omega) of each?
The two easier compressions of the system are either in financial terms (swap entropy for currency) or food-based terms (swap entropy for food access). They are not perfect but much easier to operationalize than Entropy alone, but losing rigor and risking justification of non-alignment fringe cases that would not hold under Entropy, and losing aplicability as a cell knows no value in currency and a plant’s food is questionable as category error of the light it absorbs / CO2 and nutrients. But Entropy encompasses all this sub-mode compressions.
Hi, I am a first poster and honestly was unaware of this community and am quite glad something like this exists as critical thinking is already rare as is, and assessments on the impact of AI on it at large worry me for the future of society.
I am Mec. Engineer by background and well stablished in my field for a bit over a decade now, but given that I have lived in different societal trust realities between a very low-trust city (in Latin America) and seen what a high-trust society can look like (e.g. Tokyo) this sparked me to constantly ask what is the difference, what is the reasoning behind life that could more accurately not fall to blatant exceptions like our current life preservation and survival of the fittest main pillars for morality.
Knowing I going out of my domain, I approached this from a purely engineering lens and arrived at some interesting observations of logical nature, which I am compiling in what I call the “Entropic Ultimatum”, which the TL;DR version is how to see the natural persistence of resiliency in nature and model for the best alignment with it while remaining substrate independent across scales of complexity, yes, including AI. Thus providing a meta-framework that could be used for AI Alignment and societal alignment alike post translations / projections of the model to make it measurable.
I already have a preprint on Zenodo and am more than welcome to constructive criticism or flat out invalidation of the theory if this is not something worthwhile pursuing. My intention here is to provide good grammar that can hopefully approach us to something closer to a harmony that does not take personal freedom away, while maximizing our survival at most if not all domains, realigning our sensory system that is currently living in tension with technological “hacks” (such as the very nice post regarding fitness-maximizers x adaptation-executioners in the new user guide (chocolate as a product design to profit based on evolutionary sensory work that was persisted to favor a very different ambient selection pressure)
The article can be found in this particular link:
Entropic Ultimatum: A thermodynamic framework for evaluating system resiliency
I searched for “thermodynamics” and did not find much equivalent to this in the forum, but if there is a work that is similar in nature I am more than happy to continue this discussion elsewhere or in a new post if recommended to do so.
For a quick secondary example, if you understand more or less what the framework tries to define with its variables, ask yourself the following two questions to showcase its prediction power:
1 - If a feral cat that was recently adopted by a person is fed a bowl of food. Why does it hiss and scratches the person if they try to take the bowl away? What clauses between the Control Volumes (CVs) (Person and Cat) this seems to contradict?
2 - What does it take for the cat to no longer apply “penalties” to the person? And what effects this has to the combined predictive horizon H and metabolic rate represented here by the Ssys of the person, cat and Domain of both together? Therefore, what impact it has for the Encap (Omega) of each?
The two easier compressions of the system are either in financial terms (swap entropy for currency) or food-based terms (swap entropy for food access). They are not perfect but much easier to operationalize than Entropy alone, but losing rigor and risking justification of non-alignment fringe cases that would not hold under Entropy, and losing aplicability as a cell knows no value in currency and a plant’s food is questionable as category error of the light it absorbs / CO2 and nutrients. But Entropy encompasses all this sub-mode compressions.